Update HISTORY and version

Fix a race condition in FIFO size-based compaction where concurrent threads could select the same non-L0 file (#13946 )
Summary: **Context/Summary:** Fix a race condition (illustrated below) in FIFO size-based compaction where concurrent threads could select the same non-L0 file, causing assertion failures in debug builds or "Cannot delete table file from LSM tree" errors in release builds. ``` Thread 1 Thread 2 -------- -------- FIFO size-based compaction ↓ Pick L2 file ↓ Mark: file.being_compacted = true (file.being_compacted was false) ↓ WriteManifestStart (unlock mutex) ─→ FIFO size-based compaction starts ↓ ↓ Continue manifest write... Pick SAME L2 file ↓ Mark: file.being_compacted = true (file.being_compacted was true) ❌ ↓ ↓ ↓ Unlock mutex, wait for manifest ↓ ↓ Lock mutex ←─────────────────────────────────┘ ↓ Delete L2 file ✅ ↓ Complete ─────────────────────────────→ Try delete same file ❌ ↓ ERROR: "file not in LSM tree" 🐛 BUG: Both threads pick the same file! Thread 2 doesn't properly check file.being_compacted flag ``` **Test** New test that fails before the fix and passes after Pull Request resolved: https://github.com/facebook/rocksdb/pull/13946 Reviewed By: xingbowang Differential Revision: D82279731 Pulled By: hx235 fbshipit-source-id: b426517f2d1b23dd7d4951157822a2d322fe1435
2026-07-07 14:47:40 +08:00 · 2025-09-15 14:24:48 -07:00 · 2025-09-15 14:07:59 -07:00 · 2025-09-06 18:09:37 -07:00 · 2025-09-06 18:08:38 -07:00 · 2025-09-06 18:08:25 -07:00
364 changed files with 25000 additions and 7920 deletions
@@ -4,8 +4,8 @@ runs:
  steps:
  - name: Install Maven
    run: |
-      wget --no-check-certificate https://dlcdn.apache.org/maven/maven-3/3.9.6/binaries/apache-maven-3.9.6-bin.tar.gz
-      tar zxf apache-maven-3.9.6-bin.tar.gz
-      echo "export M2_HOME=$(pwd)/apache-maven-3.9.6" >> $GITHUB_ENV
-      echo "$(pwd)/apache-maven-3.9.6/bin" >> $GITHUB_PATH
+      wget --no-check-certificate https://archive.apache.org/dist/maven/maven-3/3.9.11/binaries/apache-maven-3.9.11-bin.tar.gz
+      tar zxf apache-maven-3.9.11-bin.tar.gz
+      echo "export M2_HOME=$(pwd)/apache-maven-3.9.11" >> $GITHUB_ENV
+      echo "$(pwd)/apache-maven-3.9.11/bin" >> $GITHUB_PATH
    shell: bash
@@ -11,9 +11,9 @@ runs:
      CMAKE_BIN: C:/Program Files/CMake/bin/cmake.exe
      CTEST_BIN: C:/Program Files/CMake/bin/ctest.exe
      JAVA_HOME: C:/Program Files/BellSoft/LibericaJDK-8
-      SNAPPY_HOME: ${{ github.workspace }}/thirdparty/snappy-1.1.8
-      SNAPPY_INCLUDE: ${{ github.workspace }}/thirdparty/snappy-1.1.8;${{ github.workspace }}/thirdparty/snappy-1.1.8/build
-      SNAPPY_LIB_DEBUG: ${{ github.workspace }}/thirdparty/snappy-1.1.8/build/Debug/snappy.lib
+      SNAPPY_HOME: ${{ github.workspace }}/thirdparty/snappy-1.2.2
+      SNAPPY_INCLUDE: ${{ github.workspace }}/thirdparty/snappy-1.2.2;${{ github.workspace }}/thirdparty/snappy-1.2.2/build
+      SNAPPY_LIB_DEBUG: ${{ github.workspace }}/thirdparty/snappy-1.2.2/build/Debug/snappy.lib
    run: |-
      # NOTE: if ... Exit $LASTEXITCODE lines needed to exit and report failure
      echo ===================== Install Dependencies =====================
@@ -22,14 +22,14 @@ runs:
      mkdir $Env:THIRDPARTY_HOME
      cd $Env:THIRDPARTY_HOME
      echo "Building Snappy dependency..."
-      curl -Lo snappy-1.1.8.zip https://github.com/google/snappy/archive/refs/tags/1.1.8.zip
+      curl -Lo snappy-1.2.2.zip https://github.com/google/snappy/archive/refs/tags/1.2.2.zip
      if(!$?) { Exit $LASTEXITCODE }
-      unzip -q snappy-1.1.8.zip
+      unzip -q snappy-1.2.2.zip
      if(!$?) { Exit $LASTEXITCODE }
-      cd snappy-1.1.8
+      cd snappy-1.2.2
      mkdir build
      cd build
-      & cmake -G "$Env:CMAKE_GENERATOR" ..
+      & cmake -G "$Env:CMAKE_GENERATOR" .. -DSNAPPY_BUILD_TESTS=OFF -DSNAPPY_BUILD_BENCHMARKS=OFF
      if(!$?) { Exit $LASTEXITCODE }
      msbuild Snappy.sln -maxCpuCount -property:Configuration=Debug -property:Platform=x64
      if(!$?) { Exit $LASTEXITCODE }
@@ -38,7 +38,7 @@ runs:
      $env:Path = $env:JAVA_HOME + ";" + $env:Path
      mkdir build
      cd build
-      & cmake -G "$Env:CMAKE_GENERATOR" -DCMAKE_BUILD_TYPE=Debug -DOPTDBG=1 -DPORTABLE="$Env:CMAKE_PORTABLE" -DSNAPPY=1 -DJNI=1 ..
+      & cmake -G "$Env:CMAKE_GENERATOR" -DCMAKE_BUILD_TYPE=Debug -DOPTDBG=1 -DPORTABLE="$Env:CMAKE_PORTABLE" -DSNAPPY=1 -DXPRESS=1 -DJNI=1 ..
      if(!$?) { Exit $LASTEXITCODE }
      cd ..
      echo "Building with VS version: $Env:CMAKE_GENERATOR"
@@ -27,18 +27,6 @@ jobs:
        git config --global --add safe.directory /__w/rocksdb/rocksdb
        tools/check_format_compatible.sh
    - uses: "./.github/actions/post-steps"
-  build-linux-run-microbench:
-    if: ${{ github.repository_owner == 'facebook' }}
-    runs-on:
-      labels: 16-core-ubuntu
-    container:
-      image: zjay437/rocksdb:0.6
-      options: --shm-size=16gb
-    steps:
-    - uses: actions/checkout@v4.1.0
-    - uses: "./.github/actions/pre-steps"
-    - run: DEBUG_LEVEL=0 make -j32 run_microbench
-    - uses: "./.github/actions/post-steps"
  build-linux-non-shm:
    if: ${{ github.repository_owner == 'facebook' }}
    runs-on:
@@ -91,15 +79,6 @@ jobs:
    steps:
    - uses: actions/checkout@v4.1.0
    - uses: "./.github/actions/windows-build-steps"
-  build-windows-vs2022:
-    if: ${{ github.repository_owner == 'facebook' }}
-    runs-on: windows-2022
-    env:
-      CMAKE_GENERATOR: Visual Studio 17 2022
-      CMAKE_PORTABLE: 1
-    steps:
-    - uses: actions/checkout@v4.1.0
-    - uses: "./.github/actions/windows-build-steps"
  build-linux-arm-test-full:
    if: ${{ github.repository_owner == 'facebook' }}
    runs-on:
@@ -110,3 +89,59 @@ jobs:
      - run: sudo apt-get update && sudo apt-get install -y build-essential libgflags-dev
      - run: make V=1 J=4 -j4 check
      - uses: "./.github/actions/post-steps"
+  build-examples:
+    if: ${{ github.repository_owner == 'facebook' }}
+    runs-on:
+      labels: 4-core-ubuntu
+    container:
+      image: zjay437/rocksdb:0.6
+      options: --shm-size=16gb
+    steps:
+    - uses: actions/checkout@v4.1.0
+    - uses: "./.github/actions/pre-steps"
+    - name: Build examples
+      run: make V=1 -j4 static_lib && cd examples && make V=1 -j4
+    - uses: "./.github/actions/post-steps"
+  build-fuzzers:
+    if: ${{ github.repository_owner == 'facebook' }}
+    runs-on:
+      labels: 4-core-ubuntu
+    container:
+      image: zjay437/rocksdb:0.6
+      options: --shm-size=16gb
+    steps:
+    - uses: actions/checkout@v4.1.0
+    - uses: "./.github/actions/pre-steps"
+    - name: Build rocksdb lib
+      run: CC=clang-13 CXX=clang++-13 USE_CLANG=1 make -j4 static_lib
+    - name: Build fuzzers
+      run: cd fuzz && make sst_file_writer_fuzzer db_fuzzer db_map_fuzzer
+    - uses: "./.github/actions/post-steps"
+  build-linux-gcc-11-no_test_run:
+    if: ${{ github.repository_owner == 'facebook' }}
+    runs-on:
+      labels: 16-core-ubuntu
+    container:
+      image: zjay437/rocksdb:0.6
+      options: --shm-size=16gb
+    steps:
+    - uses: actions/checkout@v4.1.0
+    - uses: "./.github/actions/pre-steps"
+    - run: LIB_MODE=static CC=gcc-11 CXX=g++-11 V=1 make -j32 all microbench
+    - uses: "./.github/actions/post-steps"
+  build-linux-cmake-with-folly-lite-no-test:
+    if: ${{ github.repository_owner == 'facebook' }}
+    runs-on:
+      labels: 16-core-ubuntu
+    container:
+      image: zjay437/rocksdb:0.6
+      options: --shm-size=16gb
+    env:
+      CC: gcc-10
+      CXX: g++-10
+    steps:
+    - uses: actions/checkout@v4.1.0
+    - uses: "./.github/actions/pre-steps"
+    - uses: "./.github/actions/setup-folly"
+    - run: "(mkdir build && cd build && cmake -DUSE_FOLLY_LITE=1 -DWITH_GFLAGS=1 .. && make V=1 -j20)"
+    - uses: "./.github/actions/post-steps"
@@ -52,6 +52,14 @@ jobs:
      run: make check-buck-targets
    - name: Simple source code checks
      run: make check-sources
+    - name: Sanity check check_format_compatible.sh
+      run: |-
+        export TEST_TMPDIR=/dev/shm/rocksdb
+        rm -rf /dev/shm/rocksdb
+        mkdir /dev/shm/rocksdb
+        git reset --hard
+        git config --global --add safe.directory /__w/rocksdb/rocksdb
+        SANITY_CHECK=1 LONG_TEST=1 tools/check_format_compatible.sh
  # ========================= Linux With Tests ======================== #
  build-linux:
    if: ${{ github.repository_owner == 'facebook' }}
@@ -101,22 +109,6 @@ jobs:
    - uses: "./.github/actions/build-folly"
    - run: "(mkdir build && cd build && cmake -DUSE_FOLLY=1 -DWITH_GFLAGS=1 -DROCKSDB_BUILD_SHARED=0 .. && make V=1 -j20 && ctest -j20)"
    - uses: "./.github/actions/post-steps"
-  build-linux-cmake-with-folly-lite-no-test:
-    if: ${{ github.repository_owner == 'facebook' }}
-    runs-on:
-      labels: 16-core-ubuntu
-    container:
-      image: zjay437/rocksdb:0.6
-      options: --shm-size=16gb
-    env:
-      CC: gcc-10
-      CXX: g++-10
-    steps:
-    - uses: actions/checkout@v4.1.0
-    - uses: "./.github/actions/pre-steps"
-    - uses: "./.github/actions/setup-folly"
-    - run: "(mkdir build && cd build && cmake -DUSE_FOLLY_LITE=1 -DWITH_GFLAGS=1 .. && make V=1 -j20)"
-    - uses: "./.github/actions/post-steps"
  build-linux-make-with-folly:
    if: ${{ github.repository_owner == 'facebook' }}
    runs-on:
@@ -190,7 +182,7 @@ jobs:
    - uses: actions/checkout@v4.1.0
    - uses: "./.github/actions/pre-steps"
    - run: ENCRYPTED_ENV=1 ROCKSDB_DISABLE_SNAPPY=1 ROCKSDB_DISABLE_ZLIB=1 ROCKSDB_DISABLE_BZIP=1 ROCKSDB_DISABLE_LZ4=1 ROCKSDB_DISABLE_ZSTD=1 make V=1 J=32 -j32 check
-    - run: "./sst_dump --help | grep -E -q 'Supported compression types: kNoCompression$' # Verify no compiled in compression\n"
+    - run: "./sst_dump --help | grep -E -q 'Supported built-in compression types: kNoCompression$' # Verify no compiled in compression\n"
    - uses: "./.github/actions/post-steps"
  # ======================== Linux No Test Runs ======================= #
  build-linux-release:
@@ -204,63 +196,20 @@ jobs:
    - uses: actions/checkout@v4.1.0
    - run: make V=1 -j32 LIB_MODE=shared release
    - run: ls librocksdb.so
-    - run: "./db_stress --version"
+    - run: "./trace_analyzer --version" # A tool dependent on gflags that can run in release build
    - run: make clean
-    - run: make V=1 -j32 release
+    - run: USE_RTTI=1 make V=1 -j32 release
    - run: ls librocksdb.a
-    - run: "./db_stress --version"
+    - run: "./trace_analyzer --version"
    - run: make clean
    - run: apt-get remove -y libgflags-dev
    - run: make V=1 -j32 LIB_MODE=shared release
    - run: ls librocksdb.so
-    - run: if ./db_stress --version; then false; else true; fi
+    - run: if ./trace_analyzer --version; then false; else true; fi
    - run: make clean
-    - run: make V=1 -j32 release
+    - run: USE_RTTI=1 make V=1 -j32 release
    - run: ls librocksdb.a
-    - run: if ./db_stress --version; then false; else true; fi
-    - uses: "./.github/actions/post-steps"
-  build-linux-release-rtti:
-    if: ${{ github.repository_owner == 'facebook' }}
-    runs-on:
-      labels: 8-core-ubuntu
-    container:
-      image: zjay437/rocksdb:0.6
-      options: --shm-size=16gb
-    steps:
-    - uses: actions/checkout@v4.1.0
-    - run: USE_RTTI=1 DEBUG_LEVEL=0 make V=1 -j16 static_lib tools db_bench
-    - run: "./db_stress --version"
-    - run: make clean
-    - run: apt-get remove -y libgflags-dev
-    - run: USE_RTTI=1 DEBUG_LEVEL=0 make V=1 -j16 static_lib tools db_bench
-    - run: if ./db_stress --version; then false; else true; fi
-  build-examples:
-    if: ${{ github.repository_owner == 'facebook' }}
-    runs-on:
-      labels: 4-core-ubuntu
-    container:
-      image: zjay437/rocksdb:0.6
-      options: --shm-size=16gb
-    steps:
-    - uses: actions/checkout@v4.1.0
-    - uses: "./.github/actions/pre-steps"
-    - name: Build examples
-      run: make V=1 -j4 static_lib && cd examples && make V=1 -j4
-    - uses: "./.github/actions/post-steps"
-  build-fuzzers:
-    if: ${{ github.repository_owner == 'facebook' }}
-    runs-on:
-      labels: 4-core-ubuntu
-    container:
-      image: zjay437/rocksdb:0.6
-      options: --shm-size=16gb
-    steps:
-    - uses: actions/checkout@v4.1.0
-    - uses: "./.github/actions/pre-steps"
-    - name: Build rocksdb lib
-      run: CC=clang-13 CXX=clang++-13 USE_CLANG=1 make -j4 static_lib
-    - name: Build fuzzers
-      run: cd fuzz && make sst_file_writer_fuzzer db_fuzzer db_map_fuzzer
+    - run: if ./trace_analyzer --version; then false; else true; fi
    - uses: "./.github/actions/post-steps"
  build-linux-clang-no_test_run:
    if: ${{ github.repository_owner == 'facebook' }}
@@ -284,6 +233,8 @@ jobs:
    - uses: actions/checkout@v4.1.0
    - uses: "./.github/actions/pre-steps"
    - run: CC=clang-13 CXX=clang++-13 USE_CLANG=1 make -j32 all microbench
+    - run: make clean
+    - run: CC=clang-13 CXX=clang++-13 USE_CLANG=1 DEBUG_LEVEL=0 make -j32 release
    - uses: "./.github/actions/post-steps"
  build-linux-gcc-8-no_test_run:
    if: ${{ github.repository_owner == 'facebook' }}
@@ -309,18 +260,7 @@ jobs:
    - uses: "./.github/actions/pre-steps"
    - run: CC=gcc-10 CXX=g++-10 V=1 ROCKSDB_CXX_STANDARD=c++20 make -j32 all
    - uses: "./.github/actions/post-steps"
-  build-linux-gcc-11-no_test_run:
-    if: ${{ github.repository_owner == 'facebook' }}
-    runs-on:
-      labels: 16-core-ubuntu
-    container:
-      image: zjay437/rocksdb:0.6
-      options: --shm-size=16gb
-    steps:
-    - uses: actions/checkout@v4.1.0
-    - uses: "./.github/actions/pre-steps"
-    - run: LIB_MODE=static CC=gcc-11 CXX=g++-11 V=1 make -j32 all microbench
-    - uses: "./.github/actions/post-steps"
+
  # ======================== Linux Other Checks ======================= #
  build-linux-clang10-clang-analyze:
    if: ${{ github.repository_owner == 'facebook' }}
@@ -368,7 +308,7 @@ jobs:
    - run: ulimit -S -n `ulimit -H -n` && make V=1 -j8 CRASH_TEST_EXT_ARGS='--duration=960 --max_key=2500000' blackbox_crash_test_with_atomic_flush
    - uses: "./.github/actions/post-steps"
  # ======================= Linux with Sanitizers ===================== #
-  build-linux-clang10-asan:
+  build-linux-clang10-asan-ubsan:
    if: ${{ github.repository_owner == 'facebook' }}
    runs-on:
      labels: 32-core-ubuntu
@@ -378,19 +318,7 @@ jobs:
    steps:
    - uses: actions/checkout@v4.1.0
    - uses: "./.github/actions/pre-steps"
-    - run: COMPILE_WITH_ASAN=1 CC=clang-10 CXX=clang++-10 ROCKSDB_DISABLE_ALIGNED_NEW=1 USE_CLANG=1 make V=1 -j32 check
-    - uses: "./.github/actions/post-steps"
-  build-linux-clang10-ubsan:
-    if: ${{ github.repository_owner == 'facebook' }}
-    runs-on:
-      labels: 16-core-ubuntu
-    container:
-      image: zjay437/rocksdb:0.6
-      options: --shm-size=16gb
-    steps:
-    - uses: actions/checkout@v4.1.0
-    - uses: "./.github/actions/pre-steps"
-    - run: COMPILE_WITH_UBSAN=1 CC=clang-10 CXX=clang++-10 ROCKSDB_DISABLE_ALIGNED_NEW=1 USE_CLANG=1 make V=1 -j32 ubsan_check
+    - run: COMPILE_WITH_ASAN=1 COMPILE_WITH_UBSAN=1 CC=clang-10 CXX=clang++-10 ROCKSDB_DISABLE_ALIGNED_NEW=1 USE_CLANG=1 make V=1 -j40 check
    - uses: "./.github/actions/post-steps"
  build-linux-clang13-mini-tsan:
    if: ${{ github.repository_owner == 'facebook' }}
@@ -461,11 +389,11 @@ jobs:
    - uses: "./.github/actions/post-steps"
  # ======================== Windows with Tests ======================= #
  # NOTE: some windows jobs are in "nightly" to save resources
-  build-windows-vs2019:
+  build-windows-vs2022:
    if: ${{ github.repository_owner == 'facebook' }}
-    runs-on: windows-2019
+    runs-on: windows-2022
    env:
-      CMAKE_GENERATOR: Visual Studio 16 2019
+      CMAKE_GENERATOR: Visual Studio 17 2022
      CMAKE_PORTABLE: 1
    steps:
    - uses: actions/checkout@v4.1.0
@@ -476,7 +404,7 @@ jobs:
    runs-on:
      labels: 4-core-ubuntu
    container:
-      image: evolvedbinary/rocksjava:centos6_x64-be
+      image: evolvedbinary/rocksjava:centos7_x64-be
      options: --shm-size=16gb
    steps:
    # The docker image is intentionally based on an OS that has an older GLIBC version.
@@ -504,7 +432,7 @@ jobs:
    runs-on:
      labels: 4-core-ubuntu
    container:
-      image: evolvedbinary/rocksjava:centos6_x64-be
+      image: evolvedbinary/rocksjava:centos7_x64-be
      options: --shm-size=16gb
    steps:
    # The docker image is intentionally based on an OS that has an older GLIBC version.
@@ -599,7 +527,7 @@ jobs:
    runs-on:
      labels: 4-core-ubuntu
    container:
-      image: evolvedbinary/rocksjava:rockylinux8_x64-be
+      image: evolvedbinary/rocksjava:alpine3_x64-be
      options: --shm-size=16gb
    steps:
    - uses: actions/checkout@v4.1.0
@@ -88,6 +88,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
        "db/memtable_list.cc",
        "db/merge_helper.cc",
        "db/merge_operator.cc",
+        "db/multi_scan.cc",
        "db/output_validator.cc",
        "db/periodic_task_scheduler.cc",
        "db/range_del_aggregator.cc",
@@ -113,6 +114,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
        "db/write_controller.cc",
        "db/write_stall_stats.cc",
        "db/write_thread.cc",
+        "db_stress_tool/db_stress_compression_manager.cc",
        "env/composite_env.cc",
        "env/env.cc",
        "env/env_chroot.cc",
@@ -249,6 +251,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
        "trace_replay/trace_record_result.cc",
        "trace_replay/trace_replay.cc",
        "util/async_file_reader.cc",
+        "util/auto_tune_compressor.cc",
        "util/build_version.cc",
        "util/cleanable.cc",
        "util/coding.cc",
@@ -267,6 +270,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
        "util/random.cc",
        "util/rate_limiter.cc",
        "util/ribbon_config.cc",
+        "util/simple_mixed_compressor.cc",
        "util/slice.cc",
        "util/status.cc",
        "util/stderr_logger.cc",
@@ -419,6 +423,7 @@ rocks_cpp_library_wrapper(name="rocksdb_stress_lib", srcs=[
        "db_stress_tool/batched_ops_stress.cc",
        "db_stress_tool/cf_consistency_stress.cc",
        "db_stress_tool/db_stress_common.cc",
+        "db_stress_tool/db_stress_compression_manager.cc",
        "db_stress_tool/db_stress_driver.cc",
        "db_stress_tool/db_stress_filters.cc",
        "db_stress_tool/db_stress_gflags.cc",
@@ -4709,6 +4714,12 @@ cpp_unittest_wrapper(name="compressed_secondary_cache_test",
            extra_compiler_flags=[])


+cpp_unittest_wrapper(name="compression_test",
+            srcs=["util/compression_test.cc"],
+            deps=[":rocksdb_test_lib"],
+            extra_compiler_flags=[])
+
+
 cpp_unittest_wrapper(name="configurable_test",
            srcs=["options/configurable_test.cc"],
            deps=[":rocksdb_test_lib"],
@@ -5185,6 +5196,12 @@ cpp_unittest_wrapper(name="inlineskiplist_test",
            extra_compiler_flags=[])


+cpp_unittest_wrapper(name="interval_test",
+            srcs=["util/interval_test.cc"],
+            deps=[":rocksdb_test_lib"],
+            extra_compiler_flags=[])
+
+
 cpp_unittest_wrapper(name="io_posix_test",
            srcs=["env/io_posix_test.cc"],
            deps=[":rocksdb_test_lib"],
@@ -721,6 +721,7 @@ set(SOURCES
        db/memtable_list.cc
        db/merge_helper.cc
        db/merge_operator.cc
+        db/multi_scan.cc
        db/output_validator.cc
        db/periodic_task_scheduler.cc
        db/range_del_aggregator.cc
@@ -746,6 +747,7 @@ set(SOURCES
        db/write_controller.cc
        db/write_stall_stats.cc
        db/write_thread.cc
+        db_stress_tool/db_stress_compression_manager.cc
        env/composite_env.cc
        env/env.cc
        env/env_chroot.cc
@@ -874,11 +876,13 @@ set(SOURCES
        trace_replay/trace_record.cc
        trace_replay/trace_replay.cc
        util/async_file_reader.cc
+        util/auto_tune_compressor.cc
        util/cleanable.cc
        util/coding.cc
        util/compaction_job_stats_impl.cc
        util/comparator.cc
        util/compression.cc
+        util/simple_mixed_compressor.cc
        util/compression_context_cache.cc
        util/concurrent_task_limiter_impl.cc
        util/crc32c.cc
@@ -1456,6 +1460,7 @@ if(WITH_TESTS)
        util/autovector_test.cc
        util/bloom_test.cc
        util/coding_test.cc
+        util/compression_test.cc
        util/crc32c_test.cc
        util/defer_test.cc
        util/dynamic_bloom_test.cc
@@ -1,6 +1,115 @@
 # Rocksdb Change Log
 > NOTE: Entries for next release do not go here. Follow instructions in `unreleased_history/README.txt`

+## 10.6.2 (09/15/2025)
+### Bug Fixes
+* Fix a race condition in FIFO size-based compaction where concurrent threads could select the same non-L0 file, causing assertion failures in debug builds or "Cannot delete table file from LSM tree" errors in release builds.
+
+## 10.6.1 (09/05/2025)
+### New Features
+* Add the fail_if_no_udi_on_open flag in BlockBasedTableOption to control whether a missing user defined index block in a SST is a hard error or not.
+* Add new option `MultiScanArgs::max_prefetch_size` that limits the memory usage of per file pinning of prefetched blocks.
+
+## 10.6.0 (08/22/2025)
+### New Features
+* Introduce column family option `cf_allow_ingest_behind`. This option aims to replace `DBOptions::allow_ingest_behind` to enable ingest behind at the per-CF level. `DBOptions::allow_ingest_behind` is deprecated.
+* Introduce `MultiScanArgs::io_coalesce_threshold` to allow a configurable IO coalescing threshold.
+
+### Public API Changes
+* `IngestExternalFileOptions::allow_db_generated_files` now allows files ingestion of any DB generated SST file, instead of only the ones with all keys having sequence number 0.
+* `decouple_partitioned_filters = true` is now the default in BlockBasedTableOptions.
+* GetTtl() API is now available in TTL DB
+* Minimum supported version of LZ4 library is now 1.7.0 (r129 from 2015)
+* Some changes to experimental Compressor and CompressionManager APIs
+* A new Filesystem::SyncFile function is added for syncing a file that was already written, such as on file ingestion. The default implementation matches previous RocksDB behavior: re-open the file for read-write, sync it, and close it. We recommend overriding for FileSystems that do not require syncing for crash recovery or do not handle (well) re-opening for writes.
+
+### Behavior Changes
+* When `allow_ingest_behind` is enabled, compaction will no longer drop tombstones based on the absence of underlying data. Tombstones will be preserved to apply to ingested files.
+
+### Bug Fixes
+* Files in dropped column family won't be returned to the caller upon successful, offline MANIFEST iteration in `GetFileChecksumsFromCurrentManifest`.
+* Fix a bug in MultiScan that causes it to fall back to a normal scan when dictionary compression is enabled.
+* Fix a crash in iterator Prepare() when fill_cache=false
+* Fix a bug in MultiScan where incorrect results can be returned when a Scan's range is across multiple files.
+* Fixed a bug in remote compaction that may mistakenly delete live SST file(s) during the cleanup phase when no keys survive the compaction (all expired)
+* Allow a user defined index to be configured from a string.
+* Make the User Defined Index interface consistently use the user key format, fixing the previous mixed usage of internal and user key.
+
+### Performance Improvements
+* Small improvement to CPU efficiency of compression using built-in algorithms, and a dramatic efficiency improvement for LZ4HC, based on reusing data structures between invocations.
+
+## 10.5.0 (07/18/2025)
+### Public API Changes
+* DB option skip_checking_sst_file_sizes_on_db_open is deprecated, in favor of validating file size in parallel in a thread pool, when db is opened. When DB is opened, with paranoid check enabled, a file with the wrong size would fail the DB open. With paranoid check disabled, the DB open would succeed, the column family with the corrupted file would not be read or write, while the other healthy column families could be read and write normally. When max_open_files option is not set to -1, only a subset of the files will be opened and checked. The rest of the files will be opened and checked when they are accessed.
+
+### Behavior Changes
+* PessimisticTransaction::GetWaitingTxns now returns waiting transaction information even if the current transaction has timed out. This allows the information to be surfaced to users for debugging purposes once it is known that the timeout has occured.
+* A new API GetFileSize is added to FSRandomAccessFile interface class. It uses fstat vs stat on the posix implementation which is more efficient. Caller could use it to get file size faster. This function might be required in the future for FileSystem implementation outside of the RocksDB code base.
+* RocksDB now triggers eligible compactions every 12 hours when periodic compaction is configured. This solves a limitation of the compaction trigger mechanism, which would only trigger compaction after specific events like flush, compaction, or SetOptions.
+
+### Bug Fixes
+* Fix a bug in BackupEngine that can crash backup due to a null FSWritableFile passed to WritableFileWriter.
+* Fix DB::NewMultiScan iterator to respect the scan upper bound specified in ScanOptions
+
+### Performance Improvements
+* Optimized MultiScan using BlockBasedTable to coalesce I/Os and prefetch all data blocks.
+
+## 10.4.0 (06/20/2025)
+### New Features
+* Add a new CF option `memtable_avg_op_scan_flush_trigger` that supports triggering memtable flush when an iterator scans through an expensive range of keys, with the average number of skipped keys from the active memtable exceeding the threshold.
+* Vector based memtable now supports concurrent writers (DBOptions::allow_concurrent_memtable_write) #13675.
+* Add new experimental `TransactionOptions::large_txn_commit_optimize_byte_threshold` to enable optimizations for large transaction commit by transaction batch data size.
+* Add a new option `CompactionOptionsUniversal::reduce_file_locking` and if it's true, auto universal compaction picking will adjust to minimize locking of input files when bottom priority compactions are waiting to run. This can increase the likelihood of existing L0s being selected for compaction, thereby improving write stall and reducing read regression.
+* Add new `format_version=7` to aid experimental support of custom compression algorithms with CompressionManager and block-based table. This format version includes changing the format of `TableProperties::compression_name`.
+
+### Public API Changes
+* Change NewExternalTableFactory to return a unique_ptr instead of shared_ptr.
+* Add an optional min file size requirement for deletion triggered compaction. It can be specified when creating `CompactOnDeletionCollectorFactory`.
+
+### Behavior Changes
+* `TransactionOptions::large_txn_commit_optimize_threshold` now has default value 0 for disabled. `TransactionDBOptions::txn_commit_bypass_memtable_threshold` now has no effect on transactions.
+
+### Bug Fixes
+* Fix a bug where CreateColumnFamilyWithImport() could miss the SST file for the memtable flush it triggered. The exported CF then may not contain the updates in the memtable when CreateColumnFamilyWithImport() is called.
+* Fix iterator operations returning NotImplemented status if disallow_memtable_writes and paranoid_memory_checks CF options are both set.
+* Fixed handling of file checksums in IngestExternalFile() to allow providing checksums using recognized but not necessarily the DB's preferred checksum function, to ease migration between checksum functions.
+
+## 10.3.0 (05/17/2025)
+### New Features
+* Add new experimental `CompactionOptionsFIFO::allow_trivial_copy_when_change_temperature` along with `CompactionOptionsFIFO::trivial_copy_buffer_size` to allow optimizing FIFO compactions with tiering when kChangeTemperature to move files from source tier FileSystem to another tier FileSystem via trivial and direct copying raw sst file instead of reading thru the content of the SST file then rebuilding the table files.
+* Add a new field to Compaction Stats in LOG files for the pre-compression size written to each level.
+* Add new experimental `TransactionOptions::large_txn_commit_optimize_threshold` to enable optimizations for large transaction commit with per transaction threshold. `TransactionDBOptions::txn_commit_bypass_memtable_threshold` is deprecated in favor of this transaction option.
+* [internal team use only] Allow an application-defined `request_id` to be passed to RocksDB and propagated to the filesystem via IODebugContext
+
+### Bug Fixes
+* Fix a bug where transaction lock upgrade can incorrectly fail with a Deadlock status. This happens when a transaction has a non-zero timeout and tries to upgrade a shared lock that is also held by another transaction.
+* Pass wrapped WritableFileWriter pointer to ExternalTableBuilder so that the file checksum can be correctly calculated and returned by SstFileWriter for external table files.
+* Fix an infinite-loop bug in transaction locking. This can happen if a transaction reaches lock limit and its time out expires before it attempts to wait for it.
+* Fixed a potential data race with `CompressionOptions::parallel_threads > 1` and a `TablePropertiesCollector` overriding `BlockAdd()`.
+
+## 10.2.0 (04/21/2025)
+### New Features
+* Provide histogram stats `COMPACTION_PREFETCH_BYTES` to measure number of bytes for RocksDB's prefetching (as opposed to file
+system's prefetch) on SST file during compaction read
+* A new API DB::GetNewestUserDefinedTimestamp is added to return the newest user defined timestamp seen in a column family
+* Introduce API `IngestWriteBatchWithIndex()` for ingesting updates into DB while bypassing memtable writes. This improves performance when writing a large write batch to the DB.
+* Add a new CF option `memtable_op_scan_flush_trigger` that triggers a flush of the memtable if an iterator's Seek()/Next() scans over a certain number of invisible entries from the memtable.
+
+### Public API Changes
+* AdvancedColumnFamilyOptions.max_write_buffer_number_to_maintain is deleted. It's deprecated since introduction of a better option max_write_buffer_size_to_maintain since RocksDB 6.5.0.
+* Deprecated API `DB::MaxMemCompactionLevel()`.
+* Deprecated `ReadOptions::ignore_range_deletions`.
+* Deprecated API `experimental::PromoteL0()`.
+* Added arbitrary string map for additional options to be overriden for remote compactions
+* The fail_if_options_file_error option in DBOptions has been removed. The behavior now is to always return failure in any API that fails to persist the OPTIONS file.
+
+### Behavior Changes
+* Make stats `PREFETCH_BYTES_USEFUL`, `PREFETCH_HITS`, `PREFETCH_BYTES` only account for prefetching during user initiated scan
+
+### Bug Fixes
+* Fix a bug in Posix file system that the FSWritableFile created via `FileSystem::ReopenWritableFile` internally does not track the correct file size.
+* Fix a bug where tail size of remote compaction output is not persisted in primary db's manifest
+
 ## 10.1.0 (03/24/2025)
 ### New Features
 * Added a new `DBOptions.calculate_sst_write_lifetime_hint_set` setting that allows to customize which compaction styles SST write lifetime hint calculation is allowed on. Today RocksDB supports only two modes `kCompactionStyleLevel` and `kCompactionStyleUniversal`.
@@ -1357,6 +1357,9 @@ filter_bench: $(OBJ_DIR)/util/filter_bench.o $(LIBRARY)
 db_stress: $(OBJ_DIR)/db_stress_tool/db_stress.o $(STRESS_LIBRARY) $(TOOLS_LIBRARY) $(LIBRARY)
 	$(AM_LINK)

+db_stress_compression_manager: $(OBJ_DIR)/db_stress_tool/db_stress_compression_manager.o $(LIBRARY)
+	$(AM_LINK)
+
 write_stress: $(OBJ_DIR)/tools/write_stress.o $(LIBRARY)
 	$(AM_LINK)

@@ -1491,6 +1494,9 @@ db_test: $(OBJ_DIR)/db/db_test.o $(TEST_LIBRARY) $(LIBRARY)
 db_test2: $(OBJ_DIR)/db/db_test2.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)

+compression_test: $(OBJ_DIR)/util/compression_test.o $(TEST_LIBRARY) $(LIBRARY)
+	$(AM_LINK)
+
 db_logical_block_size_cache_test: $(OBJ_DIR)/db/db_logical_block_size_cache_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)

@@ -2034,6 +2040,9 @@ wide_column_serialization_test: $(OBJ_DIR)/db/wide/wide_column_serialization_tes
 wide_columns_helper_test: $(OBJ_DIR)/db/wide/wide_columns_helper_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)

+interval_test: $(OBJ_DIR)/util/interval_test.o $(TEST_LIBRARY) $(LIBRARY)
+	$(AM_LINK)
+
 #-------------------------------------------------
 # make install related stuff
 PREFIX ?= /usr/local
@@ -2144,14 +2153,14 @@ ZLIB_DOWNLOAD_BASE ?= http://zlib.net
 BZIP2_VER ?= 1.0.8
 BZIP2_SHA256 ?= ab5a03176ee106d3f0fa90e381da478ddae405918153cca248e682cd0c4a2269
 BZIP2_DOWNLOAD_BASE ?= http://sourceware.org/pub/bzip2
-SNAPPY_VER ?= 1.2.1
-SNAPPY_SHA256 ?= 736aeb64d86566d2236ddffa2865ee5d7a82d26c9016b36218fcc27ea4f09f86
+SNAPPY_VER ?= 1.2.2
+SNAPPY_SHA256 ?= 90f74bc1fbf78a6c56b3c4a082a05103b3a56bb17bca1a27e052ea11723292dc
 SNAPPY_DOWNLOAD_BASE ?= https://github.com/google/snappy/archive
-LZ4_VER ?= 1.9.4
-LZ4_SHA256 ?= 0b0e3aa07c8c063ddf40b082bdf7e37a1562bda40a0ff5272957f3e987e0e54b
+LZ4_VER ?= 1.10.0
+LZ4_SHA256 ?= 537512904744b35e232912055ccf8ec66d768639ff3abe5788d90d792ec5f48b
 LZ4_DOWNLOAD_BASE ?= https://github.com/lz4/lz4/archive
-ZSTD_VER ?= 1.5.5
-ZSTD_SHA256 ?= 98e9c3d949d1b924e28e01eccb7deed865eefebf25c2f21c702e5cd5b63b85e1
+ZSTD_VER ?= 1.5.7
+ZSTD_SHA256 ?= 37d7284556b20954e56e1ca85b80226768902e2edabd3b649e9e72c0c9012ee3
 ZSTD_DOWNLOAD_BASE ?= https://github.com/facebook/zstd/archive
 CURL_SSL_OPTS ?= --tlsv1

@@ -2372,27 +2381,27 @@ rocksdbjavastaticreleasedocker: rocksdbjavastaticosx rocksdbjavastaticdockerx86

 rocksdbjavastaticdockerx86:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_x86-be --platform linux/386 --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:centos6_x86-be /rocksdb-host/java/crossbuild/docker-build-linux.sh
+	docker run --rm --name rocksdb_linux_x86-be --platform linux/386 --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:centos7_x86-be /rocksdb-host/java/crossbuild/docker-build-linux.sh

 rocksdbjavastaticdockerx86_64:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_x64-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:centos6_x64-be /rocksdb-host/java/crossbuild/docker-build-linux.sh
+	docker run --rm --name rocksdb_linux_x64-be --platform linux/amd64 --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:centos7_x64-be /rocksdb-host/java/crossbuild/docker-build-linux.sh

 rocksdbjavastaticdockerppc64le:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_ppc64le-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:centos7_ppc64le-be /rocksdb-host/java/crossbuild/docker-build-linux.sh
+	docker run --rm --name rocksdb_linux_ppc64le-be --platform linux/ppc64le --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:centos7_ppc64le-be /rocksdb-host/java/crossbuild/docker-build-linux.sh

 rocksdbjavastaticdockerarm64v8:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_arm64v8-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:centos7_arm64v8-be /rocksdb-host/java/crossbuild/docker-build-linux.sh
+	docker run --rm --name rocksdb_linux_arm64v8-be --platform linux/aarch64 --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:centos7_arm64v8-be /rocksdb-host/java/crossbuild/docker-build-linux.sh

 rocksdbjavastaticdockers390x:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_s390x-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:ubuntu18_s390x-be /rocksdb-host/java/crossbuild/docker-build-linux.sh
+	docker run --rm --name rocksdb_linux_s390x-be --platform linux/s390x --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:ubuntu18_s390x-be /rocksdb-host/java/crossbuild/docker-build-linux.sh

 rocksdbjavastaticdockerriscv64:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_riscv64-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:ubuntu20_riscv64-be /rocksdb-host/java/crossbuild/docker-build-linux.sh
+	docker run --rm --name rocksdb_linux_riscv64-be --platform linux/riscv64 --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:ubuntu20_riscv64-be /rocksdb-host/java/crossbuild/docker-build-linux.sh

 rocksdbjavastaticdockerx86musl:
 	mkdir -p java/target
@@ -2400,19 +2409,19 @@ rocksdbjavastaticdockerx86musl:

 rocksdbjavastaticdockerx86_64musl:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_x64-musl-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:alpine3_x64-be /rocksdb-host/java/crossbuild/docker-build-linux.sh
+	docker run --rm --name rocksdb_linux_x64-musl-be --platform linux/amd64 --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:alpine3_x64-be /rocksdb-host/java/crossbuild/docker-build-linux.sh

 rocksdbjavastaticdockerppc64lemusl:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_ppc64le-musl-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:alpine3_ppc64le-be /rocksdb-host/java/crossbuild/docker-build-linux.sh
+	docker run --rm --name rocksdb_linux_ppc64le-musl-be --platform linux/ppc64le --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:alpine3_ppc64le-be /rocksdb-host/java/crossbuild/docker-build-linux.sh

 rocksdbjavastaticdockerarm64v8musl:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_arm64v8-musl-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:alpine3_arm64v8-be /rocksdb-host/java/crossbuild/docker-build-linux.sh
+	docker run --rm --name rocksdb_linux_arm64v8-musl-be --platform linux/aarch64 --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:alpine3_arm64v8-be /rocksdb-host/java/crossbuild/docker-build-linux.sh

 rocksdbjavastaticdockers390xmusl:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_s390x-musl-be --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:alpine3_s390x-be /rocksdb-host/java/crossbuild/docker-build-linux.sh
+	docker run --rm --name rocksdb_linux_s390x-musl-be --platform linux/s390x --attach stdin --attach stdout --attach stderr --volume $(HOME)/.m2:/root/.m2:ro --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) --env J=$(J) evolvedbinary/rocksjava:alpine3_s390x-be /rocksdb-host/java/crossbuild/docker-build-linux.sh

 rocksdbjavastaticpublish: rocksdbjavastaticrelease rocksdbjavastaticpublishcentral

@@ -2467,8 +2476,8 @@ jtest_run:
 jtest: rocksdbjava
 	cd java;$(MAKE) sample test

-jpmd: rocksdbjava rocksdbjavageneratepom
-	cd java;$(MAKE) pmd
+jpmd: rocksdbjavageneratepom
+	cd java;$(MAKE) java java_test pmd

 jdb_bench:
 	cd java;$(MAKE) db_bench;
@@ -2489,11 +2498,13 @@ checkout_folly:
 	fi
 	@# Pin to a particular version for public CI, so that PR authors don't
 	@# need to worry about folly breaking our integration. Update periodically
-	cd third-party/folly && git reset --hard 78286282478e1ae05b2e8cbcf0e2139eab283bea
-	@# NOTE: this hack is required for clang in some cases
-	perl -pi -e 's/int rv = syscall/int rv = (int)syscall/' third-party/folly/folly/detail/Futex.cpp
-	@# NOTE: this hack is required for gcc in some cases
-	perl -pi -e 's/(__has_include.<experimental.memory_resource>.)/__cpp_rtti && $$1/' third-party/folly/folly/memory/MemoryResource.h
+	cd third-party/folly && git reset --hard e95383b7c8b5b1e46cf47acf2f317d54f93c8268
+	@# Apparently missing include
+	perl -pi -e 's/(#include <atomic>)/$$1\n#include <cstring>/' third-party/folly/folly/lang/Exception.h
+	@# Warning-as-error on memcpy
+	perl -pi -e 's/memcpy.&ptr/memcpy((void*)&ptr/' third-party/folly/folly/lang/Exception.cpp
+	@# const mismatch
+	perl -pi -e 's/: environ/: (const char**)(environ)/' third-party/folly/folly/Subprocess.cpp
 	@# NOTE: boost source will be needed for any build including `USE_FOLLY_LITE` builds as those depend on boost headers
 	cd third-party/folly && $(PYTHON) build/fbcode_builder/getdeps.py fetch boost

@@ -55,8 +55,11 @@ fi
 # we currently depend on POSIX platform
 COMMON_FLAGS="-DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX"

-# Default to fbcode gcc on internal fb machines
-if [ -z "$ROCKSDB_NO_FBCODE" -a -d /mnt/gvfs/third-party ]; then
+# Default to fbcode gcc on Meta internal machines
+IS_META_HOST="$(hostname | grep -E '(facebook|meta).com|fbinfra.net')"
+if [ -z "$ROCKSDB_NO_FBCODE" -a "$IS_META_HOST" ]; then
+  if [ -d /mnt/gvfs/third-party ]; then
+    echo "NOTE: Using fbcode build" >&2
    FBCODE_BUILD="true"
    # If we're compiling with TSAN or shared lib, we need pic build
    PIC_BUILD=$COMPILE_WITH_TSAN
@@ -64,6 +67,11 @@ if [ -z "$ROCKSDB_NO_FBCODE" -a -d /mnt/gvfs/third-party ]; then
      PIC_BUILD=1
    fi
    source "$PWD/build_tools/fbcode_config_platform010.sh"
+  else
+    echo "************************************************************************" >&2
+    echo "WARNING: -d /mnt/gvfs/third-party failed; no fbcode build" >&2
+    echo "************************************************************************" >&2
+  fi
 fi

 # Delete existing output, if it exists
@@ -71,7 +79,9 @@ rm -f "$OUTPUT"
 touch "$OUTPUT"

 if test -z "$CC"; then
-    if [ -x "$(command -v cc)" ]; then
+    if [ "$USE_CLANG" -a -x "$(command -v clang)" ]; then
+        CC=clang
+    elif [ -x "$(command -v cc)" ]; then
        CC=cc
    elif [ -x "$(command -v clang)" ]; then
        CC=clang
@@ -81,7 +91,9 @@ if test -z "$CC"; then
 fi

 if test -z "$CXX"; then
-    if [ -x "$(command -v g++)" ]; then
+    if [ "$USE_CLANG" -a -x "$(command -v clang++)" ]; then
+        CXX=clang++
+    elif [ -x "$(command -v g++)" ]; then
        CXX=g++
    elif [ -x "$(command -v clang++)" ]; then
        CXX=clang++
@@ -91,7 +103,9 @@ if test -z "$CXX"; then
 fi

 if test -z "$AR"; then
-    if [ -x "$(command -v gcc-ar)" ]; then
+    if [ "$USE_CLANG" -a -x "$(command -v llvm-ar)" ]; then
+        AR=llvm-ar
+    elif [ -x "$(command -v gcc-ar)" ]; then
        AR=gcc-ar
    elif [ -x "$(command -v llvm-ar)" ]; then
        AR=llvm-ar
@@ -118,6 +118,9 @@ fi
 # fi
 set -e

+# Exclude third-party from formatting
+EXCLUDE=':!third-party/'
+
 uncommitted_code=`git diff HEAD`

 # If there's no uncommitted changes, we assume user are doing post-commit
@@ -137,11 +140,11 @@ then
  # should be relevant for formatting fixes.
  FORMAT_UPSTREAM_MERGE_BASE="$(git merge-base "$FORMAT_UPSTREAM" HEAD)"
  # Get the differences
-  diffs=$(git diff -U0 "$FORMAT_UPSTREAM_MERGE_BASE" | $CLANG_FORMAT_DIFF -p 1) || true
+  diffs=$(git diff -U0 "$FORMAT_UPSTREAM_MERGE_BASE" -- $EXCLUDE | $CLANG_FORMAT_DIFF -p 1) || true
  echo "Checking format of changes not yet in $FORMAT_UPSTREAM..."
 else
  # Check the format of uncommitted lines,
-  diffs=$(git diff -U0 HEAD | $CLANG_FORMAT_DIFF -p 1) || true
+  diffs=$(git diff -U0 HEAD -- $EXCLUDE | $CLANG_FORMAT_DIFF -p 1) || true
  echo "Checking format of uncommitted changes..."
 fi

@@ -187,9 +190,9 @@ fi
 # Do in-place format adjustment.
 if [ -z "$uncommitted_code" ]
 then
-  git diff -U0 "$FORMAT_UPSTREAM_MERGE_BASE" | $CLANG_FORMAT_DIFF -i -p 1
+  git diff -U0 "$FORMAT_UPSTREAM_MERGE_BASE" -- $EXCLUDE | $CLANG_FORMAT_DIFF -i -p 1
 else
-  git diff -U0 HEAD | $CLANG_FORMAT_DIFF -i -p 1
+  git diff -U0 HEAD -- $EXCLUDE | $CLANG_FORMAT_DIFF -i -p 1
 fi
 echo "Files reformatted!"

@@ -60,6 +60,8 @@ DEFINE_uint32(value_bytes, 8 * KiB, "Size of each value added.");
 DEFINE_uint32(value_bytes_estimate, 0,
              "If > 0, overrides estimated_entry_charge or "
              "min_avg_entry_charge depending on cache_type.");
+DEFINE_double(compressible_to_ratio, 0.5,
+              "Approximate size ratio that values can be compressed to.");

 DEFINE_int32(
    degenerate_hash_bits, 0,
@@ -291,10 +293,19 @@ struct KeyGen {

 Cache::ObjectPtr createValue(Random64& rnd, MemoryAllocator* alloc) {
  char* rv = AllocateBlock(FLAGS_value_bytes, alloc).release();
-  // Fill with some filler data, and take some CPU time
-  for (uint32_t i = 0; i < FLAGS_value_bytes; i += 8) {
+  // Fill with some filler data, and take some CPU time, but add redundancy
+  // as requested for compressibility.
+  uint32_t random_fill_size = std::max(
+      uint32_t{1}, std::min(FLAGS_value_bytes,
+                            static_cast<uint32_t>(FLAGS_compressible_to_ratio *
+                                                  FLAGS_value_bytes)));
+  uint32_t i = 0;
+  for (; i < random_fill_size; i += 8) {
    EncodeFixed64(rv + i, rnd.Next());
  }
+  for (; i < FLAGS_value_bytes; i++) {
+    rv[i] = rv[i % random_fill_size];
+  }
  return rv;
 }

@@ -309,16 +320,16 @@ Status SaveToFn(Cache::ObjectPtr from_obj, size_t /*from_offset*/,

 Status CreateFn(const Slice& data, CompressionType /*type*/,
                CacheTier /*source*/, Cache::CreateContext* /*context*/,
-                MemoryAllocator* /*allocator*/, Cache::ObjectPtr* out_obj,
+                MemoryAllocator* alloc, Cache::ObjectPtr* out_obj,
                size_t* out_charge) {
-  *out_obj = new char[data.size()];
+  *out_obj = AllocateBlock(data.size(), alloc).release();
  memcpy(*out_obj, data.data(), data.size());
  *out_charge = data.size();
  return Status::OK();
 };

 void DeleteFn(Cache::ObjectPtr value, MemoryAllocator* alloc) {
-  CustomDeleter{alloc}(static_cast<char*>(value));
+  CacheAllocationDeleter{alloc}(static_cast<char*>(value));
 }

 Cache::CacheItemHelper helper1_wos(CacheEntryRole::kDataBlock, DeleteFn);
@@ -16,6 +16,31 @@
 #include "util/string_util.h"

 namespace ROCKSDB_NAMESPACE {
+namespace {
+// Format of values in CompressedSecondaryCache:
+// If enable_custom_split_merge:
+//  * A chain of CacheValueChunk representing the sequence of bytes for a tagged
+//    value. The overall length of the tagged value is determined by the chain
+//    of CacheValueChunks.
+// If !enable_custom_split_merge:
+//  * A LengthPrefixedSlice (starts with varint64 size) of a tagged value.
+//
+// A tagged value has a 2-byte header before the "saved" or compressed block
+// data:
+//  * 1 byte for "source" CacheTier indicating which tier is responsible for
+//    compression/decompression.
+//  * 1 byte for compression type which is generated/used by
+//    CompressedSecondaryCache iff source == CacheTier::kVolatileCompressedTier
+//    (original entry passed in was uncompressed). Otherwise, the compression
+//    type is preserved from the entry passed in.
+constexpr uint32_t kTagSize = 2;
+
+// Size of tag + varint size prefix when applicable
+uint32_t GetHeaderSize(size_t data_size, bool enable_split_merge) {
+  return (enable_split_merge ? 0 : VarintLength(kTagSize + data_size)) +
+         kTagSize;
+}
+}  // namespace

 CompressedSecondaryCache::CompressedSecondaryCache(
    const CompressedSecondaryCacheOptions& opts)
@@ -24,7 +49,14 @@ CompressedSecondaryCache::CompressedSecondaryCache(
      cache_res_mgr_(std::make_shared<ConcurrentCacheReservationManager>(
          std::make_shared<CacheReservationManagerImpl<CacheEntryRole::kMisc>>(
              cache_))),
-      disable_cache_(opts.capacity == 0) {}
+      disable_cache_(opts.capacity == 0) {
+  auto mgr =
+      GetBuiltinCompressionManager(cache_options_.compress_format_version);
+  compressor_ = mgr->GetCompressor(cache_options_.compression_opts,
+                                   cache_options_.compression_type);
+  decompressor_ =
+      mgr->GetDecompressorOptimizeFor(cache_options_.compression_type);
+}

 CompressedSecondaryCache::~CompressedSecondaryCache() = default;

@@ -33,13 +65,9 @@ std::unique_ptr<SecondaryCacheResultHandle> CompressedSecondaryCache::Lookup(
    Cache::CreateContext* create_context, bool /*wait*/, bool advise_erase,
    Statistics* stats, bool& kept_in_sec_cache) {
  assert(helper);
-  // This is a minor optimization. Its ok to skip it in TSAN in order to
-  // avoid a false positive.
-#ifndef __SANITIZE_THREAD__
-  if (disable_cache_) {
+  if (disable_cache_.LoadRelaxed()) {
    return nullptr;
  }
-#endif

  std::unique_ptr<SecondaryCacheResultHandle> handle;
  kept_in_sec_cache = false;
@@ -55,75 +83,58 @@ std::unique_ptr<SecondaryCacheResultHandle> CompressedSecondaryCache::Lookup(
    return nullptr;
  }

-  CacheAllocationPtr* ptr{nullptr};
-  CacheAllocationPtr merged_value;
-  size_t handle_value_charge{0};
-  const char* data_ptr = nullptr;
-  CacheTier source = CacheTier::kVolatileCompressedTier;
-  CompressionType type = cache_options_.compression_type;
+  std::string merged_value;
+  Slice tagged_data;
  if (cache_options_.enable_custom_split_merge) {
    CacheValueChunk* value_chunk_ptr =
-        reinterpret_cast<CacheValueChunk*>(handle_value);
-    merged_value = MergeChunksIntoValue(value_chunk_ptr, handle_value_charge);
-    ptr = &merged_value;
-    data_ptr = ptr->get();
+        static_cast<CacheValueChunk*>(handle_value);
+    merged_value = MergeChunksIntoValue(value_chunk_ptr);
+    tagged_data = Slice(merged_value);
  } else {
-    uint32_t type_32 = static_cast<uint32_t>(type);
-    uint32_t source_32 = static_cast<uint32_t>(source);
-    ptr = reinterpret_cast<CacheAllocationPtr*>(handle_value);
-    handle_value_charge = cache_->GetCharge(lru_handle);
-    data_ptr = ptr->get();
-    data_ptr = GetVarint32Ptr(data_ptr, data_ptr + 1,
-                              static_cast<uint32_t*>(&type_32));
-    type = static_cast<CompressionType>(type_32);
-    data_ptr = GetVarint32Ptr(data_ptr, data_ptr + 1,
-                              static_cast<uint32_t*>(&source_32));
-    source = static_cast<CacheTier>(source_32);
-    uint64_t data_size = 0;
-    data_ptr = GetVarint64Ptr(data_ptr, ptr->get() + handle_value_charge,
-                              static_cast<uint64_t*>(&data_size));
-    assert(handle_value_charge > data_size);
-    handle_value_charge = data_size;
+    tagged_data = GetLengthPrefixedSlice(static_cast<char*>(handle_value));
  }
-  MemoryAllocator* allocator = cache_options_.memory_allocator.get();

-  Status s;
-  Cache::ObjectPtr value{nullptr};
-  size_t charge{0};
+  auto source = lossless_cast<CacheTier>(tagged_data[0]);
+  auto type = lossless_cast<CompressionType>(tagged_data[1]);
+
+  std::unique_ptr<char[]> uncompressed;
+  Slice saved(tagged_data.data() + kTagSize, tagged_data.size() - kTagSize);
  if (source == CacheTier::kVolatileCompressedTier) {
-    if (cache_options_.compression_type == kNoCompression ||
-        cache_options_.do_not_compress_roles.Contains(helper->role)) {
-      s = helper->create_cb(Slice(data_ptr, handle_value_charge),
-                            kNoCompression, CacheTier::kVolatileTier,
-                            create_context, allocator, &value, &charge);
-    } else {
-      UncompressionContext uncompression_context(
-          cache_options_.compression_type);
-      UncompressionInfo uncompression_info(uncompression_context,
-                                           UncompressionDict::GetEmptyDict(),
-                                           cache_options_.compression_type);
-
-      size_t uncompressed_size{0};
-      CacheAllocationPtr uncompressed =
-          UncompressData(uncompression_info, (char*)data_ptr,
-                         handle_value_charge, &uncompressed_size,
-                         cache_options_.compress_format_version, allocator);
-
-      if (!uncompressed) {
+    if (type != kNoCompression) {
+      // TODO: can we do something to avoid yet another allocation?
+      Decompressor::Args args;
+      args.compressed_data = saved;
+      args.compression_type = type;
+      Status s = decompressor_->ExtractUncompressedSize(args);
+      assert(s.ok());  // in-memory data
+      if (s.ok()) {
+        uncompressed = std::make_unique<char[]>(args.uncompressed_size);
+        s = decompressor_->DecompressBlock(args, uncompressed.get());
+        assert(s.ok());  // in-memory data
+      }
+      if (!s.ok()) {
        cache_->Release(lru_handle, /*erase_if_last_ref=*/true);
        return nullptr;
      }
-      s = helper->create_cb(Slice(uncompressed.get(), uncompressed_size),
-                            kNoCompression, CacheTier::kVolatileTier,
-                            create_context, allocator, &value, &charge);
+      saved = Slice(uncompressed.get(), args.uncompressed_size);
+      type = kNoCompression;
+      // Free temporary compressed data as early as we can. This could matter
+      // for unusually large blocks because we also have
+      // * Another compressed copy above (from lru_cache).
+      // * The uncompressed copy in `uncompressed`.
+      // * Another uncompressed copy in `result_value` below.
+      // Let's try to max out at 3 copies instead of 4.
+      merged_value = std::string();
    }
-  } else {
-    // The item was not compressed by us. Let the helper create_cb
-    // uncompress it
-    s = helper->create_cb(Slice(data_ptr, handle_value_charge), type, source,
-                          create_context, allocator, &value, &charge);
+    // Reduced as if it came from primary cache
+    source = CacheTier::kVolatileTier;
  }

+  Cache::ObjectPtr result_value = nullptr;
+  size_t result_charge = 0;
+  Status s = helper->create_cb(saved, type, source, create_context,
+                               cache_options_.memory_allocator.get(),
+                               &result_value, &result_charge);
  if (!s.ok()) {
    cache_->Release(lru_handle, /*erase_if_last_ref=*/true);
    return nullptr;
@@ -141,7 +152,8 @@ std::unique_ptr<SecondaryCacheResultHandle> CompressedSecondaryCache::Lookup(
    kept_in_sec_cache = true;
    cache_->Release(lru_handle, /*erase_if_last_ref=*/false);
  }
-  handle.reset(new CompressedSecondaryCacheResultHandle(value, charge));
+  handle.reset(
+      new CompressedSecondaryCacheResultHandle(result_value, result_charge));
  RecordTick(stats, COMPRESSED_SECONDARY_CACHE_HITS);
  return handle;
 }
@@ -164,88 +176,111 @@ bool CompressedSecondaryCache::MaybeInsertDummy(const Slice& key) {

 Status CompressedSecondaryCache::InsertInternal(
    const Slice& key, Cache::ObjectPtr value,
-    const Cache::CacheItemHelper* helper, CompressionType type,
+    const Cache::CacheItemHelper* helper, CompressionType from_type,
    CacheTier source) {
-  if (source != CacheTier::kVolatileCompressedTier &&
-      cache_options_.enable_custom_split_merge) {
-    // We don't support custom split/merge for the tiered case
-    return Status::OK();
-  }
+  bool enable_split_merge = cache_options_.enable_custom_split_merge;
+  const Cache::CacheItemHelper* internal_helper = GetHelper(enable_split_merge);

-  auto internal_helper = GetHelper(cache_options_.enable_custom_split_merge);
-  char header[20];
-  char* payload = header;
-  payload = EncodeVarint32(payload, static_cast<uint32_t>(type));
-  payload = EncodeVarint32(payload, static_cast<uint32_t>(source));
-  size_t data_size = (*helper->size_cb)(value);
-  char* data_size_ptr = payload;
-  payload = EncodeVarint64(payload, data_size);
+  // TODO: variant of size_cb that also returns a pointer to the data if
+  // already available. Saves an allocation if we keep the compressed version.
+  const size_t data_size_original = (*helper->size_cb)(value);

-  size_t header_size = payload - header;
-  size_t total_size = data_size + header_size;
-  CacheAllocationPtr ptr =
-      AllocateBlock(total_size, cache_options_.memory_allocator.get());
-  char* data_ptr = ptr.get() + header_size;
+  // Allocate enough memory for header/tag + original data because (a) we might
+  // not be attempting compression at all, and (b) we might keep the original if
+  // compression is insufficient. But we don't need the length prefix with
+  // enable_split_merge. TODO: be smarter with CacheValueChunk to save an
+  // allocation in the enable_split_merge case.
+  size_t header_size = GetHeaderSize(data_size_original, enable_split_merge);
+  CacheAllocationPtr allocation = AllocateBlock(
+      header_size + data_size_original, cache_options_.memory_allocator.get());
+  char* data_ptr = allocation.get() + header_size;
+  Slice tagged_data(data_ptr - kTagSize, data_size_original + kTagSize);
+  assert(tagged_data.data() >= allocation.get());

-  Status s = (*helper->saveto_cb)(value, 0, data_size, data_ptr);
+  Status s = (*helper->saveto_cb)(value, 0, data_size_original, data_ptr);
  if (!s.ok()) {
    return s;
  }
-  Slice val(data_ptr, data_size);

-  std::string compressed_val;
-  if (cache_options_.compression_type != kNoCompression &&
-      type == kNoCompression &&
+  std::unique_ptr<char[]> tagged_compressed_data;
+  CompressionType to_type = kNoCompression;
+  if (compressor_ && from_type == kNoCompression &&
      !cache_options_.do_not_compress_roles.Contains(helper->role)) {
-    PERF_COUNTER_ADD(compressed_sec_cache_uncompressed_bytes, data_size);
-    CompressionContext compression_context(cache_options_.compression_type,
-                                           cache_options_.compression_opts);
-    uint64_t sample_for_compression{0};
-    CompressionInfo compression_info(
-        cache_options_.compression_opts, compression_context,
-        CompressionDict::GetEmptyDict(), cache_options_.compression_type,
-        sample_for_compression);
+    assert(source == CacheTier::kVolatileCompressedTier);

-    bool success =
-        CompressData(val, compression_info,
-                     cache_options_.compress_format_version, &compressed_val);
-
-    if (!success) {
-      return Status::Corruption("Error compressing value.");
+    // TODO: consider malloc sizes for max acceptable compressed size
+    // Or maybe max_compressed_bytes_per_kb
+    size_t data_size_compressed = data_size_original - 1;
+    tagged_compressed_data =
+        std::make_unique<char[]>(data_size_compressed + kTagSize);
+    s = compressor_->CompressBlock(Slice(data_ptr, data_size_original),
+                                   tagged_compressed_data.get() + kTagSize,
+                                   &data_size_compressed, &to_type,
+                                   nullptr /*working_area*/);
+    if (!s.ok()) {
+      return s;
    }
-
-    val = Slice(compressed_val);
-    data_size = compressed_val.size();
-    payload = EncodeVarint64(data_size_ptr, data_size);
-    header_size = payload - header;
-    total_size = header_size + data_size;
-    PERF_COUNTER_ADD(compressed_sec_cache_compressed_bytes, data_size);
-
-    if (!cache_options_.enable_custom_split_merge) {
-      ptr = AllocateBlock(total_size, cache_options_.memory_allocator.get());
-      data_ptr = ptr.get() + header_size;
-      memcpy(data_ptr, compressed_val.data(), data_size);
+    PERF_COUNTER_ADD(compressed_sec_cache_uncompressed_bytes,
+                     data_size_original);
+    if (to_type == kNoCompression) {
+      // Compression rejected or otherwise aborted/failed
+      to_type = kNoCompression;
+      tagged_compressed_data.reset();
+      // TODO: consider separate counters for rejected compressions
+      PERF_COUNTER_ADD(compressed_sec_cache_compressed_bytes,
+                       data_size_original);
+    } else {
+      PERF_COUNTER_ADD(compressed_sec_cache_compressed_bytes,
+                       data_size_compressed);
+      if (enable_split_merge) {
+        // Only need tagged_data for copying into CacheValueChunks.
+        tagged_data = Slice(tagged_compressed_data.get(),
+                            data_size_compressed + kTagSize);
+        allocation.reset();
+      } else {
+        // Replace allocation with compressed version, copied from string
+        header_size = GetHeaderSize(data_size_compressed, enable_split_merge);
+        allocation = AllocateBlock(header_size + data_size_compressed,
+                                   cache_options_.memory_allocator.get());
+        data_ptr = allocation.get() + header_size;
+        // Ignore unpopulated tag on tagged_compressed_data; will only be
+        // populated on the new allocation.
+        std::memcpy(data_ptr, tagged_compressed_data.get() + kTagSize,
+                    data_size_compressed);
+        tagged_data =
+            Slice(data_ptr - kTagSize, data_size_compressed + kTagSize);
+        assert(tagged_data.data() >= allocation.get());
+      }
    }
  }

  PERF_COUNTER_ADD(compressed_sec_cache_insert_real_count, 1);
-  if (cache_options_.enable_custom_split_merge) {
+
+  // Save the tag fields
+  const_cast<char*>(tagged_data.data())[0] = lossless_cast<char>(source);
+  const_cast<char*>(tagged_data.data())[1] = lossless_cast<char>(
+      source == CacheTier::kVolatileCompressedTier ? to_type : from_type);
+
+  if (enable_split_merge) {
    size_t split_charge{0};
-    CacheValueChunk* value_chunks_head = SplitValueIntoChunks(
-        val, cache_options_.compression_type, split_charge);
-    return cache_->Insert(key, value_chunks_head, internal_helper,
-                          split_charge);
+    CacheValueChunk* value_chunks_head =
+        SplitValueIntoChunks(tagged_data, split_charge);
+    s = cache_->Insert(key, value_chunks_head, internal_helper, split_charge);
+    assert(s.ok());  // LRUCache::Insert() with handle==nullptr always OK
  } else {
+    // Save the size prefix
+    char* ptr = allocation.get();
+    ptr = EncodeVarint64(ptr, tagged_data.size());
+    assert(ptr == tagged_data.data());
 #ifdef ROCKSDB_MALLOC_USABLE_SIZE
-    size_t charge = malloc_usable_size(ptr.get());
+    size_t charge = malloc_usable_size(allocation.get());
 #else
-    size_t charge = total_size;
+    size_t charge = tagged_data.size();
 #endif
-    std::memcpy(ptr.get(), header, header_size);
-    CacheAllocationPtr* buf = new CacheAllocationPtr(std::move(ptr));
-    charge += sizeof(CacheAllocationPtr);
-    return cache_->Insert(key, buf, internal_helper, charge);
+    s = cache_->Insert(key, allocation.release(), internal_helper, charge);
+    assert(s.ok());  // LRUCache::Insert() with handle==nullptr always OK
  }
+  return Status::OK();
 }

 Status CompressedSecondaryCache::Insert(const Slice& key,
@@ -267,7 +302,17 @@ Status CompressedSecondaryCache::Insert(const Slice& key,
 Status CompressedSecondaryCache::InsertSaved(
    const Slice& key, const Slice& saved, CompressionType type = kNoCompression,
    CacheTier source = CacheTier::kVolatileTier) {
+  if (source == CacheTier::kVolatileCompressedTier) {
+    // Unexpected, would violate InsertInternal preconditions
+    assert(source != CacheTier::kVolatileCompressedTier);
+    return Status::OK();
+  }
  if (type == kNoCompression) {
+    // Not currently supported (why?)
+    return Status::OK();
+  }
+  if (cache_options_.enable_custom_split_merge) {
+    // We don't support custom split/merge for the tiered case (why?)
    return Status::OK();
  }

@@ -287,7 +332,7 @@ Status CompressedSecondaryCache::SetCapacity(size_t capacity) {
  MutexLock l(&capacity_mutex_);
  cache_options_.capacity = capacity;
  cache_->SetCapacity(capacity);
-  disable_cache_ = capacity == 0;
+  disable_cache_.StoreRelaxed(capacity == 0);
  return Status::OK();
 }

@@ -317,9 +362,14 @@ std::string CompressedSecondaryCache::GetPrintableOptions() const {
  return ret;
 }

+// FIXME: this could use a lot of attention, including:
+// * Use allocator
+// * We shouldn't be worse than non-split; be more pro-actively aware of
+// internal fragmentation
+// * Consider a unified object/chunk structure that may or may not split
+// * Optimize size overhead of chunks
 CompressedSecondaryCache::CacheValueChunk*
 CompressedSecondaryCache::SplitValueIntoChunks(const Slice& value,
-                                               CompressionType compression_type,
                                               size_t& charge) {
  assert(!value.empty());
  const char* src_ptr = value.data();
@@ -340,15 +390,14 @@ CompressedSecondaryCache::SplitValueIntoChunks(const Slice& value,
    // size, or there is no compression.
    if (upper == malloc_bin_sizes_.begin() ||
        upper == malloc_bin_sizes_.end() ||
-        *upper - predicted_chunk_size < malloc_bin_sizes_.front() ||
-        compression_type == kNoCompression) {
+        *upper - predicted_chunk_size < malloc_bin_sizes_.front()) {
      tmp_size = predicted_chunk_size;
    } else {
      tmp_size = *(--upper);
    }

    CacheValueChunk* new_chunk =
-        reinterpret_cast<CacheValueChunk*>(new char[tmp_size]);
+        static_cast<CacheValueChunk*>(static_cast<void*>(new char[tmp_size]));
    current_chunk->next = new_chunk;
    current_chunk = current_chunk->next;
    actual_chunk_size = tmp_size - sizeof(CacheValueChunk) + 1;
@@ -363,28 +412,24 @@ CompressedSecondaryCache::SplitValueIntoChunks(const Slice& value,
  return dummy_head.next;
 }

-CacheAllocationPtr CompressedSecondaryCache::MergeChunksIntoValue(
-    const void* chunks_head, size_t& charge) {
-  const CacheValueChunk* head =
-      reinterpret_cast<const CacheValueChunk*>(chunks_head);
+std::string CompressedSecondaryCache::MergeChunksIntoValue(
+    const CacheValueChunk* head) {
  const CacheValueChunk* current_chunk = head;
-  charge = 0;
+  size_t total_size = 0;
  while (current_chunk != nullptr) {
-    charge += current_chunk->size;
+    total_size += current_chunk->size;
    current_chunk = current_chunk->next;
  }

-  CacheAllocationPtr ptr =
-      AllocateBlock(charge, cache_options_.memory_allocator.get());
+  std::string result;
+  result.reserve(total_size);
  current_chunk = head;
-  size_t pos{0};
  while (current_chunk != nullptr) {
-    memcpy(ptr.get() + pos, current_chunk->data, current_chunk->size);
-    pos += current_chunk->size;
+    result.append(current_chunk->data, current_chunk->size);
    current_chunk = current_chunk->next;
  }
-
-  return ptr;
+  assert(result.size() == total_size);
+  return result;
 }

 const Cache::CacheItemHelper* CompressedSecondaryCache::GetHelper(
@@ -398,16 +443,16 @@ const Cache::CacheItemHelper* CompressedSecondaryCache::GetHelper(
            CacheValueChunk* tmp_chunk = chunks_head;
            chunks_head = chunks_head->next;
            tmp_chunk->Free();
-            obj = nullptr;
          }
        }};
    return &kHelper;
  } else {
    static const Cache::CacheItemHelper kHelper{
        CacheEntryRole::kMisc,
-        [](Cache::ObjectPtr obj, MemoryAllocator* /*alloc*/) {
-          delete static_cast<CacheAllocationPtr*>(obj);
-          obj = nullptr;
+        [](Cache::ObjectPtr obj, MemoryAllocator* alloc) {
+          if (obj != nullptr) {
+            CacheAllocationDeleter{alloc}(static_cast<char*>(obj));
+          }
        }};
    return &kHelper;
  }
@@ -418,12 +463,7 @@ size_t CompressedSecondaryCache::TEST_GetCharge(const Slice& key) {
  if (lru_handle == nullptr) {
    return 0;
  }
-
  size_t charge = cache_->GetCharge(lru_handle);
-  if (cache_->Value(lru_handle) != nullptr &&
-      !cache_options_.enable_custom_split_merge) {
-    charge -= 10;
-  }
  cache_->Release(lru_handle, /*erase_if_last_ref=*/false);
  return charge;
 }
@@ -10,13 +10,12 @@
 #include <memory>

 #include "cache/cache_reservation_manager.h"
-#include "cache/lru_cache.h"
 #include "memory/memory_allocator_impl.h"
+#include "rocksdb/advanced_compression.h"
 #include "rocksdb/secondary_cache.h"
 #include "rocksdb/slice.h"
 #include "rocksdb/status.h"
-#include "util/compression.h"
-#include "util/mutexlock.h"
+#include "util/atomic.h"

 namespace ROCKSDB_NAMESPACE {

@@ -124,14 +123,9 @@ class CompressedSecondaryCache : public SecondaryCache {
  // Split value into chunks to better fit into jemalloc bins. The chunks
  // are stored in CacheValueChunk and extra charge is needed for each chunk,
  // so the cache charge is recalculated here.
-  CacheValueChunk* SplitValueIntoChunks(const Slice& value,
-                                        CompressionType compression_type,
-                                        size_t& charge);
+  CacheValueChunk* SplitValueIntoChunks(const Slice& value, size_t& charge);

-  // After merging chunks, the extra charge for each chunk is removed, so
-  // the charge is recalculated.
-  CacheAllocationPtr MergeChunksIntoValue(const void* chunks_head,
-                                          size_t& charge);
+  std::string MergeChunksIntoValue(const CacheValueChunk* head);

  bool MaybeInsertDummy(const Slice& key);

@@ -145,9 +139,11 @@ class CompressedSecondaryCache : public SecondaryCache {
  const Cache::CacheItemHelper* GetHelper(bool enable_custom_split_merge) const;
  std::shared_ptr<Cache> cache_;
  CompressedSecondaryCacheOptions cache_options_;
+  std::unique_ptr<Compressor> compressor_;
+  std::shared_ptr<Decompressor> decompressor_;
  mutable port::Mutex capacity_mutex_;
  std::shared_ptr<ConcurrentCacheReservationManager> cache_res_mgr_;
-  bool disable_cache_;
+  RelaxedAtomic<bool> disable_cache_;
 };

 }  // namespace ROCKSDB_NAMESPACE
@@ -24,6 +24,14 @@ namespace ROCKSDB_NAMESPACE {
 using secondary_cache_test_util::GetTestingCacheTypes;
 using secondary_cache_test_util::WithCacheType;

+// Read and reset a statistic
+template <typename T>
+T Pop(T& var) {
+  T ret = var;
+  var = T();
+  return ret;
+}
+
 // 16 bytes for HCC compatibility
 const std::string key0 = "____    ____key0";
 const std::string key1 = "____    ____key1";
@@ -51,7 +59,7 @@ class CompressedSecondaryCacheTestBase : public testing::Test,

    Random rnd(301);
    // Insert and Lookup the item k1 for the first time.
-    std::string str1(rnd.RandomString(1000));
+    std::string str1 = test::CompressibleString(&rnd, 0.5, 1000);
    TestItem item1(str1.data(), str1.length());
    // A dummy handle is inserted if the item is inserted for the first time.
    ASSERT_OK(sec_cache->Insert(key1, &item1, GetHelper(), false));
@@ -68,7 +76,14 @@ class CompressedSecondaryCacheTestBase : public testing::Test,
    ASSERT_OK(sec_cache->Insert(key1, &item1, GetHelper(), false));
    ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 1);

-    ASSERT_GT(comp_sec_cache->TEST_GetCharge(key1), 1000);
+    if (sec_cache_is_compressed) {
+      ASSERT_GT(comp_sec_cache->TEST_GetCharge(key1), str1.length() / 4);
+      ASSERT_LT(comp_sec_cache->TEST_GetCharge(key1), str1.length() * 3 / 4);
+    } else {
+      ASSERT_GE(comp_sec_cache->TEST_GetCharge(key1), str1.length());
+      // NOTE: split-merge is worse (1048 vs. 1024)
+      ASSERT_LE(comp_sec_cache->TEST_GetCharge(key1), 1048U);
+    }

    std::unique_ptr<SecondaryCacheResultHandle> handle1_2 =
        sec_cache->Lookup(key1, GetHelper(), this, true, /*advise_erase=*/true,
@@ -76,10 +91,13 @@ class CompressedSecondaryCacheTestBase : public testing::Test,
    ASSERT_NE(handle1_2, nullptr);
    ASSERT_FALSE(kept_in_sec_cache);
    if (sec_cache_is_compressed) {
-      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
-                1000);
-      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes,
-                1007);
+      ASSERT_EQ(
+          Pop(get_perf_context()->compressed_sec_cache_uncompressed_bytes),
+          str1.length());
+      ASSERT_LT(get_perf_context()->compressed_sec_cache_compressed_bytes,
+                str1.length() * 3 / 4);
+      ASSERT_GT(Pop(get_perf_context()->compressed_sec_cache_compressed_bytes),
+                str1.length() / 4);
    } else {
      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);
@@ -97,7 +115,7 @@ class CompressedSecondaryCacheTestBase : public testing::Test,
    ASSERT_EQ(handle1_3, nullptr);

    // Insert and Lookup the item k2.
-    std::string str2(rnd.RandomString(1000));
+    std::string str2 = test::CompressibleString(&rnd, 0.5, 1017);
    TestItem item2(str2.data(), str2.length());
    ASSERT_OK(sec_cache->Insert(key2, &item2, GetHelper(), false));
    ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 2);
@@ -109,10 +127,13 @@ class CompressedSecondaryCacheTestBase : public testing::Test,
    ASSERT_OK(sec_cache->Insert(key2, &item2, GetHelper(), false));
    ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 2);
    if (sec_cache_is_compressed) {
-      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
-                2000);
-      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes,
-                2014);
+      ASSERT_EQ(
+          Pop(get_perf_context()->compressed_sec_cache_uncompressed_bytes),
+          str2.length());
+      ASSERT_LT(get_perf_context()->compressed_sec_cache_compressed_bytes,
+                str2.length() * 3 / 4);
+      ASSERT_GT(Pop(get_perf_context()->compressed_sec_cache_compressed_bytes),
+                str2.length() / 4);
    } else {
      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);
@@ -126,9 +147,48 @@ class CompressedSecondaryCacheTestBase : public testing::Test,
    ASSERT_NE(val2, nullptr);
    ASSERT_EQ(memcmp(val2->Buf(), item2.Buf(), item2.Size()), 0);

+    // Release handles
    std::vector<SecondaryCacheResultHandle*> handles = {handle1_2.get(),
                                                        handle2_2.get()};
    sec_cache->WaitAll(handles);
+    handle1_2.reset();
+    handle2_2.reset();
+
+    // Insert and Lookup a non-compressible item k3.
+    std::string str3 = rnd.RandomBinaryString(480);
+    TestItem item3(str3.data(), str3.length());
+    ASSERT_OK(sec_cache->Insert(key3, &item3, GetHelper(), false));
+    ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 3);
+    std::unique_ptr<SecondaryCacheResultHandle> handle3_1 =
+        sec_cache->Lookup(key3, GetHelper(), this, true, /*advise_erase=*/false,
+                          /*stats=*/nullptr, kept_in_sec_cache);
+    ASSERT_EQ(handle3_1, nullptr);
+
+    ASSERT_OK(sec_cache->Insert(key3, &item3, GetHelper(), false));
+    ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 3);
+    if (sec_cache_is_compressed) {
+      // TODO: consider a compression rejected stat?
+      ASSERT_EQ(
+          Pop(get_perf_context()->compressed_sec_cache_uncompressed_bytes),
+          str3.length());
+      ASSERT_EQ(Pop(get_perf_context()->compressed_sec_cache_compressed_bytes),
+                str3.length());
+    } else {
+      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
+      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);
+    }
+
+    std::unique_ptr<SecondaryCacheResultHandle> handle3_2 =
+        sec_cache->Lookup(key3, GetHelper(), this, true, /*advise_erase=*/false,
+                          /*stats=*/nullptr, kept_in_sec_cache);
+    ASSERT_NE(handle3_2, nullptr);
+    std::unique_ptr<TestItem> val3 =
+        std::unique_ptr<TestItem>(static_cast<TestItem*>(handle3_2->Value()));
+    ASSERT_NE(val3, nullptr);
+    ASSERT_EQ(memcmp(val3->Buf(), item3.Buf(), item3.Size()), 0);
+
+    EXPECT_GE(comp_sec_cache->TEST_GetCharge(key3), str3.length());
+    EXPECT_LE(comp_sec_cache->TEST_GetCharge(key3), 512);

    sec_cache.reset();
  }
@@ -178,8 +238,9 @@ class CompressedSecondaryCacheTestBase : public testing::Test,
      secondary_cache_opts.compression_type = CompressionType::kNoCompression;
    }

-    secondary_cache_opts.capacity = 1100;
+    secondary_cache_opts.capacity = 1400;
    secondary_cache_opts.num_shard_bits = 0;
+    secondary_cache_opts.strict_capacity_limit = true;
    std::shared_ptr<SecondaryCache> sec_cache =
        NewCompressedSecondaryCache(secondary_cache_opts);

@@ -193,7 +254,7 @@ class CompressedSecondaryCacheTestBase : public testing::Test,
    ASSERT_OK(sec_cache->Insert(key1, &item1, GetHelper(), false));

    // Insert and Lookup the second item.
-    std::string str2(rnd.RandomString(200));
+    std::string str2(rnd.RandomString(500));
    TestItem item2(str2.data(), str2.length());
    // Insert a dummy handle, k1 is not evicted.
    ASSERT_OK(sec_cache->Insert(key2, &item2, GetHelper(), false));
@@ -201,16 +262,23 @@ class CompressedSecondaryCacheTestBase : public testing::Test,
    std::unique_ptr<SecondaryCacheResultHandle> handle1 =
        sec_cache->Lookup(key1, GetHelper(), this, true, /*advise_erase=*/false,
                          /*stats=*/nullptr, kept_in_sec_cache);
-    ASSERT_EQ(handle1, nullptr);
+    ASSERT_NE(handle1, nullptr);
+    std::unique_ptr<TestItem> val1{static_cast<TestItem*>(handle1->Value())};
+    ASSERT_NE(val1, nullptr);
+    ASSERT_EQ(val1->ToString(), str1);
+    handle1.reset();

    // Insert k2 and k1 is evicted.
    ASSERT_OK(sec_cache->Insert(key2, &item2, GetHelper(), false));
+    handle1 =
+        sec_cache->Lookup(key1, GetHelper(), this, true, /*advise_erase=*/false,
+                          /*stats=*/nullptr, kept_in_sec_cache);
+    ASSERT_EQ(handle1, nullptr);
    std::unique_ptr<SecondaryCacheResultHandle> handle2 =
        sec_cache->Lookup(key2, GetHelper(), this, true, /*advise_erase=*/false,
                          /*stats=*/nullptr, kept_in_sec_cache);
    ASSERT_NE(handle2, nullptr);
-    std::unique_ptr<TestItem> val2 =
-        std::unique_ptr<TestItem>(static_cast<TestItem*>(handle2->Value()));
+    std::unique_ptr<TestItem> val2{static_cast<TestItem*>(handle2->Value())};
    ASSERT_NE(val2, nullptr);
    ASSERT_EQ(memcmp(val2->Buf(), item2.Buf(), item2.Size()), 0);

@@ -232,7 +300,7 @@ class CompressedSecondaryCacheTestBase : public testing::Test,
    // Save Fails.
    std::string str3 = rnd.RandomString(10);
    TestItem item3(str3.data(), str3.length());
-    // The Status is OK because a dummy handle is inserted.
+    // The first Status is OK because a dummy handle is inserted.
    ASSERT_OK(sec_cache->Insert(key3, &item3, GetHelperFail(), false));
    ASSERT_NOK(sec_cache->Insert(key3, &item3, GetHelperFail(), false));

@@ -265,11 +333,11 @@ class CompressedSecondaryCacheTestBase : public testing::Test,

    get_perf_context()->Reset();
    Random rnd(301);
-    std::string str1 = rnd.RandomString(1001);
+    std::string str1 = test::CompressibleString(&rnd, 0.5, 1001);
    auto item1_1 = new TestItem(str1.data(), str1.length());
    ASSERT_OK(cache->Insert(key1, item1_1, GetHelper(), str1.length()));

-    std::string str2 = rnd.RandomString(1012);
+    std::string str2 = test::CompressibleString(&rnd, 0.5, 1012);
    auto item2_1 = new TestItem(str2.data(), str2.length());
    // After this Insert, primary cache contains k2 and secondary cache contains
    // k1's dummy item.
@@ -278,7 +346,7 @@ class CompressedSecondaryCacheTestBase : public testing::Test,
    ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
    ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);

-    std::string str3 = rnd.RandomString(1024);
+    std::string str3 = test::CompressibleString(&rnd, 0.5, 1024);
    auto item3_1 = new TestItem(str3.data(), str3.length());
    // After this Insert, primary cache contains k3 and secondary cache contains
    // k1's dummy item and k2's dummy item.
@@ -297,10 +365,13 @@ class CompressedSecondaryCacheTestBase : public testing::Test,
    ASSERT_OK(cache->Insert(key2, item2_2, GetHelper(), str2.length()));
    ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 1);
    if (sec_cache_is_compressed) {
-      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
+      ASSERT_EQ(
+          Pop(get_perf_context()->compressed_sec_cache_uncompressed_bytes),
+          str1.length());
+      ASSERT_LT(get_perf_context()->compressed_sec_cache_compressed_bytes,
                str1.length());
-      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes,
-                1008);
+      ASSERT_GT(Pop(get_perf_context()->compressed_sec_cache_compressed_bytes),
+                str1.length() / 10);
    } else {
      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);
@@ -312,10 +383,13 @@ class CompressedSecondaryCacheTestBase : public testing::Test,
    ASSERT_OK(cache->Insert(key3, item3_2, GetHelper(), str3.length()));
    ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 2);
    if (sec_cache_is_compressed) {
-      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
-                str1.length() + str2.length());
-      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes,
-                2027);
+      ASSERT_EQ(
+          Pop(get_perf_context()->compressed_sec_cache_uncompressed_bytes),
+          str2.length());
+      ASSERT_LT(get_perf_context()->compressed_sec_cache_compressed_bytes,
+                str2.length());
+      ASSERT_GT(Pop(get_perf_context()->compressed_sec_cache_compressed_bytes),
+                str2.length() / 10);
    } else {
      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);
@@ -641,8 +715,7 @@ class CompressedSecondaryCacheTestBase : public testing::Test,
    size_t str_size{8500};
    std::string str = rnd.RandomString(static_cast<int>(str_size));
    size_t charge{0};
-    CacheValueChunk* chunks_head =
-        sec_cache->SplitValueIntoChunks(str, kLZ4Compression, charge);
+    CacheValueChunk* chunks_head = sec_cache->SplitValueIntoChunks(str, charge);
    ASSERT_EQ(charge, str_size + 3 * (sizeof(CacheValueChunk) - 1));

    CacheValueChunk* current_chunk = chunks_head;
@@ -688,12 +761,9 @@ class CompressedSecondaryCacheTestBase : public testing::Test,
    std::unique_ptr<CompressedSecondaryCache> sec_cache =
        std::make_unique<CompressedSecondaryCache>(
            CompressedSecondaryCacheOptions(1000, 0, true, 0.5, 0.0));
-    size_t charge{0};
-    CacheAllocationPtr value =
-        sec_cache->MergeChunksIntoValue(chunks_head, charge);
-    ASSERT_EQ(charge, size1 + size2 + size3);
-    std::string value_str{value.get(), charge};
-    ASSERT_EQ(strcmp(value_str.data(), str.data()), 0);
+    std::string value_str = sec_cache->MergeChunksIntoValue(chunks_head);
+    ASSERT_EQ(value_str.size(), size1 + size2 + size3);
+    ASSERT_EQ(value_str, str);

    while (chunks_head != nullptr) {
      CacheValueChunk* tmp_chunk = chunks_head;
@@ -725,15 +795,12 @@ class CompressedSecondaryCacheTestBase : public testing::Test,
    size_t str_size{8500};
    std::string str = rnd.RandomString(static_cast<int>(str_size));
    size_t charge{0};
-    CacheValueChunk* chunks_head =
-        sec_cache->SplitValueIntoChunks(str, kLZ4Compression, charge);
+    CacheValueChunk* chunks_head = sec_cache->SplitValueIntoChunks(str, charge);
    ASSERT_EQ(charge, str_size + 3 * (sizeof(CacheValueChunk) - 1));

-    CacheAllocationPtr value =
-        sec_cache->MergeChunksIntoValue(chunks_head, charge);
-    ASSERT_EQ(charge, str_size);
-    std::string value_str{value.get(), charge};
-    ASSERT_EQ(strcmp(value_str.data(), str.data()), 0);
+    std::string value_str = sec_cache->MergeChunksIntoValue(chunks_head);
+    ASSERT_EQ(value_str.size(), str_size);
+    ASSERT_EQ(value_str, str);

    sec_cache->GetHelper(true)->del_cb(chunks_head, /*alloc*/ nullptr);
  }
@@ -896,8 +963,8 @@ TEST_P(CompressedSecondaryCacheTestWithCompressionParam, EntryRoles) {

  std::shared_ptr<SecondaryCache> sec_cache = NewCompressedSecondaryCache(opts);

-  // Fixed seed to ensure consistent compressibility (doesn't compress)
-  std::string junk(Random(301).RandomString(1000));
+  Random rnd(301);
+  std::string junk = test::CompressibleString(&rnd, 0.5, 1000);

  for (uint32_t i = 0; i < kNumCacheEntryRoles; ++i) {
    CacheEntryRole role = static_cast<CacheEntryRole>(i);
@@ -930,9 +997,11 @@ TEST_P(CompressedSecondaryCacheTestWithCompressionParam, EntryRoles) {
        sec_cache_is_compressed_ && !do_not_compress.Contains(role);
    if (compressed) {
      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
-                1000);
-      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes,
-                1007);
+                junk.length());
+      ASSERT_LT(get_perf_context()->compressed_sec_cache_compressed_bytes,
+                junk.length() * 3 / 4);
+      ASSERT_GT(get_perf_context()->compressed_sec_cache_compressed_bytes,
+                junk.length() / 4);
    } else {
      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);
@@ -121,7 +121,14 @@ CacheWithSecondaryAdapter::~CacheWithSecondaryAdapter() {
    assert(s.ok());
    assert(placeholder_usage_ == 0);
    assert(reserved_usage_ == 0);
-    assert(pri_cache_res_->GetTotalMemoryUsed() == sec_capacity);
+    if (pri_cache_res_->GetTotalMemoryUsed() != sec_capacity) {
+      fprintf(stdout,
+              "~CacheWithSecondaryAdapter: Primary cache reservation: "
+              "%zu, Secondary cache capacity: %zu, "
+              "Secondary cache reserved: %zu\n",
+              pri_cache_res_->GetTotalMemoryUsed(), sec_capacity,
+              sec_reserved_);
+    }
  }
 #endif  // NDEBUG
 }
@@ -479,12 +486,10 @@ const char* CacheWithSecondaryAdapter::Name() const {
 // as well. At the moment, we don't have a good way of handling the case
 // where the new capacity < total cache reservations.
 void CacheWithSecondaryAdapter::SetCapacity(size_t capacity) {
-  size_t sec_capacity = static_cast<size_t>(
-      capacity * (distribute_cache_res_ ? sec_cache_res_ratio_ : 0.0));
-  size_t old_sec_capacity = 0;
-
  if (distribute_cache_res_) {
    MutexLock m(&cache_res_mutex_);
+    size_t sec_capacity = static_cast<size_t>(capacity * sec_cache_res_ratio_);
+    size_t old_sec_capacity = 0;

    Status s = secondary_cache_->GetCapacity(old_sec_capacity);
    if (!s.ok()) {
@@ -603,6 +608,7 @@ Status CacheWithSecondaryAdapter::UpdateCacheReservationRatio(
    //    cache utilization (increase in capacity - increase in share of cache
    //    reservation)
    // 3. Increase secondary cache capacity
+    assert(new_sec_reserved >= sec_reserved_);
    s = secondary_cache_->Deflate(new_sec_reserved - sec_reserved_);
    assert(s.ok());
    s = pri_cache_res_->UpdateCacheReservation(
@@ -11,18 +11,29 @@ CRASHTEST_MAKE=$(MAKE) -f crash_test.mk
 CRASHTEST_PY=$(PYTHON) -u tools/db_crashtest.py --stress_cmd=$(DB_STRESS_CMD) --cleanup_cmd='$(DB_CLEANUP_CMD)'

 .PHONY: crash_test crash_test_with_atomic_flush crash_test_with_txn \
+	crash_test_with_wc_txn crash_test_with_wp_txn crash_test_with_wup_txn \
 	crash_test_with_best_efforts_recovery crash_test_with_ts \
+	crash_test_with_multiops_wc_txn \
+	crash_test_with_multiops_wp_txn \
+	crash_test_with_multiops_wup_txn \
+	crash_test_with_optimistic_txn \
+	crash_test_with_tiered_storage \
 	blackbox_crash_test blackbox_crash_test_with_atomic_flush \
+	blackbox_crash_test_with_wc_txn blackbox_crash_test_with_wp_txn \
+	blackbox_crash_test_with_wup_txn \
 	blackbox_crash_test_with_txn blackbox_crash_test_with_ts \
 	blackbox_crash_test_with_best_efforts_recovery \
-	whitebox_crash_test whitebox_crash_test_with_atomic_flush \
-	whitebox_crash_test_with_txn whitebox_crash_test_with_ts \
 	blackbox_crash_test_with_multiops_wc_txn \
 	blackbox_crash_test_with_multiops_wp_txn \
-	crash_test_with_tiered_storage blackbox_crash_test_with_tiered_storage \
-	whitebox_crash_test_with_tiered_storage \
-	whitebox_crash_test_with_optimistic_txn \
+	blackbox_crash_test_with_multiops_wup_txn \
 	blackbox_crash_test_with_optimistic_txn \
+	blackbox_crash_test_with_tiered_storage \
+	whitebox_crash_test whitebox_crash_test_with_atomic_flush \
+	whitebox_crash_test_with_wc_txn whitebox_crash_test_with_wp_txn \
+	whitebox_crash_test_with_wup_txn \
+	whitebox_crash_test_with_txn whitebox_crash_test_with_ts \
+	whitebox_crash_test_with_optimistic_txn \
+	whitebox_crash_test_with_tiered_storage \

 crash_test: $(DB_STRESS_CMD)
 # Do not parallelize
@@ -34,10 +45,20 @@ crash_test_with_atomic_flush: $(DB_STRESS_CMD)
 	$(CRASHTEST_MAKE) whitebox_crash_test_with_atomic_flush
 	$(CRASHTEST_MAKE) blackbox_crash_test_with_atomic_flush

-crash_test_with_txn: $(DB_STRESS_CMD)
+crash_test_with_wc_txn: $(DB_STRESS_CMD)
 # Do not parallelize
-	$(CRASHTEST_MAKE) whitebox_crash_test_with_txn
-	$(CRASHTEST_MAKE) blackbox_crash_test_with_txn
+	$(CRASHTEST_MAKE) whitebox_crash_test_with_wc_txn
+	$(CRASHTEST_MAKE) blackbox_crash_test_with_wc_txn
+
+crash_test_with_wp_txn: $(DB_STRESS_CMD)
+# Do not parallelize
+	$(CRASHTEST_MAKE) whitebox_crash_test_with_wp_txn
+	$(CRASHTEST_MAKE) blackbox_crash_test_with_wp_txn
+
+crash_test_with_wup_txn: $(DB_STRESS_CMD)
+# Do not parallelize
+	$(CRASHTEST_MAKE) whitebox_crash_test_with_wup_txn
+	$(CRASHTEST_MAKE) blackbox_crash_test_with_wup_txn

 crash_test_with_optimistic_txn: $(DB_STRESS_CMD)
 # Do not parallelize
@@ -62,6 +83,9 @@ crash_test_with_multiops_wc_txn: $(DB_STRESS_CMD)
 crash_test_with_multiops_wp_txn: $(DB_STRESS_CMD)
 	$(CRASHTEST_MAKE) blackbox_crash_test_with_multiops_wp_txn

+crash_test_with_multiops_wup_txn: $(DB_STRESS_CMD)
+	$(CRASHTEST_MAKE) blackbox_crash_test_with_multiops_wup_txn
+
 blackbox_crash_test: $(DB_STRESS_CMD)
 	$(CRASHTEST_PY) --simple blackbox $(CRASH_TEST_EXT_ARGS)
 	$(CRASHTEST_PY) blackbox $(CRASH_TEST_EXT_ARGS)
@@ -69,8 +93,14 @@ blackbox_crash_test: $(DB_STRESS_CMD)
 blackbox_crash_test_with_atomic_flush: $(DB_STRESS_CMD)
 	$(CRASHTEST_PY) --cf_consistency blackbox $(CRASH_TEST_EXT_ARGS)

-blackbox_crash_test_with_txn: $(DB_STRESS_CMD)
-	$(CRASHTEST_PY) --txn blackbox $(CRASH_TEST_EXT_ARGS)
+blackbox_crash_test_with_wc_txn: $(DB_STRESS_CMD)
+	$(CRASHTEST_PY) --txn blackbox --txn_write_policy 0 $(CRASH_TEST_EXT_ARGS)
+
+blackbox_crash_test_with_wp_txn: $(DB_STRESS_CMD)
+	$(CRASHTEST_PY) --txn blackbox --txn_write_policy 1 $(CRASH_TEST_EXT_ARGS)
+
+blackbox_crash_test_with_wup_txn: $(DB_STRESS_CMD)
+	$(CRASHTEST_PY) --txn blackbox --txn_write_policy 2 $(CRASH_TEST_EXT_ARGS)

 blackbox_crash_test_with_best_efforts_recovery: $(DB_STRESS_CMD)
 	$(CRASHTEST_PY) --test_best_efforts_recovery blackbox $(CRASH_TEST_EXT_ARGS)
@@ -79,10 +109,13 @@ blackbox_crash_test_with_ts: $(DB_STRESS_CMD)
 	$(CRASHTEST_PY) --enable_ts blackbox $(CRASH_TEST_EXT_ARGS)

 blackbox_crash_test_with_multiops_wc_txn: $(DB_STRESS_CMD)
-	$(CRASHTEST_PY) --test_multiops_txn --write_policy write_committed blackbox $(CRASH_TEST_EXT_ARGS)
+	$(CRASHTEST_PY) --test_multiops_txn --txn_write_policy 0 blackbox $(CRASH_TEST_EXT_ARGS)

 blackbox_crash_test_with_multiops_wp_txn: $(DB_STRESS_CMD)
-	$(CRASHTEST_PY) --test_multiops_txn --write_policy write_prepared blackbox $(CRASH_TEST_EXT_ARGS)
+	$(CRASHTEST_PY) --test_multiops_txn --txn_write_policy 1 blackbox $(CRASH_TEST_EXT_ARGS)
+
+blackbox_crash_test_with_multiops_wup_txn: $(DB_STRESS_CMD)
+	$(CRASHTEST_PY) --test_multiops_txn --txn_write_policy 2 blackbox $(CRASH_TEST_EXT_ARGS)

 blackbox_crash_test_with_tiered_storage: $(DB_STRESS_CMD)
 	$(CRASHTEST_PY) --test_tiered_storage blackbox $(CRASH_TEST_EXT_ARGS)
@@ -104,9 +137,17 @@ whitebox_crash_test_with_atomic_flush: $(DB_STRESS_CMD)
 	$(CRASHTEST_PY) --cf_consistency whitebox  --random_kill_odd \
      $(CRASH_TEST_KILL_ODD) $(CRASH_TEST_EXT_ARGS)

-whitebox_crash_test_with_txn: $(DB_STRESS_CMD)
-	$(CRASHTEST_PY) --txn whitebox --random_kill_odd \
-      $(CRASH_TEST_KILL_ODD) $(CRASH_TEST_EXT_ARGS)
+whitebox_crash_test_with_wc_txn: $(DB_STRESS_CMD)
+	$(CRASHTEST_PY) --txn whitebox --txn_write_policy 0 \
+	  --random_kill_odd $(CRASH_TEST_KILL_ODD) $(CRASH_TEST_EXT_ARGS)
+
+whitebox_crash_test_with_wp_txn: $(DB_STRESS_CMD)
+	$(CRASHTEST_PY) --txn whitebox --txn_write_policy 1 \
+      --random_kill_odd $(CRASH_TEST_KILL_ODD) $(CRASH_TEST_EXT_ARGS)
+
+whitebox_crash_test_with_wup_txn: $(DB_STRESS_CMD)
+	$(CRASHTEST_PY) --txn whitebox --txn_write_policy 2 \
+      --random_kill_odd $(CRASH_TEST_KILL_ODD) $(CRASH_TEST_EXT_ARGS)

 whitebox_crash_test_with_ts: $(DB_STRESS_CMD)
 	$(CRASHTEST_PY) --enable_ts whitebox --random_kill_odd \
@@ -119,3 +160,8 @@ whitebox_crash_test_with_tiered_storage: $(DB_STRESS_CMD)
 whitebox_crash_test_with_optimistic_txn: $(DB_STRESS_CMD)
 	$(CRASHTEST_PY) --optimistic_txn whitebox --random_kill_odd \
      $(CRASH_TEST_KILL_ODD) $(CRASH_TEST_EXT_ARGS)
+
+# Old names DEPRECATED
+crash_test_with_txn: crash_test_with_wc_txn
+whitebox_crash_test_with_txn: whitebox_crash_test_with_wc_txn
+blackbox_crash_test_with_txn: blackbox_crash_test_with_wc_txn
@@ -42,9 +42,9 @@ Status ArenaWrappedDBIter::GetProperty(std::string prop_name,
 void ArenaWrappedDBIter::Init(
    Env* env, const ReadOptions& read_options, const ImmutableOptions& ioptions,
    const MutableCFOptions& mutable_cf_options, const Version* version,
-    const SequenceNumber& sequence, uint64_t max_sequential_skip_in_iteration,
-    uint64_t version_number, ReadCallback* read_callback,
-    ColumnFamilyHandleImpl* cfh, bool expose_blob_index, bool allow_refresh) {
+    const SequenceNumber& sequence, uint64_t version_number,
+    ReadCallback* read_callback, ColumnFamilyHandleImpl* cfh,
+    bool expose_blob_index, bool allow_refresh, ReadOnlyMemTable* active_mem) {
  read_options_ = read_options;
  if (!CheckFSFeatureSupport(env->GetFileSystem().get(),
                             FSSupportedOps::kAsyncIO)) {
@@ -52,15 +52,14 @@ void ArenaWrappedDBIter::Init(
  }
  read_options_.total_order_seek |= ioptions.prefix_seek_opt_in_only;

-  auto mem = arena_.AllocateAligned(sizeof(DBIter));
-  db_iter_ = new (mem) DBIter(env, read_options_, ioptions, mutable_cf_options,
-                              ioptions.user_comparator,
-                              /* iter */ nullptr, version, sequence, true,
-                              max_sequential_skip_in_iteration, read_callback,
-                              cfh, expose_blob_index);
+  db_iter_ = DBIter::NewIter(
+      env, read_options_, ioptions, mutable_cf_options,
+      ioptions.user_comparator, /*internal_iter=*/nullptr, version, sequence,
+      read_callback, active_mem, cfh, expose_blob_index, &arena_);

  sv_number_ = version_number;
  allow_refresh_ = allow_refresh;
+  allow_mark_memtable_for_flush_ = active_mem;
  memtable_range_tombstone_iter_ = nullptr;
 }

@@ -166,9 +165,8 @@ void ArenaWrappedDBIter::DoRefresh(const Snapshot* snapshot,
    read_callback_->Refresh(read_seq);
  }
  Init(env, read_options_, cfd->ioptions(), sv->mutable_cf_options, sv->current,
-       read_seq, sv->mutable_cf_options.max_sequential_skip_in_iterations,
-       sv->version_number, read_callback_, cfh_, expose_blob_index_,
-       allow_refresh_);
+       read_seq, sv->version_number, read_callback_, cfh_, expose_blob_index_,
+       allow_refresh_, allow_mark_memtable_for_flush_ ? sv->mem : nullptr);

  InternalIterator* internal_iter = db_impl->NewInternalIterator(
      read_options_, cfd, sv, &arena_, read_seq,
@@ -253,20 +251,26 @@ Status ArenaWrappedDBIter::Refresh(const Snapshot* snapshot) {
 }

 ArenaWrappedDBIter* NewArenaWrappedDbIterator(
-    Env* env, const ReadOptions& read_options, const ImmutableOptions& ioptions,
-    const MutableCFOptions& mutable_cf_options, const Version* version,
-    const SequenceNumber& sequence, uint64_t max_sequential_skip_in_iterations,
-    uint64_t version_number, ReadCallback* read_callback,
-    ColumnFamilyHandleImpl* cfh, bool expose_blob_index, bool allow_refresh) {
-  ArenaWrappedDBIter* iter = new ArenaWrappedDBIter();
-  iter->Init(env, read_options, ioptions, mutable_cf_options, version, sequence,
-             max_sequential_skip_in_iterations, version_number, read_callback,
-             cfh, expose_blob_index, allow_refresh);
+    Env* env, const ReadOptions& read_options, ColumnFamilyHandleImpl* cfh,
+    SuperVersion* sv, const SequenceNumber& sequence,
+    ReadCallback* read_callback, DBImpl* db_impl, bool expose_blob_index,
+    bool allow_refresh, bool allow_mark_memtable_for_flush) {
+  ArenaWrappedDBIter* db_iter = new ArenaWrappedDBIter();
+  db_iter->Init(env, read_options, cfh->cfd()->ioptions(),
+                sv->mutable_cf_options, sv->current, sequence,
+                sv->version_number, read_callback, cfh, expose_blob_index,
+                allow_refresh,
+                allow_mark_memtable_for_flush ? sv->mem : nullptr);
  if (cfh != nullptr && allow_refresh) {
-    iter->StoreRefreshInfo(cfh, read_callback, expose_blob_index);
+    db_iter->StoreRefreshInfo(cfh, read_callback, expose_blob_index);
  }

-  return iter;
+  InternalIterator* internal_iter = db_impl->NewInternalIterator(
+      db_iter->GetReadOptions(), cfh->cfd(), sv, db_iter->GetArena(), sequence,
+      /*allow_unprepared_value=*/true, db_iter);
+  db_iter->SetIterUnderDBIter(internal_iter);
+
+  return db_iter;
 }

 }  // namespace ROCKSDB_NAMESPACE
@@ -19,7 +19,6 @@
 #include "options/cf_options.h"
 #include "rocksdb/db.h"
 #include "rocksdb/iterator.h"
-#include "util/autovector.h"

 namespace ROCKSDB_NAMESPACE {

@@ -99,13 +98,19 @@ class ArenaWrappedDBIter : public Iterator {

  bool PrepareValue() override { return db_iter_->PrepareValue(); }

+  void Prepare(const MultiScanArgs& scan_opts) override {
+    db_iter_->Prepare(scan_opts);
+  }
+
+  // FIXME: we could just pass SV in for mutable cf option, version and version
+  // number, but this is used by SstFileReader which does not have a SV.
  void Init(Env* env, const ReadOptions& read_options,
            const ImmutableOptions& ioptions,
            const MutableCFOptions& mutable_cf_options, const Version* version,
-            const SequenceNumber& sequence,
-            uint64_t max_sequential_skip_in_iterations, uint64_t version_number,
+            const SequenceNumber& sequence, uint64_t version_number,
            ReadCallback* read_callback, ColumnFamilyHandleImpl* cfh,
-            bool expose_blob_index, bool allow_refresh);
+            bool expose_blob_index, bool allow_refresh,
+            ReadOnlyMemTable* active_mem);

  // Store some parameters so we can refresh the iterator at a later point
  // with these same params
@@ -128,20 +133,16 @@ class ArenaWrappedDBIter : public Iterator {
  ReadCallback* read_callback_;
  bool expose_blob_index_ = false;
  bool allow_refresh_ = true;
+  bool allow_mark_memtable_for_flush_ = true;
  // If this is nullptr, it means the mutable memtable does not contain range
  // tombstone when added under this DBIter.
  std::unique_ptr<TruncatedRangeDelIterator>* memtable_range_tombstone_iter_ =
      nullptr;
 };

-// Generate the arena wrapped iterator class.
-// `cfh` is used for reneweal. If left null, renewal will not
-// be supported.
 ArenaWrappedDBIter* NewArenaWrappedDbIterator(
-    Env* env, const ReadOptions& read_options, const ImmutableOptions& ioptions,
-    const MutableCFOptions& mutable_cf_options, const Version* version,
-    const SequenceNumber& sequence, uint64_t max_sequential_skip_in_iterations,
-    uint64_t version_number, ReadCallback* read_callback,
-    ColumnFamilyHandleImpl* cfh = nullptr, bool expose_blob_index = false,
-    bool allow_refresh = true);
+    Env* env, const ReadOptions& read_options, ColumnFamilyHandleImpl* cfh,
+    SuperVersion* sv, const SequenceNumber& sequence,
+    ReadCallback* read_callback, DBImpl* db_impl, bool expose_blob_index,
+    bool allow_refresh, bool allow_mark_memtable_for_flush);
 }  // namespace ROCKSDB_NAMESPACE
@@ -267,10 +267,9 @@ Status BlobFileBuilder::CompressBlobIfNeeded(
  // TODO: allow user CompressionOptions, including max_compressed_bytes_per_kb
  CompressionOptions opts;
  CompressionContext context(blob_compression_type_, opts);
-  constexpr uint64_t sample_for_compression = 0;

  CompressionInfo info(opts, context, CompressionDict::GetEmptyDict(),
-                       blob_compression_type_, sample_for_compression);
+                       blob_compression_type_);

  constexpr uint32_t compression_format_version = 2;

@@ -279,8 +278,8 @@ Status BlobFileBuilder::CompressBlobIfNeeded(
  {
    StopWatch stop_watch(immutable_options_->clock, immutable_options_->stats,
                         BLOB_DB_COMPRESSION_MICROS);
-    success =
-        CompressData(*blob, info, compression_format_version, compressed_blob);
+    success = OLD_CompressData(*blob, info, compression_format_version,
+                               compressed_blob);
  }

  if (!success) {
@@ -405,10 +405,9 @@ TEST_F(BlobFileBuilderTest, Compression) {

  CompressionOptions opts;
  CompressionContext context(kSnappyCompression, opts);
-  constexpr uint64_t sample_for_compression = 0;

  CompressionInfo info(opts, context, CompressionDict::GetEmptyDict(),
-                       kSnappyCompression, sample_for_compression);
+                       kSnappyCompression);

  std::string compressed_value;
  ASSERT_TRUE(Snappy_Compress(info, uncompressed_value.data(),
@@ -6,6 +6,7 @@
 #pragma once

 #include <cassert>
+#include <cstdint>
 #include <iosfwd>
 #include <memory>
 #include <string>
@@ -250,7 +250,8 @@ Status BlobFileReader::ReadFromFile(const RandomAccessFileReader* file_reader,
  Status s;

  IOOptions io_options;
-  s = file_reader->PrepareIOOptions(read_options, io_options);
+  IODebugContext dbg;
+  s = file_reader->PrepareIOOptions(read_options, io_options, &dbg);
  if (!s.ok()) {
    return s;
  }
@@ -259,13 +260,13 @@ Status BlobFileReader::ReadFromFile(const RandomAccessFileReader* file_reader,
    constexpr char* scratch = nullptr;

    s = file_reader->Read(io_options, read_offset, read_size, slice, scratch,
-                          aligned_buf);
+                          aligned_buf, &dbg);
  } else {
    buf->reset(new char[read_size]);
    constexpr AlignedBuf* aligned_scratch = nullptr;

    s = file_reader->Read(io_options, read_offset, read_size, slice, buf->get(),
-                          aligned_scratch);
+                          aligned_scratch, &dbg);
  }

  if (!s.ok()) {
@@ -334,7 +335,8 @@ Status BlobFileReader::GetBlob(
    constexpr bool for_compaction = true;

    IOOptions io_options;
-    s = file_reader_->PrepareIOOptions(read_options, io_options);
+    IODebugContext dbg;
+    s = file_reader_->PrepareIOOptions(read_options, io_options, &dbg);
    if (!s.ok()) {
      return s;
    }
@@ -463,10 +465,11 @@ void BlobFileReader::MultiGetBlob(
  PERF_COUNTER_ADD(blob_read_count, num_blobs);
  PERF_COUNTER_ADD(blob_read_byte, total_len);
  IOOptions opts;
-  s = file_reader_->PrepareIOOptions(read_options, opts);
+  IODebugContext dbg;
+  s = file_reader_->PrepareIOOptions(read_options, opts, &dbg);
  if (s.ok()) {
    s = file_reader_->MultiRead(opts, read_reqs.data(), read_reqs.size(),
-                                direct_io ? &aligned_buf : nullptr);
+                                direct_io ? &aligned_buf : nullptr, &dbg);
  }
  if (!s.ok()) {
    for (auto& req : read_reqs) {
@@ -602,9 +605,9 @@ Status BlobFileReader::UncompressBlobIfNeeded(
  {
    PERF_TIMER_GUARD(blob_decompress_time);
    StopWatch stop_watch(clock, statistics, BLOB_DB_DECOMPRESSION_MICROS);
-    output = UncompressData(info, value_slice.data(), value_slice.size(),
-                            &uncompressed_size, compression_format_version,
-                            allocator);
+    output = OLD_UncompressData(info, value_slice.data(), value_slice.size(),
+                                &uncompressed_size, compression_format_version,
+                                allocator);
  }

  TEST_SYNC_POINT_CALLBACK(
@@ -75,15 +75,14 @@ void WriteBlobFile(const ImmutableOptions& immutable_options,
  } else {
    CompressionOptions opts;
    CompressionContext context(compression, opts);
-    constexpr uint64_t sample_for_compression = 0;
    CompressionInfo info(opts, context, CompressionDict::GetEmptyDict(),
-                         compression, sample_for_compression);
+                         compression);

    constexpr uint32_t compression_format_version = 2;

    for (size_t i = 0; i < num; ++i) {
-      ASSERT_TRUE(CompressData(blobs[i], info, compression_format_version,
-                               &compressed_blobs[i]));
+      ASSERT_TRUE(OLD_CompressData(blobs[i], info, compression_format_version,
+                                   &compressed_blobs[i]));
      blobs_to_write[i] = compressed_blobs[i];
      blob_sizes[i] = compressed_blobs[i].size();
    }
@@ -77,15 +77,14 @@ void WriteBlobFile(const ImmutableOptions& immutable_options,
  } else {
    CompressionOptions opts;
    CompressionContext context(compression, opts);
-    constexpr uint64_t sample_for_compression = 0;
    CompressionInfo info(opts, context, CompressionDict::GetEmptyDict(),
-                         compression, sample_for_compression);
+                         compression);

    constexpr uint32_t compression_format_version = 2;

    for (size_t i = 0; i < num; ++i) {
-      ASSERT_TRUE(CompressData(blobs[i], info, compression_format_version,
-                               &compressed_blobs[i]));
+      ASSERT_TRUE(OLD_CompressData(blobs[i], info, compression_format_version,
+                                   &compressed_blobs[i]));
      blobs_to_write[i] = compressed_blobs[i];
      blob_sizes[i] = compressed_blobs[i].size();
    }
@@ -74,8 +74,8 @@ Status BuildTable(
    EventLogger* event_logger, int job_id, TableProperties* table_properties,
    Env::WriteLifeTimeHint write_hint, const std::string* full_history_ts_low,
    BlobFileCompletionCallback* blob_callback, Version* version,
-    uint64_t* num_input_entries, uint64_t* memtable_payload_bytes,
-    uint64_t* memtable_garbage_bytes) {
+    uint64_t* memtable_payload_bytes, uint64_t* memtable_garbage_bytes,
+    InternalStats::CompactionStats* flush_stats) {
  assert((tboptions.column_family_id ==
          TablePropertiesCollectorFactory::Context::kUnknownColumnFamily) ==
         tboptions.column_family_name.empty());
@@ -201,8 +201,7 @@ Status BuildTable(
    CompactionIterator c_iter(
        iter, ucmp, &merge, kMaxSequenceNumber, &snapshots, earliest_snapshot,
        earliest_write_conflict_snapshot, job_snapshot, snapshot_checker, env,
-        ShouldReportDetailedTime(env, ioptions.stats),
-        true /* internal key corruption is not ok */, range_del_agg.get(),
+        ShouldReportDetailedTime(env, ioptions.stats), range_del_agg.get(),
        blob_file_builder.get(), ioptions.allow_data_in_errors,
        ioptions.enforce_single_del_contracts,
        /*manual_compaction_canceled=*/kManualCompactionCanceledFalse,
@@ -218,8 +217,7 @@ Status BuildTable(
      const Slice& key = c_iter.key();
      const Slice& value = c_iter.value();
      ParsedInternalKey ikey = c_iter.ikey();
-      key_after_flush_buf.assign(key.data(), key.size());
-      Slice key_after_flush = key_after_flush_buf;
+      Slice key_after_flush = key;
      Slice value_after_flush = value;

      if (ikey.type == kTypeValuePreferredSeqno) {
@@ -237,6 +235,7 @@ Status BuildTable(
              std::min(smallest_preferred_seqno, preferred_seqno);
        } else {
          // Cannot get a useful preferred seqno, convert it to a kTypeValue.
+          key_after_flush_buf.assign(key.data(), key.size());
          UpdateInternalKey(&key_after_flush_buf, ikey.sequence, kTypeValue);
          ikey = ParsedInternalKey(ikey.user_key, ikey.sequence, kTypeValue);
          key_after_flush = key_after_flush_buf;
@@ -253,6 +252,10 @@ Status BuildTable(
      }
      builder->Add(key_after_flush, value_after_flush);

+      if (flush_stats) {
+        flush_stats->num_output_records++;
+      }
+
      s = meta->UpdateBoundaries(key_after_flush, value_after_flush,
                                 ikey.sequence, ikey.type);
      if (!s.ok()) {
@@ -284,6 +287,9 @@ Status BuildTable(
        auto tombstone = range_del_it->Tombstone();
        std::pair<InternalKey, Slice> kv = tombstone.Serialize();
        builder->Add(kv.first.Encode(), kv.second);
+        if (flush_stats) {
+          flush_stats->num_output_records++;
+        }
        InternalKey tombstone_end = tombstone.SerializeEndKey();
        meta->UpdateBoundariesForRange(kv.first, tombstone_end, tombstone.seq_,
                                       tboptions.internal_comparator);
@@ -305,9 +311,9 @@ Status BuildTable(

    TEST_SYNC_POINT("BuildTable:BeforeFinishBuildTable");
    const bool empty = builder->IsEmpty();
-    if (num_input_entries != nullptr) {
+    if (flush_stats) {
      assert(c_iter.HasNumInputEntryScanned());
-      *num_input_entries =
+      flush_stats->num_input_records =
          c_iter.NumInputEntryScanned() + num_unfragmented_tombstones;
    }
    if (!s.ok() || empty) {
@@ -334,6 +340,9 @@ Status BuildTable(
    }

    if (s.ok() && !empty) {
+      if (flush_stats) {
+        flush_stats->bytes_written_pre_comp = builder->PreCompressionSize();
+      }
      uint64_t file_size = builder->FileSize();
      meta->fd.file_size = file_size;
      meta->tail_size = builder->GetTailSize();
@@ -10,6 +10,7 @@
 #include <utility>
 #include <vector>

+#include "db/internal_stats.h"
 #include "db/range_tombstone_fragmenter.h"
 #include "db/seqno_to_time_mapping.h"
 #include "db/table_properties_collector.h"
@@ -34,7 +35,6 @@ class SnapshotChecker;
 class TableCache;
 class TableBuilder;
 class WritableFileWriter;
-class InternalStats;
 class BlobFileCompletionCallback;

 // Convenience function for NewTableBuilder on the embedded table_factory.
@@ -49,6 +49,7 @@ TableBuilder* NewTableBuilder(const TableBuilderOptions& tboptions,
 //
 // @param column_family_name Name of the column family that is also identified
 //    by column_family_id, or empty string if unknown.
+// @param flush_stats treat flush as level 0 compaction in internal stats
 Status BuildTable(
    const std::string& dbname, VersionSet* versions,
    const ImmutableDBOptions& db_options, const TableBuilderOptions& tboptions,
@@ -69,8 +70,8 @@ Status BuildTable(
    Env::WriteLifeTimeHint write_hint = Env::WLTH_NOT_SET,
    const std::string* full_history_ts_low = nullptr,
    BlobFileCompletionCallback* blob_callback = nullptr,
-    Version* version = nullptr, uint64_t* num_input_entries = nullptr,
-    uint64_t* memtable_payload_bytes = nullptr,
-    uint64_t* memtable_garbage_bytes = nullptr);
+    Version* version = nullptr, uint64_t* memtable_payload_bytes = nullptr,
+    uint64_t* memtable_garbage_bytes = nullptr,
+    InternalStats::CompactionStats* flush_stats = nullptr);

 }  // namespace ROCKSDB_NAMESPACE
@@ -24,12 +24,14 @@
 #include "rocksdb/experimental.h"
 #include "rocksdb/filter_policy.h"
 #include "rocksdb/iterator.h"
+#include "rocksdb/listener.h"
 #include "rocksdb/memtablerep.h"
 #include "rocksdb/merge_operator.h"
 #include "rocksdb/options.h"
 #include "rocksdb/perf_context.h"
 #include "rocksdb/rate_limiter.h"
 #include "rocksdb/slice_transform.h"
+#include "rocksdb/sst_file_manager.h"
 #include "rocksdb/statistics.h"
 #include "rocksdb/status.h"
 #include "rocksdb/table.h"
@@ -49,6 +51,7 @@
 #include "util/stderr_logger.h"
 #include "utilities/merge_operators.h"

+using ROCKSDB_NAMESPACE::BackgroundErrorReason;
 using ROCKSDB_NAMESPACE::BackupEngine;
 using ROCKSDB_NAMESPACE::BackupEngineOptions;
 using ROCKSDB_NAMESPACE::BackupID;
@@ -65,7 +68,9 @@ using ROCKSDB_NAMESPACE::ColumnFamilyMetaData;
 using ROCKSDB_NAMESPACE::ColumnFamilyOptions;
 using ROCKSDB_NAMESPACE::CompactionFilter;
 using ROCKSDB_NAMESPACE::CompactionFilterFactory;
+using ROCKSDB_NAMESPACE::CompactionJobInfo;
 using ROCKSDB_NAMESPACE::CompactionOptionsFIFO;
+using ROCKSDB_NAMESPACE::CompactionReason;
 using ROCKSDB_NAMESPACE::CompactRangeOptions;
 using ROCKSDB_NAMESPACE::Comparator;
 using ROCKSDB_NAMESPACE::CompressionType;
@@ -76,8 +81,11 @@ using ROCKSDB_NAMESPACE::DBOptions;
 using ROCKSDB_NAMESPACE::DbPath;
 using ROCKSDB_NAMESPACE::Env;
 using ROCKSDB_NAMESPACE::EnvOptions;
+using ROCKSDB_NAMESPACE::EventListener;
+using ROCKSDB_NAMESPACE::ExternalFileIngestionInfo;
 using ROCKSDB_NAMESPACE::FileLock;
 using ROCKSDB_NAMESPACE::FilterPolicy;
+using ROCKSDB_NAMESPACE::FlushJobInfo;
 using ROCKSDB_NAMESPACE::FlushOptions;
 using ROCKSDB_NAMESPACE::HistogramData;
 using ROCKSDB_NAMESPACE::HyperClockCacheOptions;
@@ -90,6 +98,7 @@ using ROCKSDB_NAMESPACE::Logger;
 using ROCKSDB_NAMESPACE::LRUCacheOptions;
 using ROCKSDB_NAMESPACE::MemoryAllocator;
 using ROCKSDB_NAMESPACE::MemoryUtil;
+using ROCKSDB_NAMESPACE::MemTableInfo;
 using ROCKSDB_NAMESPACE::MergeOperator;
 using ROCKSDB_NAMESPACE::NewBloomFilterPolicy;
 using ROCKSDB_NAMESPACE::NewCompactOnDeletionCollectorFactory;
@@ -113,10 +122,12 @@ using ROCKSDB_NAMESPACE::Slice;
 using ROCKSDB_NAMESPACE::SliceParts;
 using ROCKSDB_NAMESPACE::SliceTransform;
 using ROCKSDB_NAMESPACE::Snapshot;
+using ROCKSDB_NAMESPACE::SstFileManager;
 using ROCKSDB_NAMESPACE::SstFileMetaData;
 using ROCKSDB_NAMESPACE::SstFileWriter;
 using ROCKSDB_NAMESPACE::Status;
 using ROCKSDB_NAMESPACE::StderrLogger;
+using ROCKSDB_NAMESPACE::SubcompactionJobInfo;
 using ROCKSDB_NAMESPACE::TablePropertiesCollectorFactory;
 using ROCKSDB_NAMESPACE::Transaction;
 using ROCKSDB_NAMESPACE::TransactionDB;
@@ -130,6 +141,8 @@ using ROCKSDB_NAMESPACE::WriteBatch;
 using ROCKSDB_NAMESPACE::WriteBatchWithIndex;
 using ROCKSDB_NAMESPACE::WriteBufferManager;
 using ROCKSDB_NAMESPACE::WriteOptions;
+using ROCKSDB_NAMESPACE::WriteStallCondition;
+using ROCKSDB_NAMESPACE::WriteStallInfo;

 using std::unordered_set;
 using std::vector;
@@ -139,6 +152,9 @@ extern "C" {
 struct rocksdb_t {
  DB* rep;
 };
+struct rocksdb_status_ptr_t {
+  Status* rep;
+};
 struct rocksdb_backup_engine_t {
  BackupEngine* rep;
 };
@@ -226,6 +242,9 @@ struct rocksdb_cache_t {
 struct rocksdb_write_buffer_manager_t {
  std::shared_ptr<WriteBufferManager> rep;
 };
+struct rocksdb_sst_file_manager_t {
+  std::shared_ptr<SstFileManager> rep;
+};
 struct rocksdb_livefiles_t {
  std::vector<LiveFileMetaData> rep;
 };
@@ -292,6 +311,28 @@ struct rocksdb_compactionfiltercontext_t {
  CompactionFilter::Context rep;
 };

+struct rocksdb_flushjobinfo_t {
+  FlushJobInfo rep;
+};
+struct rocksdb_writestallcondition_t {
+  WriteStallCondition rep;
+};
+struct rocksdb_writestallinfo_t {
+  WriteStallInfo rep;
+};
+struct rocksdb_memtableinfo_t {
+  MemTableInfo rep;
+};
+struct rocksdb_compactionjobinfo_t {
+  CompactionJobInfo rep;
+};
+struct rocksdb_subcompactionjobinfo_t {
+  SubcompactionJobInfo rep;
+};
+struct rocksdb_externalfileingestioninfo_t {
+  ExternalFileIngestionInfo rep;
+};
+
 struct rocksdb_statistics_histogram_data_t {
  rocksdb_statistics_histogram_data_t() : rep() {}
  HistogramData rep;
@@ -884,6 +925,10 @@ void rocksdb_backup_engine_options_destroy(
  delete options;
 }

+void rocksdb_status_ptr_get_error(rocksdb_status_ptr_t* status, char** errptr) {
+  SaveError(errptr, *(status->rep));
+}
+
 rocksdb_checkpoint_t* rocksdb_checkpoint_object_create(rocksdb_t* db,
                                                       char** errptr) {
  Checkpoint* checkpoint;
@@ -2627,6 +2672,16 @@ rocksdb_iterator_t* rocksdb_writebatch_wi_create_iterator_with_base(
  return result;
 }

+rocksdb_iterator_t* rocksdb_writebatch_wi_create_iterator_with_base_readopts(
+    rocksdb_writebatch_wi_t* wbwi, rocksdb_iterator_t* base_iterator,
+    const rocksdb_readoptions_t* options) {
+  rocksdb_iterator_t* result = new rocksdb_iterator_t;
+  result->rep =
+      wbwi->rep->NewIteratorWithBase(base_iterator->rep, &options->rep);
+  delete base_iterator;
+  return result;
+}
+
 rocksdb_iterator_t* rocksdb_writebatch_wi_create_iterator_with_base_cf(
    rocksdb_writebatch_wi_t* wbwi, rocksdb_iterator_t* base_iterator,
    rocksdb_column_family_handle_t* column_family) {
@@ -2637,6 +2692,17 @@ rocksdb_iterator_t* rocksdb_writebatch_wi_create_iterator_with_base_cf(
  return result;
 }

+rocksdb_iterator_t* rocksdb_writebatch_wi_create_iterator_with_base_cf_readopts(
+    rocksdb_writebatch_wi_t* wbwi, rocksdb_iterator_t* base_iterator,
+    rocksdb_column_family_handle_t* column_family,
+    const rocksdb_readoptions_t* options) {
+  rocksdb_iterator_t* result = new rocksdb_iterator_t;
+  result->rep = wbwi->rep->NewIteratorWithBase(
+      column_family->rep, base_iterator->rep, &options->rep);
+  delete base_iterator;
+  return result;
+}
+
 char* rocksdb_writebatch_wi_get_from_batch(rocksdb_writebatch_wi_t* wbwi,
                                           const rocksdb_options_t* options,
                                           const char* key, size_t keylen,
@@ -2696,6 +2762,23 @@ char* rocksdb_writebatch_wi_get_from_batch_and_db(
  return result;
 }

+rocksdb_pinnableslice_t* rocksdb_writebatch_wi_get_pinned_from_batch_and_db(
+    rocksdb_writebatch_wi_t* wbwi, rocksdb_t* db,
+    const rocksdb_readoptions_t* options, const char* key, size_t keylen,
+    char** errptr) {
+  rocksdb_pinnableslice_t* v = new (rocksdb_pinnableslice_t);
+  Status s = wbwi->rep->GetFromBatchAndDB(db->rep, options->rep,
+                                          Slice(key, keylen), &v->rep);
+  if (!s.ok()) {
+    delete (v);
+    if (!s.IsNotFound()) {
+      SaveError(errptr, s);
+    }
+    return nullptr;
+  }
+  return v;
+}
+
 char* rocksdb_writebatch_wi_get_from_batch_and_db_cf(
    rocksdb_writebatch_wi_t* wbwi, rocksdb_t* db,
    const rocksdb_readoptions_t* options,
@@ -2717,6 +2800,24 @@ char* rocksdb_writebatch_wi_get_from_batch_and_db_cf(
  return result;
 }

+rocksdb_pinnableslice_t* rocksdb_writebatch_wi_get_pinned_from_batch_and_db_cf(
+    rocksdb_writebatch_wi_t* wbwi, rocksdb_t* db,
+    const rocksdb_readoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    size_t keylen, char** errptr) {
+  rocksdb_pinnableslice_t* v = new (rocksdb_pinnableslice_t);
+  Status s = wbwi->rep->GetFromBatchAndDB(
+      db->rep, options->rep, column_family->rep, Slice(key, keylen), &v->rep);
+  if (!s.ok()) {
+    delete (v);
+    if (!s.IsNotFound()) {
+      SaveError(errptr, s);
+    }
+    return nullptr;
+  }
+  return v;
+}
+
 void rocksdb_write_writebatch_wi(rocksdb_t* db,
                                 const rocksdb_writeoptions_t* options,
                                 rocksdb_writebatch_wi_t* wbwi, char** errptr) {
@@ -2930,6 +3031,361 @@ void rocksdb_block_based_options_set_unpartitioned_pinning_tier(
      static_cast<ROCKSDB_NAMESPACE::PinningTier>(v);
 }

+/* FlushJobInfo */
+
+const char* rocksdb_flushjobinfo_cf_name(const rocksdb_flushjobinfo_t* info,
+                                         size_t* size) {
+  *size = info->rep.cf_name.size();
+  return info->rep.cf_name.data();
+}
+
+const char* rocksdb_flushjobinfo_file_path(const rocksdb_flushjobinfo_t* info,
+                                           size_t* size) {
+  *size = info->rep.file_path.size();
+  return info->rep.file_path.data();
+}
+
+unsigned char rocksdb_flushjobinfo_triggered_writes_slowdown(
+    const rocksdb_flushjobinfo_t* info) {
+  return info->rep.triggered_writes_slowdown;
+}
+
+unsigned char rocksdb_flushjobinfo_triggered_writes_stop(
+    const rocksdb_flushjobinfo_t* info) {
+  return info->rep.triggered_writes_stop;
+}
+
+uint64_t rocksdb_flushjobinfo_largest_seqno(
+    const rocksdb_flushjobinfo_t* info) {
+  return info->rep.largest_seqno;
+}
+
+uint64_t rocksdb_flushjobinfo_smallest_seqno(
+    const rocksdb_flushjobinfo_t* info) {
+  return info->rep.smallest_seqno;
+}
+
+uint32_t rocksdb_flushjobinfo_flush_reason(const rocksdb_flushjobinfo_t* info) {
+  return static_cast<uint32_t>(info->rep.flush_reason);
+}
+
+void rocksdb_reset_status(rocksdb_status_ptr_t* status_ptr) {
+  auto ptr = status_ptr->rep;
+  *ptr = Status::OK();
+}
+
+/* CompactionJobInfo */
+
+void rocksdb_compactionjobinfo_status(const rocksdb_compactionjobinfo_t* info,
+                                      char** errptr) {
+  SaveError(errptr, info->rep.status);
+}
+
+const char* rocksdb_compactionjobinfo_cf_name(
+    const rocksdb_compactionjobinfo_t* info, size_t* size) {
+  *size = info->rep.cf_name.size();
+  return info->rep.cf_name.data();
+}
+
+size_t rocksdb_compactionjobinfo_input_files_count(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.input_files.size();
+}
+
+const char* rocksdb_compactionjobinfo_input_file_at(
+    const rocksdb_compactionjobinfo_t* info, size_t pos, size_t* size) {
+  assert(info != nullptr);
+  assert(pos < info->rep.input_files.size());
+
+  const std::string& path = info->rep.input_files[pos];
+  *size = path.size();
+  return path.data();
+}
+
+size_t rocksdb_compactionjobinfo_output_files_count(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.output_files.size();
+}
+
+const char* rocksdb_compactionjobinfo_output_file_at(
+    const rocksdb_compactionjobinfo_t* info, size_t pos, size_t* size) {
+  assert(info != nullptr);
+  assert(pos < info->rep.output_files.size());
+
+  const std::string& path = info->rep.output_files[pos];
+  *size = path.size();
+  return path.data();
+}
+
+uint64_t rocksdb_compactionjobinfo_elapsed_micros(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.elapsed_micros;
+}
+
+uint64_t rocksdb_compactionjobinfo_num_corrupt_keys(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.num_corrupt_keys;
+}
+
+int rocksdb_compactionjobinfo_base_input_level(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.base_input_level;
+}
+
+int rocksdb_compactionjobinfo_output_level(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.output_level;
+}
+
+size_t rocksdb_compactionjobinfo_num_input_files(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.num_input_files;
+}
+
+size_t rocksdb_compactionjobinfo_num_input_files_at_output_level(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.num_input_files_at_output_level;
+}
+
+uint64_t rocksdb_compactionjobinfo_input_records(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.num_input_records;
+}
+
+uint64_t rocksdb_compactionjobinfo_output_records(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.num_output_records;
+}
+
+uint64_t rocksdb_compactionjobinfo_total_input_bytes(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.total_input_bytes;
+}
+
+uint64_t rocksdb_compactionjobinfo_total_output_bytes(
+    const rocksdb_compactionjobinfo_t* info) {
+  return info->rep.stats.total_output_bytes;
+}
+
+uint32_t rocksdb_compactionjobinfo_compaction_reason(
+    const rocksdb_compactionjobinfo_t* info) {
+  return static_cast<uint32_t>(info->rep.compaction_reason);
+}
+
+/* SubcompactionJobInfo */
+
+void rocksdb_subcompactionjobinfo_status(
+    const rocksdb_subcompactionjobinfo_t* info, char** errptr) {
+  SaveError(errptr, info->rep.status);
+}
+
+const char* rocksdb_subcompactionjobinfo_cf_name(
+    const rocksdb_subcompactionjobinfo_t* info, size_t* size) {
+  *size = info->rep.cf_name.size();
+  return info->rep.cf_name.data();
+}
+
+uint64_t rocksdb_subcompactionjobinfo_thread_id(
+    const rocksdb_subcompactionjobinfo_t* info) {
+  return info->rep.thread_id;
+}
+
+int rocksdb_subcompactionjobinfo_base_input_level(
+    const rocksdb_subcompactionjobinfo_t* info) {
+  return info->rep.base_input_level;
+}
+
+int rocksdb_subcompactionjobinfo_output_level(
+    const rocksdb_subcompactionjobinfo_t* info) {
+  return info->rep.output_level;
+}
+
+uint32_t rocksdb_subcompactionjobinfo_compaction_reason(
+    const rocksdb_subcompactionjobinfo_t* info) {
+  return static_cast<uint32_t>(info->rep.compaction_reason);
+}
+
+/* ExternalFileIngestionInfo */
+
+const char* rocksdb_externalfileingestioninfo_cf_name(
+    const rocksdb_externalfileingestioninfo_t* info, size_t* size) {
+  *size = info->rep.cf_name.size();
+  return info->rep.cf_name.data();
+}
+
+const char* rocksdb_externalfileingestioninfo_internal_file_path(
+    const rocksdb_externalfileingestioninfo_t* info, size_t* size) {
+  *size = info->rep.internal_file_path.size();
+  return info->rep.internal_file_path.data();
+}
+
+/* External write stall info */
+extern ROCKSDB_LIBRARY_API const char* rocksdb_writestallinfo_cf_name(
+    const rocksdb_writestallinfo_t* info, size_t* size) {
+  *size = info->rep.cf_name.size();
+  return info->rep.cf_name.data();
+}
+
+const rocksdb_writestallcondition_t* rocksdb_writestallinfo_cur(
+    const rocksdb_writestallinfo_t* info) {
+  return reinterpret_cast<const rocksdb_writestallcondition_t*>(
+      &info->rep.condition.cur);
+}
+
+const rocksdb_writestallcondition_t* rocksdb_writestallinfo_prev(
+    const rocksdb_writestallinfo_t* info) {
+  return reinterpret_cast<const rocksdb_writestallcondition_t*>(
+      &info->rep.condition.prev);
+}
+
+const char* rocksdb_memtableinfo_cf_name(const rocksdb_memtableinfo_t* info,
+                                         size_t* size) {
+  *size = info->rep.cf_name.size();
+  return info->rep.cf_name.data();
+}
+
+uint64_t rocksdb_memtableinfo_first_seqno(const rocksdb_memtableinfo_t* info) {
+  return info->rep.first_seqno;
+}
+uint64_t rocksdb_memtableinfo_earliest_seqno(
+    const rocksdb_memtableinfo_t* info) {
+  return info->rep.earliest_seqno;
+}
+uint64_t rocksdb_memtableinfo_num_entries(const rocksdb_memtableinfo_t* info) {
+  return info->rep.num_entries;
+}
+uint64_t rocksdb_memtableinfo_num_deletes(const rocksdb_memtableinfo_t* info) {
+  return info->rep.num_deletes;
+}
+
+/* event listener */
+
+struct rocksdb_eventlistener_t : public EventListener {
+  void* state_{};
+  void (*destructor_)(void*){};
+  void (*on_flush_begin)(void*, rocksdb_t*, const rocksdb_flushjobinfo_t*){};
+  void (*on_flush_completed)(void*, rocksdb_t*,
+                             const rocksdb_flushjobinfo_t*){};
+  void (*on_compaction_begin)(void*, rocksdb_t*,
+                              const rocksdb_compactionjobinfo_t*){};
+  void (*on_compaction_completed)(void*, rocksdb_t*,
+                                  const rocksdb_compactionjobinfo_t*){};
+  void (*on_subcompaction_begin)(void*,
+                                 const rocksdb_subcompactionjobinfo_t*){};
+  void (*on_subcompaction_completed)(void*,
+                                     const rocksdb_subcompactionjobinfo_t*){};
+  void (*on_external_file_ingested)(
+      void*, rocksdb_t*, const rocksdb_externalfileingestioninfo_t*){};
+  void (*on_background_error)(void*, uint32_t, rocksdb_status_ptr_t*){};
+  void (*on_stall_conditions_changed)(void*, const rocksdb_writestallinfo_t*){};
+  void (*on_memtable_sealed)(void*, const rocksdb_memtableinfo_t*){};
+
+  rocksdb_eventlistener_t() = default;
+
+  rocksdb_eventlistener_t(const rocksdb_eventlistener_t&) = delete;
+  rocksdb_eventlistener_t& operator=(const rocksdb_eventlistener_t&) = delete;
+  rocksdb_eventlistener_t(rocksdb_eventlistener_t&&) = delete;
+  rocksdb_eventlistener_t& operator=(rocksdb_eventlistener_t&&) = delete;
+
+  void OnFlushBegin(DB* db, const FlushJobInfo& info) override {
+    rocksdb_t c_db = {db};
+    on_flush_begin(state_, &c_db,
+                   reinterpret_cast<const rocksdb_flushjobinfo_t*>(&info));
+  }
+
+  void OnFlushCompleted(DB* db, const FlushJobInfo& info) override {
+    rocksdb_t c_db = {db};
+    on_flush_completed(state_, &c_db,
+                       reinterpret_cast<const rocksdb_flushjobinfo_t*>(&info));
+  }
+
+  void OnCompactionBegin(DB* db, const CompactionJobInfo& info) override {
+    rocksdb_t c_db = {db};
+    on_compaction_begin(
+        state_, &c_db,
+        reinterpret_cast<const rocksdb_compactionjobinfo_t*>(&info));
+  }
+
+  void OnCompactionCompleted(DB* db, const CompactionJobInfo& info) override {
+    rocksdb_t c_db = {db};
+    on_compaction_completed(
+        state_, &c_db,
+        reinterpret_cast<const rocksdb_compactionjobinfo_t*>(&info));
+  }
+
+  void OnSubcompactionBegin(const SubcompactionJobInfo& info) override {
+    on_subcompaction_begin(
+        state_, reinterpret_cast<const rocksdb_subcompactionjobinfo_t*>(&info));
+  }
+
+  void OnSubcompactionCompleted(const SubcompactionJobInfo& info) override {
+    on_subcompaction_completed(
+        state_, reinterpret_cast<const rocksdb_subcompactionjobinfo_t*>(&info));
+  }
+
+  void OnExternalFileIngested(DB* db,
+                              const ExternalFileIngestionInfo& info) override {
+    rocksdb_t c_db = {db};
+    on_external_file_ingested(
+        state_, &c_db,
+        reinterpret_cast<const rocksdb_externalfileingestioninfo_t*>(&info));
+  }
+
+  void OnBackgroundError(BackgroundErrorReason reason,
+                         Status* status) override {
+    rocksdb_status_ptr_t* s = new rocksdb_status_ptr_t;
+    s->rep = status;
+    on_background_error(state_, static_cast<uint32_t>(reason), s);
+    delete s;
+  }
+
+  void OnStallConditionsChanged(const WriteStallInfo& info) override {
+    on_stall_conditions_changed(
+        state_, reinterpret_cast<const rocksdb_writestallinfo_t*>(&info));
+  }
+
+  void OnMemTableSealed(const MemTableInfo& info) override {
+    on_memtable_sealed(state_,
+                       reinterpret_cast<const rocksdb_memtableinfo_t*>(&info));
+  }
+
+  ~rocksdb_eventlistener_t() override { destructor_(state_); }
+};
+
+rocksdb_eventlistener_t* rocksdb_eventlistener_create(
+    void* state_, void (*destructor_)(void*), on_flush_begin_cb on_flush_begin,
+    on_flush_completed_cb on_flush_completed,
+    on_compaction_begin_cb on_compaction_begin,
+    on_compaction_completed_cb on_compaction_completed,
+    on_subcompaction_begin_cb on_subcompaction_begin,
+    on_subcompaction_completed_cb on_subcompaction_completed,
+    on_external_file_ingested_cb on_external_file_ingested,
+    on_background_error_cb on_background_error,
+    on_stall_conditions_changed_cb on_stall_conditions_changed,
+    on_memtable_sealed_cb on_memtable_sealed) {
+  rocksdb_eventlistener_t* et = new rocksdb_eventlistener_t;
+  et->state_ = state_;
+  et->destructor_ = destructor_;
+  et->on_flush_begin = on_flush_begin;
+  et->on_flush_completed = on_flush_completed;
+  et->on_compaction_begin = on_compaction_begin;
+  et->on_compaction_completed = on_compaction_completed;
+  et->on_subcompaction_begin = on_subcompaction_begin;
+  et->on_subcompaction_completed = on_subcompaction_completed;
+  et->on_external_file_ingested = on_external_file_ingested;
+  et->on_background_error = on_background_error;
+  et->on_stall_conditions_changed = on_stall_conditions_changed;
+  et->on_memtable_sealed = on_memtable_sealed;
+  return et;
+}
+
+void rocksdb_eventlistener_destroy(rocksdb_eventlistener_t* t) { delete t; }
+
+void rocksdb_options_add_eventlistener(rocksdb_options_t* opt,
+                                       rocksdb_eventlistener_t* t) {
+  opt->rep.listeners.emplace_back(std::shared_ptr<EventListener>(t));
+}
+
 rocksdb_cuckoo_table_options_t* rocksdb_cuckoo_options_create() {
  return new rocksdb_cuckoo_table_options_t;
 }
@@ -3183,6 +3639,11 @@ void rocksdb_options_set_write_buffer_manager(
  opt->rep.write_buffer_manager = wbm->rep;
 }

+void rocksdb_options_set_sst_file_manager(rocksdb_options_t* opt,
+                                          rocksdb_sst_file_manager_t* sfm) {
+  opt->rep.sst_file_manager = sfm->rep;
+}
+
 size_t rocksdb_options_get_write_buffer_size(rocksdb_options_t* opt) {
  return opt->rep.write_buffer_size;
 }
@@ -3295,6 +3756,26 @@ uint64_t rocksdb_options_get_periodic_compaction_seconds(
  return opt->rep.periodic_compaction_seconds;
 }

+void rocksdb_options_set_memtable_op_scan_flush_trigger(rocksdb_options_t* opt,
+                                                        uint32_t n) {
+  opt->rep.memtable_op_scan_flush_trigger = n;
+}
+
+uint32_t rocksdb_options_get_memtable_op_scan_flush_trigger(
+    rocksdb_options_t* opt) {
+  return opt->rep.memtable_op_scan_flush_trigger;
+}
+
+void rocksdb_options_set_memtable_avg_op_scan_flush_trigger(
+    rocksdb_options_t* opt, uint32_t n) {
+  opt->rep.memtable_avg_op_scan_flush_trigger = n;
+}
+
+uint32_t rocksdb_options_get_memtable_avg_op_scan_flush_trigger(
+    rocksdb_options_t* opt) {
+  return opt->rep.memtable_avg_op_scan_flush_trigger;
+}
+
 void rocksdb_options_enable_statistics(rocksdb_options_t* opt) {
  opt->rep.statistics = ROCKSDB_NAMESPACE::CreateDBStatistics();
 }
@@ -3804,16 +4285,6 @@ int rocksdb_options_get_min_write_buffer_number_to_merge(
  return opt->rep.min_write_buffer_number_to_merge;
 }

-void rocksdb_options_set_max_write_buffer_number_to_maintain(
-    rocksdb_options_t* opt, int n) {
-  opt->rep.max_write_buffer_number_to_maintain = n;
-}
-
-int rocksdb_options_get_max_write_buffer_number_to_maintain(
-    rocksdb_options_t* opt) {
-  return opt->rep.max_write_buffer_number_to_maintain;
-}
-
 void rocksdb_options_set_max_write_buffer_size_to_maintain(
    rocksdb_options_t* opt, int64_t n) {
  opt->rep.max_write_buffer_size_to_maintain = n;
@@ -4280,6 +4751,15 @@ void rocksdb_options_add_compact_on_deletion_collector_factory_del_ratio(
  opt->rep.table_properties_collector_factories.emplace_back(compact_on_del);
 }

+void rocksdb_options_add_compact_on_deletion_collector_factory_min_file_size(
+    rocksdb_options_t* opt, size_t window_size, size_t num_dels_trigger,
+    double deletion_ratio, uint64_t min_file_size) {
+  std::shared_ptr<ROCKSDB_NAMESPACE::TablePropertiesCollectorFactory>
+      compact_on_del = NewCompactOnDeletionCollectorFactory(
+          window_size, num_dels_trigger, deletion_ratio, min_file_size);
+  opt->rep.table_properties_collector_factories.emplace_back(compact_on_del);
+}
+
 void rocksdb_set_perf_level(int v) {
  PerfLevel level = static_cast<PerfLevel>(v);
  SetPerfLevel(level);
@@ -4332,6 +4812,8 @@ uint64_t rocksdb_perfcontext_metric(rocksdb_perfcontext_t* context,
      return rep->internal_recent_skipped_count;
    case rocksdb_internal_merge_count:
      return rep->internal_merge_count;
+    case rocksdb_internal_merge_point_lookup_count:
+      return rep->internal_merge_point_lookup_count;
    case rocksdb_get_snapshot_time:
      return rep->get_snapshot_time;
    case rocksdb_get_from_memtable_time:
@@ -5239,6 +5721,67 @@ ROCKSDB_LIBRARY_API void rocksdb_write_buffer_manager_set_allow_stall(
  wbm->rep->SetAllowStall(new_allow_stall);
 }

+rocksdb_sst_file_manager_t* rocksdb_sst_file_manager_create(
+    rocksdb_env_t* env) {
+  rocksdb_sst_file_manager_t* sfm = new rocksdb_sst_file_manager_t;
+  sfm->rep.reset(ROCKSDB_NAMESPACE::NewSstFileManager(env->rep));
+  return sfm;
+}
+
+void rocksdb_sst_file_manager_destroy(rocksdb_sst_file_manager_t* sfm) {
+  delete sfm;
+}
+
+void rocksdb_sst_file_manager_set_max_allowed_space_usage(
+    rocksdb_sst_file_manager_t* sfm, uint64_t max_allowed_space) {
+  sfm->rep->SetMaxAllowedSpaceUsage(max_allowed_space);
+}
+
+void rocksdb_sst_file_manager_set_compaction_buffer_size(
+    rocksdb_sst_file_manager_t* sfm, uint64_t compaction_buffer_size) {
+  sfm->rep->SetCompactionBufferSize(compaction_buffer_size);
+}
+
+bool rocksdb_sst_file_manager_is_max_allowed_space_reached(
+    rocksdb_sst_file_manager_t* sfm) {
+  return sfm->rep->IsMaxAllowedSpaceReached();
+}
+
+bool rocksdb_sst_file_manager_is_max_allowed_space_reached_including_compactions(
+    rocksdb_sst_file_manager_t* sfm) {
+  return sfm->rep->IsMaxAllowedSpaceReachedIncludingCompactions();
+}
+
+uint64_t rocksdb_sst_file_manager_get_total_size(
+    rocksdb_sst_file_manager_t* sfm) {
+  return sfm->rep->GetTotalSize();
+}
+
+int64_t rocksdb_sst_file_manager_get_delete_rate_bytes_per_second(
+    rocksdb_sst_file_manager_t* sfm) {
+  return sfm->rep->GetDeleteRateBytesPerSecond();
+}
+
+void rocksdb_sst_file_manager_set_delete_rate_bytes_per_second(
+    rocksdb_sst_file_manager_t* sfm, int64_t delete_rate) {
+  return sfm->rep->SetDeleteRateBytesPerSecond(delete_rate);
+}
+
+double rocksdb_sst_file_manager_get_max_trash_db_ratio(
+    rocksdb_sst_file_manager_t* sfm) {
+  return sfm->rep->GetMaxTrashDBRatio();
+}
+
+void rocksdb_sst_file_manager_set_max_trash_db_ratio(
+    rocksdb_sst_file_manager_t* sfm, double ratio) {
+  return sfm->rep->SetMaxTrashDBRatio(ratio);
+}
+
+uint64_t rocksdb_sst_file_manager_get_total_trash_size(
+    rocksdb_sst_file_manager_t* sfm) {
+  return sfm->rep->GetTotalTrashSize();
+}
+
 rocksdb_dbpath_t* rocksdb_dbpath_create(const char* path,
                                        uint64_t target_size) {
  rocksdb_dbpath_t* result = new rocksdb_dbpath_t;
@@ -103,6 +103,12 @@ static void CheckValue(char* err, const char* expected, char** actual,
  Free(actual);
 }

+static void CheckPinnedValue(char* err, const char* expected,
+                             const char** actual, size_t actual_length) {
+  CheckNoError(err);
+  CheckEqual(expected, *actual, actual_length);
+}
+
 static void CheckGet(rocksdb_t* db, const rocksdb_readoptions_t* options,
                     const char* key, const char* expected) {
  char* err = NULL;
@@ -1245,6 +1251,8 @@ int main(int argc, char** argv) {
    CheckCondition(count == 3);
    size_t size;
    char* value;
+    const char* pinned_value;
+    rocksdb_pinnableslice_t* p;
    value = rocksdb_writebatch_wi_get_from_batch(wbi, options, "box", 3, &size,
                                                 &err);
    CheckValue(err, "c", &value, size);
@@ -1254,9 +1262,19 @@ int main(int argc, char** argv) {
    value = rocksdb_writebatch_wi_get_from_batch_and_db(wbi, db, roptions,
                                                        "foo", 3, &size, &err);
    CheckValue(err, "hello", &value, size);
+    p = rocksdb_writebatch_wi_get_pinned_from_batch_and_db(wbi, db, roptions,
+                                                           "foo", 3, &err);
+    pinned_value = rocksdb_pinnableslice_value(p, &size);
+    CheckPinnedValue(err, "hello", &pinned_value, size);
+    rocksdb_pinnableslice_destroy(p);
    value = rocksdb_writebatch_wi_get_from_batch_and_db(wbi, db, roptions,
                                                        "box", 3, &size, &err);
    CheckValue(err, "c", &value, size);
+    p = rocksdb_writebatch_wi_get_pinned_from_batch_and_db(wbi, db, roptions,
+                                                           "box", 3, &err);
+    pinned_value = rocksdb_pinnableslice_value(p, &size);
+    CheckPinnedValue(err, "c", &pinned_value, size);
+    rocksdb_pinnableslice_destroy(p);
    rocksdb_write_writebatch_wi(db, woptions, wbi, &err);
    CheckNoError(err);
    CheckGet(db, roptions, "foo", "hello");
@@ -1362,6 +1380,46 @@ int main(int argc, char** argv) {
    rocksdb_writebatch_wi_destroy(wbi);
  }

+  StartPhase("wbwi_iter_readoptions");
+  {
+    rocksdb_readoptions_t* iter_roptions = rocksdb_readoptions_create();
+    rocksdb_readoptions_set_iterate_lower_bound(iter_roptions, "boy", 3);
+    rocksdb_readoptions_set_iterate_upper_bound(iter_roptions, "fool", 4);
+    rocksdb_iterator_t* base_iter = rocksdb_create_iterator(db, iter_roptions);
+    rocksdb_writebatch_wi_t* wbi = rocksdb_writebatch_wi_create(0, 1);
+    rocksdb_writebatch_wi_put(wbi, "bar", 3, "b",
+                              1);  // should get filtered out
+    rocksdb_writebatch_wi_put(wbi, "cat", 3, "miau", 4);
+    rocksdb_writebatch_wi_put(wbi, "gnu", 3, "muh",
+                              3);  // should get filtered out
+    rocksdb_iterator_t* iter =
+        rocksdb_writebatch_wi_create_iterator_with_base_readopts(wbi, base_iter,
+                                                                 iter_roptions);
+    CheckCondition(!rocksdb_iter_valid(iter));
+    rocksdb_iter_seek_to_first(iter);
+    CheckCondition(rocksdb_iter_valid(iter));
+    CheckIter(iter, "cat", "miau");
+    rocksdb_iter_next(iter);
+    CheckIter(iter, "foo", "hello");
+    rocksdb_iter_prev(iter);
+    CheckIter(iter, "cat", "miau");
+    rocksdb_iter_prev(iter);
+    CheckCondition(!rocksdb_iter_valid(iter));
+    rocksdb_iter_seek_to_last(iter);
+    CheckIter(iter, "foo", "hello");
+    rocksdb_iter_seek(iter, "b", 1);
+    CheckIter(iter, "cat", "miau");
+    rocksdb_iter_seek_for_prev(iter, "d", 1);
+    CheckIter(iter, "cat", "miau");
+    rocksdb_iter_seek_for_prev(iter, "fool", 3);
+    CheckIter(iter, "foo", "hello");
+    rocksdb_iter_get_error(iter, &err);
+    CheckNoError(err);
+    rocksdb_iter_destroy(iter);
+    rocksdb_writebatch_wi_destroy(wbi);
+    rocksdb_readoptions_destroy(iter_roptions);
+  }
+
  StartPhase("multiget");
  {
    const char* keys[3] = {"box", "foo", "notfound"};
@@ -1792,6 +1850,35 @@ int main(int argc, char** argv) {
    rocksdb_flush_wal(db, 1, &err);
    CheckNoError(err);

+    // Test WriteBatchWithIndex iteration with Column Family
+    rocksdb_writebatch_wi_t* wbwi = rocksdb_writebatch_wi_create(0, true);
+    rocksdb_writebatch_wi_put_cf(wbwi, handles[1], "boat", 4, "row",
+                                 3);  // should be filtered out
+    rocksdb_writebatch_wi_put_cf(wbwi, handles[1], "buffy", 5, "charmed", 7);
+    rocksdb_writebatch_wi_put_cf(wbwi, handles[1], "bus", 3, "yellow",
+                                 6);  // should be filtered out
+    rocksdb_readoptions_t* iter_roptions = rocksdb_readoptions_create();
+    rocksdb_readoptions_set_iterate_lower_bound(iter_roptions, "bu", 2);
+    rocksdb_readoptions_set_iterate_upper_bound(iter_roptions, "buffz", 5);
+    rocksdb_iterator_t* base_iter =
+        rocksdb_create_iterator_cf(db, iter_roptions, handles[1]);
+    rocksdb_iterator_t* wbwi_iter =
+        rocksdb_writebatch_wi_create_iterator_with_base_cf_readopts(
+            wbwi, base_iter, handles[1], iter_roptions);
+
+    CheckCondition(!rocksdb_iter_valid(wbwi_iter));
+    rocksdb_iter_seek_to_first(wbwi_iter);
+    CheckCondition(rocksdb_iter_valid(wbwi_iter));
+    CheckIter(wbwi_iter, "buff", "rocksdb");
+    rocksdb_iter_next(wbwi_iter);
+    CheckIter(wbwi_iter, "buffy", "charmed");
+    rocksdb_iter_next(wbwi_iter);
+    CheckCondition(!rocksdb_iter_valid(wbwi_iter));
+
+    rocksdb_iter_destroy(wbwi_iter);
+    rocksdb_writebatch_wi_destroy(wbwi);
+    rocksdb_readoptions_destroy(iter_roptions);
+
    const char* keys[3] = {"box", "box", "barfooxx"};
    const rocksdb_column_family_handle_t* get_handles[3] = {
        handles[0], handles[1], handles[1]};
@@ -2129,16 +2216,20 @@ int main(int argc, char** argv) {
    CheckCondition(100000 ==
                   rocksdb_options_get_periodic_compaction_seconds(o));

+    rocksdb_options_set_memtable_op_scan_flush_trigger(o, 100);
+    CheckCondition(100 ==
+                   rocksdb_options_get_memtable_op_scan_flush_trigger(o));
+
+    rocksdb_options_set_memtable_avg_op_scan_flush_trigger(o, 150);
+    CheckCondition(150 ==
+                   rocksdb_options_get_memtable_avg_op_scan_flush_trigger(o));
+
    rocksdb_options_set_ttl(o, 5000);
    CheckCondition(5000 == rocksdb_options_get_ttl(o));

    rocksdb_options_set_skip_stats_update_on_db_open(o, 1);
    CheckCondition(1 == rocksdb_options_get_skip_stats_update_on_db_open(o));

-    rocksdb_options_set_skip_checking_sst_file_sizes_on_db_open(o, 1);
-    CheckCondition(
-        1 == rocksdb_options_get_skip_checking_sst_file_sizes_on_db_open(o));
-
    rocksdb_options_set_max_write_buffer_number(o, 97);
    CheckCondition(97 == rocksdb_options_get_max_write_buffer_number(o));

@@ -2146,10 +2237,6 @@ int main(int argc, char** argv) {
    CheckCondition(23 ==
                   rocksdb_options_get_min_write_buffer_number_to_merge(o));

-    rocksdb_options_set_max_write_buffer_number_to_maintain(o, 64);
-    CheckCondition(64 ==
-                   rocksdb_options_get_max_write_buffer_number_to_maintain(o));
-
    rocksdb_options_set_max_write_buffer_size_to_maintain(o, 50000);
    CheckCondition(50000 ==
                   rocksdb_options_get_max_write_buffer_size_to_maintain(o));
@@ -2402,13 +2489,9 @@ int main(int argc, char** argv) {
    CheckCondition(2.0 ==
                   rocksdb_options_get_max_bytes_for_level_multiplier(copy));
    CheckCondition(1 == rocksdb_options_get_skip_stats_update_on_db_open(copy));
-    CheckCondition(
-        1 == rocksdb_options_get_skip_checking_sst_file_sizes_on_db_open(copy));
    CheckCondition(97 == rocksdb_options_get_max_write_buffer_number(copy));
    CheckCondition(23 ==
                   rocksdb_options_get_min_write_buffer_number_to_merge(copy));
-    CheckCondition(
-        64 == rocksdb_options_get_max_write_buffer_number_to_maintain(copy));
    CheckCondition(50000 ==
                   rocksdb_options_get_max_write_buffer_size_to_maintain(copy));
    CheckCondition(1 == rocksdb_options_get_enable_pipelined_write(copy));
@@ -2572,6 +2655,18 @@ int main(int argc, char** argv) {
    CheckCondition(100000 ==
                   rocksdb_options_get_periodic_compaction_seconds(o));

+    rocksdb_options_set_memtable_op_scan_flush_trigger(copy, 800);
+    CheckCondition(800 ==
+                   rocksdb_options_get_memtable_op_scan_flush_trigger(copy));
+    CheckCondition(100 ==
+                   rocksdb_options_get_memtable_op_scan_flush_trigger(o));
+
+    rocksdb_options_set_memtable_avg_op_scan_flush_trigger(copy, 900);
+    CheckCondition(
+        900 == rocksdb_options_get_memtable_avg_op_scan_flush_trigger(copy));
+    CheckCondition(150 ==
+                   rocksdb_options_get_memtable_avg_op_scan_flush_trigger(o));
+
    rocksdb_options_set_ttl(copy, 8000);
    CheckCondition(8000 == rocksdb_options_get_ttl(copy));
    CheckCondition(5000 == rocksdb_options_get_ttl(o));
@@ -2580,12 +2675,6 @@ int main(int argc, char** argv) {
    CheckCondition(0 == rocksdb_options_get_skip_stats_update_on_db_open(copy));
    CheckCondition(1 == rocksdb_options_get_skip_stats_update_on_db_open(o));

-    rocksdb_options_set_skip_checking_sst_file_sizes_on_db_open(copy, 0);
-    CheckCondition(
-        0 == rocksdb_options_get_skip_checking_sst_file_sizes_on_db_open(copy));
-    CheckCondition(
-        1 == rocksdb_options_get_skip_checking_sst_file_sizes_on_db_open(o));
-
    rocksdb_options_set_max_write_buffer_number(copy, 2000);
    CheckCondition(2000 == rocksdb_options_get_max_write_buffer_number(copy));
    CheckCondition(97 == rocksdb_options_get_max_write_buffer_number(o));
@@ -2596,12 +2685,6 @@ int main(int argc, char** argv) {
    CheckCondition(23 ==
                   rocksdb_options_get_min_write_buffer_number_to_merge(o));

-    rocksdb_options_set_max_write_buffer_number_to_maintain(copy, 128);
-    CheckCondition(
-        128 == rocksdb_options_get_max_write_buffer_number_to_maintain(copy));
-    CheckCondition(64 ==
-                   rocksdb_options_get_max_write_buffer_number_to_maintain(o));
-
    rocksdb_options_set_max_write_buffer_size_to_maintain(copy, 9000);
    CheckCondition(9000 ==
                   rocksdb_options_get_max_write_buffer_size_to_maintain(copy));
@@ -4052,6 +4135,23 @@ int main(int argc, char** argv) {
    rocksdb_cache_destroy(lru);
  }

+  StartPhase("sst_file_manager");
+  {
+    rocksdb_sst_file_manager_t* sst_file_manager;
+    sst_file_manager = rocksdb_sst_file_manager_create(env);
+    rocksdb_sst_file_manager_set_delete_rate_bytes_per_second(sst_file_manager,
+                                                              1);
+    rocksdb_sst_file_manager_set_max_trash_db_ratio(sst_file_manager, 0.75);
+
+    CheckCondition(1 ==
+                   rocksdb_sst_file_manager_get_delete_rate_bytes_per_second(
+                       sst_file_manager));
+    CheckCondition(0.75 == rocksdb_sst_file_manager_get_max_trash_db_ratio(
+                               sst_file_manager));
+
+    rocksdb_sst_file_manager_destroy(sst_file_manager);
+  }
+
  StartPhase("cancel_all_background_work");
  rocksdb_cancel_all_background_work(db, 1);

@@ -110,23 +110,48 @@ void GetInternalTblPropCollFactory(
  }
 }

+Status CheckCompressionSupportedWithManager(
+    CompressionType type, UnownedPtr<CompressionManager> mgr) {
+  if (mgr) {
+    if (!mgr->SupportsCompressionType(type)) {
+      return Status::NotSupported("Compression type " +
+                                  CompressionTypeToString(type) +
+                                  " is not recognized/supported by this "
+                                  "version of CompressionManager " +
+                                  mgr->GetId());
+    }
+  } else {
+    if (!CompressionTypeSupported(type)) {
+      if (type <= kLastBuiltinCompression) {
+        return Status::InvalidArgument("Compression type " +
+                                       CompressionTypeToString(type) +
+                                       " is not linked with the binary.");
+      } else {
+        return Status::NotSupported(
+            "Compression type " + CompressionTypeToString(type) +
+            " is not recognized/supported by built-in CompressionManager.");
+      }
+    }
+  }
+  return Status::OK();
+}
+
 Status CheckCompressionSupported(const ColumnFamilyOptions& cf_options) {
  if (!cf_options.compression_per_level.empty()) {
    for (size_t level = 0; level < cf_options.compression_per_level.size();
         ++level) {
-      if (!CompressionTypeSupported(cf_options.compression_per_level[level])) {
-        return Status::InvalidArgument(
-            "Compression type " +
-            CompressionTypeToString(cf_options.compression_per_level[level]) +
-            " is not linked with the binary.");
+      Status s = CheckCompressionSupportedWithManager(
+          cf_options.compression_per_level[level],
+          cf_options.compression_manager.get());
+      if (!s.ok()) {
+        return s;
      }
    }
  } else {
-    if (!CompressionTypeSupported(cf_options.compression)) {
-      return Status::InvalidArgument(
-          "Compression type " +
-          CompressionTypeToString(cf_options.compression) +
-          " is not linked with the binary.");
+    Status s = CheckCompressionSupportedWithManager(
+        cf_options.compression, cf_options.compression_manager.get());
+    if (!s.ok()) {
+      return s;
    }
  }
  if (cf_options.compression_opts.zstd_max_train_bytes > 0) {
@@ -255,22 +280,18 @@ ColumnFamilyOptions SanitizeCfOptions(const ImmutableDBOptions& db_options,
  }

  if (result.compaction_style == kCompactionStyleUniversal &&
-      db_options.allow_ingest_behind && result.num_levels < 3) {
+      (db_options.allow_ingest_behind || result.cf_allow_ingest_behind) &&
+      result.num_levels < 3) {
    result.num_levels = 3;
  }

  if (result.max_write_buffer_number < 2) {
    result.max_write_buffer_number = 2;
  }
-  // fall back max_write_buffer_number_to_maintain if
-  // max_write_buffer_size_to_maintain is not set
  if (result.max_write_buffer_size_to_maintain < 0) {
    result.max_write_buffer_size_to_maintain =
        result.max_write_buffer_number *
        static_cast<int64_t>(result.write_buffer_size);
-  } else if (result.max_write_buffer_size_to_maintain == 0 &&
-             result.max_write_buffer_number_to_maintain < 0) {
-    result.max_write_buffer_number_to_maintain = result.max_write_buffer_number;
  }
  // bloom filter size shouldn't exceed 1/4 of memtable size.
  if (result.memtable_prefix_bloom_size_ratio > 0.25) {
@@ -453,6 +474,21 @@ ColumnFamilyOptions SanitizeCfOptions(const ImmutableDBOptions& db_options,
    result.preclude_last_level_data_seconds = 0;
  }

+  if (read_only) {
+    if (result.memtable_op_scan_flush_trigger) {
+      ROCKS_LOG_WARN(db_options.info_log.get(),
+                     "option memtable_op_scan_flush_trigger is sanitized to "
+                     "0(disabled) for read only DB.");
+      result.memtable_op_scan_flush_trigger = 0;
+    }
+    if (result.memtable_avg_op_scan_flush_trigger) {
+      ROCKS_LOG_WARN(
+          db_options.info_log.get(),
+          "option memtable_avg_op_scan_flush_trigger is sanitized to "
+          "0(disabled) for read only DB.");
+      result.memtable_avg_op_scan_flush_trigger = 0;
+    }
+  }
  return result;
 }

@@ -577,7 +613,6 @@ ColumnFamilyData::ColumnFamilyData(
      write_buffer_manager_(write_buffer_manager),
      mem_(nullptr),
      imm_(ioptions_.min_write_buffer_number_to_merge,
-           ioptions_.max_write_buffer_number_to_maintain,
           ioptions_.max_write_buffer_size_to_maintain),
      super_version_(nullptr),
      super_version_number_(0),
@@ -1208,10 +1243,12 @@ Compaction* ColumnFamilyData::PickCompaction(
    const MutableCFOptions& mutable_options,
    const MutableDBOptions& mutable_db_options,
    const std::vector<SequenceNumber>& existing_snapshots,
-    const SnapshotChecker* snapshot_checker, LogBuffer* log_buffer) {
+    const SnapshotChecker* snapshot_checker, LogBuffer* log_buffer,
+    bool require_max_output_level) {
  auto* result = compaction_picker_->PickCompaction(
      GetName(), mutable_options, mutable_db_options, existing_snapshots,
-      snapshot_checker, current_->storage_info(), log_buffer);
+      snapshot_checker, current_->storage_info(), log_buffer,
+      require_max_output_level);
  if (result != nullptr) {
    result->FinalizeInputInfo(current_);
  }
@@ -1295,7 +1332,7 @@ Compaction* ColumnFamilyData::CompactRange(
    const InternalKey* begin, const InternalKey* end,
    InternalKey** compaction_end, bool* conflict,
    uint64_t max_file_num_to_ignore, const std::string& trim_ts) {
-  auto* result = compaction_picker_->CompactRange(
+  auto* result = compaction_picker_->PickCompactionForCompactRange(
      GetName(), mutable_cf_options, mutable_db_options,
      current_->storage_info(), input_level, output_level,
      compact_range_options, begin, end, compaction_end, conflict,
@@ -1597,6 +1634,8 @@ Status ColumnFamilyData::SetOptions(
  Status s = GetColumnFamilyOptionsFromMap(config_opts, cf_opts, options_map,
                                           &cf_opts);
  if (s.ok()) {
+    // FIXME: we should call SanitizeOptions() too or consolidate it with
+    // ValidateOptions().
    s = ValidateOptions(db_opts, cf_opts);
  }
  if (s.ok()) {
@@ -424,7 +424,8 @@ class ColumnFamilyData {
      const MutableCFOptions& mutable_options,
      const MutableDBOptions& mutable_db_options,
      const std::vector<SequenceNumber>& existing_snapshots,
-      const SnapshotChecker* snapshot_checker, LogBuffer* log_buffer);
+      const SnapshotChecker* snapshot_checker, LogBuffer* log_buffer,
+      bool require_max_output_level = false);

  // Check if the passed range overlap with any running compactions.
  // REQUIRES: DB mutex held
@@ -537,6 +538,12 @@ class ColumnFamilyData {
    assert(!ts_low.empty());
    const Comparator* ucmp = user_comparator();
    assert(ucmp);
+    // Guard against resurrected full_history_ts_low persisted in MANIFEST
+    // from previous DB sessions. This could happen if UDT was enabled and then
+    // disabled.
+    if (ucmp->timestamp_size() == 0) {
+      return;
+    }
    if (full_history_ts_low_.empty() ||
        ucmp->CompareTimestamp(ts_low, full_history_ts_low_) > 0) {
      full_history_ts_low_ = std::move(ts_low);
@@ -544,6 +551,11 @@ class ColumnFamilyData {
  }

  const std::string& GetFullHistoryTsLow() const {
+    const Comparator* ucmp = user_comparator();
+    assert(ucmp);
+    if (ucmp->timestamp_size() == 0) {
+      assert(full_history_ts_low_.empty());
+    }
    return full_history_ts_low_;
  }

@@ -588,6 +600,11 @@ class ColumnFamilyData {
    return (mem_->IsEmpty() ? 0 : 1) + imm_.NumNotFlushed();
  }

+  // thread-safe, DB mutex not needed.
+  bool AllowIngestBehind() const {
+    return ioptions_.cf_allow_ingest_behind || ioptions_.allow_ingest_behind;
+  }
+
 private:
  friend class ColumnFamilySet;
  ColumnFamilyData(
@@ -72,7 +72,6 @@ class ColumnFamilyTestBase : public testing::Test {
    env_->skip_fsync_ = true;
    dbname_ = test::PerThreadDBPath("column_family_test");
    db_options_.create_if_missing = true;
-    db_options_.fail_if_options_file_error = true;
    db_options_.env = env_;
  }

@@ -2176,7 +2175,7 @@ TEST_P(ColumnFamilyTest, FlushStaleColumnFamilies) {
  ASSERT_TRUE(has_cf2_sst);

  ASSERT_OK(Flush(0));
-  ASSERT_EQ(0, dbfull()->TEST_total_log_size());
+  ASSERT_EQ(0, dbfull()->TEST_wals_total_size());
  Close();
 }

@@ -441,6 +441,10 @@ TEST_F(CompactFilesTest, SentinelCompressionType) {
 }

 TEST_F(CompactFilesTest, CompressionWithBlockAlign) {
+  if (!Snappy_Supported()) {
+    ROCKSDB_GTEST_SKIP("Test requires Snappy support");
+    return;
+  }
  Options options;
  options.compression = CompressionType::kNoCompression;
  options.create_if_missing = true;
@@ -284,9 +284,9 @@ Compaction::Compaction(
    CompressionOptions _compression_opts, Temperature _output_temperature,
    uint32_t _max_subcompactions, std::vector<FileMetaData*> _grandparents,
    std::optional<SequenceNumber> _earliest_snapshot,
-    const SnapshotChecker* _snapshot_checker, bool _manual_compaction,
-    const std::string& _trim_ts, double _score, bool _deletion_compaction,
-    bool l0_files_might_overlap, CompactionReason _compaction_reason,
+    const SnapshotChecker* _snapshot_checker,
+    CompactionReason _compaction_reason, const std::string& _trim_ts,
+    double _score, bool l0_files_might_overlap,
    BlobGarbageCollectionPolicy _blob_garbage_collection_policy,
    double _blob_garbage_collection_age_cutoff)
    : input_vstorage_(vstorage),
@@ -304,7 +304,9 @@ Compaction::Compaction(
      output_compression_(_compression),
      output_compression_opts_(_compression_opts),
      output_temperature_(_output_temperature),
-      deletion_compaction_(_deletion_compaction),
+      deletion_compaction_(_compaction_reason == CompactionReason::kFIFOTtl ||
+                           _compaction_reason ==
+                               CompactionReason::kFIFOMaxSize),
      l0_files_might_overlap_(l0_files_might_overlap),
      inputs_(PopulateWithAtomicBoundaries(vstorage, std::move(_inputs))),
      grandparents_(std::move(_grandparents)),
@@ -321,7 +323,8 @@ Compaction::Compaction(
              ? false
              : IsBottommostLevel(output_level_, vstorage, inputs_)),
      is_full_compaction_(IsFullCompaction(vstorage, inputs_)),
-      is_manual_compaction_(_manual_compaction),
+      is_manual_compaction_(_compaction_reason ==
+                            CompactionReason::kManualCompaction),
      trim_ts_(_trim_ts),
      is_trivial_move_(false),
      compaction_reason_(_compaction_reason),
@@ -349,9 +352,6 @@ Compaction::Compaction(
                                      immutable_options_, start_level_,
                                      output_level_)) {
  MarkFilesBeingCompacted(true);
-  if (is_manual_compaction_) {
-    compaction_reason_ = CompactionReason::kManualCompaction;
-  }
  if (max_subcompactions_ == 0) {
    max_subcompactions_ = _mutable_db_options.max_subcompactions;
  }
@@ -647,6 +647,8 @@ bool Compaction::KeyNotExistsBeyondOutputLevel(
    return true;
  } else if (output_level_ != 0 &&
             cfd_->ioptions().compaction_style == kCompactionStyleLevel) {
+    // TODO: apply the optimization here to other compaction styles and
+    // compaction/flush to L0.
    // Maybe use binary search to find right entry instead of linear search?
    const Comparator* user_cmp = cfd_->user_comparator();
    for (int lvl = output_level_ + 1; lvl < number_levels_; lvl++) {
@@ -94,10 +94,9 @@ class Compaction {
             std::vector<FileMetaData*> grandparents,
             std::optional<SequenceNumber> earliest_snapshot,
             const SnapshotChecker* snapshot_checker,
-             bool manual_compaction = false, const std::string& trim_ts = "",
-             double score = -1, bool deletion_compaction = false,
+             CompactionReason compaction_reason,
+             const std::string& trim_ts = "", double score = -1,
             bool l0_files_might_overlap = true,
-             CompactionReason compaction_reason = CompactionReason::kUnknown,
             BlobGarbageCollectionPolicy blob_garbage_collection_policy =
                 BlobGarbageCollectionPolicy::kUseDefault,
             double blob_garbage_collection_age_cutoff = -1);
@@ -283,6 +282,13 @@ class Compaction {
  // are non-overlapping and can be trivially moved.
  bool is_trivial_move() const { return is_trivial_move_; }

+  bool is_trivial_copy_compaction() const {
+    return immutable_options_.compaction_style == kCompactionStyleFIFO &&
+           compaction_reason_ == CompactionReason::kChangeTemperature &&
+           mutable_cf_options_.compaction_options_fifo
+               .allow_trivial_copy_when_change_temperature;
+  }
+
  // How many total levels are there?
  int number_levels() const { return number_levels_; }

@@ -456,6 +462,11 @@ class Compaction {
                                   const int start_level,
                                   const int output_level);

+  static bool OutputToNonZeroMaxOutputLevel(int output_level,
+                                            int max_output_level) {
+    return output_level > 0 && output_level == max_output_level;
+  }
+
  // If some data cannot be safely migrated "up" the LSM tree due to a change
  // in the preclude_last_level_data_seconds setting, this indicates a sequence
  // number for the newest data that must be kept in the last level.
@@ -28,7 +28,7 @@ CompactionIterator::CompactionIterator(
    SequenceNumber earliest_snapshot,
    SequenceNumber earliest_write_conflict_snapshot,
    SequenceNumber job_snapshot, const SnapshotChecker* snapshot_checker,
-    Env* env, bool report_detailed_time, bool expect_valid_internal_key,
+    Env* env, bool report_detailed_time,
    CompactionRangeDelAggregator* range_del_agg,
    BlobFileBuilder* blob_file_builder, bool allow_data_in_errors,
    bool enforce_single_del_contracts,
@@ -42,8 +42,8 @@ CompactionIterator::CompactionIterator(
    : CompactionIterator(
          input, cmp, merge_helper, last_sequence, snapshots, earliest_snapshot,
          earliest_write_conflict_snapshot, job_snapshot, snapshot_checker, env,
-          report_detailed_time, expect_valid_internal_key, range_del_agg,
-          blob_file_builder, allow_data_in_errors, enforce_single_del_contracts,
+          report_detailed_time, range_del_agg, blob_file_builder,
+          allow_data_in_errors, enforce_single_del_contracts,
          manual_compaction_canceled,
          compaction ? std::make_unique<RealCompaction>(compaction) : nullptr,
          must_count_input_entries, compaction_filter, shutting_down, info_log,
@@ -55,7 +55,7 @@ CompactionIterator::CompactionIterator(
    SequenceNumber earliest_snapshot,
    SequenceNumber earliest_write_conflict_snapshot,
    SequenceNumber job_snapshot, const SnapshotChecker* snapshot_checker,
-    Env* env, bool report_detailed_time, bool expect_valid_internal_key,
+    Env* env, bool report_detailed_time,
    CompactionRangeDelAggregator* range_del_agg,
    BlobFileBuilder* blob_file_builder, bool allow_data_in_errors,
    bool enforce_single_del_contracts,
@@ -76,16 +76,14 @@ CompactionIterator::CompactionIterator(
      env_(env),
      clock_(env_->GetSystemClock().get()),
      report_detailed_time_(report_detailed_time),
-      expect_valid_internal_key_(expect_valid_internal_key),
      range_del_agg_(range_del_agg),
      blob_file_builder_(blob_file_builder),
      compaction_(std::move(compaction)),
      compaction_filter_(compaction_filter),
      shutting_down_(shutting_down),
      manual_compaction_canceled_(manual_compaction_canceled),
-      bottommost_level_(!compaction_ ? false
-                                     : compaction_->bottommost_level() &&
-                                           !compaction_->allow_ingest_behind()),
+      bottommost_level_(compaction_ && compaction_->bottommost_level() &&
+                        !compaction_->allow_ingest_behind()),
      // snapshots_ cannot be nullptr, but we will assert later in the body of
      // the constructor.
      visible_at_tip_(snapshots_ ? snapshots_->empty() : false),
@@ -161,6 +159,7 @@ void CompactionIterator::Next() {
      // MergeUntil stops when it encounters a corrupt key and does not
      // include them in the result, so we expect the keys here to be valid.
      if (!s.ok()) {
+        // FIXME: should fail compaction after this fatal logging.
        ROCKS_LOG_FATAL(
            info_log_, "Invalid ikey %s in compaction. %s",
            allow_data_in_errors_ ? key_.ToString(true).c_str() : "hidden",
@@ -464,18 +463,9 @@ void CompactionIterator::NextFromInput() {
    if (!pik_status.ok()) {
      iter_stats_.num_input_corrupt_records++;

-      // If `expect_valid_internal_key_` is false, return the corrupted key
-      // and let the caller decide what to do with it.
-      if (expect_valid_internal_key_) {
-        status_ = pik_status;
-        return;
-      }
-      key_ = current_key_.SetInternalKey(key_);
-      has_current_user_key_ = false;
-      current_user_key_sequence_ = kMaxSequenceNumber;
-      current_user_key_snapshot_ = 0;
-      validity_info_.SetValid(ValidContext::kParseKeyError);
-      break;
+      // Always fail compaction when encountering corrupted internal keys
+      status_ = pik_status;
+      return;
    }
    TEST_SYNC_POINT_CALLBACK("CompactionIterator:ProcessKV", &ikey_);
    if (is_range_del_) {
@@ -642,7 +632,8 @@ void CompactionIterator::NextFromInput() {
    } else if (ikey_.type == kTypeSingleDeletion) {
      // We can compact out a SingleDelete if:
      // 1) We encounter the corresponding PUT -OR- we know that this key
-      //    doesn't appear past this output level
+      //    doesn't appear past this output level and  we are not in
+      //    ingest_behind mode.
      // =AND=
      // 2) We've already returned a record in this snapshot -OR-
      //    there are no earlier earliest_write_conflict_snapshot.
@@ -731,6 +722,8 @@ void CompactionIterator::NextFromInput() {
            "CompactionIterator::NextFromInput:SingleDelete:1",
            const_cast<Compaction*>(c));
        if (last_key_seq_zeroed_) {
+          // Drop SD and the next key since they are both in the last
+          // snapshot (since last key has seqno zeroed).
          ++iter_stats_.num_record_drop_hidden;
          ++iter_stats_.num_record_drop_obsolete;
          assert(bottommost_level_);
@@ -841,7 +834,7 @@ void CompactionIterator::NextFromInput() {
        // iteration. If the next key is corrupt, we return before the
        // comparison, so the value of has_current_user_key does not matter.
        has_current_user_key_ = false;
-        if (compaction_ != nullptr &&
+        if (compaction_ != nullptr && !compaction_->allow_ingest_behind() &&
            DefinitelyInSnapshot(ikey_.sequence, earliest_snapshot_) &&
            compaction_->KeyNotExistsBeyondOutputLevel(ikey_.user_key,
                                                       &level_ptrs_) &&
@@ -854,6 +847,9 @@ void CompactionIterator::NextFromInput() {
            ++iter_stats_.num_optimized_del_drop_obsolete;
          }
        } else if (last_key_seq_zeroed_) {
+          // Sequence number zeroing requires bottommost_level_, which is
+          // false with ingest_behind.
+          assert(!compaction_->allow_ingest_behind());
          // Skip.
          ++iter_stats_.num_record_drop_hidden;
          ++iter_stats_.num_record_drop_obsolete;
@@ -870,6 +866,7 @@ void CompactionIterator::NextFromInput() {
    } else if (last_sequence != kMaxSequenceNumber &&
               (last_snapshot == current_user_key_snapshot_ ||
                last_snapshot < current_user_key_snapshot_)) {
+      // rule (A):
      // If the earliest snapshot is which this key is visible in
      // is the same as the visibility of a previous instance of the
      // same key, then this kv is not visible in any snapshot.
@@ -878,6 +875,15 @@ void CompactionIterator::NextFromInput() {
      // Note: Dropping this key will not affect TransactionDB write-conflict
      // checking since there has already been a record returned for this key
      // in this snapshot.
+      // When ingest_behind is enabled, it's ok that we drop an overwritten
+      // Delete here. The overwritting key still covers whatever that will be
+      // ingested. Note that we will not drop SingleDelete here as SingleDelte
+      // is handled entirely in its own if clause. This is important, see
+      // example: from new to old: SingleDelete_1, PUT_1, SingleDelete_2, PUT_2,
+      // where all operations are on the same key and PUT_2 is ingested with
+      // ingest_behind=true. If SingleDelete_2 is dropped due to being compacted
+      // together with PUT_1, and then PUT_1 is compacted away together with
+      // SingleDelete_1, PUT_2 can incorrectly becomes visible.
      if (last_sequence < current_user_key_sequence_) {
        ROCKS_LOG_FATAL(info_log_,
                        "key %s, last_sequence (%" PRIu64
@@ -887,12 +893,13 @@ void CompactionIterator::NextFromInput() {
        assert(false);
      }

-      ++iter_stats_.num_record_drop_hidden;  // rule (A)
+      ++iter_stats_.num_record_drop_hidden;
      AdvanceInputIter();
    } else if (compaction_ != nullptr &&
               (ikey_.type == kTypeDeletion ||
                (ikey_.type == kTypeDeletionWithTimestamp &&
                 cmp_with_history_ts_low_ < 0)) &&
+               !compaction_->allow_ingest_behind() &&
               DefinitelyInSnapshot(ikey_.sequence, earliest_snapshot_) &&
               compaction_->KeyNotExistsBeyondOutputLevel(ikey_.user_key,
                                                          &level_ptrs_)) {
@@ -928,11 +935,13 @@ void CompactionIterator::NextFromInput() {
                (ikey_.type == kTypeDeletionWithTimestamp &&
                 cmp_with_history_ts_low_ < 0)) &&
               bottommost_level_) {
+      assert(compaction_);
+      assert(!compaction_->allow_ingest_behind());  // bottommost_level_ is true
      // Handle the case where we have a delete key at the bottom most level
      // We can skip outputting the key iff there are no subsequent puts for
      // this key
-      assert(!compaction_ || compaction_->KeyNotExistsBeyondOutputLevel(
-                                 ikey_.user_key, &level_ptrs_));
+      assert(compaction_->KeyNotExistsBeyondOutputLevel(ikey_.user_key,
+                                                        &level_ptrs_));
      ParsedInternalKey next_ikey;
      AdvanceInputIter();
 #ifndef NDEBUG
@@ -974,6 +983,12 @@ void CompactionIterator::NextFromInput() {
                (compaction_ != nullptr &&
                 compaction_->KeyNotExistsBeyondOutputLevel(ikey_.user_key,
                                                            &level_ptrs_)))) {
+      // FIXME: it's possible that we are setting sequence number to 0 as
+      // preferred sequence number here. If cf_ingest_behind is enabled, this
+      // may fail ingestions since they expect all keys above the last level
+      // to have non-zero sequence number. We should probably not allow seqno
+      // zeroing here.
+      //
      // This section that attempts to swap preferred sequence number will not
      // be invoked if this is a CompactionIterator created for flush, since
      // `compaction_` will be nullptr and it's not bottommost either.
@@ -1274,11 +1289,11 @@ void CompactionIterator::PrepareOutput() {
    //
    // Can we do the same for levels above bottom level as long as
    // KeyNotExistsBeyondOutputLevel() return true?
-    if (Valid() && compaction_ != nullptr &&
-        !compaction_->allow_ingest_behind() && bottommost_level_ &&
+    if (Valid() && bottommost_level_ &&
        DefinitelyInSnapshot(ikey_.sequence, earliest_snapshot_) &&
        ikey_.type != kTypeMerge && current_key_committed_ &&
        ikey_.sequence <= preserve_seqno_after_ && !is_range_del_) {
+      assert(compaction_ != nullptr && !compaction_->allow_ingest_behind());
      if (ikey_.type == kTypeDeletion ||
          (ikey_.type == kTypeSingleDeletion && timestamp_size_ == 0)) {
        ROCKS_LOG_FATAL(
@@ -145,7 +145,8 @@ class CompactionIterator {
    }

    bool allow_ingest_behind() const override {
-      return compaction_->immutable_options().allow_ingest_behind;
+      return compaction_->immutable_options().cf_allow_ingest_behind ||
+             compaction_->immutable_options().allow_ingest_behind;
    }

    bool allow_mmap_reads() const override {
@@ -192,7 +193,7 @@ class CompactionIterator {
      SequenceNumber earliest_snapshot,
      SequenceNumber earliest_write_conflict_snapshot,
      SequenceNumber job_snapshot, const SnapshotChecker* snapshot_checker,
-      Env* env, bool report_detailed_time, bool expect_valid_internal_key,
+      Env* env, bool report_detailed_time,
      CompactionRangeDelAggregator* range_del_agg,
      BlobFileBuilder* blob_file_builder, bool allow_data_in_errors,
      bool enforce_single_del_contracts,
@@ -212,7 +213,7 @@ class CompactionIterator {
                     SequenceNumber earliest_write_conflict_snapshot,
                     SequenceNumber job_snapshot,
                     const SnapshotChecker* snapshot_checker, Env* env,
-                     bool report_detailed_time, bool expect_valid_internal_key,
+                     bool report_detailed_time,
                     CompactionRangeDelAggregator* range_del_agg,
                     BlobFileBuilder* blob_file_builder,
                     bool allow_data_in_errors,
@@ -347,7 +348,6 @@ class CompactionIterator {
  Env* env_;
  SystemClock* clock_;
  const bool report_detailed_time_;
-  const bool expect_valid_internal_key_;
  CompactionRangeDelAggregator* range_del_agg_;
  BlobFileBuilder* blob_file_builder_;
  std::unique_ptr<CompactionProxy> compaction_;
@@ -417,13 +417,15 @@ class CompactionIterator {
  // NextFromInput()).
  ParsedInternalKey ikey_;

-  // Stores whether ikey_.user_key is valid. If set to false, the user key is
-  // not compared against the current key in the underlying iterator.
+  // Stores whether current_user_key_ is valid. If so, current_user_key_
+  // stores the user key of the last key seen by the iterator.
+  // If false, treat the next key to read as a new user key.
  bool has_current_user_key_ = false;
  // If false, the iterator holds a copy of the current compaction iterator
  // output (or current key in the underlying iterator during NextFromInput()).
  bool at_next_ = false;

+  // A copy of the current internal key.
  IterKey current_key_;
  Slice current_user_key_;
  std::string curr_ts_;
@@ -433,8 +435,9 @@ class CompactionIterator {
  // True if the iterator has already returned a record for the current key.
  bool has_outputted_key_ = false;

-  // truncated the value of the next key and output it without applying any
-  // compaction rules.  This is used for outputting a put after a single delete.
+  // Truncate the value of the next key and output it without applying any
+  // compaction rules. This is an optimization for outputting a put after
+  // a single delete. See more in `NextFromInput()` under Optimization 3.
  bool clear_and_output_next_key_ = false;

  MergeOutputIterator merge_out_iter_;
@@ -294,7 +294,7 @@ class CompactionIteratorTest : public testing::TestWithParam<bool> {
        snapshots_.empty() ? kMaxSequenceNumber : snapshots_.at(0),
        earliest_write_conflict_snapshot, kMaxSequenceNumber,
        snapshot_checker_.get(), Env::Default(),
-        false /* report_detailed_time */, false, range_del_agg_.get(),
+        false /* report_detailed_time */, range_del_agg_.get(),
        nullptr /* blob_file_builder */, true /*allow_data_in_errors*/,
        true /*enforce_single_del_contracts*/,
        /*manual_compaction_canceled=*/kManualCompactionCanceledFalse_,
@@ -374,8 +374,7 @@ TEST_P(CompactionIteratorTest, EmptyResult) {
  ASSERT_FALSE(c_iter_->Valid());
 }

-// If there is a corruption after a single deletion, the corrupted key should
-// be preserved.
+// If there is a corruption after a single deletion, the compaction should fail.
 TEST_P(CompactionIteratorTest, CorruptionAfterSingleDeletion) {
  InitIterators({test::KeyStr("a", 5, kTypeSingleDeletion),
                 test::KeyStr("a", 3, kTypeValue, true),
@@ -386,14 +385,10 @@ TEST_P(CompactionIteratorTest, CorruptionAfterSingleDeletion) {
  ASSERT_EQ(test::KeyStr("a", 5, kTypeSingleDeletion),
            c_iter_->key().ToString());
  c_iter_->Next();
-  ASSERT_TRUE(c_iter_->Valid());
-  ASSERT_EQ(test::KeyStr("a", 3, kTypeValue, true), c_iter_->key().ToString());
-  c_iter_->Next();
-  ASSERT_TRUE(c_iter_->Valid());
-  ASSERT_EQ(test::KeyStr("b", 10, kTypeValue), c_iter_->key().ToString());
-  c_iter_->Next();
-  ASSERT_OK(c_iter_->status());
+  // The iterator should now fail when encountering the corrupted key
  ASSERT_FALSE(c_iter_->Valid());
+  ASSERT_FALSE(c_iter_->status().ok());
+  ASSERT_TRUE(c_iter_->status().IsCorruption());
 }

 // Tests compatibility of TimedPut and SingleDelete. TimedPut should act as if
@@ -133,10 +133,7 @@ CompactionJob::CompactionJob(
    LogBuffer* log_buffer, FSDirectory* db_directory,
    FSDirectory* output_directory, FSDirectory* blob_output_directory,
    Statistics* stats, InstrumentedMutex* db_mutex,
-    ErrorHandler* db_error_handler,
-    std::vector<SequenceNumber> existing_snapshots,
-    SequenceNumber earliest_write_conflict_snapshot,
-    const SnapshotChecker* snapshot_checker, JobContext* job_context,
+    ErrorHandler* db_error_handler, JobContext* job_context,
    std::shared_ptr<Cache> table_cache, EventLogger* event_logger,
    bool paranoid_file_checks, bool measure_io_stats, const std::string& dbname,
    CompactionJobStats* compaction_job_stats, Env::Priority thread_pri,
@@ -173,12 +170,7 @@ CompactionJob::CompactionJob(
      blob_output_directory_(blob_output_directory),
      db_mutex_(db_mutex),
      db_error_handler_(db_error_handler),
-      existing_snapshots_(std::move(existing_snapshots)),
-      earliest_snapshot_(existing_snapshots_.empty()
-                             ? kMaxSequenceNumber
-                             : existing_snapshots_.at(0)),
-      earliest_write_conflict_snapshot_(earliest_write_conflict_snapshot),
-      snapshot_checker_(snapshot_checker),
+      earliest_snapshot_(job_context->GetEarliestSnapshotSequence()),
      job_context_(job_context),
      table_cache_(std::move(table_cache)),
      event_logger_(event_logger),
@@ -193,6 +185,7 @@ CompactionJob::CompactionJob(
      bg_bottom_compaction_scheduled_(bg_bottom_compaction_scheduled) {
  assert(job_stats_ != nullptr);
  assert(log_buffer_ != nullptr);
+  assert(job_context->snapshot_context_initialized);

  const auto* cfd = compact_->compaction->column_family_data();
  ThreadStatusUtil::SetEnableTracking(db_options_.enable_thread_tracking);
@@ -224,10 +217,9 @@ void CompactionJob::ReportStartedCompaction(Compaction* compaction) {
      ThreadStatus::COMPACTION_PROP_FLAGS,
      compaction->is_manual_compaction() +
          (compaction->deletion_compaction() << 1));
-
+  auto total_input_bytes = compaction->CalculateTotalInputSize();
  ThreadStatusUtil::SetThreadOperationProperty(
-      ThreadStatus::COMPACTION_TOTAL_INPUT_BYTES,
-      compaction->CalculateTotalInputSize());
+      ThreadStatus::COMPACTION_TOTAL_INPUT_BYTES, total_input_bytes);

  IOSTATS_RESET(bytes_written);
  IOSTATS_RESET(bytes_read);
@@ -242,6 +234,16 @@ void CompactionJob::ReportStartedCompaction(Compaction* compaction) {

  job_stats_->is_manual_compaction = compaction->is_manual_compaction();
  job_stats_->is_full_compaction = compaction->is_full_compaction();
+  // populate compaction stats num_input_files and total_num_of_bytes
+  size_t num_input_files = 0;
+  for (int input_level = 0;
+       input_level < static_cast<int>(compaction->num_input_levels());
+       ++input_level) {
+    const LevelFilesBrief* flevel = compaction->input_levels(input_level);
+    num_input_files += flevel->num_files;
+  }
+  job_stats_->CompactionJobStats::num_input_files = num_input_files;
+  job_stats_->total_input_bytes = total_input_bytes;
 }

 void CompactionJob::Prepare(
@@ -666,16 +668,17 @@ void CompactionJob::GenSubcompactionBoundaries() {
               extra_num_subcompaction_threads_reserved_));
 }

-Status CompactionJob::Run() {
+void CompactionJob::InitializeCompactionRun() {
  AutoThreadOperationStageUpdater stage_updater(
      ThreadStatus::STAGE_COMPACTION_RUN);
  TEST_SYNC_POINT("CompactionJob::Run():Start");
  log_buffer_->FlushBufferToLog();
  LogCompaction();
+}

+void CompactionJob::RunSubcompactions() {
  const size_t num_threads = compact_->sub_compact_states.size();
  assert(num_threads > 0);
-  const uint64_t start_micros = db_options_.clock->NowMicros();
  compact_->compaction->GetOrInitInputTableProperties();

  // Launch a thread for each of subcompactions 1...num_threads-1
@@ -694,25 +697,43 @@ Status CompactionJob::Run() {
  for (auto& thread : thread_pool) {
    thread.join();
  }
+  RemoveEmptyOutputs();

+  ReleaseSubcompactionResources();
+  TEST_SYNC_POINT("CompactionJob::ReleaseSubcompactionResources");
+}
+
+void CompactionJob::UpdateTimingStats(uint64_t start_micros) {
  internal_stats_.SetMicros(db_options_.clock->NowMicros() - start_micros);

  for (auto& state : compact_->sub_compact_states) {
    internal_stats_.AddCpuMicros(state.compaction_job_stats.cpu_micros);
-    state.RemoveLastEmptyOutput();
  }

  RecordTimeToHistogram(stats_, COMPACTION_TIME,
                        internal_stats_.output_level_stats.micros);
  RecordTimeToHistogram(stats_, COMPACTION_CPU_TIME,
                        internal_stats_.output_level_stats.cpu_micros);
+}

-  TEST_SYNC_POINT("CompactionJob::Run:BeforeVerify");
+void CompactionJob::RemoveEmptyOutputs() {
+  for (auto& state : compact_->sub_compact_states) {
+    state.RemoveLastEmptyOutput();
+  }
+}

-  // Check if any thread encountered an error during execution
+bool CompactionJob::HasNewBlobFiles() const {
+  for (const auto& state : compact_->sub_compact_states) {
+    if (state.Current().HasBlobFileAdditions()) {
+      return true;
+    }
+  }
+  return false;
+}
+
+Status CompactionJob::CollectSubcompactionErrors() {
  Status status;
  IOStatus io_s;
-  bool wrote_new_blob_files = false;

  for (const auto& state : compact_->sub_compact_states) {
    if (!state.status.ok()) {
@@ -720,126 +741,131 @@ Status CompactionJob::Run() {
      io_s = state.io_status;
      break;
    }
-
-    if (state.Current().HasBlobFileAdditions()) {
-      wrote_new_blob_files = true;
-    }
  }

  if (io_status_.ok()) {
    io_status_ = io_s;
  }
-  if (status.ok()) {
-    constexpr IODebugContext* dbg = nullptr;

-    if (output_directory_) {
-      io_s = output_directory_->FsyncWithDirOptions(
-          IOOptions(), dbg,
-          DirFsyncOptions(DirFsyncOptions::FsyncReason::kNewFileSynced));
-    }
+  return status;
+}

-    if (io_s.ok() && wrote_new_blob_files && blob_output_directory_ &&
-        blob_output_directory_ != output_directory_) {
-      io_s = blob_output_directory_->FsyncWithDirOptions(
-          IOOptions(), dbg,
-          DirFsyncOptions(DirFsyncOptions::FsyncReason::kNewFileSynced));
-    }
+Status CompactionJob::SyncOutputDirectories() {
+  Status status;
+  IOStatus io_s;
+  constexpr IODebugContext* dbg = nullptr;
+  const bool wrote_new_blob_files = HasNewBlobFiles();
+  if (output_directory_) {
+    io_s = output_directory_->FsyncWithDirOptions(
+        IOOptions(), dbg,
+        DirFsyncOptions(DirFsyncOptions::FsyncReason::kNewFileSynced));
  }
+
+  if (io_s.ok() && wrote_new_blob_files && blob_output_directory_ &&
+      blob_output_directory_ != output_directory_) {
+    io_s = blob_output_directory_->FsyncWithDirOptions(
+        IOOptions(), dbg,
+        DirFsyncOptions(DirFsyncOptions::FsyncReason::kNewFileSynced));
+  }
+
  if (io_status_.ok()) {
    io_status_ = io_s;
  }
  if (status.ok()) {
    status = io_s;
  }
-  if (status.ok()) {
-    thread_pool.clear();
-    std::vector<const CompactionOutputs::Output*> files_output;
-    for (const auto& state : compact_->sub_compact_states) {
-      for (const auto& output : state.GetOutputs()) {
-        files_output.emplace_back(&output);
-      }
+
+  return status;
+}
+
+Status CompactionJob::VerifyOutputFiles() {
+  Status status;
+  std::vector<port::Thread> thread_pool;
+  std::vector<const CompactionOutputs::Output*> files_output;
+  for (const auto& state : compact_->sub_compact_states) {
+    for (const auto& output : state.GetOutputs()) {
+      files_output.emplace_back(&output);
    }
-    ColumnFamilyData* cfd = compact_->compaction->column_family_data();
-    std::atomic<size_t> next_file_idx(0);
-    auto verify_table = [&](Status& output_status) {
-      while (true) {
-        size_t file_idx = next_file_idx.fetch_add(1);
-        if (file_idx >= files_output.size()) {
-          break;
-        }
-        // Verify that the table is usable
-        // We set for_compaction to false and don't
-        // OptimizeForCompactionTableRead here because this is a special case
-        // after we finish the table building No matter whether
-        // use_direct_io_for_flush_and_compaction is true, we will regard this
-        // verification as user reads since the goal is to cache it here for
-        // further user reads
-        ReadOptions verify_table_read_options(Env::IOActivity::kCompaction);
-        verify_table_read_options.rate_limiter_priority =
-            GetRateLimiterPriority();
-        InternalIterator* iter = cfd->table_cache()->NewIterator(
-            verify_table_read_options, file_options_,
-            cfd->internal_comparator(), files_output[file_idx]->meta,
-            /*range_del_agg=*/nullptr,
-            compact_->compaction->mutable_cf_options(),
-            /*table_reader_ptr=*/nullptr,
-            cfd->internal_stats()->GetFileReadHist(
-                compact_->compaction->output_level()),
-            TableReaderCaller::kCompactionRefill, /*arena=*/nullptr,
-            /*skip_filters=*/false, compact_->compaction->output_level(),
-            MaxFileSizeForL0MetaPin(compact_->compaction->mutable_cf_options()),
-            /*smallest_compaction_key=*/nullptr,
-            /*largest_compaction_key=*/nullptr,
-            /*allow_unprepared_value=*/false);
-        auto s = iter->status();
+  }
+  ColumnFamilyData* cfd = compact_->compaction->column_family_data();
+  std::atomic<size_t> next_file_idx(0);
+  auto verify_table = [&](Status& output_status) {
+    while (true) {
+      size_t file_idx = next_file_idx.fetch_add(1);
+      if (file_idx >= files_output.size()) {
+        break;
+      }
+      // Verify that the table is usable
+      // We set for_compaction to false and don't
+      // OptimizeForCompactionTableRead here because this is a special case
+      // after we finish the table building No matter whether
+      // use_direct_io_for_flush_and_compaction is true, we will regard this
+      // verification as user reads since the goal is to cache it here for
+      // further user reads
+      ReadOptions verify_table_read_options(Env::IOActivity::kCompaction);
+      verify_table_read_options.rate_limiter_priority =
+          GetRateLimiterPriority();
+      InternalIterator* iter = cfd->table_cache()->NewIterator(
+          verify_table_read_options, file_options_, cfd->internal_comparator(),
+          files_output[file_idx]->meta,
+          /*range_del_agg=*/nullptr, compact_->compaction->mutable_cf_options(),
+          /*table_reader_ptr=*/nullptr,
+          cfd->internal_stats()->GetFileReadHist(
+              compact_->compaction->output_level()),
+          TableReaderCaller::kCompactionRefill, /*arena=*/nullptr,
+          /*skip_filters=*/false, compact_->compaction->output_level(),
+          MaxFileSizeForL0MetaPin(compact_->compaction->mutable_cf_options()),
+          /*smallest_compaction_key=*/nullptr,
+          /*largest_compaction_key=*/nullptr,
+          /*allow_unprepared_value=*/false);
+      auto s = iter->status();

-        if (s.ok() && paranoid_file_checks_) {
-          OutputValidator validator(cfd->internal_comparator(),
-                                    /*_enable_hash=*/true);
-          for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
-            s = validator.Add(iter->key(), iter->value());
-            if (!s.ok()) {
-              break;
-            }
-          }
-          if (s.ok()) {
-            s = iter->status();
-          }
-          if (s.ok() &&
-              !validator.CompareValidator(files_output[file_idx]->validator)) {
-            s = Status::Corruption("Paranoid checksums do not match");
+      if (s.ok() && paranoid_file_checks_) {
+        OutputValidator validator(cfd->internal_comparator(),
+                                  /*_enable_hash=*/true);
+        for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
+          s = validator.Add(iter->key(), iter->value());
+          if (!s.ok()) {
+            break;
          }
        }
-
-        delete iter;
-
-        if (!s.ok()) {
-          output_status = s;
-          break;
+        if (s.ok()) {
+          s = iter->status();
+        }
+        if (s.ok() &&
+            !validator.CompareValidator(files_output[file_idx]->validator)) {
+          s = Status::Corruption("Paranoid checksums do not match");
        }
      }
-    };
-    for (size_t i = 1; i < compact_->sub_compact_states.size(); i++) {
-      thread_pool.emplace_back(
-          verify_table, std::ref(compact_->sub_compact_states[i].status));
-    }
-    verify_table(compact_->sub_compact_states[0].status);
-    for (auto& thread : thread_pool) {
-      thread.join();
-    }

-    for (const auto& state : compact_->sub_compact_states) {
-      if (!state.status.ok()) {
-        status = state.status;
+      delete iter;
+
+      if (!s.ok()) {
+        output_status = s;
        break;
      }
    }
+  };
+  for (size_t i = 1; i < compact_->sub_compact_states.size(); i++) {
+    thread_pool.emplace_back(verify_table,
+                             std::ref(compact_->sub_compact_states[i].status));
+  }
+  verify_table(compact_->sub_compact_states[0].status);
+  for (auto& thread : thread_pool) {
+    thread.join();
  }

-  ReleaseSubcompactionResources();
-  TEST_SYNC_POINT("CompactionJob::ReleaseSubcompactionResources:0");
-  TEST_SYNC_POINT("CompactionJob::ReleaseSubcompactionResources:1");
+  for (const auto& state : compact_->sub_compact_states) {
+    if (!state.status.ok()) {
+      status = state.status;
+      break;
+    }
+  }

+  return status;
+}
+
+void CompactionJob::SetOutputTableProperties() {
  for (const auto& state : compact_->sub_compact_states) {
    for (const auto& output : state.GetOutputs()) {
      auto fn =
@@ -849,7 +875,9 @@ Status CompactionJob::Run() {
                                                     output.table_properties);
    }
  }
+}

+void CompactionJob::AggregateSubcompactionStats() {
  // Before the compaction starts, is_remote_compaction was set to true if
  // compaction_service is set. We now know whether each sub_compaction was
  // done remotely or not. Reset is_remote_compaction back to false and allow
@@ -858,61 +886,86 @@ Status CompactionJob::Run() {

  // Finish up all bookkeeping to unify the subcompaction results.
  compact_->AggregateCompactionStats(internal_stats_, *job_stats_);
+}

-  uint64_t num_input_range_del = 0;
-  bool ok = BuildStatsFromInputTableProperties(&num_input_range_del);
-  // (Sub)compactions returned ok, do sanity check on the number of input
-  // keys.
-  if (status.ok() && ok) {
-    if (job_stats_->has_num_input_records) {
-      status = VerifyInputRecordCount(num_input_range_del);
-      if (!status.ok()) {
-        ROCKS_LOG_WARN(
-            db_options_.info_log, "[%s] [JOB %d] Compaction with status: %s",
-            compact_->compaction->column_family_data()->GetName().c_str(),
-            job_context_->job_id, status.ToString().c_str());
-      }
+Status CompactionJob::VerifyCompactionRecordCounts(
+    bool stats_built_from_input_table_prop, uint64_t num_input_range_del) {
+  Status status;
+  if (stats_built_from_input_table_prop && job_stats_->has_num_input_records) {
+    status = VerifyInputRecordCount(num_input_range_del);
+    if (!status.ok()) {
+      return status;
    }
+  }
+
+  const auto& mutable_cf_options = compact_->compaction->mutable_cf_options();
+  if ((mutable_cf_options.table_factory->IsInstanceOf(
+           TableFactory::kBlockBasedTableName()) ||
+       mutable_cf_options.table_factory->IsInstanceOf(
+           TableFactory::kPlainTableName()))) {
+    status = VerifyOutputRecordCount();
+    if (!status.ok()) {
+      return status;
+    }
+  }
+  return status;
+}
+
+void CompactionJob::FinalizeCompactionRun(
+    const Status& input_status, bool stats_built_from_input_table_prop,
+    uint64_t num_input_range_del) {
+  if (stats_built_from_input_table_prop) {
    UpdateCompactionJobInputStats(internal_stats_, num_input_range_del);
  }
  UpdateCompactionJobOutputStats(internal_stats_);
-
-  // Verify number of output records
-  if (status.ok() && db_options_.compaction_verify_record_count) {
-    uint64_t total_output_num = 0;
-    for (const auto& state : compact_->sub_compact_states) {
-      for (const auto& output : state.GetOutputs()) {
-        total_output_num += output.table_properties->num_entries -
-                            output.table_properties->num_range_deletions;
-      }
-    }
-
-    uint64_t expected = internal_stats_.output_level_stats.num_output_records;
-    if (internal_stats_.has_proximal_level_output) {
-      expected += internal_stats_.proximal_level_stats.num_output_records;
-    }
-    if (expected != total_output_num) {
-      char scratch[2345];
-      compact_->compaction->Summary(scratch, sizeof(scratch));
-      std::string msg =
-          "Number of keys in compaction output SST files does not match "
-          "number of keys added. Expected " +
-          std::to_string(expected) + " but there are " +
-          std::to_string(total_output_num) +
-          " in output SST files. Compaction summary: " + scratch;
-      ROCKS_LOG_WARN(
-          db_options_.info_log, "[%s] [JOB %d] Compaction with status: %s",
-          compact_->compaction->column_family_data()->GetName().c_str(),
-          job_context_->job_id, msg.c_str());
-      status = Status::Corruption(msg);
-    }
-  }
-
  RecordCompactionIOStats();
+
  LogFlush(db_options_.info_log);
  TEST_SYNC_POINT("CompactionJob::Run():End");
-  compact_->status = status;
-  TEST_SYNC_POINT_CALLBACK("CompactionJob::Run():EndStatusSet", &status);
+  compact_->status = input_status;
+  TEST_SYNC_POINT_CALLBACK("CompactionJob::Run():EndStatusSet",
+                           const_cast<Status*>(&input_status));
+}
+
+Status CompactionJob::Run() {
+  InitializeCompactionRun();
+
+  const uint64_t start_micros = db_options_.clock->NowMicros();
+
+  RunSubcompactions();
+
+  UpdateTimingStats(start_micros);
+
+  TEST_SYNC_POINT("CompactionJob::Run:BeforeVerify");
+
+  Status status = CollectSubcompactionErrors();
+
+  if (status.ok()) {
+    status = SyncOutputDirectories();
+  }
+
+  if (status.ok()) {
+    status = VerifyOutputFiles();
+  }
+
+  if (status.ok()) {
+    SetOutputTableProperties();
+  }
+
+  AggregateSubcompactionStats();
+
+  uint64_t num_input_range_del = 0;
+  bool stats_built_from_input_table_prop =
+      BuildStatsFromInputFiles(&num_input_range_del);
+
+  if (status.ok()) {
+    status = VerifyCompactionRecordCounts(stats_built_from_input_table_prop,
+                                          num_input_range_del);
+  }
+
+  FinalizeCompactionRun(status, stats_built_from_input_table_prop,
+                        num_input_range_del);
+
  return status;
 }

@@ -1176,7 +1229,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
  // creation across both CompactionJob and CompactionServiceCompactionJob
  sub_compact->AssignRangeDelAggregator(
      std::make_unique<CompactionRangeDelAggregator>(
-          &cfd->internal_comparator(), existing_snapshots_,
+          &cfd->internal_comparator(), job_context_->snapshot_seqs,
          &full_history_ts_low_, &trim_ts_));

  // TODO: since we already use C++17, should use
@@ -1317,8 +1370,8 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
      env_, cfd->user_comparator(), cfd->ioptions().merge_operator.get(),
      compaction_filter, db_options_.info_log.get(),
      false /* internal key corruption is expected */,
-      existing_snapshots_.empty() ? 0 : existing_snapshots_.back(),
-      snapshot_checker_, compact_->compaction->level(), db_options_.stats);
+      job_context_->GetLatestSnapshotSequence(), job_context_->snapshot_checker,
+      compact_->compaction->level(), db_options_.stats);

  const auto& mutable_cf_options =
      sub_compact->compaction->mutable_cf_options();
@@ -1354,10 +1407,10 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {

  auto c_iter = std::make_unique<CompactionIterator>(
      input, cfd->user_comparator(), &merge, versions_->LastSequence(),
-      &existing_snapshots_, earliest_snapshot_,
-      earliest_write_conflict_snapshot_, job_snapshot_seq, snapshot_checker_,
-      env_, ShouldReportDetailedTime(env_, stats_),
-      /*expect_valid_internal_key=*/true, sub_compact->RangeDelAgg(),
+      &(job_context_->snapshot_seqs), earliest_snapshot_,
+      job_context_->earliest_write_conflict_snapshot, job_snapshot_seq,
+      job_context_->snapshot_checker, env_,
+      ShouldReportDetailedTime(env_, stats_), sub_compact->RangeDelAgg(),
      blob_file_builder.get(), db_options_.allow_data_in_errors,
      db_options_.enforce_single_del_contracts, manual_compaction_canceled_,
      sub_compact->compaction
@@ -1645,10 +1698,6 @@ Status CompactionJob::FinishCompactionOutputFile(
  Status s = input_status;

  // Add range tombstones
-  auto earliest_snapshot = kMaxSequenceNumber;
-  if (existing_snapshots_.size() > 0) {
-    earliest_snapshot = existing_snapshots_[0];
-  }
  if (s.ok()) {
    // Inclusive lower bound, exclusive upper bound
    std::pair<SequenceNumber, SequenceNumber> keep_seqno_range{
@@ -1674,7 +1723,7 @@ Status CompactionJob::FinishCompactionOutputFile(
      s = outputs.AddRangeDels(*sub_compact->RangeDelAgg(), comp_start_user_key,
                               comp_end_user_key, range_del_out_stats,
                               bottommost_level_, cfd->internal_comparator(),
-                               earliest_snapshot, keep_seqno_range,
+                               earliest_snapshot_, keep_seqno_range,
                               next_table_min_key, full_history_ts_low_);
    }
    RecordDroppedKeys(range_del_out_stats, &sub_compact->compaction_job_stats);
@@ -1731,14 +1780,11 @@ Status CompactionJob::FinishCompactionOutputFile(
  if (s.ok()) {
    tp = outputs.GetTableProperties();
  }
-
  if (s.ok() && current_entries == 0 && tp.num_range_deletions == 0) {
    // If there is nothing to output, no necessary to generate a sst file.
    // This happens when the output level is bottom level, at the same time
    // the sub_compact output nothing.
-    std::string fname =
-        TableFileName(sub_compact->compaction->immutable_options().cf_paths,
-                      meta->fd.GetNumber(), meta->fd.GetPathId());
+    std::string fname = GetTableFileName(meta->fd.GetNumber());

    // TODO(AR) it is not clear if there are any larger implications if
    // DeleteFile fails here
@@ -1937,6 +1983,10 @@ Status CompactionJob::OpenCompactionOutputFile(SubcompactionState* sub_compact,

  // no need to lock because VersionSet::next_file_number_ is atomic
  uint64_t file_number = versions_->NewFileNumber();
+#ifndef NDEBUG
+  TEST_SYNC_POINT_CALLBACK(
+      "CompactionJob::OpenCompactionOutputFile::NewFileNumber", &file_number);
+#endif
  std::string fname = GetTableFileName(file_number);
  // Fire events.
  ColumnFamilyData* cfd = sub_compact->compaction->column_family_data();
@@ -2100,8 +2150,7 @@ void CopyPrefix(const Slice& src, size_t prefix_length, std::string* dst) {
 }
 }  // namespace

-bool CompactionJob::BuildStatsFromInputTableProperties(
-    uint64_t* num_input_range_del) {
+bool CompactionJob::BuildStatsFromInputFiles(uint64_t* num_input_range_del) {
  assert(compact_);

  Compaction* compaction = compact_->compaction;
@@ -2293,8 +2342,8 @@ void CompactionJob::LogCompaction() {
                   cfd->GetName().c_str(), scratch);
    // build event logger report
    auto stream = event_logger_->Log();
-    stream << "job" << job_id_ << "event" << "compaction_started"
-           << "compaction_reason"
+    stream << "job" << job_id_ << "event" << "compaction_started" << "cf_name"
+           << cfd->GetName() << "compaction_reason"
           << GetCompactionReasonString(compaction->compaction_reason());
    for (size_t i = 0; i < compaction->num_input_levels(); ++i) {
      stream << ("files_L" + std::to_string(compaction->level(i)));
@@ -2306,9 +2355,10 @@ void CompactionJob::LogCompaction() {
    }
    stream << "score" << compaction->score() << "input_data_size"
           << compaction->CalculateTotalInputSize() << "oldest_snapshot_seqno"
-           << (existing_snapshots_.empty()
+           << (job_context_->snapshot_seqs.empty()
                   ? int64_t{-1}  // Use -1 for "none"
-                   : static_cast<int64_t>(existing_snapshots_[0]));
+                   : static_cast<int64_t>(
+                         job_context_->GetEarliestSnapshotSequence()));
    if (compaction->SupportsPerKeyPlacement()) {
      stream << "proximal_after_seqno" << proximal_after_seqno_;
      stream << "preserve_seqno_after" << preserve_seqno_after_;
@@ -2371,6 +2421,11 @@ Status CompactionJob::VerifyInputRecordCount(
          "number of keys processed. Expected " +
          std::to_string(expected) + " but processed " +
          std::to_string(actual) + ". Compaction summary: " + scratch;
+      ROCKS_LOG_WARN(
+          db_options_.info_log,
+          "[%s] [JOB %d] VerifyInputRecordCount() Status: %s",
+          compact_->compaction->column_family_data()->GetName().c_str(),
+          job_context_->job_id, msg.c_str());
      if (db_options_.compaction_verify_record_count) {
        return Status::Corruption(msg);
      }
@@ -2379,4 +2434,38 @@ Status CompactionJob::VerifyInputRecordCount(
  return Status::OK();
 }

+Status CompactionJob::VerifyOutputRecordCount() const {
+  uint64_t total_output_num = 0;
+  for (const auto& state : compact_->sub_compact_states) {
+    for (const auto& output : state.GetOutputs()) {
+      total_output_num += output.table_properties->num_entries -
+                          output.table_properties->num_range_deletions;
+    }
+  }
+
+  uint64_t expected = internal_stats_.output_level_stats.num_output_records;
+  if (internal_stats_.has_proximal_level_output) {
+    expected += internal_stats_.proximal_level_stats.num_output_records;
+  }
+  if (expected != total_output_num) {
+    char scratch[2345];
+    compact_->compaction->Summary(scratch, sizeof(scratch));
+    std::string msg =
+        "Number of keys in compaction output SST files does not match "
+        "number of keys added. Expected " +
+        std::to_string(expected) + " but there are " +
+        std::to_string(total_output_num) +
+        " in output SST files. Compaction summary: " + scratch;
+    ROCKS_LOG_WARN(
+        db_options_.info_log,
+        "[%s] [JOB %d] VerifyOutputRecordCount() status: %s",
+        compact_->compaction->column_family_data()->GetName().c_str(),
+        job_context_->job_id, msg.c_str());
+    if (db_options_.compaction_verify_record_count) {
+      return Status::Corruption(msg);
+    }
+  }
+  return Status::OK();
+}
+
 }  // namespace ROCKSDB_NAMESPACE
@@ -142,27 +142,27 @@ class SubcompactionState;

 class CompactionJob {
 public:
-  CompactionJob(
-      int job_id, Compaction* compaction, const ImmutableDBOptions& db_options,
-      const MutableDBOptions& mutable_db_options,
-      const FileOptions& file_options, VersionSet* versions,
-      const std::atomic<bool>* shutting_down, LogBuffer* log_buffer,
-      FSDirectory* db_directory, FSDirectory* output_directory,
-      FSDirectory* blob_output_directory, Statistics* stats,
-      InstrumentedMutex* db_mutex, ErrorHandler* db_error_handler,
-      std::vector<SequenceNumber> existing_snapshots,
-      SequenceNumber earliest_write_conflict_snapshot,
-      const SnapshotChecker* snapshot_checker, JobContext* job_context,
-      std::shared_ptr<Cache> table_cache, EventLogger* event_logger,
-      bool paranoid_file_checks, bool measure_io_stats,
-      const std::string& dbname, CompactionJobStats* compaction_job_stats,
-      Env::Priority thread_pri, const std::shared_ptr<IOTracer>& io_tracer,
-      const std::atomic<bool>& manual_compaction_canceled,
-      const std::string& db_id = "", const std::string& db_session_id = "",
-      std::string full_history_ts_low = "", std::string trim_ts = "",
-      BlobFileCompletionCallback* blob_callback = nullptr,
-      int* bg_compaction_scheduled = nullptr,
-      int* bg_bottom_compaction_scheduled = nullptr);
+  CompactionJob(int job_id, Compaction* compaction,
+                const ImmutableDBOptions& db_options,
+                const MutableDBOptions& mutable_db_options,
+                const FileOptions& file_options, VersionSet* versions,
+                const std::atomic<bool>* shutting_down, LogBuffer* log_buffer,
+                FSDirectory* db_directory, FSDirectory* output_directory,
+                FSDirectory* blob_output_directory, Statistics* stats,
+                InstrumentedMutex* db_mutex, ErrorHandler* db_error_handler,
+                JobContext* job_context, std::shared_ptr<Cache> table_cache,
+                EventLogger* event_logger, bool paranoid_file_checks,
+                bool measure_io_stats, const std::string& dbname,
+                CompactionJobStats* compaction_job_stats,
+                Env::Priority thread_pri,
+                const std::shared_ptr<IOTracer>& io_tracer,
+                const std::atomic<bool>& manual_compaction_canceled,
+                const std::string& db_id = "",
+                const std::string& db_session_id = "",
+                std::string full_history_ts_low = "", std::string trim_ts = "",
+                BlobFileCompletionCallback* blob_callback = nullptr,
+                int* bg_compaction_scheduled = nullptr,
+                int* bg_bottom_compaction_scheduled = nullptr);

  virtual ~CompactionJob();

@@ -225,7 +225,7 @@ class CompactionJob {
 private:
  friend class CompactionJobTestBase;

-  // Collect the following stats from Input Table Properties
+  // Collect the following stats from input files and table properties
  // - num_input_files_in_non_output_levels
  // - num_input_files_in_output_level
  // - bytes_read_non_output_levels
@@ -242,14 +242,14 @@ class CompactionJob {
  // num_input_range_del are calculated successfully.
  //
  // This should be called only once for compactions (not per subcompaction)
-  bool BuildStatsFromInputTableProperties(
-      uint64_t* num_input_range_del = nullptr);
+  bool BuildStatsFromInputFiles(uint64_t* num_input_range_del = nullptr);

  void UpdateCompactionJobInputStats(
      const InternalStats::CompactionStatsFull& internal_stats,
      uint64_t num_input_range_del) const;

  Status VerifyInputRecordCount(uint64_t num_input_range_del) const;
+  Status VerifyOutputRecordCount() const;

  // Generates a histogram representing potential divisions of key ranges from
  // the input. It adds the starting and/or ending keys of certain input files
@@ -278,6 +278,22 @@ class CompactionJob {
  // Release all reserved threads and update the compaction limits.
  void ReleaseSubcompactionResources();

+  void InitializeCompactionRun();
+  void RunSubcompactions();
+  void UpdateTimingStats(uint64_t start_micros);
+  void RemoveEmptyOutputs();
+  bool HasNewBlobFiles() const;
+  Status CollectSubcompactionErrors();
+  Status SyncOutputDirectories();
+  Status VerifyOutputFiles();
+  void SetOutputTableProperties();
+  void AggregateSubcompactionStats();
+  Status VerifyCompactionRecordCounts(bool stats_built_from_input_table_prop,
+                                      uint64_t num_input_range_del);
+  void FinalizeCompactionRun(const Status& status,
+                             bool stats_built_from_input_table_prop,
+                             uint64_t num_input_range_del);
+
  CompactionServiceJobStatus ProcessKeyValueCompactionWithCompactionService(
      SubcompactionState* sub_compact);

@@ -321,21 +337,8 @@ class CompactionJob {
  FSDirectory* blob_output_directory_;
  InstrumentedMutex* db_mutex_;
  ErrorHandler* db_error_handler_;
-  // If there were two snapshots with seq numbers s1 and
-  // s2 and s1 < s2, and if we find two instances of a key k1 then lies
-  // entirely within s1 and s2, then the earlier version of k1 can be safely
-  // deleted because that version is not visible in any snapshot.
-  std::vector<SequenceNumber> existing_snapshots_;

  SequenceNumber earliest_snapshot_;
-
-  // This is the earliest snapshot that could be used for write-conflict
-  // checking by a transaction.  For any user-key newer than this snapshot, we
-  // should make sure not to remove evidence that a write occurred.
-  SequenceNumber earliest_write_conflict_snapshot_;
-
-  const SnapshotChecker* const snapshot_checker_;
-
  JobContext* job_context_;

  std::shared_ptr<Cache> table_cache_;
@@ -427,8 +430,9 @@ struct CompactionServiceInput {
 // CompactionServiceOutputFile is the metadata for the output SST file
 struct CompactionServiceOutputFile {
  std::string file_name;
-  SequenceNumber smallest_seqno;
-  SequenceNumber largest_seqno;
+  uint64_t file_size{};
+  SequenceNumber smallest_seqno{};
+  SequenceNumber largest_seqno{};
  std::string smallest_internal_key;
  std::string largest_internal_key;
  uint64_t oldest_ancester_time = kUnknownOldestAncesterTime;
@@ -436,24 +440,26 @@ struct CompactionServiceOutputFile {
  uint64_t epoch_number = kUnknownEpochNumber;
  std::string file_checksum = kUnknownFileChecksum;
  std::string file_checksum_func_name = kUnknownFileChecksumFuncName;
-  uint64_t paranoid_hash;
+  uint64_t paranoid_hash{};
  bool marked_for_compaction;
  UniqueId64x2 unique_id{};
  TableProperties table_properties;
  bool is_proximal_level_output;
-  Temperature file_temperature;
+  Temperature file_temperature = Temperature::kUnknown;

  CompactionServiceOutputFile() = default;
  CompactionServiceOutputFile(
-      const std::string& name, SequenceNumber smallest, SequenceNumber largest,
-      std::string _smallest_internal_key, std::string _largest_internal_key,
-      uint64_t _oldest_ancester_time, uint64_t _file_creation_time,
-      uint64_t _epoch_number, const std::string& _file_checksum,
+      const std::string& name, uint64_t size, SequenceNumber smallest,
+      SequenceNumber largest, std::string _smallest_internal_key,
+      std::string _largest_internal_key, uint64_t _oldest_ancester_time,
+      uint64_t _file_creation_time, uint64_t _epoch_number,
+      const std::string& _file_checksum,
      const std::string& _file_checksum_func_name, uint64_t _paranoid_hash,
      bool _marked_for_compaction, UniqueId64x2 _unique_id,
      const TableProperties& _table_properties, bool _is_proximal_level_output,
      Temperature _file_temperature)
      : file_name(name),
+        file_size(size),
        smallest_seqno(smallest),
        largest_seqno(largest),
        smallest_internal_key(std::move(_smallest_internal_key)),
@@ -521,9 +527,9 @@ class CompactionServiceCompactionJob : private CompactionJob {
      const std::atomic<bool>* shutting_down, LogBuffer* log_buffer,
      FSDirectory* output_directory, Statistics* stats,
      InstrumentedMutex* db_mutex, ErrorHandler* db_error_handler,
-      std::vector<SequenceNumber> existing_snapshots,
-      std::shared_ptr<Cache> table_cache, EventLogger* event_logger,
-      const std::string& dbname, const std::shared_ptr<IOTracer>& io_tracer,
+      JobContext* job_context, std::shared_ptr<Cache> table_cache,
+      EventLogger* event_logger, const std::string& dbname,
+      const std::shared_ptr<IOTracer>& io_tracer,
      const std::atomic<bool>& manual_compaction_canceled,
      const std::string& db_id, const std::string& db_session_id,
      std::string output_path,
@@ -43,7 +43,6 @@ void VerifyInitializationOfCompactionJobStats(
  ASSERT_EQ(compaction_job_stats.elapsed_micros, 0U);

  ASSERT_EQ(compaction_job_stats.num_input_records, 0U);
-  ASSERT_EQ(compaction_job_stats.num_input_files, 0U);
  ASSERT_EQ(compaction_job_stats.num_input_files_at_output_level, 0U);

  ASSERT_EQ(compaction_job_stats.num_output_records, 0U);
@@ -52,7 +51,6 @@ void VerifyInitializationOfCompactionJobStats(
  ASSERT_TRUE(compaction_job_stats.is_manual_compaction);
  ASSERT_FALSE(compaction_job_stats.is_remote_compaction);

-  ASSERT_EQ(compaction_job_stats.total_input_bytes, 0U);
  ASSERT_EQ(compaction_job_stats.total_output_bytes, 0U);

  ASSERT_EQ(compaction_job_stats.total_input_raw_key_bytes, 0U);
@@ -232,11 +230,6 @@ class CompactionJobTestBase : public testing::Test {
    // set default for the tests
    mutable_cf_options_.target_file_size_base = 1024 * 1024;
    mutable_cf_options_.max_compaction_bytes = 10 * 1024 * 1024;
-
-    // Turn off compaction_verify_record_count MockTables
-    if (table_type == TableTypeForTest::kMockTable) {
-      db_options_.compaction_verify_record_count = false;
-    }
  }

  void SetUp() override {
@@ -465,9 +458,10 @@ class CompactionJobTestBase : public testing::Test {
      ReadOptions read_opts;
      Status s = cf_options_.table_factory->NewTableReader(
          read_opts,
-          TableReaderOptions(cfd->ioptions(), nullptr, FileOptions(),
+          TableReaderOptions(cfd->ioptions(), /*prefix_extractor=*/nullptr,
+                             /*compression_manager=*/nullptr, FileOptions(),
                             cfd_->internal_comparator(),
-                             0 /* block_protection_bytes_per_key */),
+                             /*block_protection_bytes_per_key=*/0),
          std::move(freader), file_size, &table_reader, false);
      ASSERT_OK(s);
      assert(table_reader);
@@ -600,11 +594,11 @@ class CompactionJobTestBase : public testing::Test {
      const std::vector<std::vector<FileMetaData*>>& input_files,
      const std::vector<int> input_levels,
      std::function<void(Compaction& comp)>&& verify_func,
-      const std::vector<SequenceNumber>& snapshots = {}) {
+      std::vector<SequenceNumber>&& snapshots = {}) {
    const int kLastLevel = cf_options_.num_levels - 1;
    verify_per_key_placement_ = std::move(verify_func);
    mock::KVVector empty_map;
-    RunCompaction(input_files, input_levels, {empty_map}, snapshots,
+    RunCompaction(input_files, input_levels, {empty_map}, std::move(snapshots),
                  kMaxSequenceNumber, kLastLevel, false);
  }

@@ -613,7 +607,7 @@ class CompactionJobTestBase : public testing::Test {
      const std::vector<std::vector<FileMetaData*>>& input_files,
      const std::vector<int>& input_levels,
      const std::vector<mock::KVVector>& expected_results,
-      const std::vector<SequenceNumber>& snapshots = {},
+      std::vector<SequenceNumber>&& snapshots = {},
      SequenceNumber earliest_write_conflict_snapshot = kMaxSequenceNumber,
      int output_level = 1, bool verify = true,
      std::vector<uint64_t> expected_oldest_blob_file_numbers = {},
@@ -657,7 +651,8 @@ class CompactionJobTestBase : public testing::Test {
        mutable_cf_options_.max_compaction_bytes, 0, kNoCompression,
        cfd->GetLatestMutableCFOptions().compression_opts,
        Temperature::kUnknown, max_subcompactions, grandparents,
-        /*earliest_snapshot*/ std::nullopt, /*snapshot_checker*/ nullptr, true);
+        /*earliest_snapshot*/ std::nullopt, /*snapshot_checker*/ nullptr,
+        CompactionReason::kManualCompaction);
    compaction.FinalizeInputInfo(cfd->current());

    assert(db_options_.info_log);
@@ -670,13 +665,15 @@ class CompactionJobTestBase : public testing::Test {
                ucmp_->timestamp_size() == full_history_ts_low_.size());
    const std::atomic<bool> kManualCompactionCanceledFalse{false};
    JobContext job_context(1, false /* create_superversion */);
+    job_context.InitSnapshotContext(snapshot_checker, nullptr,
+                                    earliest_write_conflict_snapshot,
+                                    std::move(snapshots));
    CompactionJob compaction_job(
        0, &compaction, db_options_, mutable_db_options_, env_options_,
        versions_.get(), &shutting_down_, &log_buffer, nullptr, nullptr,
-        nullptr, nullptr, &mutex_, &error_handler_, snapshots,
-        earliest_write_conflict_snapshot, snapshot_checker, &job_context,
-        table_cache_, &event_logger, false, false, dbname_,
-        &compaction_job_stats_, Env::Priority::USER, nullptr /* IOTracer */,
+        nullptr, nullptr, &mutex_, &error_handler_, &job_context, table_cache_,
+        &event_logger, false, false, dbname_, &compaction_job_stats_,
+        Env::Priority::USER, nullptr /* IOTracer */,
        /*manual_compaction_canceled=*/kManualCompactionCanceledFalse,
        env_->GenerateUniqueId(), DBImpl::GenerateDbSessionId(nullptr),
        full_history_ts_low_);
@@ -1674,6 +1671,7 @@ TEST_F(CompactionJobTest, ResultSerialization) {
    UniqueId64x2 id{rnd64.Uniform(UINT64_MAX), rnd64.Uniform(UINT64_MAX)};
    result.output_files.emplace_back(
        rnd.RandomString(rnd.Uniform(kStrMaxLen)) /* file_name */,
+        rnd64.Uniform(UINT64_MAX) /* file_size */,
        rnd64.Uniform(UINT64_MAX) /* smallest_seqno */,
        rnd64.Uniform(UINT64_MAX) /* largest_seqno */,
        rnd.RandomBinaryString(
@@ -2040,7 +2038,7 @@ TEST_F(CompactionJobTest, CutToAlignGrandparentBoundarySameKey) {
    snapshots.emplace_back(i);
  }
  RunCompaction({lvl0_files, lvl1_files}, input_levels,
-                {expected_file1, expected_file2}, snapshots);
+                {expected_file1, expected_file2}, std::move(snapshots));
 }

 TEST_F(CompactionJobTest, CutForMaxCompactionBytesSameKey) {
@@ -2099,7 +2097,8 @@ TEST_F(CompactionJobTest, CutForMaxCompactionBytesSameKey) {
    snapshots.emplace_back(i);
  }
  RunCompaction({lvl0_files, lvl1_files}, input_levels,
-                {expected_file1, expected_file2, expected_file3}, snapshots);
+                {expected_file1, expected_file2, expected_file3},
+                std::move(snapshots));
 }

 class CompactionJobTimestampTest : public CompactionJobTestBase {
@@ -54,6 +54,7 @@ Status CompactionOutputs::Finish(
  }
  current_output().finished = true;
  stats_.bytes_written += current_bytes;
+  stats_.bytes_written_pre_comp += builder_->PreCompressionSize();
  stats_.num_output_files = static_cast<int>(outputs_.size());

  return s;
@@ -333,11 +333,13 @@ bool CompactionPicker::AreFilesInCompaction(
  return false;
 }

-Compaction* CompactionPicker::CompactFiles(
+Compaction* CompactionPicker::PickCompactionForCompactFiles(
    const CompactionOptions& compact_options,
    const std::vector<CompactionInputFiles>& input_files, int output_level,
    VersionStorageInfo* vstorage, const MutableCFOptions& mutable_cf_options,
-    const MutableDBOptions& mutable_db_options, uint32_t output_path_id) {
+    const MutableDBOptions& mutable_db_options, uint32_t output_path_id,
+    std::optional<SequenceNumber> earliest_snapshot,
+    const SnapshotChecker* snapshot_checker) {
 #ifndef NDEBUG
  assert(input_files.size());
  // This compaction output should not overlap with a running compaction as
@@ -380,8 +382,8 @@ Compaction* CompactionPicker::CompactFiles(
      GetCompressionOptions(mutable_cf_options, vstorage, output_level),
      mutable_cf_options.default_write_temperature,
      compact_options.max_subcompactions,
-      /* grandparents */ {}, /* earliest_snapshot */ std::nullopt,
-      /* snapshot_checker */ nullptr, true);
+      /* grandparents */ {}, earliest_snapshot, snapshot_checker,
+      CompactionReason::kManualCompaction);
  RegisterCompaction(c);
  return c;
 }
@@ -601,7 +603,7 @@ void CompactionPicker::GetGrandparents(
  }
 }

-Compaction* CompactionPicker::CompactRange(
+Compaction* CompactionPicker::PickCompactionForCompactRange(
    const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
    const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
    int input_level, int output_level,
@@ -617,8 +619,8 @@ Compaction* CompactionPicker::CompactRange(
    // Universal compaction with more than one level always compacts all the
    // files together to the last level.
    assert(vstorage->num_levels() > 1);
-    int max_output_level =
-        vstorage->MaxOutputLevel(ioptions_.allow_ingest_behind);
+    int max_output_level = vstorage->MaxOutputLevel(
+        ioptions_.cf_allow_ingest_behind || ioptions_.allow_ingest_behind);
    // DBImpl::CompactRange() set output level to be the last level
    assert(output_level == max_output_level);
    // DBImpl::RunManualCompaction will make full range for universal compaction
@@ -680,10 +682,9 @@ Compaction* CompactionPicker::CompactRange(
        mutable_cf_options.default_write_temperature,
        compact_range_options.max_subcompactions,
        /* grandparents */ {}, /* earliest_snapshot */ std::nullopt,
-        /* snapshot_checker */ nullptr,
-        /* is manual */ true, trim_ts, /* score */ -1,
-        /* deletion_compaction */ false, /* l0_files_might_overlap */ true,
-        CompactionReason::kUnknown,
+        /* snapshot_checker */ nullptr, CompactionReason::kManualCompaction,
+        trim_ts, /* score */ -1,
+        /* l0_files_might_overlap */ true,
        compact_range_options.blob_garbage_collection_policy,
        compact_range_options.blob_garbage_collection_age_cutoff);

@@ -873,9 +874,8 @@ Compaction* CompactionPicker::CompactRange(
      mutable_cf_options.default_write_temperature,
      compact_range_options.max_subcompactions, std::move(grandparents),
      /* earliest_snapshot */ std::nullopt, /* snapshot_checker */ nullptr,
-      /* is manual */ true, trim_ts, /* score */ -1,
-      /* deletion_compaction */ false, /* l0_files_might_overlap */ true,
-      CompactionReason::kUnknown,
+      CompactionReason::kManualCompaction, trim_ts, /* score */ -1,
+      /* l0_files_might_overlap */ true,
      compact_range_options.blob_garbage_collection_policy,
      compact_range_options.blob_garbage_collection_age_cutoff);

@@ -65,7 +65,7 @@ class CompactionPicker {
      const MutableDBOptions& mutable_db_options,
      const std::vector<SequenceNumber>& existing_snapshots,
      const SnapshotChecker* snapshot_checker, VersionStorageInfo* vstorage,
-      LogBuffer* log_buffer) = 0;
+      LogBuffer* log_buffer, bool require_max_output_level) = 0;

  // The returned Compaction might not include the whole requested range.
  // In that case, compaction_end will be set to the next key that needs
@@ -75,7 +75,7 @@ class CompactionPicker {
  // *compaction_end should point to valid InternalKey!
  // REQUIRES: If not compacting all levels (input_level == kCompactAllLevels),
  // then levels between input_level and output_level should be empty.
-  virtual Compaction* CompactRange(
+  virtual Compaction* PickCompactionForCompactRange(
      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
      const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
      int input_level, int output_level,
@@ -117,12 +117,17 @@ class CompactionPicker {
  // Caller must provide a set of input files that has been passed through
  // `SanitizeAndConvertCompactionInputFiles` earlier. The lock should not be
  // released between that call and this one.
-  Compaction* CompactFiles(const CompactionOptions& compact_options,
-                           const std::vector<CompactionInputFiles>& input_files,
-                           int output_level, VersionStorageInfo* vstorage,
-                           const MutableCFOptions& mutable_cf_options,
-                           const MutableDBOptions& mutable_db_options,
-                           uint32_t output_path_id);
+  //
+  //  TODO - Remove default values for earliest_snapshot and snapshot_checker
+  //  and require all callers to pass them in so that DB::CompactFiles() can
+  //  also benefit from Standalone Range Tombstone Optimization
+  Compaction* PickCompactionForCompactFiles(
+      const CompactionOptions& compact_options,
+      const std::vector<CompactionInputFiles>& input_files, int output_level,
+      VersionStorageInfo* vstorage, const MutableCFOptions& mutable_cf_options,
+      const MutableDBOptions& mutable_db_options, uint32_t output_path_id,
+      std::optional<SequenceNumber> earliest_snapshot = std::nullopt,
+      const SnapshotChecker* snapshot_checker = nullptr);

  // Converts a set of compaction input file numbers into
  // a list of CompactionInputFiles.
@@ -272,23 +277,23 @@ class NullCompactionPicker : public CompactionPicker {
      const MutableDBOptions& /*mutable_db_options*/,
      const std::vector<SequenceNumber>& /*existing_snapshots*/,
      const SnapshotChecker* /*snapshot_checker*/,
-      VersionStorageInfo* /*vstorage*/, LogBuffer* /* log_buffer */) override {
+      VersionStorageInfo* /*vstorage*/, LogBuffer* /* log_buffer */,
+      bool /*require_max_output_level*/ = false) override {
    return nullptr;
  }

  // Always return "nullptr"
-  Compaction* CompactRange(const std::string& /*cf_name*/,
-                           const MutableCFOptions& /*mutable_cf_options*/,
-                           const MutableDBOptions& /*mutable_db_options*/,
-                           VersionStorageInfo* /*vstorage*/,
-                           int /*input_level*/, int /*output_level*/,
-                           const CompactRangeOptions& /*compact_range_options*/,
-                           const InternalKey* /*begin*/,
-                           const InternalKey* /*end*/,
-                           InternalKey** /*compaction_end*/,
-                           bool* /*manual_conflict*/,
-                           uint64_t /*max_file_num_to_ignore*/,
-                           const std::string& /*trim_ts*/) override {
+  Compaction* PickCompactionForCompactRange(
+      const std::string& /*cf_name*/,
+      const MutableCFOptions& /*mutable_cf_options*/,
+      const MutableDBOptions& /*mutable_db_options*/,
+      VersionStorageInfo* /*vstorage*/, int /*input_level*/,
+      int /*output_level*/,
+      const CompactRangeOptions& /*compact_range_options*/,
+      const InternalKey* /*begin*/, const InternalKey* /*end*/,
+      InternalKey** /*compaction_end*/, bool* /*manual_conflict*/,
+      uint64_t /*max_file_num_to_ignore*/,
+      const std::string& /*trim_ts*/) override {
    return nullptr;
  }

@@ -127,11 +127,9 @@ Compaction* FIFOCompactionPicker::PickTTLCompaction(
      mutable_cf_options.compression_opts,
      mutable_cf_options.default_write_temperature,
      /* max_subcompactions */ 0, {}, /* earliest_snapshot */ std::nullopt,
-      /* snapshot_checker */ nullptr,
-      /* is manual */ false,
+      /* snapshot_checker */ nullptr, CompactionReason::kFIFOTtl,
      /* trim_ts */ "", vstorage->CompactionScore(0),
-      /* is deletion compaction */ true, /* l0_files_might_overlap */ true,
-      CompactionReason::kFIFOTtl);
+      /* l0_files_might_overlap */ true);
  return c;
 }

@@ -200,11 +198,10 @@ Compaction* FIFOCompactionPicker::PickSizeCompaction(
            mutable_cf_options.default_write_temperature,
            0 /* max_subcompactions */, {},
            /* earliest_snapshot */ std::nullopt,
-            /* snapshot_checker */ nullptr, /* is manual */ false,
+            /* snapshot_checker */ nullptr,
+            CompactionReason::kFIFOReduceNumFiles,
            /* trim_ts */ "", vstorage->CompactionScore(0),
-            /* is deletion compaction */ false,
-            /* l0_files_might_overlap */ true,
-            CompactionReason::kFIFOReduceNumFiles);
+            /* l0_files_might_overlap */ true);
        return c;
      }
    }
@@ -261,6 +258,9 @@ Compaction* FIFOCompactionPicker::PickSizeCompaction(
    // better serves a major type of FIFO use cases where smaller keys are
    // associated with older data.
    for (const auto& f : last_level_files) {
+      if (f->being_compacted) {
+        continue;
+      }
      total_size -= f->fd.file_size;
      inputs[0].files.push_back(f);
      char tmp_fsize[16];
@@ -297,11 +297,9 @@ Compaction* FIFOCompactionPicker::PickSizeCompaction(
      mutable_cf_options.compression_opts,
      mutable_cf_options.default_write_temperature,
      /* max_subcompactions */ 0, {}, /* earliest_snapshot */ std::nullopt,
-      /* snapshot_checker */ nullptr,
-      /* is manual */ false,
+      /* snapshot_checker */ nullptr, CompactionReason::kFIFOMaxSize,
      /* trim_ts */ "", vstorage->CompactionScore(0),
-      /* is deletion compaction */ true,
-      /* l0_files_might_overlap */ true, CompactionReason::kFIFOMaxSize);
+      /* l0_files_might_overlap */ true);
  return c;
 }

@@ -416,10 +414,9 @@ Compaction* FIFOCompactionPicker::PickTemperatureChangeCompaction(
      mutable_cf_options.compression, mutable_cf_options.compression_opts,
      compaction_target_temp,
      /* max_subcompactions */ 0, {}, /* earliest_snapshot */ std::nullopt,
-      /* snapshot_checker */ nullptr,
-      /* is manual */ false, /* trim_ts */ "", vstorage->CompactionScore(0),
-      /* is deletion compaction */ false, /* l0_files_might_overlap */ true,
-      CompactionReason::kChangeTemperature);
+      /* snapshot_checker */ nullptr, CompactionReason::kChangeTemperature,
+      /* trim_ts */ "", vstorage->CompactionScore(0),
+      /* l0_files_might_overlap */ true);
  return c;
 }

@@ -428,7 +425,7 @@ Compaction* FIFOCompactionPicker::PickCompaction(
    const MutableDBOptions& mutable_db_options,
    const std::vector<SequenceNumber>& /* existing_snapshots */,
    const SnapshotChecker* /* snapshot_checker */, VersionStorageInfo* vstorage,
-    LogBuffer* log_buffer) {
+    LogBuffer* log_buffer, bool /* require_max_output_level*/) {
  Compaction* c = nullptr;
  if (mutable_cf_options.ttl > 0) {
    c = PickTTLCompaction(cf_name, mutable_cf_options, mutable_db_options,
@@ -446,7 +443,7 @@ Compaction* FIFOCompactionPicker::PickCompaction(
  return c;
 }

-Compaction* FIFOCompactionPicker::CompactRange(
+Compaction* FIFOCompactionPicker::PickCompactionForCompactRange(
    const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
    const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
    int input_level, int output_level,
@@ -23,18 +23,17 @@ class FIFOCompactionPicker : public CompactionPicker {
      const MutableDBOptions& mutable_db_options,
      const std::vector<SequenceNumber>& /* existing_snapshots */,
      const SnapshotChecker* /* snapshot_checker */,
-      VersionStorageInfo* version, LogBuffer* log_buffer) override;
+      VersionStorageInfo* version, LogBuffer* log_buffer,
+      bool /* require_max_output_level*/ = false) override;

-  Compaction* CompactRange(const std::string& cf_name,
-                           const MutableCFOptions& mutable_cf_options,
-                           const MutableDBOptions& mutable_db_options,
-                           VersionStorageInfo* vstorage, int input_level,
-                           int output_level,
-                           const CompactRangeOptions& compact_range_options,
-                           const InternalKey* begin, const InternalKey* end,
-                           InternalKey** compaction_end, bool* manual_conflict,
-                           uint64_t max_file_num_to_ignore,
-                           const std::string& trim_ts) override;
+  Compaction* PickCompactionForCompactRange(
+      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
+      const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
+      int input_level, int output_level,
+      const CompactRangeOptions& compact_range_options,
+      const InternalKey* begin, const InternalKey* end,
+      InternalKey** compaction_end, bool* manual_conflict,
+      uint64_t max_file_num_to_ignore, const std::string& trim_ts) override;

  // The maximum allowed output level.  Always returns 0.
  int MaxOutputLevel() const override { return 0; }
@@ -145,7 +145,6 @@ class LevelCompactionBuilder {
  int parent_index_ = -1;
  int base_index_ = -1;
  double start_level_score_ = 0;
-  bool is_manual_ = false;
  bool is_l0_trivial_move_ = false;
  CompactionInputFiles start_level_inputs_;
  std::vector<CompactionInputFiles> compaction_inputs_;
@@ -561,9 +560,8 @@ Compaction* LevelCompactionBuilder::GetCompaction() {
      mutable_cf_options_.default_write_temperature,
      /* max_subcompactions */ 0, std::move(grandparents_),
      /* earliest_snapshot */ std::nullopt, /* snapshot_checker */ nullptr,
-      is_manual_,
-      /* trim_ts */ "", start_level_score_, false /* deletion_compaction */,
-      l0_files_might_overlap, compaction_reason_);
+      compaction_reason_,
+      /* trim_ts */ "", start_level_score_, l0_files_might_overlap);

  // If it's level 0 compaction, make sure we don't execute any other level 0
  // compactions in parallel
@@ -978,7 +976,7 @@ Compaction* LevelCompactionPicker::PickCompaction(
    const MutableDBOptions& mutable_db_options,
    const std::vector<SequenceNumber>& /*existing_snapshots */,
    const SnapshotChecker* /*snapshot_checker*/, VersionStorageInfo* vstorage,
-    LogBuffer* log_buffer) {
+    LogBuffer* log_buffer, bool /* require_max_output_level*/) {
  LevelCompactionBuilder builder(cf_name, vstorage, this, log_buffer,
                                 mutable_cf_options, ioptions_,
                                 mutable_db_options);
@@ -25,7 +25,8 @@ class LevelCompactionPicker : public CompactionPicker {
      const MutableDBOptions& mutable_db_options,
      const std::vector<SequenceNumber>& /* existing_snapshots */,
      const SnapshotChecker* /* snapshot_checker */,
-      VersionStorageInfo* vstorage, LogBuffer* log_buffer) override;
+      VersionStorageInfo* vstorage, LogBuffer* log_buffer,
+      bool /*require_max_output_level*/ = false) override;

  bool NeedsCompaction(const VersionStorageInfo* vstorage) const override;
 };
@@ -544,41 +544,48 @@ TEST_F(CompactionPickerTest, NeedsCompactionUniversal) {
 }

 TEST_F(CompactionPickerTest, CompactionUniversalIngestBehindReservedLevel) {
-  const uint64_t kFileSize = 100000;
-  NewVersionStorage(3 /* num_levels */, kCompactionStyleUniversal);
-  ioptions_.allow_ingest_behind = true;
-  ioptions_.num_levels = 3;
-  UniversalCompactionPicker universal_compaction_picker(ioptions_, &icmp_);
-  UpdateVersionStorageInfo();
-  // must return false when there's no files.
-  ASSERT_EQ(universal_compaction_picker.NeedsCompaction(vstorage_.get()),
-            false);
+  for (bool cf_option : {false, true}) {
+    SCOPED_TRACE("cf_option = " + std::to_string(cf_option));
+    const uint64_t kFileSize = 100000;
+    NewVersionStorage(3 /* num_levels */, kCompactionStyleUniversal);
+    if (cf_option) {
+      ioptions_.cf_allow_ingest_behind = true;
+    } else {
+      ioptions_.allow_ingest_behind = true;
+    }
+    ioptions_.num_levels = 3;
+    UniversalCompactionPicker universal_compaction_picker(ioptions_, &icmp_);
+    UpdateVersionStorageInfo();
+    // must return false when there's no files.
+    ASSERT_EQ(universal_compaction_picker.NeedsCompaction(vstorage_.get()),
+              false);

-  NewVersionStorage(3, kCompactionStyleUniversal);
+    NewVersionStorage(3, kCompactionStyleUniversal);

-  Add(0, 1U, "150", "200", kFileSize, 0, 500, 550);
-  Add(0, 2U, "201", "250", kFileSize, 0, 401, 450);
-  Add(0, 4U, "260", "300", kFileSize, 0, 260, 300);
-  Add(1, 5U, "100", "151", kFileSize, 0, 200, 251);
-  Add(1, 3U, "301", "350", kFileSize, 0, 101, 150);
-  Add(2, 6U, "120", "200", kFileSize, 0, 20, 100);
+    Add(0, 1U, "150", "200", kFileSize, 0, 500, 550);
+    Add(0, 2U, "201", "250", kFileSize, 0, 401, 450);
+    Add(0, 4U, "260", "300", kFileSize, 0, 260, 300);
+    Add(1, 5U, "100", "151", kFileSize, 0, 200, 251);
+    Add(1, 3U, "301", "350", kFileSize, 0, 101, 150);
+    Add(2, 6U, "120", "200", kFileSize, 0, 20, 100);

-  UpdateVersionStorageInfo();
+    UpdateVersionStorageInfo();

-  std::unique_ptr<Compaction> compaction(
-      universal_compaction_picker.PickCompaction(
-          cf_name_, mutable_cf_options_, mutable_db_options_,
-          /*existing_snapshots=*/{}, /* snapshot_checker */ nullptr,
-          vstorage_.get(), &log_buffer_));
+    std::unique_ptr<Compaction> compaction(
+        universal_compaction_picker.PickCompaction(
+            cf_name_, mutable_cf_options_, mutable_db_options_,
+            /*existing_snapshots=*/{}, /* snapshot_checker */ nullptr,
+            vstorage_.get(), &log_buffer_));

-  // output level should be the one above the bottom-most
-  ASSERT_EQ(1, compaction->output_level());
+    // output level should be the one above the bottom-most
+    ASSERT_EQ(1, compaction->output_level());

-  // input should not include the reserved level
-  const std::vector<CompactionInputFiles>* inputs = compaction->inputs();
-  for (const auto& compaction_input : *inputs) {
-    if (!compaction_input.empty()) {
-      ASSERT_LT(compaction_input.level, 2);
+    // input should not include the reserved level
+    const std::vector<CompactionInputFiles>* inputs = compaction->inputs();
+    for (const auto& compaction_input : *inputs) {
+      if (!compaction_input.empty()) {
+        ASSERT_LT(compaction_input.level, 2);
+      }
    }
  }
 }
@@ -1134,10 +1141,15 @@ TEST_F(CompactionPickerTest, FIFOToCold1) {
    fifo_options_.max_table_files_size = kMaxSize;
    fifo_options_.file_temperature_age_thresholds = {
        {Temperature::kCold, kColdThreshold}};
+    fifo_options_.allow_trivial_copy_when_change_temperature = true;
+    fifo_options_.trivial_copy_buffer_size = 16 * 1024 * 1024;
    mutable_cf_options_.compaction_options_fifo = fifo_options_;
    mutable_cf_options_.level0_file_num_compaction_trigger = 100;
    mutable_cf_options_.max_compaction_bytes = kFileSize * 100;
-    FIFOCompactionPicker fifo_compaction_picker(ioptions_, &icmp_);
+
+    auto copiedIOptions = ioptions_;
+    copiedIOptions.compaction_style = kCompactionStyleFIFO;
+    FIFOCompactionPicker fifo_compaction_picker(copiedIOptions, &icmp_);

    int64_t current_time = 0;
    ASSERT_OK(Env::Default()->GetCurrentTime(&current_time));
@@ -1186,7 +1198,10 @@ TEST_F(CompactionPickerTest, FIFOToColdMaxCompactionSize) {
    mutable_cf_options_.compaction_options_fifo = fifo_options_;
    mutable_cf_options_.level0_file_num_compaction_trigger = 100;
    mutable_cf_options_.max_compaction_bytes = kFileSize * 9;
-    FIFOCompactionPicker fifo_compaction_picker(ioptions_, &icmp_);
+
+    auto copiedIOptions = ioptions_;
+    copiedIOptions.compaction_style = kCompactionStyleFIFO;
+    FIFOCompactionPicker fifo_compaction_picker(copiedIOptions, &icmp_);

    int64_t current_time = 0;
    ASSERT_OK(Env::Default()->GetCurrentTime(&current_time));
@@ -1253,7 +1268,10 @@ TEST_F(CompactionPickerTest, FIFOToColdWithExistingCold) {
    mutable_cf_options_.compaction_options_fifo = fifo_options_;
    mutable_cf_options_.level0_file_num_compaction_trigger = 100;
    mutable_cf_options_.max_compaction_bytes = kFileSize * 100;
-    FIFOCompactionPicker fifo_compaction_picker(ioptions_, &icmp_);
+
+    auto copiedIOptions = ioptions_;
+    copiedIOptions.compaction_style = kCompactionStyleFIFO;
+    FIFOCompactionPicker fifo_compaction_picker(copiedIOptions, &icmp_);

    int64_t current_time = 0;
    ASSERT_OK(Env::Default()->GetCurrentTime(&current_time));
@@ -1318,7 +1336,10 @@ TEST_F(CompactionPickerTest, FIFOToColdWithHotBetweenCold) {
    mutable_cf_options_.compaction_options_fifo = fifo_options_;
    mutable_cf_options_.level0_file_num_compaction_trigger = 100;
    mutable_cf_options_.max_compaction_bytes = kFileSize * 100;
-    FIFOCompactionPicker fifo_compaction_picker(ioptions_, &icmp_);
+
+    auto copiedIOptions = ioptions_;
+    copiedIOptions.compaction_style = kCompactionStyleFIFO;
+    FIFOCompactionPicker fifo_compaction_picker(copiedIOptions, &icmp_);

    int64_t current_time = 0;
    ASSERT_OK(Env::Default()->GetCurrentTime(&current_time));
@@ -1385,7 +1406,10 @@ TEST_F(CompactionPickerTest, FIFOToHotAndWarm) {
    mutable_cf_options_.compaction_options_fifo = fifo_options_;
    mutable_cf_options_.level0_file_num_compaction_trigger = 100;
    mutable_cf_options_.max_compaction_bytes = kFileSize * 100;
-    FIFOCompactionPicker fifo_compaction_picker(ioptions_, &icmp_);
+
+    auto copiedIOptions = ioptions_;
+    copiedIOptions.compaction_style = kCompactionStyleFIFO;
+    FIFOCompactionPicker fifo_compaction_picker(copiedIOptions, &icmp_);

    int64_t current_time = 0;
    ASSERT_OK(Env::Default()->GetCurrentTime(&current_time));
@@ -2672,13 +2696,14 @@ TEST_F(CompactionPickerTest, CompactRangeMaxCompactionBytes) {
  bool manual_conflict = false;
  InternalKey manual_end;
  InternalKey* manual_end_ptr = &manual_end;
-  std::unique_ptr<Compaction> compaction(level_compaction_picker.CompactRange(
-      cf_name_, mutable_cf_options_, mutable_db_options_, vstorage_.get(),
-      /*input_level=*/1, /*output_level=*/2,
-      /*compact_range_options*/ {}, /*begin=*/nullptr, /*end=*/nullptr,
-      &manual_end_ptr, &manual_conflict,
-      /*max_file_num_to_ignore=*/std::numeric_limits<uint64_t>::max(),
-      /*trim_ts=*/""));
+  std::unique_ptr<Compaction> compaction(
+      level_compaction_picker.PickCompactionForCompactRange(
+          cf_name_, mutable_cf_options_, mutable_db_options_, vstorage_.get(),
+          /*input_level=*/1, /*output_level=*/2,
+          /*compact_range_options*/ {}, /*begin=*/nullptr, /*end=*/nullptr,
+          &manual_end_ptr, &manual_conflict,
+          /*max_file_num_to_ignore=*/std::numeric_limits<uint64_t>::max(),
+          /*trim_ts=*/""));
  ASSERT_TRUE(compaction.get() != nullptr);
  ASSERT_EQ(1U, compaction->num_input_levels());
  ASSERT_EQ(2, compaction->output_level());
@@ -3610,7 +3635,7 @@ TEST_F(CompactionPickerTest, UniversalMarkedManualCompaction) {
  bool manual_conflict = false;
  InternalKey* manual_end = nullptr;
  std::unique_ptr<Compaction> compaction(
-      universal_compaction_picker.CompactRange(
+      universal_compaction_picker.PickCompactionForCompactRange(
          cf_name_, mutable_cf_options_, mutable_db_options_, vstorage_.get(),
          ColumnFamilyData::kCompactAllLevels, 6, CompactRangeOptions(),
          nullptr, nullptr, &manual_end, &manual_conflict,
@@ -3814,9 +3839,10 @@ TEST_F(CompactionPickerU64TsTest, Overlap) {
  std::vector<CompactionInputFiles> input_files;
  ASSERT_OK(level_compaction_picker.GetCompactionInputsFromFileNumbers(
      &input_files, &input, vstorage_.get(), CompactionOptions()));
-  std::unique_ptr<Compaction> comp1(level_compaction_picker.CompactFiles(
-      CompactionOptions(), input_files, level, vstorage_.get(),
-      mutable_cf_options_, mutable_db_options_, /*output_path_id=*/0));
+  std::unique_ptr<Compaction> comp1(
+      level_compaction_picker.PickCompactionForCompactFiles(
+          CompactionOptions(), input_files, level, vstorage_.get(),
+          mutable_cf_options_, mutable_db_options_, /*output_path_id=*/0));

  {
    // [600, ts=50000] to [600, ts=50000] is the range to check.
@@ -3925,9 +3951,10 @@ TEST_P(PerKeyPlacementCompactionPickerTest, OverlapWithNormalCompaction) {
  ASSERT_OK(level_compaction_picker.GetCompactionInputsFromFileNumbers(
      &input_files, &input_set, vstorage_.get(), comp_options));

-  std::unique_ptr<Compaction> comp1(level_compaction_picker.CompactFiles(
-      comp_options, input_files, 5, vstorage_.get(), mutable_cf_options_,
-      mutable_db_options_, 0));
+  std::unique_ptr<Compaction> comp1(
+      level_compaction_picker.PickCompactionForCompactFiles(
+          comp_options, input_files, 5, vstorage_.get(), mutable_cf_options_,
+          mutable_db_options_, 0));

  input_set.clear();
  input_files.clear();
@@ -3971,9 +3998,10 @@ TEST_P(PerKeyPlacementCompactionPickerTest, NormalCompactionOverlap) {
  ASSERT_OK(level_compaction_picker.GetCompactionInputsFromFileNumbers(
      &input_files, &input_set, vstorage_.get(), comp_options));

-  std::unique_ptr<Compaction> comp1(level_compaction_picker.CompactFiles(
-      comp_options, input_files, 6, vstorage_.get(), mutable_cf_options_,
-      mutable_db_options_, 0));
+  std::unique_ptr<Compaction> comp1(
+      level_compaction_picker.PickCompactionForCompactFiles(
+          comp_options, input_files, 6, vstorage_.get(), mutable_cf_options_,
+          mutable_db_options_, 0));

  input_set.clear();
  input_files.clear();
@@ -4013,9 +4041,10 @@ TEST_P(PerKeyPlacementCompactionPickerTest,
  ASSERT_OK(universal_compaction_picker.GetCompactionInputsFromFileNumbers(
      &input_files, &input_set, vstorage_.get(), comp_options));

-  std::unique_ptr<Compaction> comp1(universal_compaction_picker.CompactFiles(
-      comp_options, input_files, 5, vstorage_.get(), mutable_cf_options_,
-      mutable_db_options_, 0));
+  std::unique_ptr<Compaction> comp1(
+      universal_compaction_picker.PickCompactionForCompactFiles(
+          comp_options, input_files, 5, vstorage_.get(), mutable_cf_options_,
+          mutable_db_options_, 0));

  input_set.clear();
  input_files.clear();
@@ -4060,9 +4089,10 @@ TEST_P(PerKeyPlacementCompactionPickerTest, NormalCompactionOverlapUniversal) {
  ASSERT_OK(universal_compaction_picker.GetCompactionInputsFromFileNumbers(
      &input_files, &input_set, vstorage_.get(), comp_options));

-  std::unique_ptr<Compaction> comp1(universal_compaction_picker.CompactFiles(
-      comp_options, input_files, 6, vstorage_.get(), mutable_cf_options_,
-      mutable_db_options_, 0));
+  std::unique_ptr<Compaction> comp1(
+      universal_compaction_picker.PickCompactionForCompactFiles(
+          comp_options, input_files, 6, vstorage_.get(), mutable_cf_options_,
+          mutable_db_options_, 0));

  input_set.clear();
  input_files.clear();
@@ -4108,9 +4138,10 @@ TEST_P(PerKeyPlacementCompactionPickerTest, ProximalOverlapUniversal) {
  ASSERT_OK(universal_compaction_picker.GetCompactionInputsFromFileNumbers(
      &input_files, &input_set, vstorage_.get(), comp_options));

-  std::unique_ptr<Compaction> comp1(universal_compaction_picker.CompactFiles(
-      comp_options, input_files, 6, vstorage_.get(), mutable_cf_options_,
-      mutable_db_options_, 0));
+  std::unique_ptr<Compaction> comp1(
+      universal_compaction_picker.PickCompactionForCompactFiles(
+          comp_options, input_files, 6, vstorage_.get(), mutable_cf_options_,
+          mutable_db_options_, 0));

  input_set.clear();
  input_files.clear();
@@ -4159,9 +4190,10 @@ TEST_P(PerKeyPlacementCompactionPickerTest, LastLevelOnlyOverlapUniversal) {
  ASSERT_OK(universal_compaction_picker.GetCompactionInputsFromFileNumbers(
      &input_files, &input_set, vstorage_.get(), comp_options));

-  std::unique_ptr<Compaction> comp1(universal_compaction_picker.CompactFiles(
-      comp_options, input_files, 6, vstorage_.get(), mutable_cf_options_,
-      mutable_db_options_, 0));
+  std::unique_ptr<Compaction> comp1(
+      universal_compaction_picker.PickCompactionForCompactFiles(
+          comp_options, input_files, 6, vstorage_.get(), mutable_cf_options_,
+          mutable_db_options_, 0));

  // cannot compact file 41 if the preclude_last_level feature is on, otherwise
  // compact file 41 is okay.
@@ -4217,9 +4249,10 @@ TEST_P(PerKeyPlacementCompactionPickerTest,
  ASSERT_OK(universal_compaction_picker.GetCompactionInputsFromFileNumbers(
      &input_files, &input_set, vstorage_.get(), comp_options));

-  std::unique_ptr<Compaction> comp1(universal_compaction_picker.CompactFiles(
-      comp_options, input_files, 6, vstorage_.get(), mutable_cf_options_,
-      mutable_db_options_, 0));
+  std::unique_ptr<Compaction> comp1(
+      universal_compaction_picker.PickCompactionForCompactFiles(
+          comp_options, input_files, 6, vstorage_.get(), mutable_cf_options_,
+          mutable_db_options_, 0));

  ASSERT_TRUE(comp1);
  ASSERT_EQ(comp1->GetProximalLevel(), Compaction::kInvalidLevel);
@@ -4235,9 +4268,10 @@ TEST_P(PerKeyPlacementCompactionPickerTest,
  ASSERT_FALSE(universal_compaction_picker.FilesRangeOverlapWithCompaction(
      input_files, 5, Compaction::kInvalidLevel));

-  std::unique_ptr<Compaction> comp2(universal_compaction_picker.CompactFiles(
-      comp_options, input_files, 5, vstorage_.get(), mutable_cf_options_,
-      mutable_db_options_, 0));
+  std::unique_ptr<Compaction> comp2(
+      universal_compaction_picker.PickCompactionForCompactFiles(
+          comp_options, input_files, 5, vstorage_.get(), mutable_cf_options_,
+          mutable_db_options_, 0));
  ASSERT_TRUE(comp2);
  ASSERT_EQ(Compaction::kInvalidLevel, comp2->GetProximalLevel());
 }
@@ -4273,9 +4307,10 @@ TEST_P(PerKeyPlacementCompactionPickerTest,
  ASSERT_OK(universal_compaction_picker.GetCompactionInputsFromFileNumbers(
      &input_files, &input_set, vstorage_.get(), comp_options));

-  std::unique_ptr<Compaction> comp1(universal_compaction_picker.CompactFiles(
-      comp_options, input_files, 5, vstorage_.get(), mutable_cf_options_,
-      mutable_db_options_, 0));
+  std::unique_ptr<Compaction> comp1(
+      universal_compaction_picker.PickCompactionForCompactFiles(
+          comp_options, input_files, 5, vstorage_.get(), mutable_cf_options_,
+          mutable_db_options_, 0));

  ASSERT_TRUE(comp1);
  ASSERT_EQ(comp1->GetProximalLevel(), Compaction::kInvalidLevel);
@@ -4293,9 +4328,10 @@ TEST_P(PerKeyPlacementCompactionPickerTest,
                    vstorage_.get(), mutable_cf_options_, ioptions_, 6, 6)));

  if (!enable_per_key_placement_) {
-    std::unique_ptr<Compaction> comp2(universal_compaction_picker.CompactFiles(
-        comp_options, input_files, 6, vstorage_.get(), mutable_cf_options_,
-        mutable_db_options_, 0));
+    std::unique_ptr<Compaction> comp2(
+        universal_compaction_picker.PickCompactionForCompactFiles(
+            comp_options, input_files, 6, vstorage_.get(), mutable_cf_options_,
+            mutable_db_options_, 0));
    ASSERT_TRUE(comp2);
    ASSERT_EQ(Compaction::kInvalidLevel, comp2->GetProximalLevel());
  }
@@ -4333,9 +4369,10 @@ TEST_P(PerKeyPlacementCompactionPickerTest,
  ASSERT_OK(universal_compaction_picker.GetCompactionInputsFromFileNumbers(
      &input_files, &input_set, vstorage_.get(), comp_options));

-  std::unique_ptr<Compaction> comp1(universal_compaction_picker.CompactFiles(
-      comp_options, input_files, 5, vstorage_.get(), mutable_cf_options_,
-      mutable_db_options_, 0));
+  std::unique_ptr<Compaction> comp1(
+      universal_compaction_picker.PickCompactionForCompactFiles(
+          comp_options, input_files, 5, vstorage_.get(), mutable_cf_options_,
+          mutable_db_options_, 0));

  ASSERT_TRUE(comp1);
  ASSERT_EQ(comp1->GetProximalLevel(), Compaction::kInvalidLevel);
@@ -4353,9 +4390,10 @@ TEST_P(PerKeyPlacementCompactionPickerTest,
                                        ioptions_, 6, 6)));

  // 2 compactions can be run in parallel
-  std::unique_ptr<Compaction> comp2(universal_compaction_picker.CompactFiles(
-      comp_options, input_files, 6, vstorage_.get(), mutable_cf_options_,
-      mutable_db_options_, 0));
+  std::unique_ptr<Compaction> comp2(
+      universal_compaction_picker.PickCompactionForCompactFiles(
+          comp_options, input_files, 6, vstorage_.get(), mutable_cf_options_,
+          mutable_db_options_, 0));
  ASSERT_TRUE(comp2);
  if (enable_per_key_placement_) {
    ASSERT_NE(Compaction::kInvalidLevel, comp2->GetProximalLevel());
@@ -38,7 +38,8 @@ class UniversalCompactionBuilder {
      const MutableDBOptions& mutable_db_options,
      const std::vector<SequenceNumber>& existing_snapshots,
      const SnapshotChecker* snapshot_checker, VersionStorageInfo* vstorage,
-      UniversalCompactionPicker* picker, LogBuffer* log_buffer)
+      UniversalCompactionPicker* picker, LogBuffer* log_buffer,
+      bool require_max_output_level)
      : ioptions_(ioptions),
        icmp_(icmp),
        cf_name_(cf_name),
@@ -46,7 +47,10 @@ class UniversalCompactionBuilder {
        mutable_db_options_(mutable_db_options),
        vstorage_(vstorage),
        picker_(picker),
-        log_buffer_(log_buffer) {
+        log_buffer_(log_buffer),
+        require_max_output_level_(require_max_output_level),
+        allow_ingest_behind_(ioptions.cf_allow_ingest_behind ||
+                             ioptions.allow_ingest_behind) {
    assert(icmp_);
    const auto* ucmp = icmp_->user_comparator();
    assert(ucmp);
@@ -102,6 +106,174 @@ class UniversalCompactionBuilder {
    bool level_has_marked_standalone_rangedel;
  };

+  unsigned int GetMaxNumFilesToCompactBasedOnMaxReadAmp(
+      const int file_num_compaction_trigger, const unsigned int ratio,
+      int* num_sr_not_compacted_output, int* max_num_runs_output) const {
+    assert(num_sr_not_compacted_output);
+    assert(max_num_runs_output);
+    int max_num_runs =
+        mutable_cf_options_.compaction_options_universal.max_read_amp;
+    if (max_num_runs < 0) {
+      // any value < -1 is not valid
+      assert(max_num_runs == -1);
+      // By default, fall back to `level0_file_num_compaction_trigger`
+      max_num_runs = file_num_compaction_trigger;
+    } else if (max_num_runs == 0) {
+      if (mutable_cf_options_.compaction_options_universal.stop_style ==
+          kCompactionStopStyleTotalSize) {
+        // 0 means auto-tuning by RocksDB. We estimate max num run based on
+        // max_run_size, size_ratio and write buffer size:
+        // Assume the size of the lowest level size is equal to
+        // write_buffer_size. Each subsequent level is the max size without
+        // triggering size_ratio compaction. `max_num_runs` is the minimum
+        // number of levels required such that the target size of the
+        // largest level is at least `max_run_size_`.
+        max_num_runs = 1;
+        double cur_level_max_size =
+            static_cast<double>(mutable_cf_options_.write_buffer_size);
+        double total_run_size = 0;
+        while (cur_level_max_size < static_cast<double>(max_run_size_)) {
+          // This loop should not take too many iterations since
+          // cur_level_max_size at least doubles each iteration.
+          total_run_size += cur_level_max_size;
+          cur_level_max_size = (100.0 + ratio) / 100.0 * total_run_size;
+          ++max_num_runs;
+        }
+      } else {
+        // TODO: implement the auto-tune logic for this stop style
+        max_num_runs = file_num_compaction_trigger;
+      }
+    } else {
+      // max_num_runs > 0, it's the limit on the number of sorted run
+    }
+
+    // Get the total number of sorted runs that are not being compacted
+    int num_sr_not_compacted = 0;
+    for (size_t i = 0; i < sorted_runs_.size(); i++) {
+      if (sorted_runs_[i].being_compacted == false &&
+          !sorted_runs_[i].level_has_marked_standalone_rangedel) {
+        num_sr_not_compacted++;
+      }
+    }
+
+    *num_sr_not_compacted_output = num_sr_not_compacted;
+    *max_num_runs_output = max_num_runs;
+
+    if (num_sr_not_compacted > max_num_runs) {
+      return num_sr_not_compacted - max_num_runs + 1;
+    } else {
+      return 0;
+    }
+  }
+
+  Compaction* MaybePickPeriodicCompaction(Compaction* const prev_picked_c) {
+    if (prev_picked_c != nullptr ||
+        vstorage_->FilesMarkedForPeriodicCompaction().empty()) {
+      return prev_picked_c;
+    }
+    // Always need to do a full compaction for periodic compaction.
+    Compaction* c = PickPeriodicCompaction();
+    TEST_SYNC_POINT_CALLBACK("PostPickPeriodicCompaction", c);
+    if (c != nullptr) {
+      ROCKS_LOG_BUFFER(log_buffer_,
+                       "[%s] Universal: picked for periodic compaction\n",
+                       cf_name_.c_str());
+    }
+    return c;
+  }
+
+  Compaction* MaybePickSizeAmpCompaction(Compaction* const prev_picked_c,
+                                         int file_num_compaction_trigger) {
+    if (prev_picked_c != nullptr ||
+        sorted_runs_.size() <
+            static_cast<size_t>(file_num_compaction_trigger)) {
+      return prev_picked_c;
+    }
+    Compaction* c = PickCompactionToReduceSizeAmp();
+    if (c != nullptr) {
+      TEST_SYNC_POINT("PickCompactionToReduceSizeAmpReturnNonnullptr");
+      ROCKS_LOG_BUFFER(log_buffer_,
+                       "[%s] Universal: picked for size amp compaction \n",
+                       cf_name_.c_str());
+    }
+    return c;
+  }
+
+  Compaction* MaybePickCompactionToReduceSortedRunsBasedFileRatio(
+      Compaction* const prev_picked_c, int file_num_compaction_trigger,
+      unsigned int ratio) {
+    if (prev_picked_c != nullptr ||
+        sorted_runs_.size() <
+            static_cast<size_t>(file_num_compaction_trigger)) {
+      return prev_picked_c;
+    }
+    Compaction* c = PickCompactionToReduceSortedRuns(ratio, UINT_MAX);
+    if (c != nullptr) {
+      TEST_SYNC_POINT("PickCompactionToReduceSortedRunsReturnNonnullptr");
+      ROCKS_LOG_BUFFER(log_buffer_,
+                       "[%s] Universal: picked for size ratio compaction to "
+                       "reduce sorted run\n",
+                       cf_name_.c_str());
+    }
+    return c;
+  }
+
+  Compaction* MaybePickCompactionToReduceSortedRuns(
+      Compaction* const prev_picked_c, int file_num_compaction_trigger,
+      unsigned int ratio) {
+    if (prev_picked_c != nullptr ||
+        sorted_runs_.size() <
+            static_cast<size_t>(file_num_compaction_trigger)) {
+      return prev_picked_c;
+    }
+
+    int num_sr_not_compacted = 0;
+    int max_num_runs = 0;
+    const unsigned int max_num_files_to_compact =
+        GetMaxNumFilesToCompactBasedOnMaxReadAmp(file_num_compaction_trigger,
+                                                 ratio, &num_sr_not_compacted,
+                                                 &max_num_runs);
+    if (max_num_files_to_compact == 0) {
+      ROCKS_LOG_BUFFER(
+          log_buffer_,
+          "[%s] Universal: skipping compaction to reduce sorted run, num "
+          "sorted runs not "
+          "being compacted -- %u, max num runs allowed -- %d, max_run_size "
+          "-- %" PRIu64 "\n",
+          cf_name_.c_str(), num_sr_not_compacted, max_num_runs, max_run_size_);
+      return nullptr;
+    }
+
+    Compaction* c =
+        PickCompactionToReduceSortedRuns(UINT_MAX, max_num_files_to_compact);
+    if (c != nullptr) {
+      ROCKS_LOG_BUFFER(log_buffer_,
+                       "[%s] Universal: picked for sorted run num compaction "
+                       "to reduce sorted run, to "
+                       "compact file num -- %u, max num runs allowed"
+                       "-- %d, max_run_size -- %" PRIu64 "\n",
+                       cf_name_.c_str(), max_num_files_to_compact, max_num_runs,
+                       max_run_size_);
+    }
+    return c;
+  }
+
+  Compaction* MaybePickDeleteTriggeredCompaction(
+      Compaction* const prev_picked_c) {
+    if (prev_picked_c != nullptr) {
+      return prev_picked_c;
+    }
+    Compaction* c = PickDeleteTriggeredCompaction();
+    if (c != nullptr) {
+      TEST_SYNC_POINT("PickDeleteTriggeredCompactionReturnNonnullptr");
+      ROCKS_LOG_BUFFER(
+          log_buffer_,
+          "[%s] Universal: picked for delete triggered compaction\n",
+          cf_name_.c_str());
+    }
+    return c;
+  }
+
  // Pick Universal compaction to limit read amplification
  Compaction* PickCompactionToReduceSortedRuns(
      unsigned int ratio, unsigned int max_number_of_files_to_compact);
@@ -249,6 +421,12 @@ class UniversalCompactionBuilder {
    return num_l0_to_exclude;
  }

+  bool MeetsOutputLevelRequirements(int output_level) const {
+    return !require_max_output_level_ ||
+           Compaction::OutputToNonZeroMaxOutputLevel(
+               output_level, vstorage_->MaxOutputLevel(allow_ingest_behind_));
+  }
+
  const ImmutableOptions& ioptions_;
  const InternalKeyComparator* icmp_;
  double score_;
@@ -270,6 +448,8 @@ class UniversalCompactionBuilder {
  // marked for compaction. This is only populated when snapshot info is
  // populated.
  std::map<uint64_t, size_t> file_marked_for_compaction_to_sorted_run_index_;
+  bool require_max_output_level_;
+  bool allow_ingest_behind_;

  std::vector<UniversalCompactionBuilder::SortedRun> CalculateSortedRuns(
      const VersionStorageInfo& vstorage, int last_level,
@@ -420,10 +600,11 @@ Compaction* UniversalCompactionPicker::PickCompaction(
    const MutableDBOptions& mutable_db_options,
    const std::vector<SequenceNumber>& existing_snapshots,
    const SnapshotChecker* snapshot_checker, VersionStorageInfo* vstorage,
-    LogBuffer* log_buffer) {
+    LogBuffer* log_buffer, bool require_max_output_level) {
  UniversalCompactionBuilder builder(
      ioptions_, icmp_, cf_name, mutable_cf_options, mutable_db_options,
-      existing_snapshots, snapshot_checker, vstorage, this, log_buffer);
+      existing_snapshots, snapshot_checker, vstorage, this, log_buffer,
+      require_max_output_level);
  return builder.PickCompaction();
 }

@@ -554,13 +735,20 @@ bool UniversalCompactionBuilder::ShouldSkipMarkedFile(
 Compaction* UniversalCompactionBuilder::PickCompaction() {
  const int kLevel0 = 0;
  score_ = vstorage_->CompactionScore(kLevel0);
-  int max_output_level =
-      vstorage_->MaxOutputLevel(ioptions_.allow_ingest_behind);
+  const int max_output_level = vstorage_->MaxOutputLevel(allow_ingest_behind_);
+  const int file_num_compaction_trigger =
+      mutable_cf_options_.level0_file_num_compaction_trigger;
+  const unsigned int ratio =
+      mutable_cf_options_.compaction_options_universal.size_ratio;
+
+  if (max_output_level == 0 &&
+      !MeetsOutputLevelRequirements(0 /* output_level */)) {
+    return nullptr;
+  }
+
  max_run_size_ = 0;
  sorted_runs_ =
      CalculateSortedRuns(*vstorage_, max_output_level, &max_run_size_);
-  int file_num_compaction_trigger =
-      mutable_cf_options_.level0_file_num_compaction_trigger;

  if (sorted_runs_.size() == 0 ||
      (vstorage_->FilesMarkedForPeriodicCompaction().empty() &&
@@ -572,6 +760,7 @@ Compaction* UniversalCompactionBuilder::PickCompaction() {
        "UniversalCompactionBuilder::PickCompaction:Return", nullptr);
    return nullptr;
  }
+
  VersionStorageInfo::LevelSummaryStorage tmp;
  ROCKS_LOG_BUFFER_MAX_SZ(
      log_buffer_, 3072,
@@ -579,127 +768,22 @@ Compaction* UniversalCompactionBuilder::PickCompaction() {
      cf_name_.c_str(), sorted_runs_.size(), vstorage_->LevelSummary(&tmp));

  Compaction* c = nullptr;
-  // Periodic compaction has higher priority than other type of compaction
-  // because it's a hard requirement.
-  if (!vstorage_->FilesMarkedForPeriodicCompaction().empty()) {
-    // Always need to do a full compaction for periodic compaction.
-    c = PickPeriodicCompaction();
-    TEST_SYNC_POINT_CALLBACK("PostPickPeriodicCompaction", c);
-  }

-  if (c == nullptr &&
-      sorted_runs_.size() >= static_cast<size_t>(file_num_compaction_trigger)) {
-    // Check for size amplification.
-    if ((c = PickCompactionToReduceSizeAmp()) != nullptr) {
-      TEST_SYNC_POINT("PickCompactionToReduceSizeAmpReturnNonnullptr");
-      ROCKS_LOG_BUFFER(log_buffer_, "[%s] Universal: compacting for size amp\n",
-                       cf_name_.c_str());
-    } else {
-      // Size amplification is within limits. Try reducing read
-      // amplification while maintaining file size ratios.
-      unsigned int ratio =
-          mutable_cf_options_.compaction_options_universal.size_ratio;
-
-      if ((c = PickCompactionToReduceSortedRuns(ratio, UINT_MAX)) != nullptr) {
-        TEST_SYNC_POINT("PickCompactionToReduceSortedRunsReturnNonnullptr");
-        ROCKS_LOG_BUFFER(log_buffer_,
-                         "[%s] Universal: compacting for size ratio\n",
-                         cf_name_.c_str());
-      } else {
-        // Size amplification and file size ratios are within configured limits.
-        // If max read amplification exceeds configured limits, then force
-        // compaction to reduce the number sorted runs without looking at file
-        // size ratios.
-
-        // This is guaranteed by NeedsCompaction()
-        assert(sorted_runs_.size() >=
-               static_cast<size_t>(file_num_compaction_trigger));
-        int max_num_runs =
-            mutable_cf_options_.compaction_options_universal.max_read_amp;
-        if (max_num_runs < 0) {
-          // any value < -1 is not valid
-          assert(max_num_runs == -1);
-          // By default, fall back to `level0_file_num_compaction_trigger`
-          max_num_runs = file_num_compaction_trigger;
-        } else if (max_num_runs == 0) {
-          if (mutable_cf_options_.compaction_options_universal.stop_style ==
-              kCompactionStopStyleTotalSize) {
-            // 0 means auto-tuning by RocksDB. We estimate max num run based on
-            // max_run_size, size_ratio and write buffer size:
-            // Assume the size of the lowest level size is equal to
-            // write_buffer_size. Each subsequent level is the max size without
-            // triggering size_ratio compaction. `max_num_runs` is the minimum
-            // number of levels required such that the target size of the
-            // largest level is at least `max_run_size_`.
-            max_num_runs = 1;
-            double cur_level_max_size =
-                static_cast<double>(mutable_cf_options_.write_buffer_size);
-            double total_run_size = 0;
-            while (cur_level_max_size < static_cast<double>(max_run_size_)) {
-              // This loop should not take too many iterations since
-              // cur_level_max_size at least doubles each iteration.
-              total_run_size += cur_level_max_size;
-              cur_level_max_size = (100.0 + ratio) / 100.0 * total_run_size;
-              ++max_num_runs;
-            }
-          } else {
-            // TODO: implement the auto-tune logic for this stop style
-            max_num_runs = file_num_compaction_trigger;
-          }
-        } else {
-          // max_num_runs > 0, it's the limit on the number of sorted run
-        }
-        // Get the total number of sorted runs that are not being compacted
-        int num_sr_not_compacted = 0;
-        for (size_t i = 0; i < sorted_runs_.size(); i++) {
-          if (sorted_runs_[i].being_compacted == false &&
-              !sorted_runs_[i].level_has_marked_standalone_rangedel) {
-            num_sr_not_compacted++;
-          }
-        }
-
-        // The number of sorted runs that are not being compacted is greater
-        // than the maximum allowed number of sorted runs
-        if (num_sr_not_compacted > max_num_runs) {
-          unsigned int num_files = num_sr_not_compacted - max_num_runs + 1;
-          if ((c = PickCompactionToReduceSortedRuns(UINT_MAX, num_files)) !=
-              nullptr) {
-            ROCKS_LOG_BUFFER(log_buffer_,
-                             "[%s] Universal: compacting for file num, to "
-                             "compact file num -- %u, max num runs allowed"
-                             "-- %d, max_run_size -- %" PRIu64 "\n",
-                             cf_name_.c_str(), num_files, max_num_runs,
-                             max_run_size_);
-          }
-        } else {
-          ROCKS_LOG_BUFFER(
-              log_buffer_,
-              "[%s] Universal: skipping compaction for file num, num runs not "
-              "being compacted -- %u, max num runs allowed -- %d, max_run_size "
-              "-- %" PRIu64 "\n",
-              cf_name_.c_str(), num_sr_not_compacted, max_num_runs,
-              max_run_size_);
-        }
-      }
-    }
-  }
-
-  if (c == nullptr) {
-    if ((c = PickDeleteTriggeredCompaction()) != nullptr) {
-      TEST_SYNC_POINT("PickDeleteTriggeredCompactionReturnNonnullptr");
-      ROCKS_LOG_BUFFER(log_buffer_,
-                       "[%s] Universal: delete triggered compaction\n",
-                       cf_name_.c_str());
-    }
-  }
+  c = MaybePickPeriodicCompaction(c);
+  c = MaybePickSizeAmpCompaction(c, file_num_compaction_trigger);
+  c = MaybePickCompactionToReduceSortedRunsBasedFileRatio(
+      c, file_num_compaction_trigger, ratio);
+  c = MaybePickCompactionToReduceSortedRuns(c, file_num_compaction_trigger,
+                                            ratio);
+  c = MaybePickDeleteTriggeredCompaction(c);

  if (c == nullptr) {
    TEST_SYNC_POINT_CALLBACK(
        "UniversalCompactionBuilder::PickCompaction:Return", nullptr);
    return nullptr;
  }
-  assert(c->output_level() <=
-         vstorage_->MaxOutputLevel(ioptions_.allow_ingest_behind));
+  assert(c->output_level() <= vstorage_->MaxOutputLevel(allow_ingest_behind_));
+  assert(MeetsOutputLevelRequirements(c->output_level()));

  if (mutable_cf_options_.compaction_options_universal.allow_trivial_move ==
          true &&
@@ -825,14 +909,16 @@ Compaction* UniversalCompactionBuilder::PickCompactionToReduceSortedRuns(
      if (sr->being_compacted) {
        ROCKS_LOG_BUFFER(log_buffer_,
                         "[%s] Universal: %s"
-                         "[%d] being compacted, skipping",
+                         "[%d] being compacted, skipping for compaction to "
+                         "reduce sorted runs",
                         cf_name_.c_str(), file_num_buf, loop);
      } else if (sr->level_has_marked_standalone_rangedel) {
-        ROCKS_LOG_BUFFER(log_buffer_,
-                         "[%s] Universal: %s"
-                         "[%d] has standalone range tombstone files marked for "
-                         "compaction, skipping",
-                         cf_name_.c_str(), file_num_buf, loop);
+        ROCKS_LOG_BUFFER(
+            log_buffer_,
+            "[%s] Universal: %s"
+            "[%d] has standalone range tombstone files marked for "
+            "compaction, skipping for compaction to reduce sorted runs",
+            cf_name_.c_str(), file_num_buf, loop);
      }

      sr = nullptr;
@@ -845,7 +931,8 @@ Compaction* UniversalCompactionBuilder::PickCompactionToReduceSortedRuns(
      char file_num_buf[kFormatFileNumberBufSize];
      sr->Dump(file_num_buf, sizeof(file_num_buf), true);
      ROCKS_LOG_BUFFER(log_buffer_,
-                       "[%s] Universal: Possible candidate %s[%d].",
+                       "[%s] Universal: Possible candidate for compaction to "
+                       "reduce sorted runs %s[%d].",
                       cf_name_.c_str(), file_num_buf, loop);
    }

@@ -937,8 +1024,7 @@ Compaction* UniversalCompactionBuilder::PickCompactionToReduceSortedRuns(
  int start_level = sorted_runs_[start_index].level;
  int output_level;
  // last level is reserved for the files ingested behind
-  int max_output_level =
-      vstorage_->MaxOutputLevel(ioptions_.allow_ingest_behind);
+  int max_output_level = vstorage_->MaxOutputLevel(allow_ingest_behind_);
  if (first_index_after == sorted_runs_.size()) {
    output_level = max_output_level;
  } else if (sorted_runs_[first_index_after].level == 0) {
@@ -947,6 +1033,10 @@ Compaction* UniversalCompactionBuilder::PickCompactionToReduceSortedRuns(
    output_level = sorted_runs_[first_index_after].level - 1;
  }

+  if (!MeetsOutputLevelRequirements(output_level)) {
+    return nullptr;
+  }
+
  std::vector<CompactionInputFiles> inputs(max_output_level + 1);
  for (size_t i = 0; i < inputs.size(); ++i) {
    inputs[i].level = start_level + static_cast<int>(i);
@@ -1006,10 +1096,9 @@ Compaction* UniversalCompactionBuilder::PickCompactionToReduceSortedRuns(
                        mutable_cf_options_.default_write_temperature,
                        /* max_subcompactions */ 0, grandparents,
                        /* earliest_snapshot */ std::nullopt,
-                        /* snapshot_checker */ nullptr,
-                        /* is manual */ false, /* trim_ts */ "", score_,
-                        false /* deletion_compaction */,
-                        /* l0_files_might_overlap */ true, compaction_reason);
+                        /* snapshot_checker */ nullptr, compaction_reason,
+                        /* trim_ts */ "", score_,
+                        /* l0_files_might_overlap */ true);
 }

 // Look at overall size amplification. If size amplification
@@ -1039,18 +1128,19 @@ Compaction* UniversalCompactionBuilder::PickCompactionToReduceSizeAmp() {
      char file_num_buf[kFormatFileNumberBufSize];
      sr->Dump(file_num_buf, sizeof(file_num_buf), true);
      if (sr->being_compacted) {
-        ROCKS_LOG_BUFFER(
-            log_buffer_,
-            "[%s] Universal: stopping at sorted run undergoing compaction: "
-            "%s[%" ROCKSDB_PRIszt "]",
-            cf_name_.c_str(), file_num_buf, start_index - 1);
+        ROCKS_LOG_BUFFER(log_buffer_,
+                         "[%s] Universal: stopping for size amp compaction at "
+                         "sorted run undergoing compaction: "
+                         "%s[%" ROCKSDB_PRIszt "]",
+                         cf_name_.c_str(), file_num_buf, start_index - 1);
      } else if (sr->level_has_marked_standalone_rangedel) {
-        ROCKS_LOG_BUFFER(
-            log_buffer_,
-            "[%s] Universal: stopping at sorted run that has standalone range "
-            "tombstone files marked for compaction: "
-            "%s[%" ROCKSDB_PRIszt "]",
-            cf_name_.c_str(), file_num_buf, start_index - 1);
+        ROCKS_LOG_BUFFER(log_buffer_,
+                         "[%s] Universal: stopping for size amp compaction at "
+                         "sorted run that has "
+                         "standalone range "
+                         "tombstone files marked for compaction: "
+                         "%s[%" ROCKSDB_PRIszt "]",
+                         cf_name_.c_str(), file_num_buf, start_index - 1);
      }
      break;
    }
@@ -1066,11 +1156,12 @@ Compaction* UniversalCompactionBuilder::PickCompactionToReduceSizeAmp() {
  {
    const size_t num_l0_to_exclude = MightExcludeNewL0sToReduceWriteStop(
        num_l0_files, end_index, start_index, candidate_size);
-    ROCKS_LOG_BUFFER(log_buffer_,
-                     "[%s] Universal: Excluding %" ROCKSDB_PRIszt
-                     " latest L0 files to reduce potential write stop "
-                     "triggered by `level0_stop_writes_trigger`",
-                     cf_name_.c_str(), num_l0_to_exclude);
+    ROCKS_LOG_BUFFER(
+        log_buffer_,
+        "[%s] Universal: Excluding for size amp compaction %" ROCKSDB_PRIszt
+        " latest L0 files to reduce potential write stop "
+        "triggered by `level0_stop_writes_trigger`",
+        cf_name_.c_str(), num_l0_to_exclude);
  }

  {
@@ -1088,18 +1179,18 @@ Compaction* UniversalCompactionBuilder::PickCompactionToReduceSizeAmp() {

  // size amplification = percentage of additional size
  if (candidate_size * 100 < ratio * base_sr_size) {
-    ROCKS_LOG_BUFFER(
-        log_buffer_,
-        "[%s] Universal: size amp not needed. newer-files-total-size %" PRIu64
-        " earliest-file-size %" PRIu64,
-        cf_name_.c_str(), candidate_size, base_sr_size);
+    ROCKS_LOG_BUFFER(log_buffer_,
+                     "[%s] Universal: size amp compction not needed. "
+                     "newer-files-total-size %" PRIu64
+                     " earliest-file-size %" PRIu64,
+                     cf_name_.c_str(), candidate_size, base_sr_size);
    return nullptr;
  } else {
-    ROCKS_LOG_BUFFER(
-        log_buffer_,
-        "[%s] Universal: size amp needed. newer-files-total-size %" PRIu64
-        " earliest-file-size %" PRIu64,
-        cf_name_.c_str(), candidate_size, base_sr_size);
+    ROCKS_LOG_BUFFER(log_buffer_,
+                     "[%s] Universal: size amp compaction needed. "
+                     "newer-files-total-size %" PRIu64
+                     " earliest-file-size %" PRIu64,
+                     cf_name_.c_str(), candidate_size, base_sr_size);
  }
  // Since incremental compaction can't include more than second last
  // level, it can introduce penalty, compared to full compaction. We
@@ -1354,10 +1445,9 @@ Compaction* UniversalCompactionBuilder::PickIncrementalForReduceSizeAmp(
      /* max_subcompactions */ 0, /* grandparents */ {},
      /* earliest_snapshot */ std::nullopt,
      /* snapshot_checker */ nullptr,
-      /* is manual */ false,
-      /* trim_ts */ "", score_, false /* deletion_compaction */,
-      /* l0_files_might_overlap */ true,
-      CompactionReason::kUniversalSizeAmplification);
+      CompactionReason::kUniversalSizeAmplification,
+      /* trim_ts */ "", score_,
+      /* l0_files_might_overlap */ true);
 }

 // Pick files marked for compaction. Typically, files are marked by
@@ -1426,8 +1516,7 @@ Compaction* UniversalCompactionBuilder::PickDeleteTriggeredCompaction() {
      return nullptr;
    }

-    int max_output_level =
-        vstorage_->MaxOutputLevel(ioptions_.allow_ingest_behind);
+    int max_output_level = vstorage_->MaxOutputLevel(allow_ingest_behind_);
    // Pick the first non-empty level after the start_level
    for (output_level = start_level + 1; output_level <= max_output_level;
         output_level++) {
@@ -1450,6 +1539,10 @@ Compaction* UniversalCompactionBuilder::PickDeleteTriggeredCompaction() {
    }
    assert(output_level <= max_output_level);

+    if (!MeetsOutputLevelRequirements(output_level)) {
+      return nullptr;
+    }
+
    if (output_level != 0) {
      if (start_level == 0) {
        if (!picker_->GetOverlappingL0Files(vstorage_, &start_level_inputs,
@@ -1503,11 +1596,9 @@ Compaction* UniversalCompactionBuilder::PickDeleteTriggeredCompaction() {
      GetCompressionOptions(mutable_cf_options_, vstorage_, output_level),
      mutable_cf_options_.default_write_temperature,
      /* max_subcompactions */ 0, grandparents, earliest_snapshot_,
-      snapshot_checker_,
-      /* is manual */ false,
-      /* trim_ts */ "", score_, false /* deletion_compaction */,
-      /* l0_files_might_overlap */ true,
-      CompactionReason::kFilesMarkedForCompaction);
+      snapshot_checker_, CompactionReason::kFilesMarkedForCompaction,
+      /* trim_ts */ "", score_,
+      /* l0_files_might_overlap */ true);
 }

 Compaction* UniversalCompactionBuilder::PickCompactionToOldest(
@@ -1528,8 +1619,7 @@ Compaction* UniversalCompactionBuilder::PickCompactionWithSortedRunRange(
  uint32_t path_id =
      GetPathId(ioptions_, mutable_cf_options_, estimated_total_size);
  int start_level = sorted_runs_[start_index].level;
-  int max_output_level =
-      vstorage_->MaxOutputLevel(ioptions_.allow_ingest_behind);
+  int max_output_level = vstorage_->MaxOutputLevel(allow_ingest_behind_);
  std::vector<CompactionInputFiles> inputs(max_output_level + 1);
  for (size_t i = 0; i < inputs.size(); ++i) {
    inputs[i].level = start_level + static_cast<int>(i);
@@ -1574,6 +1664,10 @@ Compaction* UniversalCompactionBuilder::PickCompactionWithSortedRunRange(
    output_level = sorted_runs_[end_index + 1].level - 1;
  }

+  if (!MeetsOutputLevelRequirements(output_level)) {
+    return nullptr;
+  }
+
  // intra L0 compactions outputs could have overlap
  if (output_level != 0 && picker_->FilesRangeOverlapWithCompaction(
                               inputs, output_level,
@@ -1599,10 +1693,9 @@ Compaction* UniversalCompactionBuilder::PickCompactionWithSortedRunRange(
      mutable_cf_options_.default_write_temperature,
      /* max_subcompactions */ 0, /* grandparents */ {},
      /* earliest_snapshot */ std::nullopt,
-      /* snapshot_checker */ nullptr,
-      /* is manual */ false,
-      /* trim_ts */ "", score_, false /* deletion_compaction */,
-      /* l0_files_might_overlap */ true, compaction_reason);
+      /* snapshot_checker */ nullptr, compaction_reason,
+      /* trim_ts */ "", score_,
+      /* l0_files_might_overlap */ true);
 }

 Compaction* UniversalCompactionBuilder::PickPeriodicCompaction() {
@@ -18,12 +18,15 @@ class UniversalCompactionPicker : public CompactionPicker {
  UniversalCompactionPicker(const ImmutableOptions& ioptions,
                            const InternalKeyComparator* icmp)
      : CompactionPicker(ioptions, icmp) {}
+
+  // If `require_max_output_level` is true, only pick compaction
+  // with max output level or return nullptr if no such compaction exists.
  Compaction* PickCompaction(
      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
      const MutableDBOptions& mutable_db_options,
      const std::vector<SequenceNumber>& existing_snapshots,
      const SnapshotChecker* snapshot_checker, VersionStorageInfo* vstorage,
-      LogBuffer* log_buffer) override;
+      LogBuffer* log_buffer, bool require_max_output_level = false) override;
  int MaxOutputLevel() const override { return NumberLevels() - 1; }

  bool NeedsCompaction(const VersionStorageInfo* vstorage) const override;
@@ -41,7 +41,7 @@ CompactionJob::ProcessKeyValueCompactionWithCompactionService(
  }

  compaction_input.cf_name = compaction->column_family_data()->GetName();
-  compaction_input.snapshots = existing_snapshots_;
+  compaction_input.snapshots = job_context_->snapshot_seqs;
  compaction_input.has_begin = sub_compact->start.has_value();
  compaction_input.begin =
      compaction_input.has_begin ? sub_compact->start->ToString() : "";
@@ -74,10 +74,14 @@ CompactionJob::ProcessKeyValueCompactionWithCompactionService(
      compaction->column_family_data()->GetName().c_str(), job_id_,
      compaction_input.output_level, input_files_oss.str().c_str());
  CompactionServiceJobInfo info(
-      dbname_, db_id_, db_session_id_, GetCompactionId(sub_compact),
+      dbname_, db_id_, db_session_id_,
+      compaction->column_family_data()->GetID(),
+      compaction->column_family_data()->GetName(), GetCompactionId(sub_compact),
      thread_pri_, compaction->compaction_reason(),
      compaction->is_full_compaction(), compaction->is_manual_compaction(),
-      compaction->bottommost_level());
+      compaction->bottommost_level(), compaction->start_level(),
+      compaction->output_level());
+
  CompactionServiceScheduleResponse response =
      db_options_.compaction_service->Schedule(info, compaction_input_binary);
  switch (response.status) {
@@ -111,7 +115,7 @@ CompactionJob::ProcessKeyValueCompactionWithCompactionService(
  }

  std::string debug_str_before_wait =
-      compaction->input_version()->DebugString();
+      compaction->input_version()->DebugString(/*hex=*/true);

  ROCKS_LOG_INFO(db_options_.info_log,
                 "[%s] [JOB %d] Waiting for remote compaction...",
@@ -122,13 +126,14 @@ CompactionJob::ProcessKeyValueCompactionWithCompactionService(
                                           &compaction_result_binary);

  if (compaction_status != CompactionServiceJobStatus::kSuccess) {
-    ROCKS_LOG_ERROR(db_options_.info_log,
-                    "[%s] [JOB %d] Wait() status is not kSuccess. "
-                    "\nDebugString Before Wait():\n%s"
-                    "\nDebugString After Wait():\n%s",
-                    compaction->column_family_data()->GetName().c_str(),
-                    job_id_, debug_str_before_wait.c_str(),
-                    compaction->input_version()->DebugString().c_str());
+    ROCKS_LOG_ERROR(
+        db_options_.info_log,
+        "[%s] [JOB %d] Wait() status is not kSuccess. "
+        "\nDebugString Before Wait():\n%s"
+        "\nDebugString After Wait():\n%s",
+        compaction->column_family_data()->GetName().c_str(), job_id_,
+        debug_str_before_wait.c_str(),
+        compaction->input_version()->DebugString(/*hex=*/true).c_str());
  }

  if (compaction_status == CompactionServiceJobStatus::kUseLocal) {
@@ -216,18 +221,24 @@ CompactionJob::ProcessKeyValueCompactionWithCompactionService(
    }

    FileMetaData meta;
-    uint64_t file_size;
-    // FIXME: file_size should be part of CompactionServiceOutputFile so that
-    // we don't get DB corruption if the full file size has not been propagated
-    // back to the caller through the file system (which could have metadata
-    // lag or caching bugs).
-    s = fs_->GetFileSize(tgt_file, IOOptions(), &file_size, nullptr);
+    uint64_t file_size = file.file_size;
+
+    // TODO - Clean this up in the next release.
+    // For backward compatibility - in case the remote worker does not populate
+    // the file_size yet. If missing, continue to populate this from the file
+    // system.
+    if (file_size == 0) {
+      s = fs_->GetFileSize(tgt_file, IOOptions(), &file_size, nullptr);
+    }
+
    if (!s.ok()) {
      sub_compact->status = s;
      db_options_.compaction_service->OnInstallation(
          response.scheduled_job_id, CompactionServiceJobStatus::kFailure);
      return CompactionServiceJobStatus::kFailure;
    }
+    assert(file_size > 0);
+
    meta.fd = FileDescriptor(file_num, compaction->output_path_id(), file_size,
                             file.smallest_seqno, file.largest_seqno);
    meta.smallest.DecodeFrom(file.smallest_internal_key);
@@ -240,7 +251,8 @@ CompactionJob::ProcessKeyValueCompactionWithCompactionService(
    meta.marked_for_compaction = file.marked_for_compaction;
    meta.unique_id = file.unique_id;
    meta.temperature = file.file_temperature;
-
+    meta.tail_size =
+        FileMetaData::CalculateTailSize(file_size, file.table_properties);
    auto cfd = compaction->column_family_data();
    CompactionOutputs* compaction_outputs =
        sub_compact->Outputs(file.is_proximal_level_output);
@@ -292,9 +304,9 @@ CompactionServiceCompactionJob::CompactionServiceCompactionJob(
    VersionSet* versions, const std::atomic<bool>* shutting_down,
    LogBuffer* log_buffer, FSDirectory* output_directory, Statistics* stats,
    InstrumentedMutex* db_mutex, ErrorHandler* db_error_handler,
-    std::vector<SequenceNumber> existing_snapshots,
-    std::shared_ptr<Cache> table_cache, EventLogger* event_logger,
-    const std::string& dbname, const std::shared_ptr<IOTracer>& io_tracer,
+    JobContext* job_context, std::shared_ptr<Cache> table_cache,
+    EventLogger* event_logger, const std::string& dbname,
+    const std::shared_ptr<IOTracer>& io_tracer,
    const std::atomic<bool>& manual_compaction_canceled,
    const std::string& db_id, const std::string& db_session_id,
    std::string output_path,
@@ -303,9 +315,8 @@ CompactionServiceCompactionJob::CompactionServiceCompactionJob(
    : CompactionJob(job_id, compaction, db_options, mutable_db_options,
                    file_options, versions, shutting_down, log_buffer, nullptr,
                    output_directory, nullptr, stats, db_mutex,
-                    db_error_handler, std::move(existing_snapshots),
-                    kMaxSequenceNumber, nullptr, nullptr,
-                    std::move(table_cache), event_logger,
+                    db_error_handler, job_context, std::move(table_cache),
+                    event_logger,
                    compaction->mutable_cf_options().paranoid_file_checks,
                    compaction->mutable_cf_options().report_bg_io_stats, dbname,
                    &(compaction_service_result->stats), Env::Priority::USER,
@@ -415,14 +426,14 @@ Status CompactionServiceCompactionJob::Run() {
    for (const auto& output_file : sub_compact->GetOutputs()) {
      auto& meta = output_file.meta;
      compaction_result_->output_files.emplace_back(
-          MakeTableFileName(meta.fd.GetNumber()), meta.fd.smallest_seqno,
-          meta.fd.largest_seqno, meta.smallest.Encode().ToString(),
-          meta.largest.Encode().ToString(), meta.oldest_ancester_time,
-          meta.file_creation_time, meta.epoch_number, meta.file_checksum,
-          meta.file_checksum_func_name, output_file.validator.GetHash(),
-          meta.marked_for_compaction, meta.unique_id,
-          *output_file.table_properties, output_file.is_proximal_level,
-          meta.temperature);
+          MakeTableFileName(meta.fd.GetNumber()), meta.fd.GetFileSize(),
+          meta.fd.smallest_seqno, meta.fd.largest_seqno,
+          meta.smallest.Encode().ToString(), meta.largest.Encode().ToString(),
+          meta.oldest_ancester_time, meta.file_creation_time, meta.epoch_number,
+          meta.file_checksum, meta.file_checksum_func_name,
+          output_file.validator.GetHash(), meta.marked_for_compaction,
+          meta.unique_id, *output_file.table_properties,
+          output_file.is_proximal_level, meta.temperature);
    }
  }

@@ -522,6 +533,10 @@ static std::unordered_map<std::string, OptionTypeInfo>
         {offsetof(struct CompactionServiceOutputFile, file_name),
          OptionType::kEncodedString, OptionVerificationType::kNormal,
          OptionTypeFlags::kNone}},
+        {"file_size",
+         {offsetof(struct CompactionServiceOutputFile, file_size),
+          OptionType::kUInt64T, OptionVerificationType::kNormal,
+          OptionTypeFlags::kNone}},
        {"smallest_seqno",
         {offsetof(struct CompactionServiceOutputFile, smallest_seqno),
          OptionType::kUInt64T, OptionVerificationType::kNormal,
@@ -7,6 +7,7 @@
 #include "port/stack_trace.h"
 #include "rocksdb/utilities/options_util.h"
 #include "table/unique_id_impl.h"
+#include "utilities/merge_operators/string_append/stringappend.h"

 namespace ROCKSDB_NAMESPACE {

@@ -21,10 +22,10 @@ class MyTestCompactionService : public CompactionService {
      : db_path_(std::move(db_path)),
        options_(options),
        statistics_(statistics),
-        start_info_("na", "na", "na", 0, Env::TOTAL, CompactionReason::kUnknown,
-                    false, false, false),
-        wait_info_("na", "na", "na", 0, Env::TOTAL, CompactionReason::kUnknown,
-                   false, false, false),
+        start_info_("na", "na", "na", 0, "na", 0, Env::TOTAL,
+                    CompactionReason::kUnknown, false, false, false, -1, -1),
+        wait_info_("na", "na", "na", 0, "na", 0, Env::TOTAL,
+                   CompactionReason::kUnknown, false, false, false, -1, -1),
        listeners_(listeners),
        table_properties_collector_factories_(
            std::move(table_properties_collector_factories)) {}
@@ -84,6 +85,7 @@ class MyTestCompactionService : public CompactionService {
    options_override.table_factory = options_.table_factory;
    options_override.sst_partitioner_factory = options_.sst_partitioner_factory;
    options_override.statistics = statistics_;
+    options_override.info_log = options_.info_log;
    if (!listeners_.empty()) {
      options_override.listeners = listeners_;
    }
@@ -277,8 +279,17 @@ TEST_F(CompactionServiceTest, BasicCompactions) {
  Statistics* primary_statistics = GetPrimaryStatistics();
  Statistics* compactor_statistics = GetCompactorStatistics();

+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "BlockBasedTable::PrefetchTail::TaiSizeNotRecorded",
+      [&](void* /* arg */) {
+        // Trigger assertion to verify precise tail prefetch size calculation
+        assert(false);
+      });
+
+  SyncPoint::GetInstance()->EnableProcessing();
  GenerateTestData();
  ASSERT_OK(dbfull()->TEST_WaitForCompact());
+  SyncPoint::GetInstance()->DisableProcessing();
  VerifyTestData();

  auto my_cs = GetCompactionService();
@@ -380,6 +391,7 @@ TEST_F(CompactionServiceTest, BasicCompactions) {
  ASSERT_FALSE(result.stats.is_full_compaction);

  Close();
+  SyncPoint::GetInstance()->DisableProcessing();
 }

 TEST_F(CompactionServiceTest, ManualCompaction) {
@@ -423,6 +435,121 @@ TEST_F(CompactionServiceTest, ManualCompaction) {
  ASSERT_OK(result.status);
  ASSERT_TRUE(result.stats.is_manual_compaction);
  ASSERT_TRUE(result.stats.is_remote_compaction);
+
+  auto info = my_cs->GetCompactionInfoForStart();
+  ASSERT_EQ(0, info.cf_id);
+  ASSERT_EQ(kDefaultColumnFamilyName, info.cf_name);
+
+  info = my_cs->GetCompactionInfoForWait();
+  ASSERT_EQ(0, info.cf_id);
+  ASSERT_EQ(kDefaultColumnFamilyName, info.cf_name);
+
+  // Test non-default CF
+  ASSERT_OK(
+      db_->CompactRange(CompactRangeOptions(), handles_[1], nullptr, nullptr));
+  my_cs->GetResult(&result);
+  ASSERT_OK(result.status);
+  ASSERT_TRUE(result.stats.is_manual_compaction);
+  ASSERT_TRUE(result.stats.is_remote_compaction);
+
+  info = my_cs->GetCompactionInfoForStart();
+  ASSERT_EQ(handles_[1]->GetID(), info.cf_id);
+  ASSERT_EQ(handles_[1]->GetName(), info.cf_name);
+
+  info = my_cs->GetCompactionInfoForWait();
+  ASSERT_EQ(handles_[1]->GetID(), info.cf_id);
+  ASSERT_EQ(handles_[1]->GetName(), info.cf_name);
+}
+
+TEST_F(CompactionServiceTest, StandaloneDeleteRangeTombstoneOptimization) {
+  Options options = CurrentOptions();
+  options.compaction_style = CompactionStyle::kCompactionStyleUniversal;
+  ReopenWithCompactionService(&options);
+
+  size_t num_files_after_filtered = 0;
+  SyncPoint::GetInstance()->SetCallBack(
+      "VersionSet::MakeInputIterator:NewCompactionMergingIterator",
+      [&](void* arg) {
+        num_files_after_filtered = *static_cast<size_t*>(arg);
+      });
+
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  std::vector<std::string> files;
+  {
+    // Writes first version of data in range partitioned files.
+    SstFileWriter sst_file_writer(EnvOptions(), options);
+    std::string file1 = dbname_ + "file1.sst";
+    ASSERT_OK(sst_file_writer.Open(file1));
+    ASSERT_OK(sst_file_writer.Put("a", "a1"));
+    ASSERT_OK(sst_file_writer.Put("b", "b1"));
+    ExternalSstFileInfo file1_info;
+    ASSERT_OK(sst_file_writer.Finish(&file1_info));
+    files.push_back(std::move(file1));
+
+    std::string file2 = dbname_ + "file2.sst";
+    ASSERT_OK(sst_file_writer.Open(file2));
+    ASSERT_OK(sst_file_writer.Put("x", "x1"));
+    ASSERT_OK(sst_file_writer.Put("y", "y1"));
+    ExternalSstFileInfo file2_info;
+    ASSERT_OK(sst_file_writer.Finish(&file2_info));
+    files.push_back(std::move(file2));
+  }
+
+  IngestExternalFileOptions ifo;
+  ASSERT_OK(db_->IngestExternalFile(files, ifo));
+  ASSERT_EQ(Get("a"), "a1");
+  ASSERT_EQ(Get("b"), "b1");
+  ASSERT_EQ(Get("x"), "x1");
+  ASSERT_EQ(Get("y"), "y1");
+  ASSERT_EQ(2, NumTableFilesAtLevel(6));
+
+  auto my_cs = GetCompactionService();
+  uint64_t comp_num = my_cs->GetCompactionNum();
+
+  {
+    // Atomically delete old version of data with one range delete file.
+    // And a new batch of range partitioned files with new version of data.
+    files.clear();
+    SstFileWriter sst_file_writer(EnvOptions(), options);
+    std::string file2 = dbname_ + "file2.sst";
+    ASSERT_OK(sst_file_writer.Open(file2));
+    ASSERT_OK(sst_file_writer.DeleteRange("a", "z"));
+    ExternalSstFileInfo file2_info;
+    ASSERT_OK(sst_file_writer.Finish(&file2_info));
+    files.push_back(std::move(file2));
+
+    std::string file3 = dbname_ + "file3.sst";
+    ASSERT_OK(sst_file_writer.Open(file3));
+    ASSERT_OK(sst_file_writer.Put("a", "a2"));
+    ASSERT_OK(sst_file_writer.Put("b", "b2"));
+    ExternalSstFileInfo file3_info;
+    ASSERT_OK(sst_file_writer.Finish(&file3_info));
+    files.push_back(std::move(file3));
+
+    std::string file4 = dbname_ + "file4.sst";
+    ASSERT_OK(sst_file_writer.Open(file4));
+    ASSERT_OK(sst_file_writer.Put("x", "x2"));
+    ASSERT_OK(sst_file_writer.Put("y", "y2"));
+    ExternalSstFileInfo file4_info;
+    ASSERT_OK(sst_file_writer.Finish(&file4_info));
+    files.push_back(std::move(file4));
+  }
+
+  ASSERT_OK(db_->IngestExternalFile(files, ifo));
+  ASSERT_OK(db_->WaitForCompact(WaitForCompactOptions()));
+  ASSERT_GE(my_cs->GetCompactionNum(), comp_num + 1);
+
+  CompactionServiceResult result;
+  my_cs->GetResult(&result);
+  ASSERT_OK(result.status);
+  ASSERT_TRUE(result.stats.is_manual_compaction);
+  ASSERT_TRUE(result.stats.is_remote_compaction);
+
+  ASSERT_EQ(num_files_after_filtered, 1);
+
+  Close();
+  SyncPoint::GetInstance()->DisableProcessing();
 }

 TEST_F(CompactionServiceTest, CompactionOutputFileIOError) {
@@ -757,6 +884,79 @@ TEST_F(CompactionServiceTest, VerifyInputRecordCount) {
  SyncPoint::GetInstance()->ClearAllCallBacks();
 }

+TEST_F(CompactionServiceTest, EmptyResult) {
+  Options options = CurrentOptions();
+  options.disable_auto_compactions = true;
+  ReopenWithCompactionService(&options);
+  GenerateTestData();
+
+  auto my_cs = GetCompactionService();
+
+  uint64_t comp_num = my_cs->GetCompactionNum();
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  ASSERT_GE(my_cs->GetCompactionNum(), comp_num + 1);
+
+  // Delete range to cover entire range
+  ASSERT_OK(db_->DeleteRange(WriteOptions(), "key", "keyz"));
+  ASSERT_OK(Flush());
+
+  // In this unit test, both remote compaction and primary db instance are
+  // running in the same process, so NewFileNumber will never have a collision.
+  // In the real-world remote compactions, when the compaction is indeed running
+  // in another process, this is not going to be the case.
+  // To simulate the SST file with the same name created in the tmp directory,
+  // override the file number in remote compaction to re-use old SST file
+  // number.
+  bool need_to_override_file_number = false;
+  SyncPoint::GetInstance()->SetCallBack(
+      "DBImplSecondary::OpenAndCompact::BeforeLoadingOptions:0",
+      [&](void*) { need_to_override_file_number = true; });
+
+  SyncPoint::GetInstance()->SetCallBack(
+      "CompactionJob::OpenCompactionOutputFile::NewFileNumber",
+      [&](void* file_number) {
+        if (need_to_override_file_number) {
+          auto n = static_cast<uint64_t*>(file_number);
+          ColumnFamilyMetaData cf_meta;
+          db_->GetColumnFamilyMetaData(&cf_meta);
+          for (const auto& level : cf_meta.levels) {
+            for (const auto& file : level.files) {
+              // Use one of the existing file name
+              *n = test::GetFileNumber(file.name);
+              need_to_override_file_number = false;
+              return;
+            }
+          }
+        }
+      });
+
+  // Inject failure, so that the remote compaction fails after
+  // ProcessKeyValueCompaction()
+  SyncPoint::GetInstance()->SetCallBack(
+      "DBImplSecondary::CompactWithoutInstallation::End", [&](void* status) {
+        // override job status
+        auto s = static_cast<Status*>(status);
+        *s = Status::Aborted("MyTestCompactionService failed to compact!");
+      });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  // Compaction should fail and SST files in the primary db should exist
+  {
+    ASSERT_NOK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+    ColumnFamilyMetaData meta;
+    db_->GetColumnFamilyMetaData(&meta);
+    for (const auto& level : meta.levels) {
+      for (const auto& file : level.files) {
+        std::string fname = file.db_path + "/" + file.name;
+        ASSERT_OK(db_->GetEnv()->FileExists(fname));
+      }
+    }
+  }
+  Close();
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+}
+
 TEST_F(CompactionServiceTest, CorruptedOutput) {
  Options options = CurrentOptions();
  options.disable_auto_compactions = true;
@@ -890,6 +1090,12 @@ TEST_F(CompactionServiceTest, TruncatedOutput) {
  Slice end(end_str);
  uint64_t comp_num = my_cs->GetCompactionNum();

+  // Skip calculating tail size to avoid crashing due to truncated file size
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "FileMetaData::CalculateTailSize", [&](void* arg) {
+        bool* skip = static_cast<bool*>(arg);
+        *skip = true;
+      });
  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
      "CompactionServiceCompactionJob::Run:0", [&](void* arg) {
        CompactionServiceResult* compaction_result =
@@ -906,7 +1112,7 @@ TEST_F(CompactionServiceTest, TruncatedOutput) {
          ASSERT_OK(s);
          ASSERT_GT(file_size, 0);

-          ASSERT_OK(test::TruncateFile(env_, file_name, file_size / 2));
+          ASSERT_OK(test::TruncateFile(env_, file_name, file_size / 4));
        }
      });
  SyncPoint::GetInstance()->EnableProcessing();
@@ -1181,6 +1387,32 @@ TEST_F(CompactionServiceTest, CompactionFilter) {
  ASSERT_GE(my_cs->GetCompactionNum(), 1);
 }

+TEST_F(CompactionServiceTest, MergeOperator) {
+  Options options = CurrentOptions();
+  options.merge_operator.reset(new StringAppendOperator(','));
+  ReopenWithCompactionService(&options);
+  GenerateTestData();
+  ASSERT_OK(dbfull()->TEST_WaitForCompact());
+  for (int i = 0; i < 200; i++) {
+    ASSERT_OK(db_->Merge(WriteOptions(), Key(i),
+                         "merge_op_append_" + std::to_string(i)));
+  }
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  // verify result
+  for (int i = 0; i < 200; i++) {
+    auto result = Get(Key(i));
+    if (i % 2) {
+      ASSERT_EQ(result, "value" + std::to_string(i) + ",merge_op_append_" +
+                            std::to_string(i));
+    } else {
+      ASSERT_EQ(result, "value_new" + std::to_string(i) + ",merge_op_append_" +
+                            std::to_string(i));
+    }
+  }
+  auto my_cs = GetCompactionService();
+  ASSERT_GE(my_cs->GetCompactionNum(), 1);
+}
+
 TEST_F(CompactionServiceTest, Snapshot) {
  Options options = CurrentOptions();
  ReopenWithCompactionService(&options);
@@ -1250,17 +1482,31 @@ TEST_F(CompactionServiceTest, PrecludeLastLevel) {

  // Verify Output Stats
  auto my_cs = GetCompactionService();
-  CompactionServiceResult result;
-  my_cs->GetResult(&result);
-  ASSERT_OK(result.status);
-  ASSERT_GT(result.internal_stats.output_level_stats.cpu_micros, 0);
-  ASSERT_GT(result.internal_stats.output_level_stats.micros, 0);
-  ASSERT_EQ(result.internal_stats.output_level_stats.num_output_records +
-                result.internal_stats.proximal_level_stats.num_output_records,
-            kNumTrigger * kNumKeys);
-  ASSERT_EQ(result.internal_stats.output_level_stats.num_output_files +
-                result.internal_stats.proximal_level_stats.num_output_files,
-            2);
+  {
+    CompactionServiceResult result;
+    my_cs->GetResult(&result);
+    ASSERT_OK(result.status);
+    ASSERT_GT(result.internal_stats.output_level_stats.cpu_micros, 0);
+    ASSERT_GT(result.internal_stats.output_level_stats.micros, 0);
+    ASSERT_EQ(result.internal_stats.output_level_stats.num_output_records +
+                  result.internal_stats.proximal_level_stats.num_output_records,
+              kNumTrigger * kNumKeys);
+    ASSERT_EQ(result.internal_stats.output_level_stats.num_output_files +
+                  result.internal_stats.proximal_level_stats.num_output_files,
+              2);
+
+    CompactionServiceJobInfo info = my_cs->GetCompactionInfoForStart();
+    ASSERT_EQ(0, info.base_input_level);
+    ASSERT_EQ(kNumLevels - 1, info.output_level);
+  }
+  SyncPoint::GetInstance()->DisableProcessing();
+  // Disable Preclude feature and run full compaction to the bottommost level
+  {
+    ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+    CompactionServiceJobInfo info = my_cs->GetCompactionInfoForStart();
+    ASSERT_EQ(kNumLevels - 2, info.base_input_level);
+    ASSERT_EQ(kNumLevels - 1, info.output_level);
+  }
 }

 TEST_F(CompactionServiceTest, ConcurrentCompaction) {
@@ -1330,12 +1576,17 @@ TEST_F(CompactionServiceTest, CompactionInfo) {
  ASSERT_EQ(true, info.is_manual_compaction);
  ASSERT_EQ(false, info.is_full_compaction);
  ASSERT_EQ(true, info.bottommost_level);
+  ASSERT_EQ(1, info.base_input_level);
+  ASSERT_EQ(2, info.output_level);
  info = my_cs->GetCompactionInfoForWait();
  ASSERT_EQ(Env::USER, info.priority);
  ASSERT_EQ(CompactionReason::kManualCompaction, info.compaction_reason);
  ASSERT_EQ(true, info.is_manual_compaction);
  ASSERT_EQ(false, info.is_full_compaction);
  ASSERT_EQ(true, info.bottommost_level);
+  ASSERT_EQ(1, info.base_input_level);
+  ASSERT_EQ(2, info.output_level);
+  ASSERT_EQ(kDefaultColumnFamilyName, info.cf_name);

  // Test priority BOTTOM
  env_->SetBackgroundThreads(1, Env::BOTTOM);
@@ -1367,18 +1618,24 @@ TEST_F(CompactionServiceTest, CompactionInfo) {
  ASSERT_EQ(false, info.is_full_compaction);
  ASSERT_EQ(true, info.bottommost_level);
  ASSERT_EQ(Env::BOTTOM, info.priority);
+  ASSERT_EQ(0, info.base_input_level);
+  ASSERT_EQ(db_->NumberLevels() - 1, info.output_level);
  info = my_cs->GetCompactionInfoForWait();
  ASSERT_EQ(Env::BOTTOM, info.priority);
  ASSERT_EQ(CompactionReason::kLevelL0FilesNum, info.compaction_reason);
  ASSERT_EQ(false, info.is_manual_compaction);
  ASSERT_EQ(false, info.is_full_compaction);
  ASSERT_EQ(true, info.bottommost_level);
+  ASSERT_EQ(0, info.base_input_level);
+  ASSERT_EQ(db_->NumberLevels() - 1, info.output_level);

  // Test Non-Bottommost Level
  options.num_levels = 4;
  ReopenWithCompactionService(&options);
  my_cs =
      static_cast_with_check<MyTestCompactionService>(GetCompactionService());
+  int compaction_num = my_cs->GetCompactionNum();
+  ASSERT_EQ(0, compaction_num);

  for (int i = 0; i < options.level0_file_num_compaction_trigger; i++) {
    for (int j = 0; j < 10; j++) {
@@ -1387,16 +1644,22 @@ TEST_F(CompactionServiceTest, CompactionInfo) {
    }
    ASSERT_OK(Flush());
  }
-
  ASSERT_OK(dbfull()->TEST_WaitForCompact());
+
+  // This is trivial move. Done locally.
+  ASSERT_EQ(0, my_cs->GetCompactionNum());
  info = my_cs->GetCompactionInfoForStart();
  ASSERT_EQ(false, info.is_manual_compaction);
  ASSERT_EQ(false, info.is_full_compaction);
  ASSERT_EQ(false, info.bottommost_level);
+  ASSERT_EQ(-1, info.base_input_level);
+  ASSERT_EQ(-1, info.output_level);
  info = my_cs->GetCompactionInfoForWait();
  ASSERT_EQ(false, info.is_manual_compaction);
  ASSERT_EQ(false, info.is_full_compaction);
  ASSERT_EQ(false, info.bottommost_level);
+  ASSERT_EQ(-1, info.base_input_level);
+  ASSERT_EQ(-1, info.output_level);

  // Test Full Compaction + Bottommost Level
  options.num_levels = 6;
@@ -1411,7 +1674,10 @@ TEST_F(CompactionServiceTest, CompactionInfo) {
    }
    ASSERT_OK(Flush());
  }
+  MoveFilesToLevel(options.num_levels - 1);

+  // Force final level compaction
+  // base_input_level == output_level == last_level
  CompactRangeOptions cro;
  cro.bottommost_level_compaction = BottommostLevelCompaction::kForce;
  ASSERT_OK(db_->CompactRange(cro, nullptr, nullptr));
@@ -1423,10 +1689,15 @@ TEST_F(CompactionServiceTest, CompactionInfo) {
  ASSERT_EQ(true, info.bottommost_level);
  ASSERT_EQ(CompactionReason::kManualCompaction, info.compaction_reason);
  info = my_cs->GetCompactionInfoForWait();
+  ASSERT_EQ(options.num_levels - 1, info.base_input_level);
+  ASSERT_EQ(options.num_levels - 1, info.output_level);
  ASSERT_EQ(true, info.is_manual_compaction);
  ASSERT_EQ(true, info.is_full_compaction);
  ASSERT_EQ(true, info.bottommost_level);
  ASSERT_EQ(CompactionReason::kManualCompaction, info.compaction_reason);
+  ASSERT_EQ(options.num_levels - 1, info.base_input_level);
+  ASSERT_EQ(options.num_levels - 1, info.output_level);
+  ASSERT_EQ("0,0,0,0,0,1", FilesPerLevel());
 }

 TEST_F(CompactionServiceTest, FallbackLocalAuto) {
@@ -106,7 +106,11 @@ class SubcompactionState {
    subcompaction_job_info.subcompaction_job_id = static_cast<int>(sub_job_id);
    subcompaction_job_info.base_input_level = c->start_level();
    subcompaction_job_info.output_level = c->output_level();
+    subcompaction_job_info.compaction_reason = c->compaction_reason();
+    subcompaction_job_info.compression = c->output_compression();
    subcompaction_job_info.stats = compaction_job_stats;
+    subcompaction_job_info.blob_compression_type =
+        c->mutable_cf_options().blob_compression_type;
  }

  SubcompactionState() = delete;
@@ -225,6 +225,8 @@ TEST_F(TieredCompactionTest, SequenceBasedTieredStorageUniversal) {
    flush_stats.micros = 1;
    flush_stats.bytes_written = bytes_per_file;
    flush_stats.num_output_files = 1;
+    flush_stats.num_input_records = kNumKeys;
+    flush_stats.num_output_records = kNumKeys;
    expect_stats[0].Add(flush_stats);
  }
  ASSERT_OK(dbfull()->TEST_WaitForCompact());
@@ -1080,6 +1082,8 @@ TEST_F(TieredCompactionTest, SequenceBasedTieredStorageLevel) {
    flush_stats.micros = 1;
    flush_stats.bytes_written = bytes_per_file;
    flush_stats.num_output_files = 1;
+    flush_stats.num_input_records = kNumKeys;
+    flush_stats.num_output_records = kNumKeys;
    expect_stats[0].Add(flush_stats);
  }
  ASSERT_OK(dbfull()->TEST_WaitForCompact());
@@ -93,9 +93,10 @@ Status VerifySstFileChecksumInternal(const Options& options,
          nullptr /* file_read_hist */, ioptions.rate_limiter.get()));
  const bool kImmortal = true;
  auto reader_options = TableReaderOptions(
-      ioptions, options.prefix_extractor, env_options, internal_comparator,
-      options.block_protection_bytes_per_key, false /* skip_filters */,
-      !kImmortal, false /* force_direct_prefetch */, -1 /* level */);
+      ioptions, options.prefix_extractor, options.compression_manager.get(),
+      env_options, internal_comparator, options.block_protection_bytes_per_key,
+      false /* skip_filters */, !kImmortal, false /* force_direct_prefetch */,
+      -1 /* level */);
  reader_options.largest_seqno = largest_seqno;
  s = options.table_factory->NewTableReader(
      read_options, reader_options, std::move(file_reader), file_size,
@@ -556,6 +556,74 @@ TEST_F(CorruptionTest, TableFileFooterNotMagic) {
  ASSERT_TRUE(s.ToString().find(".sst") != std::string::npos);
 }

+TEST_F(CorruptionTest, DBOpenWithWrongFileSize) {
+  // Validate that when paranoid flag is true, DB::Open() fails if one of the
+  // file corrupted. Validate that when paranoid flag is false, DB::Open()
+  // succeed if one of the file corrupted, and the healthy file is readable.
+  CloseDb();
+
+  const std::string test_cf_name = "test_cf";
+  std::vector<ColumnFamilyDescriptor> cf_descs;
+  cf_descs.emplace_back(kDefaultColumnFamilyName, ColumnFamilyOptions());
+  cf_descs.emplace_back(test_cf_name, ColumnFamilyOptions());
+
+  {
+    options_.create_missing_column_families = true;
+    std::vector<ColumnFamilyHandle*> cfhs;
+    ASSERT_OK(DB::Open(options_, dbname_, cf_descs, &cfhs, &db_));
+    assert(db_ != nullptr);  // suppress false clang-analyze report
+
+    ASSERT_OK(db_->Put(WriteOptions(), cfhs[0], "k", "v"));
+    ASSERT_OK(db_->Put(WriteOptions(), cfhs[1], "k1", "v1"));
+    ASSERT_OK(db_->Put(WriteOptions(), cfhs[0], "k2", "v2"));
+    for (auto* cfh : cfhs) {
+      delete cfh;
+    }
+    DBImpl* dbi = static_cast_with_check<DBImpl>(db_);
+    ASSERT_OK(dbi->TEST_FlushMemTable());
+
+    // ********************************************
+    // Corrupt the file by making the file bigger
+    std::vector<LiveFileMetaData> metadata;
+    db_->GetLiveFilesMetaData(&metadata);
+    std::string filename = dbname_ + metadata[0].name;
+    const auto& fs = options_.env->GetFileSystem();
+    {
+      std::unique_ptr<FSWritableFile> f;
+      ASSERT_OK(fs->ReopenWritableFile(filename, FileOptions(), &f, nullptr));
+      ASSERT_OK(f->Append("blahblah", IOOptions(), nullptr));
+      ASSERT_OK(f->Close(IOOptions(), nullptr));
+    }
+    CloseDb();
+  }
+
+  // DB failed to open due to one of the file is corrupted, as paranoid flag is
+  // true
+  options_.paranoid_checks = true;
+  std::vector<ColumnFamilyHandle*> cfhs;
+  auto s = DB::Open(options_, dbname_, cf_descs, &cfhs, &db_);
+  ASSERT_TRUE(s.IsCorruption());
+  ASSERT_TRUE(s.ToString().find("file size mismatch") != std::string::npos);
+
+  // DB opened successfully, as paranoid flag is false, validate the one that is
+  // healthy is still accessible
+  options_.paranoid_checks = false;
+  ASSERT_OK(DB::Open(options_, dbname_, cf_descs, &cfhs, &db_));
+  assert(db_ != nullptr);  // suppress false clang-analyze report
+
+  std::string v;
+  ASSERT_OK(db_->Get(ReadOptions(), cfhs[1], "k1", &v));
+  ASSERT_EQ(v, "v1");
+
+  // Validate the default column family is corrupted
+  Check(0, 0);
+  s = db_->Get(ReadOptions(), cfhs[0], "k1", &v);
+  ASSERT_TRUE(s.IsCorruption());
+
+  delete cfhs[1];
+  delete cfhs[0];
+}
+
 TEST_F(CorruptionTest, TableFileWrongSize) {
  Build(100);
  DBImpl* dbi = static_cast_with_check<DBImpl>(db_);
@@ -579,13 +647,16 @@ TEST_F(CorruptionTest, TableFileWrongSize) {
  // DB actually accepts this without paranoid checks, relying on size
  // recorded in manifest to locate the SST footer.
  options_.paranoid_checks = false;
-  options_.skip_checking_sst_file_sizes_on_db_open = false;
  Reopen();
-  Check(100, 100);
+  // As footer could not be extraced, file is completely unreadable
+  Check(0, 0);
+  std::string v;
+  auto s = db_->Get(ReadOptions(), "k1", &v);
+  ASSERT_TRUE(s.IsCorruption());

  // But reports the issue with paranoid checks
  options_.paranoid_checks = true;
-  Status s = TryReopen();
+  s = TryReopen();
  ASSERT_TRUE(s.IsCorruption());
  ASSERT_TRUE(s.ToString().find("file size mismatch") != std::string::npos);

@@ -851,9 +922,6 @@ TEST_F(CorruptionTest, ParanoidFileChecksOnCompact) {
  options.env = env_.get();
  options.paranoid_file_checks = true;
  options.create_if_missing = true;
-  // Skip verifying record count against TableProperties for
-  // MockTables
-  options.compaction_verify_record_count = false;
  Status s;
  for (const auto& mode : corruption_modes) {
    delete db_;
@@ -5086,6 +5086,7 @@ TEST_F(DBBasicTest, DisallowMemtableWrite) {
  options_allow.create_if_missing = true;
  Options options_disallow = options_allow;
  options_disallow.disallow_memtable_writes = true;
+  options_disallow.paranoid_memory_checks = true;

  DestroyAndReopen(options_allow);
  // CFs allowing and disallowing memtable write
@@ -5125,6 +5126,11 @@ TEST_F(DBBasicTest, DisallowMemtableWrite) {
  EXPECT_EQ(Get(2, "b2"), "2");
  EXPECT_EQ(Get(3, "b3"), "NOT_FOUND");

+  std::unique_ptr<Iterator> iter(
+      dbfull()->NewIterator(ReadOptions(), handles_[3]));
+  iter->Seek("a3");
+  ASSERT_OK(iter->status());
+  iter.reset();
  // When the DB is re-opened with WAL entries for a CF that is newly setting
  // disallow_memtable_writes, we detect that and fail the open gracefully.
  ASSERT_EQ(TryReopenWithColumnFamilies(
@@ -506,6 +506,8 @@ TEST_P(DBBlockCacheTest1, WarmCacheWithBlocksDuringFlush) {
  table_options.prepopulate_block_cache =
      BlockBasedTableOptions::PrepopulateBlockCache::kFlushOnly;
  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  // Include a compression dictionary block
+  options.compression_opts.max_dict_bytes = 123;
  DestroyAndReopen(options);

  std::string value(kValueSize, 'a');
@@ -537,6 +539,9 @@ TEST_P(DBBlockCacheTest1, WarmCacheWithBlocksDuringFlush) {
                options.statistics->getTickerCount(BLOCK_CACHE_FILTER_HIT));
    }
    ASSERT_EQ(0, options.statistics->getTickerCount(BLOCK_CACHE_FILTER_MISS));
+
+    // Including compression dict
+    ASSERT_EQ(0, options.statistics->getTickerCount(BLOCK_CACHE_MISS));
  }

  // Verify compaction not counted
@@ -824,68 +829,78 @@ TEST_F(DBBlockCacheTest, CacheCompressionDict) {
  const int kNumEntriesPerFile = 128;
  const int kNumBytesPerEntry = 1024;

-  // Try all the available libraries that support dictionary compression
-  std::vector<CompressionType> compression_types;
-  if (Zlib_Supported()) {
-    compression_types.push_back(kZlibCompression);
-  }
-  if (LZ4_Supported()) {
-    compression_types.push_back(kLZ4Compression);
-    compression_types.push_back(kLZ4HCCompression);
-  }
-  if (ZSTD_Supported()) {
-    compression_types.push_back(kZSTD);
-  }
+  std::vector<CompressionType> dict_compressions =
+      GetSupportedDictCompressions();
  Random rnd(301);
-  for (auto compression_type : compression_types) {
-    Options options = CurrentOptions();
-    options.bottommost_compression = compression_type;
-    options.bottommost_compression_opts.max_dict_bytes = 4096;
-    options.bottommost_compression_opts.enabled = true;
-    options.create_if_missing = true;
-    options.num_levels = 2;
-    options.statistics = ROCKSDB_NAMESPACE::CreateDBStatistics();
-    options.target_file_size_base = kNumEntriesPerFile * kNumBytesPerEntry;
-    BlockBasedTableOptions table_options;
-    table_options.cache_index_and_filter_blocks = true;
-    table_options.block_cache.reset(new MockCache());
-    options.table_factory.reset(NewBlockBasedTableFactory(table_options));
-    DestroyAndReopen(options);
+  // Format version before and after compression handling changes
+  for (int format_version : {6, 7}) {
+    // Test all supported compression types because (at least historically)
+    // dictionary compression could be enabled and a dictionary block saved
+    // but ignored by some compression types. Ensure we at least don't crash
+    // or return corruption for those.
+    for (auto compression_type : GetSupportedCompressions()) {
+      // Extra handling checks only for types actually supporting dictionary
+      // compression.
+      bool dict_supported =
+          std::count(dict_compressions.begin(), dict_compressions.end(),
+                     compression_type) > 0;

-    RecordCacheCountersForCompressionDict(options);
+      Options options = CurrentOptions();
+      options.bottommost_compression = compression_type;
+      options.bottommost_compression_opts.max_dict_bytes = 4096;
+      options.bottommost_compression_opts.enabled = true;
+      options.create_if_missing = true;
+      options.num_levels = 2;
+      options.statistics = ROCKSDB_NAMESPACE::CreateDBStatistics();
+      options.target_file_size_base = kNumEntriesPerFile * kNumBytesPerEntry;
+      BlockBasedTableOptions table_options;
+      table_options.cache_index_and_filter_blocks = true;
+      table_options.block_cache.reset(new MockCache());
+      table_options.format_version = format_version;
+      options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+      DestroyAndReopen(options);

-    for (int i = 0; i < kNumFiles; ++i) {
-      ASSERT_EQ(i, NumTableFilesAtLevel(0, 0));
-      for (int j = 0; j < kNumEntriesPerFile; ++j) {
-        std::string value = rnd.RandomString(kNumBytesPerEntry);
-        ASSERT_OK(Put(Key(j * kNumFiles + i), value.c_str()));
+      RecordCacheCountersForCompressionDict(options);
+
+      for (int i = 0; i < kNumFiles; ++i) {
+        ASSERT_EQ(i, NumTableFilesAtLevel(0, 0));
+        for (int j = 0; j < kNumEntriesPerFile; ++j) {
+          std::string value = rnd.RandomString(kNumBytesPerEntry);
+          ASSERT_OK(Put(Key(j * kNumFiles + i), value.c_str()));
+        }
+        ASSERT_OK(Flush());
+      }
+      ASSERT_OK(dbfull()->TEST_WaitForCompact());
+      ASSERT_EQ(0, NumTableFilesAtLevel(0));
+      ASSERT_EQ(kNumFiles, NumTableFilesAtLevel(1));
+
+      if (dict_supported) {
+        // Compression dictionary blocks are preloaded.
+        CheckCacheCountersForCompressionDict(
+            options, kNumFiles /* expected_compression_dict_misses */,
+            0 /* expected_compression_dict_hits */,
+            kNumFiles /* expected_compression_dict_inserts */);
+      }
+
+      // Seek to a key in a file. It should cause the SST's dictionary
+      // meta-block to be read.
+      RecordCacheCounters(options);
+      RecordCacheCountersForCompressionDict(options);
+      ReadOptions read_options;
+      ASSERT_NE("NOT_FOUND", Get(Key(kNumFiles * kNumEntriesPerFile - 1)));
+
+      if (dict_supported) {
+        // Two block hits: index and dictionary since they are prefetched
+        // One block missed/added: data block
+        CheckCacheCounters(options, 1 /* expected_misses */,
+                           2 /* expected_hits */, 1 /* expected_inserts */,
+                           0 /* expected_failures */);
+        CheckCacheCountersForCompressionDict(
+            options, 0 /* expected_compression_dict_misses */,
+            1 /* expected_compression_dict_hits */,
+            0 /* expected_compression_dict_inserts */);
      }
-      ASSERT_OK(Flush());
    }
-    ASSERT_OK(dbfull()->TEST_WaitForCompact());
-    ASSERT_EQ(0, NumTableFilesAtLevel(0));
-    ASSERT_EQ(kNumFiles, NumTableFilesAtLevel(1));
-
-    // Compression dictionary blocks are preloaded.
-    CheckCacheCountersForCompressionDict(
-        options, kNumFiles /* expected_compression_dict_misses */,
-        0 /* expected_compression_dict_hits */,
-        kNumFiles /* expected_compression_dict_inserts */);
-
-    // Seek to a key in a file. It should cause the SST's dictionary meta-block
-    // to be read.
-    RecordCacheCounters(options);
-    RecordCacheCountersForCompressionDict(options);
-    ReadOptions read_options;
-    ASSERT_NE("NOT_FOUND", Get(Key(kNumFiles * kNumEntriesPerFile - 1)));
-    // Two block hits: index and dictionary since they are prefetched
-    // One block missed/added: data block
-    CheckCacheCounters(options, 1 /* expected_misses */, 2 /* expected_hits */,
-                       1 /* expected_inserts */, 0 /* expected_failures */);
-    CheckCacheCountersForCompressionDict(
-        options, 0 /* expected_compression_dict_misses */,
-        1 /* expected_compression_dict_hits */,
-        0 /* expected_compression_dict_inserts */);
  }
 }

@@ -74,6 +74,43 @@ class CompactionStatsCollector : public EventListener {
  std::vector<std::atomic<int>> compaction_completed_;
 };

+class DeletionTriggeredCompactionWithMinFileSizeTestListener
+    : public EventListener {
+ public:
+  explicit DeletionTriggeredCompactionWithMinFileSizeTestListener(
+      uint64_t min_file_size)
+      : min_file_size_(min_file_size) {}
+
+  void OnCompactionBegin(DB* db, const CompactionJobInfo& ci) override {
+    ASSERT_EQ(ci.compaction_reason,
+              CompactionReason::kFilesMarkedForCompaction);
+
+    auto env = db->GetEnv();
+    const std::vector<DbPath>& db_paths = db->GetOptions().db_paths;
+    for (const auto& file : ci.input_file_infos) {
+      uint64_t file_size = GetSstFileSize(env, db_paths, file.file_number);
+
+      // Assert that the file size respects the minimum threshold
+      ASSERT_GE(file_size, min_file_size_);
+    }
+  }
+
+ private:
+  static uint64_t GetSstFileSize(Env* env, const std::vector<DbPath>& db_paths,
+                                 uint64_t file_number) {
+    uint32_t path_id = 0;  // since only one path
+    std::string sst_file_name = TableFileName(db_paths, file_number, path_id);
+    uint64_t file_size = 0;
+    Status s = env->GetFileSize(sst_file_name, &file_size);
+    if (!s.ok()) {
+      return 0;
+    }
+    return file_size;
+  }
+
+  uint64_t min_file_size_;
+};
+
 class DBCompactionTest : public DBTestBase {
 public:
  DBCompactionTest()
@@ -455,6 +492,72 @@ TEST_P(DBCompactionTestWithParam, CompactionDeletionTrigger) {
  }
 }
 #endif  // !defined(ROCKSDB_VALGRIND_RUN) || defined(ROCKSDB_FULL_VALGRIND_RUN)
+TEST_F(DBCompactionTest, UniversalReduceFileLockingRepickNothing) {
+  const int kFileNumCompactionTrigger = 3;
+
+  Options options = CurrentOptions();
+  options.compaction_options_universal.reduce_file_locking = true;
+  // Set `max_background_jobs` to be 3 to allow low and bottom priority thread
+  // to run compaction together
+  options.max_background_jobs = 3;
+  Env::Default()->SetBackgroundThreads(1, Env::Priority::BOTTOM);
+  options.num_levels = 3;
+  options.compaction_style = kCompactionStyleUniversal;
+  options.level0_file_num_compaction_trigger = kFileNumCompactionTrigger;
+  options.compaction_options_universal.max_size_amplification_percent = 1;
+
+  DestroyAndReopen(options);
+
+  // Need to get a token to enable compaction parallelism up to
+  // `max_background_compactions` jobs.
+  auto pressure_token =
+      dbfull()->TEST_write_controler().GetCompactionPressureToken();
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
+      {// Wait for the full (bottom-priority) compaction to be pre-picked as an
+       // intent (that is allowing files to be picked by other compactions and
+       // will pick later when the bottom-priority thread is available to
+       // execute the compaction) before triggering the low-priority compaction.
+       {"DBImpl::BackgroundCompaction:ForwardToBottomPriPool",
+        "LowPriCompaction"},
+       // Wait for low-priority compaction to start before
+       // repicking for the full compaction intent (bottom-priority), enabling
+       // them to run in parallel.
+       {"DBImpl::BackgroundCompaction:NonTrivial",
+        "DBImpl::BGWorkBottomCompaction"}});
+
+  bool bottom_pri_compaction_attempt_repick = false;
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::BackgroundCompaction():AfterPickCompactionBottomPri",
+      [&](void* arg) {
+        bottom_pri_compaction_attempt_repick = true;
+        Compaction* c = static_cast<Compaction*>(arg);
+        // Verify the intended full compaction for bottom priority thread does
+        // not get to run (i.e, output to bottommost level) since when it
+        // repicks its files, some of the the intended input files are already
+        // compacted by the low priority thread
+        assert(c == nullptr);
+      });
+
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  for (int i = 0; i < kFileNumCompactionTrigger; ++i) {
+    if (i == 0) {
+      ASSERT_OK(Put("file_locked_for_bottom_pri_compaction", "value"));
+    } else {
+      ASSERT_OK(
+          Put("file_not_locked_for_bottom_pri_compaction" + std::to_string(i),
+              "value"));
+    }
+    ASSERT_OK(Flush());
+  }
+
+  TEST_SYNC_POINT("LowPriCompaction");
+  ASSERT_OK(Put("a_new_file_to_pick_for_low_pri_compaction", "value"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(dbfull()->TEST_WaitForCompact());
+  ASSERT_TRUE(bottom_pri_compaction_attempt_repick);
+}

 TEST_F(DBCompactionTest, SkipStatsUpdateTest) {
  // This test verify UpdateAccumulatedStats is not on
@@ -1305,6 +1408,89 @@ TEST_F(DBCompactionTest, RecoverDuringMemtableCompaction) {
  } while (ChangeOptions());
 }

+TEST_F(DBCompactionTest, CompactionWithDeletionsAndMinFileSize) {
+  const uint64_t kMinFileSize = 32 * 1024;  // 32KB
+  const int kDeletionTriggerCount = 50;
+  const int kInitialKeyCount = 100;
+  const int kAdditionalKeyCount = 50;
+  const int kValueSize = 1024;
+  const int kSmallValueSize = 512;
+  const int kSeed = 301;
+
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleLevel;
+  options.write_buffer_size = 1024 * 1024;  // 1MB
+  options.level0_file_num_compaction_trigger = 100;
+
+  options.table_properties_collector_factories = {
+      NewCompactOnDeletionCollectorFactory(
+          kInitialKeyCount /* sliding window size */, kDeletionTriggerCount,
+          0.5 /* deletion ratio */, kMinFileSize)};
+  auto listener =
+      new DeletionTriggeredCompactionWithMinFileSizeTestListener(kMinFileSize);
+  options.listeners.emplace_back(listener);
+
+  DestroyAndReopen(options);
+  Random rnd(kSeed);
+
+  // Create a large file that will be subject to DTC later
+  for (int i = 0; i < kInitialKeyCount; i++) {
+    ASSERT_OK(Put(Key(i), rnd.RandomString(kValueSize)));
+  }
+  ASSERT_OK(Flush());
+
+  std::vector<LiveFileMetaData> initial_metadata;
+  db_->GetLiveFilesMetaData(&initial_metadata);
+  ASSERT_EQ(initial_metadata.size(), 1);
+
+  // Create small files that should not trigger compaction
+  ASSERT_OK(Put("small_file_key1", rnd.RandomString(kSmallValueSize)));
+  ASSERT_OK(Put("small_file_key2", rnd.RandomString(kSmallValueSize)));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Delete("small_file_key1"));
+  ASSERT_OK(Flush());
+
+  // Create a file with enough deletions and size to trigger DTC
+  // Delete keys from the large file to reach deletion threshold
+  for (int i = 0; i < kDeletionTriggerCount; i++) {
+    ASSERT_OK(Delete(Key(i)));
+  }
+
+  // Add new keys to ensure the deletion file meets the min_file_size threshold
+  for (int i = kInitialKeyCount; i < kInitialKeyCount + kAdditionalKeyCount;
+       i++) {
+    ASSERT_OK(Put(Key(i), rnd.RandomString(kValueSize)));
+  }
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(dbfull()->TEST_WaitForCompact());
+
+  // Verify file count after compaction
+  ASSERT_EQ(NumTableFilesAtLevel(0), 2);  // Small file and deletion file
+  ASSERT_EQ(NumTableFilesAtLevel(1), 1);  // Compacted large file
+
+  // Verify deleted keys are gone
+  for (int i = 0; i < kDeletionTriggerCount; i++) {
+    std::string value;
+    ASSERT_TRUE(db_->Get(ReadOptions(), Key(i), &value).IsNotFound());
+  }
+
+  // Verify non-deleted keys from large file are still accessible
+  for (int i = kDeletionTriggerCount; i < kInitialKeyCount; i++) {
+    std::string value;
+    ASSERT_OK(db_->Get(ReadOptions(), Key(i), &value));
+    ASSERT_EQ(value.size(), kValueSize);
+  }
+
+  // Verify new keys are accessible
+  for (int i = kInitialKeyCount; i < kInitialKeyCount + kAdditionalKeyCount;
+       i++) {
+    std::string value;
+    ASSERT_OK(db_->Get(ReadOptions(), Key(i), &value));
+    ASSERT_EQ(value.size(), kValueSize);
+  }
+}
+
 TEST_P(DBCompactionTestWithParam, TrivialMoveOneFile) {
  int32_t trivial_move = 0;
  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
@@ -2824,46 +3010,99 @@ TEST_F(DBCompactionTest, L0_CompactionBug_Issue44_b) {
 }

 TEST_F(DBCompactionTest, ManualAutoRace) {
-  CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
-      {{"DBImpl::BGWorkCompaction", "DBCompactionTest::ManualAutoRace:1"},
-       {"DBImpl::RunManualCompaction:WaitScheduled",
-        "BackgroundCallCompaction:0"}});
+  const int kNumL0FilesTrigger = 4;
+  // Verify that the auto compaction is retried after the conflicting exclusive
+  // manual compaction finishes for:
+  // 1. Non-bottom-priority compactions (tested with level compaction)
+  // 2. Bottom-priority compactions (tested with universal compaction)
+  for (auto compaction_style :
+       {kCompactionStyleLevel, kCompactionStyleUniversal}) {
+    Env::Default()->SetBackgroundThreads(
+        compaction_style == kCompactionStyleUniversal ? 2 : 0,
+        Env::Priority::BOTTOM);
+    for (auto universal_reduce_file_locking : {false, true}) {
+      if (compaction_style != kCompactionStyleUniversal &&
+          universal_reduce_file_locking) {
+        continue;
+      }

-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+      Options options = CurrentOptions();
+      options.num_levels = 3;
+      options.level0_file_num_compaction_trigger = kNumL0FilesTrigger;
+      options.compaction_style = compaction_style;
+      options.compaction_options_universal.reduce_file_locking =
+          universal_reduce_file_locking;

-  ASSERT_OK(Put(1, "foo", ""));
-  ASSERT_OK(Put(1, "bar", ""));
-  ASSERT_OK(Flush(1));
-  ASSERT_OK(Put(1, "foo", ""));
-  ASSERT_OK(Put(1, "bar", ""));
-  // Generate four files in CF 0, which should trigger an auto compaction
-  ASSERT_OK(Put("foo", ""));
-  ASSERT_OK(Put("bar", ""));
-  ASSERT_OK(Flush());
-  ASSERT_OK(Put("foo", ""));
-  ASSERT_OK(Put("bar", ""));
-  ASSERT_OK(Flush());
-  ASSERT_OK(Put("foo", ""));
-  ASSERT_OK(Put("bar", ""));
-  ASSERT_OK(Flush());
-  ASSERT_OK(Put("foo", ""));
-  ASSERT_OK(Put("bar", ""));
-  ASSERT_OK(Flush());
+      DestroyAndReopen(options);
+      CreateAndReopenWithCF({"exclusive_manual_compaction_cf"}, options);

-  // The auto compaction is scheduled but waited until here
-  TEST_SYNC_POINT("DBCompactionTest::ManualAutoRace:1");
-  // The auto compaction will wait until the manual compaction is registerd
-  // before processing so that it will be cancelled.
-  CompactRangeOptions cro;
-  cro.exclusive_manual_compaction = true;
-  ASSERT_OK(dbfull()->CompactRange(cro, handles_[1], nullptr, nullptr));
-  ASSERT_EQ("0,1", FilesPerLevel(1));
+      // Set up sync points to ensure that the auto compaction
+      // encounters a conflict from exclusive manual compaction before the auto
+      // compaction gets to pick files, This will trigger a retry later.
+      //
+      // Specifically, the sync points are set up as following:
+      // 1. Wait until background low-pri scheduled (not picking files yet) or
+      // bottom-pri scheduled (not repicking files yet) for
+      // `universal_reduce_file_locking = true` before triggering
+      // CompactRange()
+      //
+      // 2. Wait until the triggered CompactRange()
+      // registers its compaction and creates conflict before the auto
+      // compaction picks or repicks files for the background compaction.
+      if (compaction_style == kCompactionStyleLevel ||
+          !universal_reduce_file_locking) {
+        ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
+            {{"DBImpl::BGWorkCompaction", "DBCompactionTest::ManualAutoRace:1"},
+             {"DBImpl::RunManualCompaction:WaitScheduled",
+              "BackgroundCallCompaction:0"}});
+      } else {
+        ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
+            {{"DBImpl::BackgroundCompaction:ForwardToBottomPriPool",
+              "DBCompactionTest::ManualAutoRace:1"},
+             {"DBImpl::RunManualCompaction:WaitScheduled",
+              "BackgroundCallCompaction:0:BottomPri"}});
+      }

-  // Eventually the cancelled compaction will be rescheduled and executed.
-  ASSERT_OK(dbfull()->TEST_WaitForCompact());
-  ASSERT_EQ("0,1", FilesPerLevel(0));
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
+      bool encounter_conflict = false;
+      ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+          "DBImpl::BackgroundCompaction()::Conflict",
+          [&](void* /*arg*/) { encounter_conflict = true; });
+      ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
+      // Generate files in CF 1 for exclusive CompactRange()
+      ASSERT_OK(Put(1, "foo", ""));
+      ASSERT_OK(Put(1, "bar", ""));
+      ASSERT_OK(Flush(1));
+      ASSERT_OK(Put(1, "foo", ""));
+      ASSERT_OK(Put(1, "bar", ""));
+      // Generate files in CF0 to trigger full compaction
+      for (int i = 0; i < kNumL0FilesTrigger; ++i) {
+        ASSERT_OK(Put("foo", ""));
+        ASSERT_OK(Put("bar", ""));
+        ASSERT_OK(Flush());
+      }
+
+      TEST_SYNC_POINT("DBCompactionTest::ManualAutoRace:1");
+      CompactRangeOptions cro;
+      cro.exclusive_manual_compaction = true;
+      ASSERT_OK(dbfull()->CompactRange(cro, handles_[1], nullptr, nullptr));
+      ASSERT_EQ(compaction_style == kCompactionStyleLevel ? "0,1" : "0,0,1",
+                FilesPerLevel(1));
+
+      ASSERT_OK(dbfull()->TEST_WaitForCompact());
+
+      ASSERT_TRUE(encounter_conflict);
+
+      // Verify that the auto compaction is eventually executed after the
+      // exclusive CompactRange() finishes.
+      ASSERT_EQ(compaction_style == kCompactionStyleLevel ? "0,1" : "0,0,1",
+                FilesPerLevel(0));
+
+      ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
+      ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
+    }
+    Env::Default()->SetBackgroundThreads(0, Env::Priority::BOTTOM);
+  }
 }

 TEST_P(DBCompactionTestWithParam, ManualCompaction) {
@@ -3986,41 +4225,51 @@ TEST_P(DBCompactionTestWithParam, IntraL0CompactionDoesNotObsoleteDeletions) {
 TEST_P(DBCompactionTestWithParam, FullCompactionInBottomPriThreadPool) {
  const int kNumFilesTrigger = 3;
  Env::Default()->SetBackgroundThreads(1, Env::Priority::BOTTOM);
-  for (bool use_universal_compaction : {false, true}) {
-    Options options = CurrentOptions();
-    if (use_universal_compaction) {
-      options.compaction_style = kCompactionStyleUniversal;
-    } else {
-      options.compaction_style = kCompactionStyleLevel;
-      options.level_compaction_dynamic_level_bytes = true;
+  for (auto compaction_style :
+       {kCompactionStyleLevel, kCompactionStyleUniversal}) {
+    for (auto universal_reduce_file_locking : {false, true}) {
+      if (compaction_style != kCompactionStyleUniversal &&
+          universal_reduce_file_locking) {
+        continue;
+      }
+      Options options = CurrentOptions();
+      options.compaction_style = compaction_style;
+      if (compaction_style == kCompactionStyleLevel) {
+        options.level_compaction_dynamic_level_bytes = true;
+      } else {
+        options.compaction_options_universal.reduce_file_locking =
+            universal_reduce_file_locking;
+        // Trigger compaction if size amplification exceeds 110%
+        options.compaction_options_universal.max_size_amplification_percent =
+            110;
+      }
+      options.num_levels = 4;
+      options.write_buffer_size = 100 << 10;     // 100KB
+      options.target_file_size_base = 32 << 10;  // 32KB
+      options.level0_file_num_compaction_trigger = kNumFilesTrigger;
+
+      DestroyAndReopen(options);
+
+      int num_bottom_pri_compactions = 0;
+      SyncPoint::GetInstance()->SetCallBack(
+          "DBImpl::BGWorkBottomCompaction",
+          [&](void* /*arg*/) { ++num_bottom_pri_compactions; });
+      SyncPoint::GetInstance()->EnableProcessing();
+
+      Random rnd(301);
+      for (int num = 0; num < kNumFilesTrigger; num++) {
+        ASSERT_EQ(NumSortedRuns(), num);
+        int key_idx = 0;
+        GenerateNewFile(&rnd, &key_idx);
+      }
+      ASSERT_OK(dbfull()->TEST_WaitForCompact());
+
+      ASSERT_EQ(1, num_bottom_pri_compactions);
+
+      // Verify that size amplification did occur
+      ASSERT_EQ(NumSortedRuns(), 1);
+      ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
    }
-    options.num_levels = 4;
-    options.write_buffer_size = 100 << 10;     // 100KB
-    options.target_file_size_base = 32 << 10;  // 32KB
-    options.level0_file_num_compaction_trigger = kNumFilesTrigger;
-    // Trigger compaction if size amplification exceeds 110%
-    options.compaction_options_universal.max_size_amplification_percent = 110;
-    DestroyAndReopen(options);
-
-    int num_bottom_pri_compactions = 0;
-    SyncPoint::GetInstance()->SetCallBack(
-        "DBImpl::BGWorkBottomCompaction",
-        [&](void* /*arg*/) { ++num_bottom_pri_compactions; });
-    SyncPoint::GetInstance()->EnableProcessing();
-
-    Random rnd(301);
-    for (int num = 0; num < kNumFilesTrigger; num++) {
-      ASSERT_EQ(NumSortedRuns(), num);
-      int key_idx = 0;
-      GenerateNewFile(&rnd, &key_idx);
-    }
-    ASSERT_OK(dbfull()->TEST_WaitForCompact());
-
-    ASSERT_EQ(1, num_bottom_pri_compactions);
-
-    // Verify that size amplification did occur
-    ASSERT_EQ(NumSortedRuns(), 1);
-    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
  }
  Env::Default()->SetBackgroundThreads(0, Env::Priority::BOTTOM);
 }
@@ -6619,7 +6868,11 @@ INSTANTIATE_TEST_CASE_P(RoundRobinSubcompactionsAgainstPressureToken,
                        RoundRobinSubcompactionsAgainstPressureToken,
                        testing::Bool());

-TEST_P(RoundRobinSubcompactionsAgainstResources, SubcompactionsUsingResources) {
+// FIXME: the test is flaky and failing the assertion
+// ASSERT_EQ(actual_reserved_threads, expected_reserved_threads);
+// It's likely a test set up issue, fix if we are to use RoubdRobin compaction.
+TEST_P(RoundRobinSubcompactionsAgainstResources,
+       DISABLED_SubcompactionsUsingResources) {
  const int kKeysPerBuffer = 200;
  Options options = CurrentOptions();
  options.num_levels = 4;
@@ -6634,7 +6887,7 @@ TEST_P(RoundRobinSubcompactionsAgainstResources, SubcompactionsUsingResources) {
  // compaction is enough to make post-compaction L1 size less than
  // the maximum size (this test assumes only one round-robin compaction
  // is triggered by kLevelMaxLevelSize)
-  options.max_compaction_bytes = 100000000;
+  options.max_compaction_bytes = std::numeric_limits<uint64_t>::max();

  DestroyAndReopen(options);
  env_->SetBackgroundThreads(total_low_pri_threads_, Env::LOW);
@@ -6667,41 +6920,33 @@ TEST_P(RoundRobinSubcompactionsAgainstResources, SubcompactionsUsingResources) {
        // More than 10 files are selected for round-robin under auto
        // compaction. The number of planned subcompaction is restricted by
        // the minimum number between available threads and compaction limits
-        ASSERT_EQ(num_planned_subcompactions - options.max_subcompactions,
-                  std::min(total_low_pri_threads_, max_compaction_limits_) - 1);
+        auto actual_reserved_threads =
+            num_planned_subcompactions - options.max_subcompactions;
+        auto expected_reserved_threads =
+            std::min(total_low_pri_threads_, max_compaction_limits_) - 1;
+        ASSERT_EQ(actual_reserved_threads, expected_reserved_threads);
        num_planned_subcompactions_verified = true;
      });
-  SyncPoint::GetInstance()->LoadDependency(
-      {{"RoundRobinSubcompactionsAgainstResources:0",
-        "BackgroundCallCompaction:0"},
-       {"CompactionJob::AcquireSubcompactionResources:0",
-        "RoundRobinSubcompactionsAgainstResources:1"},
-       {"RoundRobinSubcompactionsAgainstResources:2",
-        "CompactionJob::AcquireSubcompactionResources:1"},
-       {"CompactionJob::ReleaseSubcompactionResources:0",
-        "RoundRobinSubcompactionsAgainstResources:3"},
-       {"RoundRobinSubcompactionsAgainstResources:4",
-        "CompactionJob::ReleaseSubcompactionResources:1"}});
+
+  int acquire_count = 0;
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "CompactionJob::AcquireSubcompactionResources:0",
+      [&](void* /*arg*/) { acquire_count++; });
+  int release_count = 0;
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "CompactionJob::ReleaseSubcompactionResources",
+      [&](void* /*arg*/) { release_count++; });
+
  SyncPoint::GetInstance()->EnableProcessing();

-  ASSERT_OK(dbfull()->TEST_WaitForCompact());
-  ASSERT_OK(dbfull()->EnableAutoCompaction({dbfull()->DefaultColumnFamily()}));
-  TEST_SYNC_POINT("RoundRobinSubcompactionsAgainstResources:0");
-  TEST_SYNC_POINT("RoundRobinSubcompactionsAgainstResources:1");
  auto pressure_token =
      dbfull()->TEST_write_controler().GetCompactionPressureToken();
-
-  TEST_SYNC_POINT("RoundRobinSubcompactionsAgainstResources:2");
-  TEST_SYNC_POINT("RoundRobinSubcompactionsAgainstResources:3");
-  // We can reserve more threads now except one is being used
-  ASSERT_EQ(total_low_pri_threads_ - 1,
-            env_->ReserveThreads(total_low_pri_threads_, Env::Priority::LOW));
-  ASSERT_EQ(
-      total_low_pri_threads_ - 1,
-      env_->ReleaseThreads(total_low_pri_threads_ - 1, Env::Priority::LOW));
-  TEST_SYNC_POINT("RoundRobinSubcompactionsAgainstResources:4");
+  ASSERT_OK(dbfull()->EnableAutoCompaction({dbfull()->DefaultColumnFamily()}));
  ASSERT_OK(dbfull()->TEST_WaitForCompact());
+
  ASSERT_TRUE(num_planned_subcompactions_verified);
+  ASSERT_EQ(acquire_count, release_count);
+
  SyncPoint::GetInstance()->DisableProcessing();
  SyncPoint::GetInstance()->ClearAllCallBacks();
 }
@@ -6883,6 +7128,70 @@ TEST_F(DBCompactionTest, PartialManualCompaction) {
  ASSERT_OK(dbfull()->CompactRange(cro, nullptr, nullptr));
 }

+TEST_F(DBCompactionTest, ConcurrentFIFOPickingSameFileBug) {
+  Options opts = CurrentOptions();
+  opts.compaction_style = CompactionStyle::kCompactionStyleLevel;
+  opts.num_levels = 3;
+  opts.disable_auto_compactions = true;
+  opts.max_background_jobs = 3;
+
+  DestroyAndReopen(opts);
+
+  ASSERT_OK(Put("k1", "v1"));
+  ASSERT_OK(Flush());
+
+  // Create a non-L0 SST file for multi-level FIFO size-based compaction later
+  MoveFilesToLevel(2);
+
+  Options opts_new(opts);
+  opts_new.compaction_style = CompactionStyle::kCompactionStyleFIFO;
+  opts_new.max_open_files = -1;
+  // Set a low threshold to trigger multi-level size-based compaction
+  opts_new.compaction_options_fifo.max_table_files_size = 1;
+
+  Reopen(opts_new);
+
+  const CompactRangeOptions cro;
+  const Slice begin_key("k1");
+  const Slice end_key("k2");
+
+  std::unique_ptr<port::Thread> concurrent_compaction;
+
+  bool within_first_compaction = true;
+  SyncPoint::GetInstance()->SetCallBack(
+      "VersionSet::LogAndApply:WriteManifestStart", [&](void* /*arg*/) {
+        if (!within_first_compaction) {
+          return;
+        }
+        within_first_compaction = false;
+
+        // To allow the second/concurrent compaction to still see the non-L0
+        // SST file and coerce the bug of picking that file
+        SyncPoint::GetInstance()->LoadDependency({
+            {"DBImpl::BackgroundCompaction:BeforeCompaction",
+             "VersionSet::LogAndApply:WriteManifest"},
+        });
+
+        concurrent_compaction.reset(new port::Thread([&]() {
+          // Before the fix, the second CompactRange() will either fail the
+          // assertion of double file picking `being_compacted !=
+          // inputs_[i][j]->being_compacted` in debug mode or cause LSM shape
+          // corruption "Cannot delete table file XXX from level 2 since it is
+          // not in the LSM tree" in release mode
+          Status s = db_->CompactRange(cro, &begin_key, &end_key);
+          ASSERT_OK(s);
+        }));
+      });
+
+  SyncPoint::GetInstance()->EnableProcessing();
+  Status s = db_->CompactRange(cro, &begin_key, &end_key);
+  SyncPoint::GetInstance()->DisableProcessing();
+
+  ASSERT_OK(s);
+
+  concurrent_compaction->join();
+}
+
 TEST_F(DBCompactionTest, ManualCompactionFailsInReadOnlyMode) {
  // Regression test for bug where manual compaction hangs forever when the DB
  // is in read-only mode. Verify it now at least returns, despite failing.
@@ -9434,104 +9743,134 @@ TEST_F(DBCompactionTest, CompactionWithChecksumHandoffManifest2) {
 }

 TEST_F(DBCompactionTest, FIFOChangeTemperature) {
-  for (bool write_time_default : {false, true}) {
-    SCOPED_TRACE("write time default? " + std::to_string(write_time_default));
+  for (bool should_allow_trivial_copy : {false, true}) {
+    for (bool write_time_default : {false, true}) {
+      int32_t before_compaction_calls = 0;
+      int32_t after_compaction_calls = 0;
+      if (should_allow_trivial_copy) {
+        ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+            "DBImpl::BackgroundCompaction:TriviaCopyBeforeCompaction",
+            [&](void*) { ++before_compaction_calls; });
+        ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+            "DBImpl::BackgroundCompaction:TriviaCopyAfterCompaction",
+            [&](void*) { ++after_compaction_calls; });
+      } else {
+        ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+            "DBImpl::BackgroundCompaction:BeforeCompaction",
+            [&](void*) { ++before_compaction_calls; });

-    Options options = CurrentOptions();
-    options.compaction_style = kCompactionStyleFIFO;
-    options.num_levels = 1;
-    options.max_open_files = -1;
-    options.level0_file_num_compaction_trigger = 2;
-    options.create_if_missing = true;
-    CompactionOptionsFIFO fifo_options;
-    fifo_options.file_temperature_age_thresholds = {{Temperature::kCold, 1000}};
-    fifo_options.max_table_files_size = 100000000;
-    options.compaction_options_fifo = fifo_options;
-    env_->SetMockSleep();
-    if (write_time_default) {
-      options.default_write_temperature = Temperature::kWarm;
+        ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+            "DBImpl::BackgroundCompaction:AfterCompaction",
+            [&](void*) { ++after_compaction_calls; });
+      }
+
+      SCOPED_TRACE("write time default? " + std::to_string(write_time_default));
+
+      Options options = CurrentOptions();
+      options.compaction_style = kCompactionStyleFIFO;
+      options.num_levels = 1;
+      options.max_open_files = -1;
+      options.level0_file_num_compaction_trigger = 2;
+      options.create_if_missing = true;
+      CompactionOptionsFIFO fifo_options;
+      fifo_options.file_temperature_age_thresholds = {
+          {Temperature::kCold, 1000}};
+      fifo_options.max_table_files_size = 100000000;
+      fifo_options.allow_trivial_copy_when_change_temperature =
+          should_allow_trivial_copy;
+      fifo_options.trivial_copy_buffer_size = 4096;
+      options.compaction_options_fifo = fifo_options;
+      env_->SetMockSleep();
+      if (write_time_default) {
+        options.default_write_temperature = Temperature::kWarm;
+      }
+      // Should be ignored (TODO: fail?)
+      options.last_level_temperature = Temperature::kHot;
+      Reopen(options);
+
+      int total_cold = 0;
+      int total_warm = 0;
+      int total_hot = 0;
+      int total_unknown = 0;
+      ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+          "NewWritableFile::FileOptions.temperature", [&](void* arg) {
+            Temperature temperature = *(static_cast<Temperature*>(arg));
+            if (temperature == Temperature::kCold) {
+              total_cold++;
+            } else if (temperature == Temperature::kWarm) {
+              total_warm++;
+            } else if (temperature == Temperature::kHot) {
+              total_hot++;
+            } else {
+              assert(temperature == Temperature::kUnknown);
+              total_unknown++;
+            }
+          });
+      ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
+      // The file system does not support checksum handoff. The check
+      // will be ignored.
+      ASSERT_OK(Put(Key(0), "value1"));
+      env_->MockSleepForSeconds(800);
+      ASSERT_OK(Put(Key(2), "value2"));
+      ASSERT_OK(Flush());
+
+      ASSERT_OK(Put(Key(0), "value1"));
+      ASSERT_OK(Put(Key(2), "value2"));
+      ASSERT_OK(Flush());
+
+      // First two L0 files both become eligible for temperature change
+      // compaction They should be compacted one-by-one.
+      ASSERT_OK(Put(Key(0), "value1"));
+      env_->MockSleepForSeconds(1200);
+      ASSERT_OK(Put(Key(2), "value2"));
+      ASSERT_OK(Flush());
+      ASSERT_OK(dbfull()->TEST_WaitForCompact());
+
+      if (write_time_default) {
+        // Also test dynamic option change
+        ASSERT_OK(db_->SetOptions({{"default_write_temperature", "kHot"}}));
+      }
+
+      ASSERT_OK(Put(Key(0), "value1"));
+      env_->MockSleepForSeconds(800);
+      ASSERT_OK(Put(Key(2), "value2"));
+      ASSERT_OK(Flush());
+
+      ASSERT_OK(dbfull()->TEST_WaitForCompact());
+
+      ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
+
+      ColumnFamilyMetaData metadata;
+      db_->GetColumnFamilyMetaData(&metadata);
+      ASSERT_EQ(4, metadata.file_count);
+      if (write_time_default) {
+        ASSERT_EQ(Temperature::kHot, metadata.levels[0].files[0].temperature);
+        ASSERT_EQ(Temperature::kWarm, metadata.levels[0].files[1].temperature);
+        // Includes obsolete/deleted files moved to cold
+        ASSERT_EQ(total_warm, 3);
+        ASSERT_EQ(total_hot, 1);
+        // Includes non-SST DB files
+        ASSERT_GT(total_unknown, 0);
+      } else {
+        ASSERT_EQ(Temperature::kUnknown,
+                  metadata.levels[0].files[0].temperature);
+        ASSERT_EQ(Temperature::kUnknown,
+                  metadata.levels[0].files[1].temperature);
+        ASSERT_EQ(total_warm, 0);
+        ASSERT_EQ(total_hot, 0);
+        // Includes non-SST DB files
+        ASSERT_GT(total_unknown, 4);
+      }
+      ASSERT_EQ(Temperature::kCold, metadata.levels[0].files[2].temperature);
+      ASSERT_EQ(Temperature::kCold, metadata.levels[0].files[3].temperature);
+      ASSERT_EQ(2, total_cold);
+
+      ASSERT_EQ(2, before_compaction_calls);
+      ASSERT_EQ(2, after_compaction_calls);
+
+      Destroy(options);
    }
-    // Should be ignored (TODO: fail?)
-    options.last_level_temperature = Temperature::kHot;
-    Reopen(options);
-
-    int total_cold = 0;
-    int total_warm = 0;
-    int total_hot = 0;
-    int total_unknown = 0;
-    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
-        "NewWritableFile::FileOptions.temperature", [&](void* arg) {
-          Temperature temperature = *(static_cast<Temperature*>(arg));
-          if (temperature == Temperature::kCold) {
-            total_cold++;
-          } else if (temperature == Temperature::kWarm) {
-            total_warm++;
-          } else if (temperature == Temperature::kHot) {
-            total_hot++;
-          } else {
-            assert(temperature == Temperature::kUnknown);
-            total_unknown++;
-          }
-        });
-    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
-
-    // The file system does not support checksum handoff. The check
-    // will be ignored.
-    ASSERT_OK(Put(Key(0), "value1"));
-    env_->MockSleepForSeconds(800);
-    ASSERT_OK(Put(Key(2), "value2"));
-    ASSERT_OK(Flush());
-
-    ASSERT_OK(Put(Key(0), "value1"));
-    ASSERT_OK(Put(Key(2), "value2"));
-    ASSERT_OK(Flush());
-
-    // First two L0 files both become eligible for temperature change compaction
-    // They should be compacted one-by-one.
-    ASSERT_OK(Put(Key(0), "value1"));
-    env_->MockSleepForSeconds(1200);
-    ASSERT_OK(Put(Key(2), "value2"));
-    ASSERT_OK(Flush());
-    ASSERT_OK(dbfull()->TEST_WaitForCompact());
-
-    if (write_time_default) {
-      // Also test dynamic option change
-      ASSERT_OK(db_->SetOptions({{"default_write_temperature", "kHot"}}));
-    }
-
-    ASSERT_OK(Put(Key(0), "value1"));
-    env_->MockSleepForSeconds(800);
-    ASSERT_OK(Put(Key(2), "value2"));
-    ASSERT_OK(Flush());
-
-    ASSERT_OK(dbfull()->TEST_WaitForCompact());
-
-    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
-
-    ColumnFamilyMetaData metadata;
-    db_->GetColumnFamilyMetaData(&metadata);
-    ASSERT_EQ(4, metadata.file_count);
-    if (write_time_default) {
-      ASSERT_EQ(Temperature::kHot, metadata.levels[0].files[0].temperature);
-      ASSERT_EQ(Temperature::kWarm, metadata.levels[0].files[1].temperature);
-      // Includes obsolete/deleted files moved to cold
-      ASSERT_EQ(total_warm, 3);
-      ASSERT_EQ(total_hot, 1);
-      // Includes non-SST DB files
-      ASSERT_GT(total_unknown, 0);
-    } else {
-      ASSERT_EQ(Temperature::kUnknown, metadata.levels[0].files[0].temperature);
-      ASSERT_EQ(Temperature::kUnknown, metadata.levels[0].files[1].temperature);
-      ASSERT_EQ(total_warm, 0);
-      ASSERT_EQ(total_hot, 0);
-      // Includes non-SST DB files
-      ASSERT_GT(total_unknown, 4);
-    }
-    ASSERT_EQ(Temperature::kCold, metadata.levels[0].files[2].temperature);
-    ASSERT_EQ(Temperature::kCold, metadata.levels[0].files[3].temperature);
-    ASSERT_EQ(2, total_cold);
-
-    Destroy(options);
  }
 }

@@ -9976,55 +10315,60 @@ TEST_F(DBCompactionTest, BottomPriCompactionCountsTowardConcurrencyLimit) {

  env_->SetBackgroundThreads(1, Env::Priority::BOTTOM);

-  Options options = CurrentOptions();
-  options.level0_file_num_compaction_trigger = kNumL0Files;
-  options.num_levels = kNumLevels;
-  DestroyAndReopen(options);
+  for (bool universal_reduce_file_locking : {false, true}) {
+    Options options = CurrentOptions();
+    options.level0_file_num_compaction_trigger = kNumL0Files;
+    options.num_levels = kNumLevels;
+    options.compaction_style = kCompactionStyleUniversal;
+    options.compaction_options_universal.reduce_file_locking =
+        universal_reduce_file_locking;
+    DestroyAndReopen(options);

-  // Setup last level to be non-empty since it's a bit unclear whether
-  // compaction to an empty level would be considered "bottommost".
-  ASSERT_OK(Put(Key(0), "val"));
-  ASSERT_OK(Flush());
-  MoveFilesToLevel(kNumLevels - 1);
-
-  SyncPoint::GetInstance()->LoadDependency(
-      {{"DBImpl::BGWorkBottomCompaction",
-        "DBCompactionTest::BottomPriCompactionCountsTowardConcurrencyLimit:"
-        "PreTriggerCompaction"},
-       {"DBCompactionTest::BottomPriCompactionCountsTowardConcurrencyLimit:"
-        "PostTriggerCompaction",
-        "BackgroundCallCompaction:0"}});
-  SyncPoint::GetInstance()->EnableProcessing();
-
-  port::Thread compact_range_thread([&] {
-    CompactRangeOptions cro;
-    cro.bottommost_level_compaction = BottommostLevelCompaction::kForce;
-    cro.exclusive_manual_compaction = false;
-    ASSERT_OK(dbfull()->CompactRange(cro, nullptr, nullptr));
-  });
-
-  // Sleep in the low-pri thread so any newly scheduled compaction will be
-  // queued. Otherwise it might finish before we check its existence.
-  test::SleepingBackgroundTask sleeping_task_low;
-  env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task_low,
-                 Env::Priority::LOW);
-  sleeping_task_low.WaitUntilSleeping();
-
-  TEST_SYNC_POINT(
-      "DBCompactionTest::BottomPriCompactionCountsTowardConcurrencyLimit:"
-      "PreTriggerCompaction");
-  for (int i = 0; i < kNumL0Files; ++i) {
+    // Setup last level to be non-empty since it's a bit unclear whether
+    // compaction to an empty level would be considered "bottommost".
    ASSERT_OK(Put(Key(0), "val"));
    ASSERT_OK(Flush());
-  }
-  ASSERT_EQ(0u, env_->GetThreadPoolQueueLen(Env::Priority::LOW));
-  TEST_SYNC_POINT(
-      "DBCompactionTest::BottomPriCompactionCountsTowardConcurrencyLimit:"
-      "PostTriggerCompaction");
+    MoveFilesToLevel(kNumLevels - 1);

-  sleeping_task_low.WakeUp();
-  sleeping_task_low.WaitUntilDone();
-  compact_range_thread.join();
+    SyncPoint::GetInstance()->LoadDependency(
+        {{"DBImpl::BGWorkBottomCompaction",
+          "DBCompactionTest::BottomPriCompactionCountsTowardConcurrencyLimit:"
+          "PreTriggerCompaction"},
+         {"DBCompactionTest::BottomPriCompactionCountsTowardConcurrencyLimit:"
+          "PostTriggerCompaction",
+          "BackgroundCallCompaction:0"}});
+    SyncPoint::GetInstance()->EnableProcessing();
+
+    port::Thread compact_range_thread([&] {
+      CompactRangeOptions cro;
+      cro.bottommost_level_compaction = BottommostLevelCompaction::kForce;
+      cro.exclusive_manual_compaction = false;
+      ASSERT_OK(dbfull()->CompactRange(cro, nullptr, nullptr));
+    });
+
+    // Sleep in the low-pri thread so any newly scheduled compaction will be
+    // queued. Otherwise it might finish before we check its existence.
+    test::SleepingBackgroundTask sleeping_task_low;
+    env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask,
+                   &sleeping_task_low, Env::Priority::LOW);
+    sleeping_task_low.WaitUntilSleeping();
+
+    TEST_SYNC_POINT(
+        "DBCompactionTest::BottomPriCompactionCountsTowardConcurrencyLimit:"
+        "PreTriggerCompaction");
+    for (int i = 0; i < kNumL0Files; ++i) {
+      ASSERT_OK(Put(Key(0), "val"));
+      ASSERT_OK(Flush());
+    }
+    ASSERT_EQ(0u, env_->GetThreadPoolQueueLen(Env::Priority::LOW));
+    TEST_SYNC_POINT(
+        "DBCompactionTest::BottomPriCompactionCountsTowardConcurrencyLimit:"
+        "PostTriggerCompaction");
+
+    sleeping_task_low.WakeUp();
+    sleeping_task_low.WaitUntilDone();
+    compact_range_thread.join();
+  }
 }

 TEST_F(DBCompactionTest, BottommostFileCompactionAllowIngestBehind) {
@@ -10937,6 +11281,59 @@ TEST_F(DBCompactionTest, RecordNewestKeyTimeForTtlCompaction) {
  ASSERT_OK(dbfull()->TEST_WaitForCompact());
  ASSERT_EQ(NumTableFilesAtLevel(0), 0);
 }
+
+class PeriodicCompactionListener : public EventListener {
+ public:
+  explicit PeriodicCompactionListener() {}
+  void OnCompactionBegin(DB* /*db*/, const CompactionJobInfo& ci) override {
+    if (ci.compaction_reason == CompactionReason::kPeriodicCompaction) {
+      ++num_periodic_compactions;
+    }
+  }
+
+  std::atomic<int> num_periodic_compactions = 0;
+};
+
+TEST_F(DBCompactionTest, PeriodicTask) {
+  // Tests that when no trigger event is fired (flush/compaction/setoptions),
+  // periodic compaction is still triggered by a scheduled periodic function.
+  auto mock_clock = std::make_shared<MockSystemClock>(env_->GetSystemClock());
+  mock_clock->SetCurrentTime(100);
+  mock_clock->InstallTimedWaitFixCallback();
+  auto mock_env = std::make_unique<CompositeEnvWrapper>(env_, mock_clock);
+  SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::StartPeriodicTaskScheduler:Init", [&](void* arg) {
+        auto periodic_task_scheduler_ptr =
+            static_cast<PeriodicTaskScheduler*>(arg);
+        periodic_task_scheduler_ptr->TEST_OverrideTimer(mock_clock.get());
+      });
+
+  Options options;
+  options.env = mock_env.get();
+  options.compaction_style = kCompactionStyleUniversal;
+  options.statistics = CreateDBStatistics();
+  int kPeriodicCompactionSeconds = 7 * 24 * 60 * 60;  // 1 week
+  options.periodic_compaction_seconds = kPeriodicCompactionSeconds;
+  options.num_levels = 50;
+  auto listener = std::make_shared<PeriodicCompactionListener>();
+  options.listeners.push_back(listener);
+  ASSERT_OK(TryReopen(options));
+
+  Random* rnd = Random::GetTLSInstance();
+  for (int k = 0; k < 10; ++k) {
+    ASSERT_OK(Put(Key(k), rnd->RandomString(100)));
+  }
+  ASSERT_OK(Flush());
+  ASSERT_OK(db_->CompactRange({}, nullptr, nullptr));
+  ASSERT_EQ(1, NumTableFilesAtLevel(49));
+
+  dbfull()->TEST_WaitForPeriodicTaskRun(
+      [&] { mock_clock->MockSleepForSeconds(kPeriodicCompactionSeconds + 1); });
+  ASSERT_OK(db_->WaitForCompact({}));
+
+  ASSERT_EQ(listener->num_periodic_compactions, 1);
+  Close();
+}
 }  // namespace ROCKSDB_NAMESPACE

 int main(int argc, char** argv) {
@@ -17,9 +17,10 @@ class DBEncryptionTest : public DBTestBase {
 public:
  DBEncryptionTest()
      : DBTestBase("db_encryption_test", /*env_do_fsync=*/true) {}
-  Env* GetTargetEnv() {
+  Env* GetNonEncryptedEnv() {
    if (encrypted_env_ != nullptr) {
-      return (static_cast<EnvWrapper*>(encrypted_env_))->target();
+      return (static_cast_with_check<CompositeEnvWrapper>(encrypted_env_))
+          ->env_target();
    } else {
      return env_;
    }
@@ -38,7 +39,7 @@ TEST_F(DBEncryptionTest, CheckEncrypted) {
  auto status = env_->GetChildren(dbname_, &fileNames);
  ASSERT_OK(status);

-  Env* target = GetTargetEnv();
+  Env* target = GetNonEncryptedEnv();
  int hits = 0;
  for (auto it = fileNames.begin(); it != fileNames.end(); ++it) {
    if (*it == "LOCK") {
@@ -89,7 +90,7 @@ TEST_F(DBEncryptionTest, CheckEncrypted) {
 }

 TEST_F(DBEncryptionTest, ReadEmptyFile) {
-  auto defaultEnv = GetTargetEnv();
+  auto defaultEnv = GetNonEncryptedEnv();

  // create empty file for reading it back in later
  auto envOptions = EnvOptions(CurrentOptions());
@@ -116,6 +117,40 @@ TEST_F(DBEncryptionTest, ReadEmptyFile) {
  ASSERT_TRUE(data.empty());
 }

+TEST_F(DBEncryptionTest, NotSupportedGetFileSize) {
+  // Validate envrypted env does not support GetFileSize.
+  // The goal of the test is to validate the encrypted env/fs does not support
+  // GetFileSize API on FSRandomAccessFile interface.
+  // This test combined with the rest of the integration tests validate that
+  // the new API GetFileSize on FSRandomAccessFile interface is not required to
+  // be supported for database to work properly.
+  // The GetFileSize API is used in ReadFooterFromFile() API to get the file
+  // size. When GetFileSize API is not supported, the ReadFooterFromFile() API
+  // will use FileSystem GetFileSize API as fallback. Refer to the
+  // EncryptedRandomAccessFile class definition for more details.
+  if (!encrypted_env_) {
+    return;
+  }
+
+  auto fs = encrypted_env_->GetFileSystem();
+
+  // create empty file for reading it back in later
+  auto filePath = dbname_ + "/empty.empty";
+
+  // Create empty file
+  CreateFile(fs.get(), filePath, "", false);
+
+  // Open it for reading footer
+  std::unique_ptr<FSRandomAccessFile> randomAccessFile;
+  auto status = fs->NewRandomAccessFile(filePath, FileOptions(),
+                                        &randomAccessFile, nullptr);
+  ASSERT_OK(status);
+
+  uint64_t fileSize;
+  status = randomAccessFile->GetFileSize(&fileSize);
+  ASSERT_TRUE(status.IsNotSupported());
+}
+
 }  // namespace ROCKSDB_NAMESPACE

 int main(int argc, char** argv) {
@@ -75,11 +75,9 @@ Status DBImpl::GetLiveFiles(std::vector<std::string>& ret,

  ret.emplace_back(CurrentFileName(""));
  ret.emplace_back(DescriptorFileName("", versions_->manifest_file_number()));
-  // The OPTIONS file number is zero in read-write mode when OPTIONS file
-  // writing failed and the DB was configured with
-  // `fail_if_options_file_error == false`. In read-only mode the OPTIONS file
-  // number is zero when no OPTIONS file exist at all. In those cases we do not
-  // record any OPTIONS file in the live file list.
+  // In read-only mode the OPTIONS file number is zero when no OPTIONS file
+  // exist at all. In this cases we do not record any OPTIONS file in the live
+  // file list.
  if (versions_->options_file_number() != 0) {
    ret.emplace_back(OptionsFileName("", versions_->options_file_number()));
  }
@@ -185,14 +183,14 @@ Status DBImpl::GetSortedWalFilesImpl(VectorWalPtr& files, bool need_seqnos) {
  return s;
 }

-Status DBImpl::GetCurrentWalFile(std::unique_ptr<WalFile>* current_log_file) {
+Status DBImpl::GetCurrentWalFile(std::unique_ptr<WalFile>* current_wal_file) {
  uint64_t current_logfile_number;
  {
    InstrumentedMutexLock l(&mutex_);
-    current_logfile_number = logfile_number_;
+    current_logfile_number = cur_wal_number_;
  }

-  return wal_manager_.GetLiveWalFile(current_logfile_number, current_log_file);
+  return wal_manager_.GetLiveWalFile(current_logfile_number, current_wal_file);
 }

 Status DBImpl::GetLiveFilesStorageInfo(
@@ -332,7 +330,7 @@ Status DBImpl::GetLiveFilesStorageInfo(
  const uint64_t options_size = versions_->options_file_size_;
  const uint64_t min_log_num = MinLogNumberToKeep();
  // Ensure consistency with manifest for track_and_verify_wals_in_manifest
-  const uint64_t max_log_num = logfile_number_;
+  const uint64_t max_log_num = cur_wal_number_;

  mutex_.Unlock();

@@ -369,11 +367,9 @@ Status DBImpl::GetLiveFilesStorageInfo(
    }
  }

-  // The OPTIONS file number is zero in read-write mode when OPTIONS file
-  // writing failed and the DB was configured with
-  // `fail_if_options_file_error == false`. In read-only mode the OPTIONS file
-  // number is zero when no OPTIONS file exist at all. In those cases we do not
-  // record any OPTIONS file in the live file list.
+  // In read-only mode the OPTIONS file number is zero when no OPTIONS file
+  // exist at all. In this cases we do not record any OPTIONS file in the live
+  // file list.
  if (options_number != 0) {
    results.emplace_back();
    LiveFileStorageInfo& info = results.back();
@@ -3504,6 +3504,63 @@ TEST_F(DBFlushTest, DBStuckAfterAtomicFlushError) {
  ASSERT_OK(dbfull()->TEST_WaitForBackgroundWork());
  ASSERT_EQ(1, NumTableFilesAtLevel(0));
 }
+
+TEST_F(DBFlushTest, VerifyOutputRecordCount) {
+  for (bool use_plain_table : {false, true}) {
+    Options options = CurrentOptions();
+    options.flush_verify_memtable_count = true;
+    options.merge_operator = MergeOperators::CreateStringAppendOperator();
+    DestroyAndReopen(options);
+    // Verify flush output record count verification in different table
+    // formats
+    if (use_plain_table) {
+      options.table_factory.reset(NewPlainTableFactory());
+    }
+
+    // Verify that flush output record count verification does not produce false
+    // positives.
+    ASSERT_OK(Merge("k0", "v1"));
+    ASSERT_OK(Put("k1", "v1"));
+    ASSERT_OK(Put("k2", "v1"));
+    ASSERT_OK(SingleDelete("k2"));
+    ASSERT_OK(Delete("k2"));
+    ASSERT_OK(Delete("k3"));
+    ASSERT_OK(db_->DeleteRange(WriteOptions(), "k1", "k3"));
+    ASSERT_OK(Flush());
+
+    // Verify that flush output record count verification catch corruption
+    DestroyAndReopen(options);
+    if (use_plain_table) {
+      ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+          "PlainTableBuilder::Add::skip",
+          [&](void* skip) { *(bool*)skip = true; });
+
+    } else {
+      ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+          "BlockBasedTableBuilder::Add::skip",
+          [&](void* skip) { *(bool*)skip = true; });
+    }
+    SyncPoint::GetInstance()->EnableProcessing();
+    const char* expect =
+        "Number of keys in flush output SST files does not match";
+
+    // 1. During DB open flush
+    ASSERT_OK(Put("k1", "v1"));
+    ASSERT_OK(Put("k2", "v1"));
+    Status s = TryReopen(options);
+    ASSERT_TRUE(s.IsCorruption());
+    ASSERT_TRUE(std::strstr(s.getState(), expect));
+
+    // 2. During regular flush
+    DestroyAndReopen(options);
+    ASSERT_OK(Put("k1", "v1"));
+    ASSERT_OK(Put("k2", "v1"));
+    s = Flush();
+    ASSERT_TRUE(s.IsCorruption());
+    ASSERT_TRUE(std::strstr(s.getState(), expect));
+    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
+  }
+}
 }  // namespace ROCKSDB_NAMESPACE

 int main(int argc, char** argv) {
@@ -74,6 +74,7 @@
 #include "options/cf_options.h"
 #include "options/options_helper.h"
 #include "options/options_parser.h"
+#include "util/udt_util.h"
 #ifdef ROCKSDB_JEMALLOC
 #include "port/jemalloc_helper.h"
 #endif
@@ -169,7 +170,6 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname,
               bool read_only)
    : dbname_(dbname),
      own_info_log_(options.info_log == nullptr),
-      init_logger_creation_s_(),
      initial_db_options_(SanitizeOptions(dbname, options, read_only,
                                          &init_logger_creation_s_)),
      env_(initial_db_options_.env),
@@ -185,7 +185,6 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname,
      mutex_(stats_, immutable_db_options_.clock, DB_MUTEX_WAIT_MICROS,
             immutable_db_options_.use_adaptive_mutex),
 #endif  // COERCE_CONTEXT_SWITCH
-      default_cf_handle_(nullptr),
      error_handler_(this, immutable_db_options_, &mutex_),
      event_logger_(immutable_db_options_.info_log.get()),
      max_total_in_memory_state_(0),
@@ -194,45 +193,15 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname,
          file_options_, immutable_db_options_)),
      seq_per_batch_(seq_per_batch),
      batch_per_txn_(batch_per_txn),
-      next_job_id_(1),
-      shutting_down_(false),
-      reject_new_background_jobs_(false),
-      db_lock_(nullptr),
-      manual_compaction_paused_(false),
      bg_cv_(&mutex_),
-      logfile_number_(0),
-      log_dir_synced_(false),
-      log_empty_(true),
-      persist_stats_cf_handle_(nullptr),
-      log_sync_cv_(&log_write_mutex_),
-      total_log_size_(0),
-      is_snapshot_supported_(true),
+      wal_sync_cv_(&wal_write_mutex_),
      write_buffer_manager_(immutable_db_options_.write_buffer_manager.get()),
      write_thread_(immutable_db_options_),
      nonmem_write_thread_(immutable_db_options_),
      write_controller_(mutable_db_options_.delayed_write_rate),
-      last_batch_group_size_(0),
-      unscheduled_flushes_(0),
-      unscheduled_compactions_(0),
-      bg_bottom_compaction_scheduled_(0),
-      bg_compaction_scheduled_(0),
-      num_running_compactions_(0),
-      bg_flush_scheduled_(0),
-      num_running_flushes_(0),
-      bg_purge_scheduled_(0),
-      disable_delete_obsolete_files_(0),
-      pending_purge_obsolete_files_(0),
      delete_obsolete_files_last_run_(immutable_db_options_.clock->NowMicros()),
-      has_unpersisted_data_(false),
-      unable_to_release_oldest_log_(false),
-      num_running_ingest_file_(0),
      wal_manager_(immutable_db_options_, file_options_, io_tracer_,
                   seq_per_batch),
-      bg_work_paused_(0),
-      bg_compaction_paused_(0),
-      refitting_level_(false),
-      opened_successfully_(false),
-      periodic_task_scheduler_(),
      two_write_queues_(options.two_write_queues),
      manual_wal_flush_(options.manual_wal_flush),
      // last_sequencee_ is always maintained by the main queue that also writes
@@ -250,14 +219,11 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname,
      // requires a custom gc for compaction, we use that to set use_custom_gc_
      // as well.
      use_custom_gc_(seq_per_batch),
-      shutdown_initiated_(false),
      own_sfm_(options.sst_file_manager == nullptr),
-      closed_(false),
      atomic_flush_install_cv_(&mutex_),
      blob_callback_(immutable_db_options_.sst_file_manager.get(), &mutex_,
                     &error_handler_, &event_logger_,
-                     immutable_db_options_.listeners, dbname_),
-      lock_wal_count_(0) {
+                     immutable_db_options_.listeners, dbname_) {
  // !batch_per_trx_ implies seq_per_batch_ because it is only unset for
  // WriteUnprepared, which should use seq_per_batch_.
  assert(batch_per_txn_ || seq_per_batch_);
@@ -287,6 +253,9 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname,
  periodic_task_functions_.emplace(
      PeriodicTaskType::kRecordSeqnoTime,
      [this]() { this->RecordSeqnoToTimeMapping(); });
+  periodic_task_functions_.emplace(
+      PeriodicTaskType::kTriggerCompaction,
+      [this]() { this->TriggerPeriodicCompaction(); });

  versions_.reset(new VersionSet(
      dbname_, &immutable_db_options_, file_options_, table_cache_.get(),
@@ -636,8 +605,8 @@ Status DBImpl::CloseHelper() {
    mutex_.Lock();
  }
  {
-    InstrumentedMutexLock lock(&log_write_mutex_);
-    for (auto l : logs_to_free_) {
+    InstrumentedMutexLock lock(&wal_write_mutex_);
+    for (auto l : wals_to_free_) {
      delete l;
    }
    for (auto& log : logs_) {
@@ -821,7 +790,8 @@ Status DBImpl::StartPeriodicTaskScheduler() {
    Status s = periodic_task_scheduler_.Register(
        PeriodicTaskType::kDumpStats,
        periodic_task_functions_.at(PeriodicTaskType::kDumpStats),
-        mutable_db_options_.stats_dump_period_sec);
+        mutable_db_options_.stats_dump_period_sec,
+        /*run_immediately=*/true);
    if (!s.ok()) {
      return s;
    }
@@ -830,7 +800,8 @@ Status DBImpl::StartPeriodicTaskScheduler() {
    Status s = periodic_task_scheduler_.Register(
        PeriodicTaskType::kPersistStats,
        periodic_task_functions_.at(PeriodicTaskType::kPersistStats),
-        mutable_db_options_.stats_persist_period_sec);
+        mutable_db_options_.stats_persist_period_sec,
+        /*run_immediately=*/true);
    if (!s.ok()) {
      return s;
    }
@@ -838,7 +809,15 @@ Status DBImpl::StartPeriodicTaskScheduler() {

  Status s = periodic_task_scheduler_.Register(
      PeriodicTaskType::kFlushInfoLog,
-      periodic_task_functions_.at(PeriodicTaskType::kFlushInfoLog));
+      periodic_task_functions_.at(PeriodicTaskType::kFlushInfoLog),
+      /*run_immediately=*/true);
+
+  if (s.ok()) {
+    s = periodic_task_scheduler_.Register(
+        PeriodicTaskType::kTriggerCompaction,
+        periodic_task_functions_.at(PeriodicTaskType::kTriggerCompaction),
+        /*run_immediately=*/false);
+  }

  return s;
 }
@@ -889,7 +868,7 @@ Status DBImpl::RegisterRecordSeqnoTimeWorker() {
    s = periodic_task_scheduler_.Register(
        PeriodicTaskType::kRecordSeqnoTime,
        periodic_task_functions_.at(PeriodicTaskType::kRecordSeqnoTime),
-        seqno_time_cadence);
+        seqno_time_cadence, /*run_immediately=*/true);
  }

  return s;
@@ -1180,11 +1159,11 @@ Status DBImpl::TablesRangeTombstoneSummary(ColumnFamilyHandle* column_family,

 void DBImpl::ScheduleBgLogWriterClose(JobContext* job_context) {
  mutex_.AssertHeld();
-  if (!job_context->logs_to_free.empty()) {
-    for (auto l : job_context->logs_to_free) {
+  if (!job_context->wals_to_free.empty()) {
+    for (auto l : job_context->wals_to_free) {
      AddToLogsToFreeQueue(l);
    }
-    job_context->logs_to_free.clear();
+    job_context->wals_to_free.clear();
  }
 }

@@ -1399,7 +1378,7 @@ Status DBImpl::SetDBOptions(
        s = periodic_task_scheduler_.Register(
            PeriodicTaskType::kDumpStats,
            periodic_task_functions_.at(PeriodicTaskType::kDumpStats),
-            new_options.stats_dump_period_sec);
+            new_options.stats_dump_period_sec, /*run_immediately=*/true);
      }
      if (new_options.max_total_wal_size !=
          mutable_db_options_.max_total_wal_size) {
@@ -1414,7 +1393,7 @@ Status DBImpl::SetDBOptions(
          s = periodic_task_scheduler_.Register(
              PeriodicTaskType::kPersistStats,
              periodic_task_functions_.at(PeriodicTaskType::kPersistStats),
-              new_options.stats_persist_period_sec);
+              new_options.stats_persist_period_sec, /*run_immediately=*/true);
        }
      }
      mutex_.Lock();
@@ -1443,7 +1422,7 @@ Status DBImpl::SetDBOptions(
        WriteThread::Writer w;
        write_thread_.EnterUnbatched(&w, &mutex_);
        if (wal_other_option_changed ||
-            total_log_size_ > GetMaxTotalWalSize()) {
+            wals_total_size_.LoadRelaxed() > GetMaxTotalWalSize()) {
          Status purge_wal_status = SwitchWAL(&write_context);
          if (!purge_wal_status.ok()) {
            ROCKS_LOG_WARN(immutable_db_options_.info_log,
@@ -1470,14 +1449,9 @@ Status DBImpl::SetDBOptions(
    ROCKS_LOG_INFO(immutable_db_options_.info_log, "SetDBOptions() succeeded");
    new_options.Dump(immutable_db_options_.info_log.get());
    if (!persist_options_status.ok()) {
-      if (immutable_db_options_.fail_if_options_file_error) {
-        s = Status::IOError(
-            "SetDBOptions() succeeded, but unable to persist options",
-            persist_options_status.ToString());
-      }
-      ROCKS_LOG_WARN(immutable_db_options_.info_log,
-                     "Unable to persist options in SetDBOptions() -- %s",
-                     persist_options_status.ToString().c_str());
+      s = Status::IOError(
+          "SetDBOptions() succeeded, but unable to persist options",
+          persist_options_status.ToString());
    }
  } else {
    ROCKS_LOG_WARN(immutable_db_options_.info_log, "SetDBOptions failed");
@@ -1512,8 +1486,8 @@ Status DBImpl::FlushWAL(const WriteOptions& write_options, bool sync) {
  if (manual_wal_flush_) {
    IOStatus io_s;
    {
-      // We need to lock log_write_mutex_ since logs_ might change concurrently
-      InstrumentedMutexLock wl(&log_write_mutex_);
+      // We need to lock wal_write_mutex_ since logs_ might change concurrently
+      InstrumentedMutexLock wl(&wal_write_mutex_);
      log::Writer* cur_log_writer = logs_.back().writer;
      io_s = cur_log_writer->WriteBuffer(write_options);
    }
@@ -1540,7 +1514,7 @@ Status DBImpl::FlushWAL(const WriteOptions& write_options, bool sync) {
 }

 bool DBImpl::WALBufferIsEmpty() {
-  InstrumentedMutexLock l(&log_write_mutex_);
+  InstrumentedMutexLock l(&wal_write_mutex_);
  log::Writer* cur_log_writer = logs_.back().writer;
  auto res = cur_log_writer->BufferIsEmpty();
  return res;
@@ -1548,7 +1522,7 @@ bool DBImpl::WALBufferIsEmpty() {

 Status DBImpl::GetOpenWalSizes(std::map<uint64_t, uint64_t>& number_to_size) {
  assert(number_to_size.empty());
-  InstrumentedMutexLock l(&log_write_mutex_);
+  InstrumentedMutexLock l(&wal_write_mutex_);
  for (auto& log : logs_) {
    auto* open_file = log.writer->file();
    if (open_file) {
@@ -1590,15 +1564,15 @@ IOStatus DBImpl::SyncWalImpl(bool include_current_wal,
  uint64_t up_to_number;

  {
-    InstrumentedMutexLock l(&log_write_mutex_);
+    InstrumentedMutexLock l(&wal_write_mutex_);
    assert(!logs_.empty());

-    maybe_active_number = logfile_number_;
+    maybe_active_number = cur_wal_number_;
    up_to_number =
        include_current_wal ? maybe_active_number : maybe_active_number - 1;

    while (logs_.front().number <= up_to_number && logs_.front().IsSyncing()) {
-      log_sync_cv_.Wait();
+      wal_sync_cv_.Wait();
    }
    // First check that logs are safe to sync in background.
    if (include_current_wal &&
@@ -1622,7 +1596,7 @@ IOStatus DBImpl::SyncWalImpl(bool include_current_wal,
      }
    }

-    need_wal_dir_sync = !log_dir_synced_;
+    need_wal_dir_sync = !wal_dir_synced_;
  }

  if (include_current_wal) {
@@ -1695,7 +1669,7 @@ IOStatus DBImpl::SyncWalImpl(bool include_current_wal,
                             /*arg=*/nullptr);
  }
  {
-    InstrumentedMutexLock l(&log_write_mutex_);
+    InstrumentedMutexLock l(&wal_write_mutex_);
    for (auto* wal : wals_internally_closed) {
      // We can only modify the state of log::Writer under the mutex
      bool was_closed = wal->PublishIfClosed();
@@ -1812,9 +1786,9 @@ Status DBImpl::UnlockWAL() {

 void DBImpl::MarkLogsSynced(uint64_t up_to, bool synced_dir,
                            VersionEdit* synced_wals) {
-  log_write_mutex_.AssertHeld();
-  if (synced_dir && logfile_number_ == up_to) {
-    log_dir_synced_ = true;
+  wal_write_mutex_.AssertHeld();
+  if (synced_dir && cur_wal_number_ == up_to) {
+    wal_dir_synced_ = true;
  }
  for (auto it = logs_.begin(); it != logs_.end() && it->number <= up_to;) {
    auto& wal = *it;
@@ -1836,7 +1810,7 @@ void DBImpl::MarkLogsSynced(uint64_t up_to, bool synced_dir,
          (immutable_db_options_.background_close_inactive_wals &&
           wal.GetPreSyncSize() == wal.writer->file()->GetFlushedSize())) {
        // Fully synced
-        logs_to_free_.push_back(wal.ReleaseWriter());
+        wals_to_free_.push_back(wal.ReleaseWriter());
        it = logs_.erase(it);
      } else {
        wal.FinishSync();
@@ -1849,17 +1823,17 @@ void DBImpl::MarkLogsSynced(uint64_t up_to, bool synced_dir,
      ++it;
    }
  }
-  log_sync_cv_.SignalAll();
+  wal_sync_cv_.SignalAll();
 }

 void DBImpl::MarkLogsNotSynced(uint64_t up_to) {
-  log_write_mutex_.AssertHeld();
+  wal_write_mutex_.AssertHeld();
  for (auto it = logs_.begin(); it != logs_.end() && it->number <= up_to;
       ++it) {
    auto& wal = *it;
    wal.FinishSync();
  }
-  log_sync_cv_.SignalAll();
+  wal_sync_cv_.SignalAll();
 }

 SequenceNumber DBImpl::GetLatestSequenceNumber() const {
@@ -1895,6 +1869,69 @@ Status DBImpl::GetFullHistoryTsLow(ColumnFamilyHandle* column_family,
  return Status::OK();
 }

+Status DBImpl::GetNewestUserDefinedTimestamp(ColumnFamilyHandle* column_family,
+                                             std::string* newest_timestamp) {
+  if (newest_timestamp == nullptr) {
+    return Status::InvalidArgument("newest_timestamp is nullptr");
+  }
+  ColumnFamilyData* cfd = nullptr;
+  if (column_family == nullptr) {
+    cfd = default_cf_handle_->cfd();
+  } else {
+    auto cfh = static_cast_with_check<ColumnFamilyHandleImpl>(column_family);
+    assert(cfh != nullptr);
+    cfd = cfh->cfd();
+  }
+  assert(cfd != nullptr && cfd->user_comparator() != nullptr);
+  if (cfd->user_comparator()->timestamp_size() == 0) {
+    return Status::InvalidArgument(
+        "Timestamp is not enabled in this column family");
+  }
+  if (cfd->ioptions().persist_user_defined_timestamps) {
+    return Status::NotSupported(
+        "GetNewestUserDefinedTimestamp doesn't support the case when user"
+        "defined timestamps are persisted.");
+  }
+
+  Status status;
+  // Acquire SuperVersion
+  SuperVersion* sv = GetAndRefSuperVersion(cfd);
+  {
+    InstrumentedMutexLock l(&mutex_);
+    bool enter_write_thread = sv->mem == cfd->mem();
+    WriteThread::Writer w;
+    // Enter write thread to read the mutable memtable to avoid racing access
+    // with concurrent writes. No need to enter nonmem_write_thread_ since this
+    // call only care about memtable writes, not WAL writes.
+    if (enter_write_thread) {
+      write_thread_.EnterUnbatched(&w, &mutex_);
+      WaitForPendingWrites();
+    }
+    *newest_timestamp = sv->mem->GetNewestUDT().ToString();
+    assert(!newest_timestamp->empty() || sv->mem->IsEmpty());
+    if (enter_write_thread) {
+      write_thread_.ExitUnbatched(&w);
+    }
+  }
+  // Read from immutable memtables if nothing found in mutable memtable.
+  if (newest_timestamp->empty()) {
+    *newest_timestamp = sv->imm->GetNewestUDT().ToString();
+  }
+  // Read from SST files if no result can be found in memtables.
+  if (newest_timestamp->empty() && sv->current->GetSstFilesSize() != 0) {
+    // full_history_ts_low is used to track the exclusive upperbound of
+    // flushed user defined timestamp. So we can use it to deduce the newest
+    // timestamp in the SST files that the column family has seen.
+    Slice full_history_ts_low = sv->full_history_ts_low;
+    if (!full_history_ts_low.empty()) {
+      GetU64CutoffTsFromFullHistoryTsLow(&full_history_ts_low,
+                                         newest_timestamp);
+    }
+  }
+  ReturnAndCleanupSuperVersion(cfd, sv);
+  return status;
+}
+
 InternalIterator* DBImpl::NewInternalIterator(const ReadOptions& read_options,
                                              Arena* arena,
                                              SequenceNumber sequence,
@@ -1928,10 +1965,10 @@ void DBImpl::BackgroundCallPurge() {
  TEST_SYNC_POINT("DBImpl::BackgroundCallPurge:beforeMutexLock");
  mutex_.Lock();

-  while (!logs_to_free_queue_.empty()) {
-    assert(!logs_to_free_queue_.empty());
-    log::Writer* log_writer = *(logs_to_free_queue_.begin());
-    logs_to_free_queue_.pop_front();
+  while (!wals_to_free_queue_.empty()) {
+    assert(!wals_to_free_queue_.empty());
+    log::Writer* log_writer = *(wals_to_free_queue_.begin());
+    wals_to_free_queue_.pop_front();
    mutex_.Unlock();
    delete log_writer;
    mutex_.Lock();
@@ -3597,7 +3634,7 @@ Status DBImpl::CreateColumnFamilyImpl(const ReadOptions& read_options,
    edit.AddColumnFamily(column_family_name);
    uint32_t new_id = versions_->GetColumnFamilySet()->GetNextColumnFamilyID();
    edit.SetColumnFamily(new_id);
-    edit.SetLogNumber(logfile_number_);
+    edit.SetLogNumber(cur_wal_number_);
    edit.SetComparatorName(cf_options.comparator->Name());
    edit.SetPersistUserDefinedTimestamps(
        cf_options.persist_user_defined_timestamps);
@@ -3796,6 +3833,14 @@ bool DBImpl::KeyMayExist(const ReadOptions& read_options,
  return s.ok() || s.IsIncomplete();
 }

+std::unique_ptr<MultiScan> DBImpl::NewMultiScan(
+    const ReadOptions& _read_options, ColumnFamilyHandle* column_family,
+    const MultiScanArgs& scan_opts) {
+  std::unique_ptr<MultiScan> ms_iter = std::make_unique<MultiScan>(
+      _read_options, scan_opts, this, column_family);
+  return ms_iter;
+}
+
 Iterator* DBImpl::NewIterator(const ReadOptions& _read_options,
                              ColumnFamilyHandle* column_family) {
  if (_read_options.io_activity != Env::IOActivity::kUnknown &&
@@ -3852,11 +3897,14 @@ Iterator* DBImpl::NewIterator(const ReadOptions& _read_options,

    auto iter = new ForwardIterator(this, read_options, cfd, sv,
                                    /* allow_unprepared_value */ true);
-    result = NewDBIterator(
-        env_, read_options, cfd->ioptions(), sv->mutable_cf_options,
-        cfd->user_comparator(), iter, sv->current, kMaxSequenceNumber,
-        sv->mutable_cf_options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */, cfh);
+    // TODO(cbi): Add support for `memtable_op_scan_flush_trigger` for tailing
+    // iterator. This requires refreshing DBIter's pointer to active_mem when
+    // tailing iterator refreshes to new memtable internally.
+    result = DBIter::NewIter(env_, read_options, cfd->ioptions(),
+                             sv->mutable_cf_options, cfd->user_comparator(),
+                             iter, sv->current, kMaxSequenceNumber,
+                             /*read_callback=*/nullptr, /*active_mem=*/nullptr,
+                             cfh, /*expose_blob_index=*/false);
  } else {
    // Note: no need to consider the special case of
    // last_seq_same_as_publish_seq_==false since NewIterator is overridden in
@@ -3934,18 +3982,9 @@ ArenaWrappedDBIter* DBImpl::NewIteratorImpl(
  // Laying out the iterators in the order of being accessed makes it more
  // likely that any iterator pointer is close to the iterator it points to so
  // that they are likely to be in the same cache line and/or page.
-  ArenaWrappedDBIter* db_iter = NewArenaWrappedDbIterator(
-      env_, read_options, cfh->cfd()->ioptions(), sv->mutable_cf_options,
-      sv->current, snapshot,
-      sv->mutable_cf_options.max_sequential_skip_in_iterations,
-      sv->version_number, read_callback, cfh, expose_blob_index, allow_refresh);
-
-  InternalIterator* internal_iter = NewInternalIterator(
-      db_iter->GetReadOptions(), cfh->cfd(), sv, db_iter->GetArena(), snapshot,
-      /* allow_unprepared_value */ true, db_iter);
-  db_iter->SetIterUnderDBIter(internal_iter);
-
-  return db_iter;
+  return NewArenaWrappedDbIterator(
+      env_, read_options, cfh, sv, snapshot, read_callback, this,
+      expose_blob_index, allow_refresh, /*allow_mark_memtable_for_flush=*/true);
 }

 std::unique_ptr<Iterator> DBImpl::NewCoalescingIterator(
@@ -4069,14 +4108,12 @@ Status DBImpl::NewIterators(
      auto iter = new ForwardIterator(this, read_options, cf_sv_pair.cfd,
                                      cf_sv_pair.super_version,
                                      /* allow_unprepared_value */ true);
-      iterators->push_back(
-          NewDBIterator(env_, read_options, cf_sv_pair.cfd->ioptions(),
-                        cf_sv_pair.super_version->mutable_cf_options,
-                        cf_sv_pair.cfd->user_comparator(), iter,
-                        cf_sv_pair.super_version->current, kMaxSequenceNumber,
-                        cf_sv_pair.super_version->mutable_cf_options
-                            .max_sequential_skip_in_iterations,
-                        nullptr /*read_callback*/, cf_sv_pair.cfh));
+      iterators->push_back(DBIter::NewIter(
+          env_, read_options, cf_sv_pair.cfd->ioptions(),
+          cf_sv_pair.super_version->mutable_cf_options,
+          cf_sv_pair.cfd->user_comparator(), iter,
+          cf_sv_pair.super_version->current, kMaxSequenceNumber,
+          nullptr /*read_callback*/, /*active_mem=*/nullptr, cf_sv_pair.cfh));
    }
  } else {
    for (const auto& cf_sv_pair : cf_sv_pairs) {
@@ -4308,7 +4345,7 @@ void DBImpl::ReleaseSnapshot(const Snapshot* s) {
    CfdList cf_scheduled;
    if (oldest_snapshot > bottommost_files_mark_threshold_) {
      for (auto* cfd : *versions_->GetColumnFamilySet()) {
-        if (!cfd->ioptions().allow_ingest_behind) {
+        if (!cfd->AllowIngestBehind()) {
          cfd->current()->storage_info()->UpdateOldestSnapshot(
              oldest_snapshot, /*allow_ingest_behind=*/false);
          if (!cfd->current()
@@ -4328,8 +4365,7 @@ void DBImpl::ReleaseSnapshot(const Snapshot* s) {
      // inaccurate.
      SequenceNumber new_bottommost_files_mark_threshold = kMaxSequenceNumber;
      for (auto* cfd : *versions_->GetColumnFamilySet()) {
-        if (CfdListContains(cf_scheduled, cfd) ||
-            cfd->ioptions().allow_ingest_behind) {
+        if (CfdListContains(cf_scheduled, cfd) || cfd->AllowIngestBehind()) {
          continue;
        }
        new_bottommost_files_mark_threshold = std::min(
@@ -4992,7 +5028,6 @@ void DBImpl::GetColumnFamilyMetaData(ColumnFamilyHandle* column_family,
  assert(column_family);
  auto* cfd =
      static_cast_with_check<ColumnFamilyHandleImpl>(column_family)->cfd();
-  auto* sv = GetAndRefSuperVersion(cfd);
  {
    // Without mutex, Version::GetColumnFamilyMetaData will have data race
    // with Compaction::MarkFilesBeingCompacted. One solution is to use mutex,
@@ -5004,9 +5039,8 @@ void DBImpl::GetColumnFamilyMetaData(ColumnFamilyHandle* column_family,
    // DB::GetColumnFamilyMetaData is not called frequently, the regression
    // should not be big. We still need to keep an eye on it.
    InstrumentedMutexLock l(&mutex_);
-    sv->current->GetColumnFamilyMetaData(cf_meta);
+    cfd->current()->GetColumnFamilyMetaData(cf_meta);
  }
-  ReturnAndCleanupSuperVersion(cfd, sv);
 }

 void DBImpl::GetAllColumnFamilyMetaData(
@@ -5020,85 +5054,6 @@ void DBImpl::GetAllColumnFamilyMetaData(
  }
 }

-Status DBImpl::CheckConsistency() {
-  mutex_.AssertHeld();
-  std::vector<LiveFileMetaData> metadata;
-  versions_->GetLiveFilesMetaData(&metadata);
-  TEST_SYNC_POINT("DBImpl::CheckConsistency:AfterGetLiveFilesMetaData");
-
-  std::string corruption_messages;
-
-  if (immutable_db_options_.skip_checking_sst_file_sizes_on_db_open) {
-    // Instead of calling GetFileSize() for each expected file, call
-    // GetChildren() for the DB directory and check that all expected files
-    // are listed, without checking their sizes.
-    // Since sst files might be in different directories, do it for each
-    // directory separately.
-    std::map<std::string, std::vector<std::string>> files_by_directory;
-    for (const auto& md : metadata) {
-      // md.name has a leading "/". Remove it.
-      std::string fname = md.name;
-      if (!fname.empty() && fname[0] == '/') {
-        fname = fname.substr(1);
-      }
-      files_by_directory[md.db_path].push_back(fname);
-    }
-
-    IOOptions io_opts;
-    io_opts.do_not_recurse = true;
-    for (const auto& dir_files : files_by_directory) {
-      std::string directory = dir_files.first;
-      std::vector<std::string> existing_files;
-      Status s = fs_->GetChildren(directory, io_opts, &existing_files,
-                                  /*IODebugContext*=*/nullptr);
-      if (!s.ok()) {
-        corruption_messages +=
-            "Can't list files in " + directory + ": " + s.ToString() + "\n";
-        continue;
-      }
-      std::sort(existing_files.begin(), existing_files.end());
-
-      for (const std::string& fname : dir_files.second) {
-        if (!std::binary_search(existing_files.begin(), existing_files.end(),
-                                fname) &&
-            !std::binary_search(existing_files.begin(), existing_files.end(),
-                                Rocks2LevelTableFileName(fname))) {
-          corruption_messages +=
-              "Missing sst file " + fname + " in " + directory + "\n";
-        }
-      }
-    }
-  } else {
-    for (const auto& md : metadata) {
-      // md.name has a leading "/".
-      std::string file_path = md.db_path + md.name;
-
-      uint64_t fsize = 0;
-      TEST_SYNC_POINT("DBImpl::CheckConsistency:BeforeGetFileSize");
-      Status s = env_->GetFileSize(file_path, &fsize);
-      if (!s.ok() &&
-          env_->GetFileSize(Rocks2LevelTableFileName(file_path), &fsize).ok()) {
-        s = Status::OK();
-      }
-      if (!s.ok()) {
-        corruption_messages +=
-            "Can't access " + md.name + ": " + s.ToString() + "\n";
-      } else if (fsize != md.size) {
-        corruption_messages += "Sst file size mismatch: " + file_path +
-                               ". Size recorded in manifest " +
-                               std::to_string(md.size) + ", actual size " +
-                               std::to_string(fsize) + "\n";
-      }
-    }
-  }
-
-  if (corruption_messages.size() == 0) {
-    return Status::OK();
-  } else {
-    return Status::Corruption(corruption_messages);
-  }
-}
-
 Status DBImpl::GetDbIdentity(std::string& identity) const {
  identity.assign(db_id_);
  return Status::OK();
@@ -5435,12 +5390,7 @@ Status DBImpl::WriteOptionsFile(const WriteOptions& write_options,
  if (!s.ok()) {
    ROCKS_LOG_WARN(immutable_db_options_.info_log,
                   "Unnable to persist options -- %s", s.ToString().c_str());
-    if (immutable_db_options_.fail_if_options_file_error) {
-      s = Status::IOError("Unable to persist options.", s.ToString().c_str());
-    } else {
-      // Ignore error
-      s = Status::OK();
-    }
+    s = Status::IOError("Unable to persist options.", s.ToString().c_str());
  }

  // Restore lock if appropriate
@@ -5810,10 +5760,6 @@ Status DBImpl::IngestExternalFiles(
  for (const auto& arg : args) {
    const IngestExternalFileOptions& ingest_opts = arg.options;
    if (ingest_opts.ingest_behind) {
-      if (!immutable_db_options_.allow_ingest_behind) {
-        return Status::InvalidArgument(
-            "can't ingest_behind file in DB with allow_ingest_behind=false");
-      }
      auto ucmp = arg.column_family->GetComparator();
      assert(ucmp);
      if (ucmp->timestamp_size() > 0) {
@@ -5821,6 +5767,14 @@ Status DBImpl::IngestExternalFiles(
            "Column family with user-defined "
            "timestamps enabled doesn't support ingest behind.");
      }
+
+      if (!static_cast<ColumnFamilyHandleImpl*>(arg.column_family)
+               ->cfd()
+               ->AllowIngestBehind()) {
+        return Status::InvalidArgument(
+            "Can't ingest_behind file in ColumnFamily %s with "
+            "cf_allow_ingest_behind=false");
+      }
    }
    if (arg.atomic_replace_range.has_value()) {
      if (ingest_opts.ingest_behind) {
@@ -6055,18 +6009,19 @@ Status DBImpl::IngestExternalFiles(
      // mutex when persisting MANIFEST file, and the snapshots taken during
      // that period will not be stable if VersionSet last seqno is updated
      // before LogAndApply.
-      int consumed_seqno_count =
-          ingestion_jobs[0].ConsumedSequenceNumbersCount();
+      SequenceNumber max_assigned_seqno =
+          ingestion_jobs[0].MaxAssignedSequenceNumber();
      for (size_t i = 1; i != num_cfs; ++i) {
-        consumed_seqno_count =
-            std::max(consumed_seqno_count,
-                     ingestion_jobs[i].ConsumedSequenceNumbersCount());
+        max_assigned_seqno = std::max(
+            max_assigned_seqno, ingestion_jobs[i].MaxAssignedSequenceNumber());
      }
-      if (consumed_seqno_count > 0) {
+      if (max_assigned_seqno > 0) {
        const SequenceNumber last_seqno = versions_->LastSequence();
-        versions_->SetLastAllocatedSequence(last_seqno + consumed_seqno_count);
-        versions_->SetLastPublishedSequence(last_seqno + consumed_seqno_count);
-        versions_->SetLastSequence(last_seqno + consumed_seqno_count);
+        if (max_assigned_seqno > last_seqno) {
+          versions_->SetLastAllocatedSequence(max_assigned_seqno);
+          versions_->SetLastPublishedSequence(max_assigned_seqno);
+          versions_->SetLastSequence(max_assigned_seqno);
+        }
      }
    }

@@ -6865,6 +6820,35 @@ void DBImpl::RecordSeqnoToTimeMapping() {
  sv_context.Clean();
 }

+void DBImpl::TriggerPeriodicCompaction() {
+  TEST_SYNC_POINT("DBImpl::TriggerPeriodicCompaction:StartRunning");
+  {
+    InstrumentedMutexLock l(&mutex_);
+    ROCKS_LOG_INFO(immutable_db_options_.info_log,
+                   "Running the periodic task to trigger compactions.");
+
+    for (ColumnFamilyData* cfd : *versions_->GetColumnFamilySet()) {
+      if (cfd->IsDropped()) {
+        continue;
+      }
+      if (cfd->GetLatestCFOptions().periodic_compaction_seconds &&
+          !cfd->queued_for_compaction()) {
+        cfd->current()->storage_info()->ComputeCompactionScore(
+            cfd->ioptions(), cfd->GetLatestMutableCFOptions());
+        EnqueuePendingCompaction(cfd);
+        if (cfd->queued_for_compaction()) {
+          ROCKS_LOG_INFO(immutable_db_options_.info_log,
+                         "Periodic task to trigger compaction queued Column "
+                         "family [%s] for compaction.",
+                         cfd->GetName().c_str());
+        }
+      }
+    }
+    MaybeScheduleFlushOrCompaction();
+    bg_cv_.SignalAll();
+  }
+}
+
 void DBImpl::TrackOrUntrackFiles(
    const std::vector<std::string>& existing_data_files, bool track) {
  auto sfm = static_cast_with_check<SstFileManagerImpl>(
@@ -173,10 +173,10 @@ struct DBOpenLogRecordReadReporter : public log::Reader::Reporter {

  void OldLogRecord(size_t bytes) override;

-  uint64_t GetCorruptedLogNumber() const { return corrupted_log_number_; }
+  uint64_t GetCorruptedLogNumber() const { return corrupted_wal_number_; }

 private:
-  uint64_t corrupted_log_number_ = kMaxSequenceNumber;
+  uint64_t corrupted_wal_number_ = kMaxSequenceNumber;
 };

 // While DB is the public interface of RocksDB, and DBImpl is the actual
@@ -256,6 +256,10 @@ class DBImpl : public DB {
  Status WriteWithCallback(const WriteOptions& options, WriteBatch* updates,
                           UserWriteCallback* user_write_cb) override;

+  Status IngestWriteBatchWithIndex(
+      const WriteOptions& options,
+      std::shared_ptr<WriteBatchWithIndex> wbwi) override;
+
  using DB::Get;
  Status Get(const ReadOptions& _read_options,
             ColumnFamilyHandle* column_family, const Slice& key,
@@ -379,6 +383,11 @@ class DBImpl : public DB {
                      const std::vector<ColumnFamilyHandle*>& column_families,
                      std::vector<Iterator*>* iterators) override;

+  using DB::NewMultiScan;
+  std::unique_ptr<MultiScan> NewMultiScan(
+      const ReadOptions& _read_options, ColumnFamilyHandle* column_family,
+      const MultiScanArgs& scan_opts) override;
+
  const Snapshot* GetSnapshot() override;
  void ReleaseSnapshot(const Snapshot* snapshot) override;

@@ -497,6 +506,9 @@ class DBImpl : public DB {
  Status GetFullHistoryTsLow(ColumnFamilyHandle* column_family,
                             std::string* ts_low) override;

+  Status GetNewestUserDefinedTimestamp(ColumnFamilyHandle* column_family,
+                                       std::string* newest_timestamp) override;
+
  Status GetDbIdentity(std::string& identity) const override;

  virtual Status GetDbIdentityFromIdentityFile(const IOOptions& opts,
@@ -530,11 +542,11 @@ class DBImpl : public DB {

  // Get the known flushed sizes of WALs that might still be written to
  // or have pending sync.
-  // NOTE: unlike alive_log_files_, this function includes WALs that might
+  // NOTE: unlike alive_wal_files_, this function includes WALs that might
  // be obsolete (but not obsolete to a pending Checkpoint) and not yet fully
  // synced.
  Status GetOpenWalSizes(std::map<uint64_t, uint64_t>& number_to_size);
-  Status GetCurrentWalFile(std::unique_ptr<WalFile>* current_log_file) override;
+  Status GetCurrentWalFile(std::unique_ptr<WalFile>* current_wal_file) override;
  Status GetCreationTimeOfOldestFile(uint64_t* creation_time) override;

  Status GetUpdatesSince(
@@ -792,10 +804,6 @@ class DBImpl : public DB {
  // being detected.
  const Snapshot* GetSnapshotForWriteConflictBoundary();

-  // checks if all live files exist on file system and that their file sizes
-  // match to our in-memory records
-  virtual Status CheckConsistency();
-
  // max_file_num_to_ignore allows bottom level compaction to filter out newly
  // compacted SST files. Setting max_file_num_to_ignore to kMaxUint64 will
  // disable the filtering
@@ -1068,7 +1076,7 @@ class DBImpl : public DB {

  void AddToLogsToFreeQueue(log::Writer* log_writer) {
    mutex_.AssertHeld();
-    logs_to_free_queue_.push_back(log_writer);
+    wals_to_free_queue_.push_back(log_writer);
  }

  void AddSuperVersionsToFreeQueue(SuperVersion* sv) {
@@ -1078,10 +1086,7 @@ class DBImpl : public DB {
  void SetSnapshotChecker(SnapshotChecker* snapshot_checker);

  // Fill JobContext with snapshot information needed by flush and compaction.
-  void GetSnapshotContext(JobContext* job_context,
-                          std::vector<SequenceNumber>* snapshot_seqs,
-                          SequenceNumber* earliest_write_conflict_snapshot,
-                          SnapshotChecker** snapshot_checker);
+  void InitSnapshotContext(JobContext* job_context);

  // Not thread-safe.
  void SetRecoverableStatePreReleaseCallback(PreReleaseCallback* callback);
@@ -1133,7 +1138,7 @@ class DBImpl : public DB {
  bool TEST_UnableToReleaseOldestLog() { return unable_to_release_oldest_log_; }

  bool TEST_IsLogGettingFlushed() {
-    return alive_log_files_.begin()->getting_flushed;
+    return alive_wal_files_.begin()->getting_flushed;
  }

  Status TEST_SwitchMemtable(ColumnFamilyData* cfd = nullptr);
@@ -1213,7 +1218,9 @@ class DBImpl : public DB {

  uint64_t TEST_LogfileNumber();

-  uint64_t TEST_total_log_size() const { return total_log_size_; }
+  uint64_t TEST_wals_total_size() const {
+    return wals_total_size_.LoadRelaxed();
+  }

  void TEST_GetAllBlockCaches(std::unordered_set<const Cache*>* cache_set);

@@ -1275,6 +1282,12 @@ class DBImpl : public DB {
  // For the background timer job
  void RecordSeqnoToTimeMapping();

+  // Compactions rely on an event triggers like flush/compaction/SetOptions.
+  // We need to trigger periodic compactions even when there is no such trigger.
+  // This function checks and schedules available compactions and will run
+  // periodically.
+  void TriggerPeriodicCompaction();
+
  // REQUIRES: DB mutex held
  std::pair<SequenceNumber, uint64_t> GetSeqnoToTimeSample() const;

@@ -1371,16 +1384,19 @@ class DBImpl : public DB {

  // State below is protected by mutex_
  // With two_write_queues enabled, some of the variables that accessed during
-  // WriteToWAL need different synchronization: log_empty_, alive_log_files_,
-  // logs_, logfile_number_. Refer to the definition of each variable below for
+  // WriteToWAL need different synchronization: wal_empty_, alive_wal_files_,
+  // logs_, cur_wal_number_. Refer to the definition of each variable below for
  // more description.
  //
+  // Protects access to most ColumnFamilyData methods, see more in comment for
+  // each method.
+  //
  // `mutex_` can be a hot lock in some workloads, so it deserves dedicated
  // cachelines.
  mutable CacheAlignedInstrumentedMutex mutex_;

-  ColumnFamilyHandleImpl* default_cf_handle_;
-  InternalStats* default_cf_internal_stats_;
+  ColumnFamilyHandleImpl* default_cf_handle_ = nullptr;
+  InternalStats* default_cf_internal_stats_ = nullptr;

  // table_cache_ provides its own synchronization
  std::shared_ptr<Cache> table_cache_;
@@ -1392,7 +1408,7 @@ class DBImpl : public DB {

  // only used for dynamically adjusting max_total_wal_size. it is a sum of
  // [write_buffer_size * max_write_buffer_number] over all column families
-  std::atomic<uint64_t> max_total_in_memory_state_;
+  std::atomic<uint64_t> max_total_in_memory_state_ = 0;

  // The options to access storage files
  const FileOptions file_options_;
@@ -1419,14 +1435,14 @@ class DBImpl : public DB {

  // Each flush or compaction gets its own job id. this counter makes sure
  // they're unique
-  std::atomic<int> next_job_id_;
+  std::atomic<int> next_job_id_ = 1;

-  std::atomic<bool> shutting_down_;
+  std::atomic<bool> shutting_down_ = false;

  // No new background jobs can be queued if true. This is used to prevent new
  // background jobs from being queued after WaitForCompact() completes waiting
  // all background jobs then attempts to close when close_db_ option is true.
-  bool reject_new_background_jobs_;
+  bool reject_new_background_jobs_ = false;

  // RecoveryContext struct stores the context about version edits along
  // with corresponding column_family_data and column_family_options.
@@ -1524,11 +1540,11 @@ class DBImpl : public DB {
  // ingests `wbwi` is done.
  // @param memtable_updated Whether the same write that ingests wbwi has
  // updated memtable. This is useful for determining whether to set bg
-  // error when IngestWBWI fails.
-  Status IngestWBWI(std::shared_ptr<WriteBatchWithIndex> wbwi,
-                    const WBWIMemTable::SeqnoRange& assigned_seqno,
-                    uint64_t min_prep_log, SequenceNumber last_seqno,
-                    bool memtable_updated, bool ignore_missing_cf);
+  // error when IngestWBWIAsMemtable fails.
+  Status IngestWBWIAsMemtable(std::shared_ptr<WriteBatchWithIndex> wbwi,
+                              const WBWIMemTable::SeqnoRange& assigned_seqno,
+                              uint64_t min_prep_log, SequenceNumber last_seqno,
+                              bool memtable_updated, bool ignore_missing_cf);

  // If disable_memtable is set the application logic must guarantee that the
  // batch will still be skipped from memtable during the recovery. An excption
@@ -1558,18 +1574,17 @@ class DBImpl : public DB {
  Status WriteImpl(const WriteOptions& options, WriteBatch* updates,
                   WriteCallback* callback = nullptr,
                   UserWriteCallback* user_write_cb = nullptr,
-                   uint64_t* log_used = nullptr, uint64_t log_ref = 0,
+                   uint64_t* wal_used = nullptr, uint64_t log_ref = 0,
                   bool disable_memtable = false, uint64_t* seq_used = nullptr,
                   size_t batch_cnt = 0,
                   PreReleaseCallback* pre_release_callback = nullptr,
                   PostMemTableCallback* post_memtable_callback = nullptr,
-                   std::shared_ptr<WriteBatchWithIndex> wbwi = nullptr,
-                   uint64_t min_prep_log = 0);
+                   std::shared_ptr<WriteBatchWithIndex> wbwi = nullptr);

  Status PipelinedWriteImpl(const WriteOptions& options, WriteBatch* updates,
                            WriteCallback* callback = nullptr,
                            UserWriteCallback* user_write_cb = nullptr,
-                            uint64_t* log_used = nullptr, uint64_t log_ref = 0,
+                            uint64_t* wal_used = nullptr, uint64_t log_ref = 0,
                            bool disable_memtable = false,
                            uint64_t* seq_used = nullptr);

@@ -1596,7 +1611,7 @@ class DBImpl : public DB {
  Status WriteImplWALOnly(
      WriteThread* write_thread, const WriteOptions& options,
      WriteBatch* updates, WriteCallback* callback,
-      UserWriteCallback* user_write_cb, uint64_t* log_used,
+      UserWriteCallback* user_write_cb, uint64_t* wal_used,
      const uint64_t log_ref, uint64_t* seq_used, const size_t sub_batch_cnt,
      PreReleaseCallback* pre_release_callback, const AssignOrder assign_order,
      const PublishLastSeq publish_last_seq, const bool disable_memtable);
@@ -1757,9 +1772,9 @@ class DBImpl : public DB {
    }
  };

-  struct LogFileNumberSize {
-    explicit LogFileNumberSize(uint64_t _number) : number(_number) {}
-    LogFileNumberSize() {}
+  struct WalFileNumberSize {
+    explicit WalFileNumberSize(uint64_t _number) : number(_number) {}
+    WalFileNumberSize() {}
    void AddSize(uint64_t new_size) { size += new_size; }
    uint64_t number;
    uint64_t size = 0;
@@ -1781,6 +1796,13 @@ class DBImpl : public DB {
      if (writer->file()) {
        // TODO: plumb Env::IOActivity, Env::IOPriority
        s = writer->WriteBuffer(WriteOptions());
+        if (attempt_truncate_size < SIZE_MAX &&
+            attempt_truncate_size < writer->file()->GetFileSize()) {
+          Status s2 = writer->file()->writable_file()->Truncate(
+              attempt_truncate_size, IOOptions{}, nullptr);
+          // This is just a best effort attempt
+          s2.PermitUncheckedError();
+        }
      }
      delete writer;
      writer = nullptr;
@@ -1813,6 +1835,11 @@ class DBImpl : public DB {
      getting_synced = false;
    }

+    void SetAttemptTruncateSize(uint64_t size) {
+      assert(attempt_truncate_size == SIZE_MAX);
+      attempt_truncate_size = size;
+    }
+
    uint64_t number;
    // Visual Studio doesn't support deque's member to be noncopyable because
    // of a std::unique_ptr as a member.
@@ -1825,15 +1852,20 @@ class DBImpl : public DB {
    // to be persisted even if appends happen during sync so it can be used for
    // tracking the synced size in MANIFEST.
    uint64_t pre_sync_size = 0;
+    // When < SIZE_MAX, attempt to truncate the WAL to this size on close,
+    // because a bad entry was written to it beyond that point and it likely
+    // won't be recoverable with the bad entry.
+    uint64_t attempt_truncate_size = SIZE_MAX;
  };

-  struct LogContext {
-    explicit LogContext(bool need_sync = false)
-        : need_log_sync(need_sync), need_log_dir_sync(need_sync) {}
-    bool need_log_sync = false;
-    bool need_log_dir_sync = false;
+  struct WalContext {
+    explicit WalContext(bool need_sync = false)
+        : need_wal_sync(need_sync), need_wal_dir_sync(need_sync) {}
+    bool need_wal_sync = false;
+    bool need_wal_dir_sync = false;
    log::Writer* writer = nullptr;
-    LogFileNumberSize* log_file_number_size = nullptr;
+    WalFileNumberSize* wal_file_number_size = nullptr;
+    uint64_t prev_size = SIZE_MAX;
  };

  // PurgeFileInfo is a structure to hold information of files to be deleted in
@@ -1925,12 +1957,19 @@ class DBImpl : public DB {
  };
  struct PrepickedCompaction {
    // background compaction takes ownership of `compaction`.
+    // TODO(hx235): consider using std::shared_ptr for easier ownership
+    // management
    Compaction* compaction;
    // caller retains ownership of `manual_compaction_state` as it is reused
    // across background compactions.
    ManualCompactionState* manual_compaction_state;  // nullptr if non-manual
    // task limiter token is requested during compaction picking.
    std::unique_ptr<TaskLimiterToken> task_token;
+    // If true, `compaction` is picked temporarily to express compaction intent
+    // and will be released before re-picking a real compaction based on the
+    // updated LSM shape when thread associated with `compaction` is ready to
+    // run
+    bool need_repick;
  };

  struct CompactionArg {
@@ -2021,14 +2060,13 @@ class DBImpl : public DB {
  // Flush the in-memory write buffer to storage.  Switches to a new
  // log-file/memtable and writes a new descriptor iff successful. Then
  // installs a new super version for the column family.
-  Status FlushMemTableToOutputFile(
-      ColumnFamilyData* cfd, const MutableCFOptions& mutable_cf_options,
-      bool* madeProgress, JobContext* job_context, FlushReason flush_reason,
-      SuperVersionContext* superversion_context,
-      std::vector<SequenceNumber>& snapshot_seqs,
-      SequenceNumber earliest_write_conflict_snapshot,
-      SnapshotChecker* snapshot_checker, LogBuffer* log_buffer,
-      Env::Priority thread_pri);
+  Status FlushMemTableToOutputFile(ColumnFamilyData* cfd,
+                                   const MutableCFOptions& mutable_cf_options,
+                                   bool* madeProgress, JobContext* job_context,
+                                   FlushReason flush_reason,
+                                   SuperVersionContext* superversion_context,
+                                   LogBuffer* log_buffer,
+                                   Env::Priority thread_pri);

  // Flush the memtables of (multiple) column families to multiple files on
  // persistent storage.
@@ -2041,10 +2079,10 @@ class DBImpl : public DB {
      JobContext* job_context, LogBuffer* log_buffer, Env::Priority thread_pri);

  // REQUIRES: log_numbers are sorted in ascending order
-  // corrupted_log_found is set to true if we recover from a corrupted log file.
+  // corrupted_wal_found is set to true if we recover from a corrupted log file.
  Status RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
                         SequenceNumber* next_sequence, bool read_only,
-                         bool is_retry, bool* corrupted_log_found,
+                         bool is_retry, bool* corrupted_wal_found,
                         RecoveryContext* recovery_ctx);

  void SetupLogFilesRecovery(
@@ -2152,12 +2190,12 @@ class DBImpl : public DB {
  // log file to its actual size, thereby freeing preallocated space.
  // Return success even if truncate fails
  Status GetLogSizeAndMaybeTruncate(uint64_t wal_number, bool truncate,
-                                    LogFileNumberSize* log);
+                                    WalFileNumberSize* log);

-  // Restore alive_log_files_ and total_log_size_ after recovery.
+  // Restore alive_wal_files_ and wals_total_size_ after recovery.
  // It needs to run only when there's no flush during recovery
  // (e.g. avoid_flush_during_recovery=true). May also trigger flush
-  // in case total_log_size > max_total_wal_size.
+  // in case wals_total_size > max_total_wal_size.
  Status RestoreAliveLogFiles(const std::vector<uint64_t>& log_numbers);

  // num_bytes: for slowdown case, delay time is calculated based on
@@ -2306,7 +2344,7 @@ class DBImpl : public DB {

  // REQUIRES: mutex locked
  Status PreprocessWrite(const WriteOptions& write_options,
-                         LogContext* log_context, WriteContext* write_context);
+                         WalContext* log_context, WriteContext* write_context);

  // Merge write batches in the write group into merged_batch.
  // Returns OK if merge is successful.
@@ -2317,20 +2355,21 @@ class DBImpl : public DB {

  IOStatus WriteToWAL(const WriteBatch& merged_batch,
                      const WriteOptions& write_options,
-                      log::Writer* log_writer, uint64_t* log_used,
+                      log::Writer* log_writer, uint64_t* wal_used,
                      uint64_t* log_size,
-                      LogFileNumberSize& log_file_number_size,
+                      WalFileNumberSize& wal_file_number_size,
                      SequenceNumber sequence);

-  IOStatus WriteToWAL(const WriteThread::WriteGroup& write_group,
-                      log::Writer* log_writer, uint64_t* log_used,
-                      bool need_log_sync, bool need_log_dir_sync,
-                      SequenceNumber sequence,
-                      LogFileNumberSize& log_file_number_size);
+  IOStatus WriteGroupToWAL(const WriteThread::WriteGroup& write_group,
+                           log::Writer* log_writer, uint64_t* wal_used,
+                           bool need_wal_sync, bool need_wal_dir_sync,
+                           SequenceNumber sequence,
+                           WalFileNumberSize& wal_file_number_size);

-  IOStatus ConcurrentWriteToWAL(const WriteThread::WriteGroup& write_group,
-                                uint64_t* log_used,
-                                SequenceNumber* last_sequence, size_t seq_inc);
+  IOStatus ConcurrentWriteGroupToWAL(const WriteThread::WriteGroup& write_group,
+                                     uint64_t* wal_used,
+                                     SequenceNumber* last_sequence,
+                                     size_t seq_inc);

  // Used by WriteImpl to update bg_error_ if paranoid check is enabled.
  // Caller must hold mutex_.
@@ -2344,7 +2383,7 @@ class DBImpl : public DB {
  void WALIOStatusCheck(const IOStatus& status);

  // Used by WriteImpl to update bg_error_ in case of memtable insert error.
-  void MemTableInsertStatusCheck(const Status& memtable_insert_status);
+  void HandleMemTableInsertFailure(const Status& nonok_memtable_insert_status);

  Status CompactFilesImpl(const CompactionOptions& compact_options,
                          ColumnFamilyData* cfd, Version* version,
@@ -2429,6 +2468,8 @@ class DBImpl : public DB {
                         bool* flush_rescheduled_to_retain_udt,
                         Env::Priority thread_pri);

+  Compaction* CreateIntendedCompactionForwardedToBottomPriorityPool(
+      Compaction* c);
  bool EnoughRoomForCompaction(ColumnFamilyData* cfd,
                               const std::vector<CompactionInputFiles>& inputs,
                               bool* sfm_bookkeeping, LogBuffer* log_buffer);
@@ -2546,7 +2587,7 @@ class DBImpl : public DB {
  bool ShouldntRunManualCompaction(ManualCompactionState* m);
  bool HaveManualCompaction(ColumnFamilyData* cfd);
  bool MCOverlap(ManualCompactionState* m, ManualCompactionState* m1);
-  void UpdateDeletionCompactionStats(const std::unique_ptr<Compaction>& c);
+  void UpdateFIFOCompactionStatus(const std::unique_ptr<Compaction>& c);

  // May open and read table files for table property.
  // Should not be called while holding mutex_.
@@ -2696,8 +2737,13 @@ class DBImpl : public DB {
      const std::vector<ColumnFamilyHandle*>& column_families,
      ErrorIteratorFuncType error_iterator_func);

+  bool ShouldPickCompaction(bool is_prepicked,
+                            const PrepickedCompaction* prepicked_compaction);
+
+  void ResetBottomPriCompactionIntent(ColumnFamilyData* cfd,
+                                      std::unique_ptr<Compaction>& c);
  // Lock over the persistent DB state.  Non-nullptr iff successfully acquired.
-  FileLock* db_lock_;
+  FileLock* db_lock_ = nullptr;

  // Guards changes to DB and CF options to ensure consistency between
  // * In-memory options objects
@@ -2711,20 +2757,20 @@ class DBImpl : public DB {
  // Guards reads and writes to in-memory stats_history_.
  InstrumentedMutex stats_history_mutex_;

-  // In addition to mutex_, log_write_mutex_ protects writes to logs_ and
-  // logfile_number_. With two_write_queues it also protects alive_log_files_,
-  // and log_empty_. Refer to the definition of each variable below for more
+  // In addition to mutex_, wal_write_mutex_ protects writes to logs_ and
+  // cur_wal_number_. With two_write_queues it also protects alive_wal_files_,
+  // and wal_empty_. Refer to the definition of each variable below for more
  // details.
-  // Note: to avoid deadlock, if needed to acquire both log_write_mutex_ and
-  // mutex_, the order should be first mutex_ and then log_write_mutex_.
-  InstrumentedMutex log_write_mutex_;
+  // Note: to avoid deadlock, if needed to acquire both wal_write_mutex_ and
+  // mutex_, the order should be first mutex_ and then wal_write_mutex_.
+  InstrumentedMutex wal_write_mutex_;

  // If zero, manual compactions are allowed to proceed. If non-zero, manual
  // compactions may still be running, but will quickly fail with
  // `Status::Incomplete`. The value indicates how many threads have paused
  // manual compactions. It is accessed in read mode outside the DB mutex in
  // compaction code paths.
-  std::atomic<int> manual_compaction_paused_;
+  std::atomic<int> manual_compaction_paused_ = false;

  // This condition variable is signaled on these conditions:
  // * whenever bg_compaction_scheduled_ goes down to 0
@@ -2740,106 +2786,114 @@ class DBImpl : public DB {
  // * whenever SetOptions successfully updates options.
  // * whenever a column family is dropped.
  InstrumentedCondVar bg_cv_;
-  // Writes are protected by locking both mutex_ and log_write_mutex_, and reads
-  // must be under either mutex_ or log_write_mutex_. Since after ::Open,
-  // logfile_number_ is currently updated only in write_thread_, it can be read
+
+  ColumnFamilyHandleImpl* persist_stats_cf_handle_ = nullptr;
+
+  bool persistent_stats_cfd_exists_ = true;
+
+  // Writes are protected by locking both mutex_ and wal_write_mutex_, and reads
+  // must be under either mutex_ or wal_write_mutex_. Since after ::Open,
+  // cur_wal_number_ is currently updated only in write_thread_, it can be read
  // from the same write_thread_ without any locks.
-  uint64_t logfile_number_;
+  uint64_t cur_wal_number_ = 0;
+
  // Log files that we can recycle. Must be protected by db mutex_.
-  std::deque<uint64_t> log_recycle_files_;
+  std::deque<uint64_t> wal_recycle_files_;
+
  // The minimum log file number taht can be recycled, if log recycling is
  // enabled. This is used to ensure that log files created by previous
  // instances of the database are not recycled, as we cannot be sure they
  // were created in the recyclable format.
-  uint64_t min_log_number_to_recycle_;
-  // Protected by log_write_mutex_.
-  bool log_dir_synced_;
-  // Without two_write_queues, read and writes to log_empty_ are protected by
+  uint64_t min_wal_number_to_recycle_ = 0;
+
+  // Protected by wal_write_mutex_.
+  bool wal_dir_synced_ = false;
+
+  // Without two_write_queues, read and writes to wal_empty_ are protected by
  // mutex_. Since it is currently updated/read only in write_thread_, it can be
  // accessed from the same write_thread_ without any locks. With
  // two_write_queues writes, where it can be updated in different threads,
-  // read and writes are protected by log_write_mutex_ instead. This is to avoid
-  // expensive mutex_ lock during WAL write, which update log_empty_.
-  bool log_empty_;
-
-  ColumnFamilyHandleImpl* persist_stats_cf_handle_;
-
-  bool persistent_stats_cfd_exists_ = true;
+  // read and writes are protected by wal_write_mutex_ instead. This is to avoid
+  // expensive mutex_ lock during WAL write, which update wal_empty_.
+  bool wal_empty_ = true;

  // The current WAL file and those that have not been found obsolete from
  // memtable flushes. A WAL not on this list might still be pending writer
-  // flush and/or sync and close and might still be in logs_. alive_log_files_
-  // is protected by mutex_ and log_write_mutex_ with details as follows:
+  // flush and/or sync and close and might still be in logs_. alive_wal_files_
+  // is protected by mutex_ and wal_write_mutex_ with details as follows:
  // 1. read by FindObsoleteFiles() which can be called in either application
-  //    thread or RocksDB bg threads, both mutex_ and log_write_mutex_ are
+  //    thread or RocksDB bg threads, both mutex_ and wal_write_mutex_ are
  //    held.
-  // 2. pop_front() by FindObsoleteFiles(), both mutex_ and log_write_mutex_
+  // 2. pop_front() by FindObsoleteFiles(), both mutex_ and wal_write_mutex_
  //    are held.
  // 3. push_back() by DBImpl::Open() and DBImpl::RestoreAliveLogFiles()
  //    (actually called by Open()), only mutex_ is held because at this point,
  //    the DB::Open() call has not returned success to application, and the
  //    only other thread(s) that can conflict are bg threads calling
-  //    FindObsoleteFiles() which ensure that both mutex_ and log_write_mutex_
-  //    are held when accessing alive_log_files_.
+  //    FindObsoleteFiles() which ensure that both mutex_ and wal_write_mutex_
+  //    are held when accessing alive_wal_files_.
  // 4. read by DBImpl::Open() is protected by mutex_.
-  // 5. push_back() by SwitchMemtable(). Both mutex_ and log_write_mutex_ are
+  // 5. push_back() by SwitchMemtable(). Both mutex_ and wal_write_mutex_ are
  //    held. This is done by the write group leader. Note that in the case of
  //    two-write-queues, another WAL-only write thread can be writing to the
  //    WAL concurrently. See 9.
-  // 6. read by SwitchWAL() with both mutex_ and log_write_mutex_ held. This is
+  // 6. read by SwitchWAL() with both mutex_ and wal_write_mutex_ held. This is
  //    done by write group leader.
  // 7. read by ConcurrentWriteToWAL() by the write group leader in the case of
-  //    two-write-queues. Only log_write_mutex_ is held to protect concurrent
+  //    two-write-queues. Only wal_write_mutex_ is held to protect concurrent
  //    pop_front() by FindObsoleteFiles().
-  // 8. read by PreprocessWrite() by the write group leader. log_write_mutex_
+  // 8. read by PreprocessWrite() by the write group leader. wal_write_mutex_
  //    is held to protect the data structure from concurrent pop_front() by
  //    FindObsoleteFiles().
  // 9. read by ConcurrentWriteToWAL() by a WAL-only write thread in the case
-  //    of two-write-queues. Only log_write_mutex_ is held. This suffices to
+  //    of two-write-queues. Only wal_write_mutex_ is held. This suffices to
  //    protect the data structure from concurrent push_back() by current
  //    write group leader as well as pop_front() by FindObsoleteFiles().
-  std::deque<LogFileNumberSize> alive_log_files_;
+  std::deque<WalFileNumberSize> alive_wal_files_;
+
+  // Total size of all "alive" WALs (for easy access without synchronization)
+  RelaxedAtomic<uint64_t> wals_total_size_{0};

  // Log files that aren't fully synced, and the current log file.
  // Synchronization:
  // 1. read by FindObsoleteFiles() which can be called either in application
-  //    thread or RocksDB bg threads. log_write_mutex_ is always held, while
+  //    thread or RocksDB bg threads. wal_write_mutex_ is always held, while
  //    some reads are performed without mutex_.
-  // 2. pop_front() by FindObsoleteFiles() with only log_write_mutex_ held.
-  // 3. read by DBImpl::Open() with both mutex_ and log_write_mutex_.
-  // 4. emplace_back() by DBImpl::Open() with both mutex_ and log_write_mutex.
+  // 2. pop_front() by FindObsoleteFiles() with only wal_write_mutex_ held.
+  // 3. read by DBImpl::Open() with both mutex_ and wal_write_mutex_.
+  // 4. emplace_back() by DBImpl::Open() with both mutex_ and wal_write_mutex.
  //    Note that at this point, DB::Open() has not returned success to
  //    application, thus the only other thread(s) that can conflict are bg
  //    threads calling FindObsoleteFiles(). See 1.
-  // 5. iteration and clear() from CloseHelper() always hold log_write_mutex
+  // 5. iteration and clear() from CloseHelper() always hold wal_write_mutex
  //    and mutex_.
  // 6. back() called by APIs FlushWAL() and LockWAL() are protected by only
-  //    log_write_mutex_. These two can be called by application threads after
+  //    wal_write_mutex_. These two can be called by application threads after
  //    DB::Open() returns success to applications.
-  // 7. read by SyncWAL(), another API, protected by only log_write_mutex_.
+  // 7. read by SyncWAL(), another API, protected by only wal_write_mutex_.
  // 8. read by MarkLogsNotSynced() and MarkLogsSynced() are protected by
-  //    log_write_mutex_.
-  // 9. erase() by MarkLogsSynced() protected by log_write_mutex_.
-  // 10. read by SyncClosedWals() protected by only log_write_mutex_. This can
+  //    wal_write_mutex_.
+  // 9. erase() by MarkLogsSynced() protected by wal_write_mutex_.
+  // 10. read by SyncClosedWals() protected by only wal_write_mutex_. This can
  //     happen in bg flush threads after DB::Open() returns success to
  //     applications.
  // 11. reads, e.g. front(), iteration, and back() called by PreprocessWrite()
-  //     holds only the log_write_mutex_. This is done by the write group
+  //     holds only the wal_write_mutex_. This is done by the write group
  //     leader. A bg thread calling FindObsoleteFiles() or MarkLogsSynced()
-  //     can happen concurrently. This is fine because log_write_mutex_ is used
+  //     can happen concurrently. This is fine because wal_write_mutex_ is used
  //     by all parties. See 2, 5, 9.
  // 12. reads, empty(), back() called by SwitchMemtable() hold both mutex_ and
-  //     log_write_mutex_. This happens in the write group leader.
+  //     wal_write_mutex_. This happens in the write group leader.
  // 13. emplace_back() by SwitchMemtable() hold both mutex_ and
-  //     log_write_mutex_. This happens in the write group leader. Can conflict
+  //     wal_write_mutex_. This happens in the write group leader. Can conflict
  //     with bg threads calling FindObsoleteFiles(), MarkLogsSynced(),
  //     SyncClosedWals(), etc. as well as application threads calling
  //     FlushWAL(), SyncWAL(), LockWAL(). This is fine because all parties
-  //     require at least log_write_mutex_.
+  //     require at least wal_write_mutex_.
  // 14. iteration called in WriteToWAL(write_group) protected by
-  //     log_write_mutex_. This is done by write group leader when
+  //     wal_write_mutex_. This is done by write group leader when
  //     two-write-queues is disabled and write needs to sync logs.
-  // 15. back() called in ConcurrentWriteToWAL() protected by log_write_mutex_.
+  // 15. back() called in ConcurrentWriteToWAL() protected by wal_write_mutex_.
  //     This can be done by the write group leader if two-write-queues is
  //     enabled. It can also be done by another WAL-only write thread.
  //
@@ -2856,23 +2910,22 @@ class DBImpl : public DB {
  std::deque<LogWriterNumber> logs_;

  // Signaled when getting_synced becomes false for some of the logs_.
-  InstrumentedCondVar log_sync_cv_;
+  InstrumentedCondVar wal_sync_cv_;
  // This is the app-level state that is written to the WAL but will be used
  // only during recovery. Using this feature enables not writing the state to
  // memtable on normal writes and hence improving the throughput. Each new
  // write of the state will replace the previous state entirely even if the
  // keys in the two consecutive states do not overlap.
-  // It is protected by log_write_mutex_ when two_write_queues_ is enabled.
+  // It is protected by wal_write_mutex_ when two_write_queues_ is enabled.
  // Otherwise only the heaad of write_thread_ can access it.
  WriteBatch cached_recoverable_state_;
  std::atomic<bool> cached_recoverable_state_empty_ = {true};
-  std::atomic<uint64_t> total_log_size_;

  // If this is non-empty, we need to delete these log files in background
-  // threads. Protected by log_write_mutex_.
-  autovector<log::Writer*> logs_to_free_;
+  // threads. Protected by wal_write_mutex_.
+  autovector<log::Writer*> wals_to_free_;

-  bool is_snapshot_supported_;
+  bool is_snapshot_supported_ = true;

  std::map<uint64_t, std::map<std::string, uint64_t>> stats_history_;

@@ -2896,7 +2949,7 @@ class DBImpl : public DB {
  // sleep if it uses up the quota.
  // Note: This is to protect memtable and compaction. If the batch only writes
  // to the WAL its size need not to be included in this.
-  uint64_t last_batch_group_size_;
+  uint64_t last_batch_group_size_ = 0;

  FlushScheduler flush_scheduler_;

@@ -2955,32 +3008,32 @@ class DBImpl : public DB {
  std::unordered_set<uint64_t> files_grabbed_for_purge_;

  // A queue to store log writers to close. Protected by db mutex_.
-  std::deque<log::Writer*> logs_to_free_queue_;
+  std::deque<log::Writer*> wals_to_free_queue_;

  std::deque<SuperVersion*> superversions_to_free_queue_;

-  int unscheduled_flushes_;
+  int unscheduled_flushes_ = 0;

-  int unscheduled_compactions_;
+  int unscheduled_compactions_ = 0;

  // count how many background compactions are running or have been scheduled in
  // the BOTTOM pool
-  int bg_bottom_compaction_scheduled_;
+  int bg_bottom_compaction_scheduled_ = 0;

  // count how many background compactions are running or have been scheduled
-  int bg_compaction_scheduled_;
+  int bg_compaction_scheduled_ = 0;

  // stores the number of compactions are currently running
-  int num_running_compactions_;
+  int num_running_compactions_ = 0;

  // number of background memtable flush jobs, submitted to the HIGH pool
-  int bg_flush_scheduled_;
+  int bg_flush_scheduled_ = 0;

  // stores the number of flushes are currently running
-  int num_running_flushes_;
+  int num_running_flushes_ = 0;

  // number of background obsolete file purge jobs, submitted to the HIGH pool
-  int bg_purge_scheduled_;
+  int bg_purge_scheduled_ = 0;

  std::deque<ManualCompactionState*> manual_compaction_dequeue_;

@@ -2990,11 +3043,11 @@ class DBImpl : public DB {
  // This enables two different threads to call
  // EnableFileDeletions() and DisableFileDeletions()
  // without any synchronization
-  int disable_delete_obsolete_files_;
+  int disable_delete_obsolete_files_ = 0;

  // Number of times FindObsoleteFiles has found deletable files and the
  // corresponding call to PurgeObsoleteFiles has not yet finished.
-  int pending_purge_obsolete_files_;
+  int pending_purge_obsolete_files_ = 0;

  // last time when DeleteObsoleteFiles with full scan was executed. Originally
  // initialized with startup time.
@@ -3006,12 +3059,12 @@ class DBImpl : public DB {
  // The mutex used by switch_cv_. mutex_ should be acquired beforehand.
  std::mutex switch_mutex_;
  // Number of threads intending to write to memtable
-  std::atomic<size_t> pending_memtable_writes_ = {};
+  std::atomic<size_t> pending_memtable_writes_{0};

  // A flag indicating whether the current rocksdb database has any
  // data that is not yet persisted into either WAL or SST file.
  // Used when disableWAL is true.
-  std::atomic<bool> has_unpersisted_data_;
+  std::atomic<bool> has_unpersisted_data_{false};

  // if an attempt was made to flush all column families that
  // the oldest log depends on but uncommitted data in the oldest
@@ -3019,26 +3072,26 @@ class DBImpl : public DB {
  // We must attempt to free the dependent memtables again
  // at a later time after the transaction in the oldest
  // log is fully commited.
-  bool unable_to_release_oldest_log_;
+  bool unable_to_release_oldest_log_{false};

  // Number of running IngestExternalFile() or CreateColumnFamilyWithImport()
  // calls.
  // REQUIRES: mutex held
-  int num_running_ingest_file_;
+  int num_running_ingest_file_ = 0;

  WalManager wal_manager_;

  // A value of > 0 temporarily disables scheduling of background work
-  int bg_work_paused_;
+  int bg_work_paused_ = 0;

  // A value of > 0 temporarily disables scheduling of background compaction
-  int bg_compaction_paused_;
+  int bg_compaction_paused_ = 0;

  // Guard against multiple concurrent refitting
-  bool refitting_level_;
+  bool refitting_level_ = false;

  // Indicate DB was opened successfully
-  bool opened_successfully_;
+  bool opened_successfully_ = false;

  // The min threshold to triggere bottommost compaction for removing
  // garbages, among all column families.
@@ -3084,13 +3137,13 @@ class DBImpl : public DB {
  // error recovery from going on in parallel. The latter, shutting_down_,
  // is set a little later during the shutdown after scheduling memtable
  // flushes
-  std::atomic<bool> shutdown_initiated_;
+  std::atomic<bool> shutdown_initiated_{false};
  // Flag to indicate whether sst_file_manager object was allocated in
  // DB::Open() or passed to us
  bool own_sfm_;

  // Flag to check whether Close() has been called on this DB
-  bool closed_;
+  bool closed_ = false;
  // save the closing status, for re-calling the close()
  Status closing_status_;
  // mutex for DB::Close()
@@ -3126,7 +3179,7 @@ class DBImpl : public DB {

  // The number of LockWAL called without matching UnlockWAL call.
  // See also lock_wal_write_token_
-  uint32_t lock_wal_count_;
+  uint32_t lock_wal_count_ = 0;
 };

 class GetWithTimestampReadCallback : public ReadCallback {
@@ -13,12 +13,14 @@
 #include "db/db_impl/db_impl.h"
 #include "db/error_handler.h"
 #include "db/event_helpers.h"
+#include "file/file_util.h"
 #include "file/sst_file_manager_impl.h"
 #include "logging/logging.h"
 #include "monitoring/iostats_context_imp.h"
 #include "monitoring/perf_context_imp.h"
 #include "monitoring/thread_status_updater.h"
 #include "monitoring/thread_status_util.h"
+#include "options/options_helper.h"
 #include "rocksdb/file_system.h"
 #include "rocksdb/io_status.h"
 #include "rocksdb/options.h"
@@ -143,10 +145,7 @@ IOStatus DBImpl::SyncClosedWals(const WriteOptions& write_options,
 Status DBImpl::FlushMemTableToOutputFile(
    ColumnFamilyData* cfd, const MutableCFOptions& mutable_cf_options,
    bool* made_progress, JobContext* job_context, FlushReason flush_reason,
-    SuperVersionContext* superversion_context,
-    std::vector<SequenceNumber>& snapshot_seqs,
-    SequenceNumber earliest_write_conflict_snapshot,
-    SnapshotChecker* snapshot_checker, LogBuffer* log_buffer,
+    SuperVersionContext* superversion_context, LogBuffer* log_buffer,
    Env::Priority thread_pri) {
  mutex_.AssertHeld();
  assert(cfd);
@@ -168,7 +167,7 @@ Status DBImpl::FlushMemTableToOutputFile(
  // had not been committed yet. Make sure we sync them to keep the persisted
  // WAL state at least as new as the persisted SST state.
  const bool needs_to_sync_closed_wals =
-      logfile_number_ > 0 &&
+      cur_wal_number_ > 0 &&
      (versions_->GetColumnFamilySet()->NumberOfColumnFamilies() > 1 ||
       allow_2pc());

@@ -210,7 +209,6 @@ Status DBImpl::FlushMemTableToOutputFile(
  FlushJob flush_job(
      dbname_, cfd, immutable_db_options_, mutable_cf_options, max_memtable_id,
      file_options_for_compaction_, versions_.get(), &mutex_, &shutting_down_,
-      snapshot_seqs, earliest_write_conflict_snapshot, snapshot_checker,
      job_context, flush_reason, log_buffer, directories_.GetDbDir(),
      GetDataDir(cfd, 0U),
      GetCompressionFlush(cfd->ioptions(), mutable_cf_options), stats_,
@@ -224,7 +222,7 @@ Status DBImpl::FlushMemTableToOutputFile(
  bool need_cancel = false;
  IOStatus log_io_s = IOStatus::OK();
  if (needs_to_sync_closed_wals) {
-    // SyncClosedWals() may unlock and re-lock the log_write_mutex multiple
+    // SyncClosedWals() may unlock and re-lock the wal_write_mutex multiple
    // times.
    VersionEdit synced_wals;
    bool error_recovery_in_prog = error_handler_.IsRecoveryInProgress();
@@ -395,11 +393,8 @@ Status DBImpl::FlushMemTablesToOutputFiles(
        bg_flush_args, made_progress, job_context, log_buffer, thread_pri);
  }
  assert(bg_flush_args.size() == 1);
-  std::vector<SequenceNumber> snapshot_seqs;
-  SequenceNumber earliest_write_conflict_snapshot;
-  SnapshotChecker* snapshot_checker;
-  GetSnapshotContext(job_context, &snapshot_seqs,
-                     &earliest_write_conflict_snapshot, &snapshot_checker);
+  InitSnapshotContext(job_context);
+
  const auto& bg_flush_arg = bg_flush_args[0];
  ColumnFamilyData* cfd = bg_flush_arg.cfd_;
  // intentional infrequent copy for each flush
@@ -410,8 +405,7 @@ Status DBImpl::FlushMemTablesToOutputFiles(
  FlushReason flush_reason = bg_flush_arg.flush_reason_;
  Status s = FlushMemTableToOutputFile(
      cfd, mutable_cf_options_copy, made_progress, job_context, flush_reason,
-      superversion_context, snapshot_seqs, earliest_write_conflict_snapshot,
-      snapshot_checker, log_buffer, thread_pri);
+      superversion_context, log_buffer, thread_pri);
  return s;
 }

@@ -446,12 +440,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
  }
 #endif /* !NDEBUG */

-  std::vector<SequenceNumber> snapshot_seqs;
-  SequenceNumber earliest_write_conflict_snapshot;
-  SnapshotChecker* snapshot_checker;
-  GetSnapshotContext(job_context, &snapshot_seqs,
-                     &earliest_write_conflict_snapshot, &snapshot_checker);
-
+  InitSnapshotContext(job_context);
  autovector<FSDirectory*> distinct_output_dirs;
  autovector<std::string> distinct_output_dir_paths;
  std::vector<std::unique_ptr<FlushJob>> jobs;
@@ -485,8 +474,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
    jobs.emplace_back(new FlushJob(
        dbname_, cfd, immutable_db_options_, mutable_cf_options,
        max_memtable_id, file_options_for_compaction_, versions_.get(), &mutex_,
-        &shutting_down_, snapshot_seqs, earliest_write_conflict_snapshot,
-        snapshot_checker, job_context, flush_reason, log_buffer,
+        &shutting_down_, job_context, flush_reason, log_buffer,
        directories_.GetDbDir(), data_dir,
        GetCompressionFlush(cfd->ioptions(), mutable_cf_options), stats_,
        &event_logger_, mutable_cf_options.report_bg_io_stats,
@@ -512,7 +500,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
                       job_context->job_id, flush_reason);
  }

-  if (logfile_number_ > 0) {
+  if (cur_wal_number_ > 0) {
    // TODO (yanqin) investigate whether we should sync the closed logs for
    // single column family case.
    VersionEdit synced_wals;
@@ -528,7 +516,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(

    if (!log_io_s.ok() && !log_io_s.IsShutdownInProgress() &&
        !log_io_s.IsColumnFamilyDropped()) {
-      if (total_log_size_ > 0) {
+      if (wals_total_size_.LoadRelaxed() > 0) {
        error_handler_.SetBGError(log_io_s, BackgroundErrorReason::kFlush);
      } else {
        // If the WAL is empty, we use different error reason
@@ -1123,8 +1111,7 @@ Status DBImpl::CompactRangeInternal(const CompactRangeOptions& options,
      cfd->NumberLevels() > 1) {
    // Always compact all files together.
    final_output_level = cfd->NumberLevels() - 1;
-    // if bottom most level is reserved
-    if (immutable_db_options_.allow_ingest_behind) {
+    if (cfd->AllowIngestBehind()) {
      final_output_level--;
    }
    s = RunManualCompaction(cfd, ColumnFamilyData::kCompactAllLevels,
@@ -1393,6 +1380,9 @@ Status DBImpl::CompactFiles(const CompactionOptions& compact_options,
  TEST_SYNC_POINT_CALLBACK("TestCompactFiles:PausingManualCompaction:3",
                           static_cast<void*>(const_cast<std::atomic<int>*>(
                               &manual_compaction_paused_)));
+  TEST_SYNC_POINT_CALLBACK("TestCancelCompactFiles:SuccessfulCompaction",
+                           static_cast<void*>(const_cast<std::atomic<int>*>(
+                               &manual_compaction_paused_)));
  {
    InstrumentedMutexLock l(&mutex_);
    auto* current = cfd->current();
@@ -1445,7 +1435,12 @@ Status DBImpl::CompactFilesImpl(
  if (shutting_down_.load(std::memory_order_acquire)) {
    return Status::ShutdownInProgress();
  }
-  if (manual_compaction_paused_.load(std::memory_order_acquire) > 0) {
+
+  // triggered by DisableManualCompactions or by user-set canceled flag in
+  // CompactionOptions
+  if (manual_compaction_paused_.load(std::memory_order_acquire) > 0 ||
+      (compact_options.canceled &&
+       compact_options.canceled->load(std::memory_order_acquire))) {
    return Status::Incomplete(Status::SubCode::kManualCompactionPaused);
  }

@@ -1464,7 +1459,7 @@ Status DBImpl::CompactFilesImpl(
    }
  }

-  if (cfd->ioptions().allow_ingest_behind &&
+  if (cfd->AllowIngestBehind() &&
      output_level >= cfd->ioptions().num_levels - 1) {
    return Status::InvalidArgument(
        "Exceed the maximum output level defined by "
@@ -1504,7 +1499,7 @@ Status DBImpl::CompactFilesImpl(

  std::unique_ptr<Compaction> c;
  assert(cfd->compaction_picker());
-  c.reset(cfd->compaction_picker()->CompactFiles(
+  c.reset(cfd->compaction_picker()->PickCompactionForCompactFiles(
      compact_options, input_files, output_level, version->storage_info(),
      cfd->GetLatestMutableCFOptions(), mutable_db_options_, output_path_id));
  // we already sanitized the set of input files and checked for conflicts
@@ -1516,11 +1511,7 @@ Status DBImpl::CompactFilesImpl(
  // deletion compaction currently not allowed in CompactFiles.
  assert(!c->deletion_compaction());

-  std::vector<SequenceNumber> snapshot_seqs;
-  SequenceNumber earliest_write_conflict_snapshot;
-  SnapshotChecker* snapshot_checker;
-  GetSnapshotContext(job_context, &snapshot_seqs,
-                     &earliest_write_conflict_snapshot, &snapshot_checker);
+  InitSnapshotContext(job_context);

  std::unique_ptr<std::list<uint64_t>::iterator> pending_outputs_inserted_elem(
      new std::list<uint64_t>::iterator(
@@ -1534,7 +1525,6 @@ Status DBImpl::CompactFilesImpl(
      log_buffer, directories_.GetDbDir(),
      GetDataDir(c->column_family_data(), c->output_path_id()),
      GetDataDir(c->column_family_data(), 0), stats_, &mutex_, &error_handler_,
-      snapshot_seqs, earliest_write_conflict_snapshot, snapshot_checker,
      job_context, table_cache_, &event_logger_,
      c->mutable_cf_options().paranoid_file_checks,
      c->mutable_cf_options().report_bg_io_stats, dbname_,
@@ -1858,10 +1848,9 @@ Status DBImpl::ReFitLevel(ColumnFamilyData* cfd, int level, int target_level) {
        0 /* max_subcompactions, not applicable */,
        {} /* grandparents, not applicable */,
        std::nullopt /* earliest_snapshot */, nullptr /* snapshot_checker */,
-        false /* is manual */, "" /* trim_ts */, -1 /* score, not applicable */,
-        false /* is deletion compaction, not applicable */,
-        false /* l0_files_might_overlap, not applicable */,
-        CompactionReason::kRefitLevel));
+        CompactionReason::kRefitLevel, "" /* trim_ts */,
+        -1 /* score, not applicable */,
+        false /* l0_files_might_overlap, not applicable */));
    cfd->compaction_picker()->RegisterCompaction(c.get());
    TEST_SYNC_POINT("DBImpl::ReFitLevel:PostRegisterCompaction");
    VersionEdit edit;
@@ -2187,6 +2176,7 @@ Status DBImpl::RunManualCompaction(
        // Don't throttle manual compaction, only count outstanding tasks.
        assert(false);
      }
+      ca->prepicked_compaction->need_repick = false;
      manual.incomplete = false;
      if (compaction->bottommost_level() &&
          env_->GetBackgroundThreads(Env::Priority::BOTTOM) > 0) {
@@ -2716,6 +2706,10 @@ Status DBImpl::WaitUntilFlushWouldNotStallWrites(ColumnFamilyData* cfd,
 // Finish waiting when ALL column families finish flushing memtables.
 // resuming_from_bg_err indicates whether the caller is trying to resume from
 // background error or in normal processing.
+// Note that the wait finishes when the flush result is installed to column
+// families' Versions and persisted in MANIFEST. It doesn't wait until
+// SuperVersion to reflect the flush result, except for the case when
+// flush_reason is `kExternalFileIngestion`.
 Status DBImpl::WaitForFlushMemTables(
    const autovector<ColumnFamilyData*>& cfds,
    const autovector<const uint64_t*>& flush_memtable_ids,
@@ -3409,6 +3403,10 @@ void DBImpl::BackgroundCallCompaction(PrepickedCompaction* prepicked_compaction,
  bool made_progress = false;
  JobContext job_context(next_job_id_.fetch_add(1), true);
  TEST_SYNC_POINT("BackgroundCallCompaction:0");
+  if (bg_thread_pri == Env::Priority::BOTTOM) {
+    TEST_SYNC_POINT("BackgroundCallCompaction:0:BottomPri");
+  }
+
  LogBuffer log_buffer(InfoLogLevel::INFO_LEVEL,
                       immutable_db_options_.info_log.get());
  {
@@ -3644,34 +3642,54 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
                 : m->manual_end->DebugString(true).c_str()));
      }
    }
-  } else if (!is_prepicked && !compaction_queue_.empty()) {
+  } else if (ShouldPickCompaction(is_prepicked, prepicked_compaction)) {
+    bool need_repick = is_prepicked && prepicked_compaction->need_repick;
    if (HasExclusiveManualCompaction()) {
-      // Can't compact right now, but try again later
      TEST_SYNC_POINT("DBImpl::BackgroundCompaction()::Conflict");

-      // Stay in the compaction queue.
-      unscheduled_compactions_++;
+      // TODO(hx235): Resolve conflict between intended
+      // bottom-priority compaction (requiring repick, i.e., need_repick = true)
+      // and exclusive manual compaction by releasing the intended
+      // bottom-priority compaction.
+      if (!need_repick) {
+        // Can't compact right now, but try again later
+        //
+        // Increase `unscheduled_compactions_` directly so we
+        // don't need to
+        // dequeue and enqueue the CFD again in the compaction queue and thus
+        // keep the CFD's position in the queue
+        unscheduled_compactions_++;

-      return Status::OK();
+        return Status::OK();
+      }
    }

-    auto cfd = PickCompactionFromQueue(&task_token, log_buffer);
-    if (cfd == nullptr) {
-      // Can't find any executable task from the compaction queue.
-      // All tasks have been throttled by compaction thread limiter.
-      ++unscheduled_compactions_;
-      return Status::Busy();
-    }
+    ColumnFamilyData* cfd = nullptr;

-    // We unreference here because the following code will take a Ref() on
-    // this cfd if it is going to use it (Compaction class holds a
-    // reference).
-    // This will all happen under a mutex so we don't have to be afraid of
-    // somebody else deleting it.
-    if (cfd->UnrefAndTryDelete()) {
-      // This was the last reference of the column family, so no need to
-      // compact.
-      return Status::OK();
+    if (!need_repick) {
+      cfd = PickCompactionFromQueue(&task_token, log_buffer);
+      if (cfd == nullptr) {
+        // Can't find any executable task from the compaction queue.
+        // All tasks have been throttled by compaction thread limiter.
+        ++unscheduled_compactions_;
+        return Status::Busy();
+      }
+
+      // We unreference here because the following code will take a Ref() on
+      // this cfd if it is going to use it (Compaction class holds a
+      // reference).
+      // This will all happen under a mutex so we don't have to be afraid of
+      // somebody else deleting it.
+      if (cfd->UnrefAndTryDelete()) {
+        // This was the last reference of the column family, so no need to
+        // compact.
+        return Status::OK();
+      }
+    } else {
+      cfd = c->column_family_data();
+      assert(cfd);
+      ResetBottomPriCompactionIntent(cfd, c);
+      assert(c == nullptr);
    }

    // Pick up latest mutable CF Options and use it throughout the
@@ -3685,21 +3703,24 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
      // compaction is not necessary. Need to make sure mutex is held
      // until we make a copy in the following code
      TEST_SYNC_POINT("DBImpl::BackgroundCompaction():BeforePickCompaction");
-      SnapshotChecker* snapshot_checker = nullptr;
-      std::vector<SequenceNumber> snapshot_seqs;
      // This info is not useful for other scenarios, so save querying existing
      // snapshots for those cases.
      if (cfd->ioptions().compaction_style == kCompactionStyleUniversal &&
          cfd->user_comparator()->timestamp_size() == 0) {
-        SequenceNumber earliest_write_conflict_snapshot;
-        GetSnapshotContext(job_context, &snapshot_seqs,
-                           &earliest_write_conflict_snapshot,
-                           &snapshot_checker);
+        InitSnapshotContext(job_context);
        assert(is_snapshot_supported_ || snapshots_.empty());
      }
-      c.reset(cfd->PickCompaction(mutable_cf_options, mutable_db_options_,
-                                  snapshot_seqs, snapshot_checker, log_buffer));
-      TEST_SYNC_POINT("DBImpl::BackgroundCompaction():AfterPickCompaction");
+      c.reset(cfd->PickCompaction(
+          mutable_cf_options, mutable_db_options_, job_context->snapshot_seqs,
+          job_context->snapshot_checker, log_buffer,
+          thread_pri == Env::Priority::BOTTOM /* require_max_output_level */));
+      if (thread_pri == Env::Priority::LOW) {
+        TEST_SYNC_POINT("DBImpl::BackgroundCompaction():AfterPickCompaction");
+      } else if (thread_pri == Env::Priority::BOTTOM) {
+        TEST_SYNC_POINT_CALLBACK(
+            "DBImpl::BackgroundCompaction():AfterPickCompactionBottomPri",
+            c.get());
+      }

      if (c != nullptr) {
        bool enough_room = EnoughRoomForCompaction(
@@ -3713,7 +3734,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
              ->storage_info()
              ->ComputeCompactionScore(c->immutable_options(),
                                       c->mutable_cf_options());
-          AddToCompactionQueue(cfd);
+          EnqueuePendingCompaction(cfd);

          c.reset();
          // Don't need to sleep here, because BackgroundCallCompaction
@@ -3735,16 +3756,21 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
          // options take effect.
          // 3) When we Pick a new compaction, we "remove" those files being
          // compacted from the calculation, which then influences compaction
-          // score. Here we check if we need the new compaction even without the
-          // files that are currently being compacted. If we need another
-          // compaction, we might be able to execute it in parallel, so we add
-          // it to the queue and schedule a new thread.
-          if (cfd->NeedsCompaction()) {
-            // Yes, we need more compactions!
-            AddToCompactionQueue(cfd);
-            MaybeScheduleFlushOrCompaction();
-          }
+          // score. Inside EnqueuePendingCompaction(),  we check if we need
+          // the new compaction even without the files that are currently being
+          // compacted. If we need another compaction, we might be able to
+          // execute it in parallel, so we add it to the queue and schedule a
+          // new thread.
+          EnqueuePendingCompaction(cfd);
+          MaybeScheduleFlushOrCompaction();
        }
+      } else if (is_prepicked) {
+        ROCKS_LOG_BUFFER(
+            log_buffer,
+            "[%s] Pre-picked compaction repicked files for compaction as "
+            "required, "
+            "but upon re-evaluation, no compaction was found necessary \n",
+            cfd->GetName().c_str());
      }
    }
  }
@@ -3786,11 +3812,251 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
                     c->column_family_data()->GetName().c_str(),
                     c->num_input_files(0));
    if (status.ok() && io_s.ok()) {
-      UpdateDeletionCompactionStats(c);
+      UpdateFIFOCompactionStatus(c);
    }
    *made_progress = true;
    TEST_SYNC_POINT_CALLBACK("DBImpl::BackgroundCompaction:AfterCompaction",
                             c->column_family_data());
+  } else if (c->is_trivial_copy_compaction()) {
+    TEST_SYNC_POINT_CALLBACK(
+        "DBImpl::BackgroundCompaction:TriviaCopyBeforeCompaction",
+        c->column_family_data());
+    assert(c->num_input_files(1) == 0);
+    assert(c->column_family_data()->ioptions().compaction_style ==
+           kCompactionStyleFIFO);
+    assert(c->compaction_reason() == CompactionReason::kChangeTemperature);
+
+    compaction_job_stats.num_input_files = c->num_input_files(0);
+
+    NotifyOnCompactionBegin(c->column_family_data(), c.get(), status,
+                            compaction_job_stats, job_context->job_id);
+
+    std::vector<FileMetaData> out_files;
+    for (const auto& in_file : *c->inputs(0)) {
+      const uint64_t out_file_number = versions_->NewFileNumber();
+      const std::string in_fname =
+          TableFileName(c->immutable_options().cf_paths,
+                        in_file->fd.GetNumber(), in_file->fd.GetPathId());
+      const std::string out_fname =
+          TableFileName(c->immutable_options().cf_paths, out_file_number,
+                        c->output_path_id());
+
+      // TODO (mikechuang): Currently skip calling
+      // EventHelpers::NotifyTableFileCreationStarted for the trivial copy.
+      // Since it's a trivial copy we should ideally use the exact
+      // TableProperties from the input file but that will break some existing
+      // stress tests. For now skip the listener call for the FIFO
+      // kChangeTemperature trivial copy move.
+
+      int64_t tmp_current_time = 0;
+      auto get_time_status =
+          immutable_db_options_.clock->GetCurrentTime(&tmp_current_time);
+      if (!get_time_status.ok()) {
+        ROCKS_LOG_BUFFER(log_buffer,
+                         "[%s] WARNING: Failed to get current time %s "
+                         "status=%s",
+                         c->column_family_data()->GetName().c_str(),
+                         get_time_status.ToString().c_str());
+      }
+      uint64_t out_file_creation_time = static_cast<uint64_t>(tmp_current_time);
+
+      FileOptions copied_file_options = file_options_;
+      copied_file_options.temperature = c->output_temperature();
+      std::unique_ptr<WritableFileWriter> dest_writer;
+      {
+        std::unique_ptr<FSWritableFile> dest_file;
+        IOStatus writable_file_io_status =
+            immutable_db_options_.fs.get()->NewWritableFile(
+                out_fname, copied_file_options, &dest_file, nullptr /* dbg */);
+        TEST_SYNC_POINT_CALLBACK(
+            "NewWritableFile::FileOptions.temperature",
+            const_cast<Temperature*>(&copied_file_options.temperature));
+        if (!writable_file_io_status.ok()) {
+          io_s = writable_file_io_status;
+          ROCKS_LOG_BUFFER(
+              log_buffer,
+              "[%s] Error: Abort trivial copy compaction, failed to open "
+              "NewWritableFile %s\n"
+              " out_fname=%s, temperature=%s, io_status=%s",
+              c->column_family_data()->GetName().c_str(), out_fname.c_str(),
+              temperature_to_string[c->output_temperature()].c_str(),
+              io_s.ToString().c_str());
+          break;
+        }
+
+        FileTypeSet tmp_set = immutable_db_options_.checksum_handoff_file_types;
+        dest_writer.reset(new WritableFileWriter(
+            std::move(dest_file), out_fname, copied_file_options,
+            immutable_db_options_.clock, io_tracer_,
+            immutable_db_options_.stats, Histograms::SST_WRITE_MICROS,
+            c->immutable_options().listeners,
+            immutable_db_options_.file_checksum_gen_factory.get(),
+            tmp_set.Contains(FileType::kTableFile), false));
+      }
+
+      ROCKS_LOG_BUFFER(
+          log_buffer,
+          "[%s] Started copying from: %s\n"
+          " temperature=%s, to: %s, temperature=%s, buffer_size=%" PRIu64,
+          c->column_family_data()->GetName().c_str(), in_fname.c_str(),
+          temperature_to_string[in_file->temperature].c_str(),
+          out_fname.c_str(),
+          temperature_to_string[c->output_temperature()].c_str(),
+          c->mutable_cf_options()
+              .compaction_options_fifo.trivial_copy_buffer_size);
+      // Add IO_LOW HINT for compaction
+      IOOptions copy_files_compaction_io_options;
+      copy_files_compaction_io_options.rate_limiter_priority =
+          Env::IOPriority::IO_LOW;
+      copy_files_compaction_io_options.type = IOType::kData;
+      copy_files_compaction_io_options.io_activity =
+          Env::IOActivity::kCompaction;
+
+      IOStatus copy_file_io_status = CopyFile(
+          immutable_db_options_.fs.get() /* fileSystem */,
+          in_fname /* source */, in_file->temperature /* src_temp_hint */,
+          dest_writer /* dest_writer */, 0 /* size */, true /* use_fsync */,
+          io_tracer_ /* io_tracer*/,
+          c->mutable_cf_options()
+              .compaction_options_fifo
+              .trivial_copy_buffer_size /* max_read_buffer_size
+                                         */
+          ,
+          copy_files_compaction_io_options /* readIOOptions */,
+          copy_files_compaction_io_options /* writeIOOptions */);
+      if (dest_writer) {
+        IOOptions close_files_compaction_io_options;
+        close_files_compaction_io_options.rate_limiter_priority =
+            Env::IOPriority::IO_LOW;
+        close_files_compaction_io_options.type = IOType::kData;
+        close_files_compaction_io_options.io_activity =
+            Env::IOActivity::kCompaction;
+        // Close the dest_write
+        io_s = dest_writer->Close(close_files_compaction_io_options);
+        if (!io_s.ok()) {
+          ROCKS_LOG_BUFFER(
+              log_buffer,
+              "[%s] Failed to close the writer. Failed to copy from: %s\n"
+              " temperature=%s, to=%s, temperature=%s, io_status=%s",
+              c->column_family_data()->GetName().c_str(), in_fname.c_str(),
+              temperature_to_string[in_file->temperature].c_str(),
+              out_fname.c_str(),
+              temperature_to_string[c->output_temperature()].c_str(),
+              io_s.ToString().c_str());
+          break;
+        }
+      }
+
+      io_s = copy_file_io_status;
+
+      if (!io_s.ok()) {
+        ROCKS_LOG_BUFFER(log_buffer,
+                         "[%s] Failed to copy from: %s\n"
+                         " temperature=%s, to=%s, temperature=%s, io_status=%s",
+                         c->column_family_data()->GetName().c_str(),
+                         in_fname.c_str(),
+                         temperature_to_string[in_file->temperature].c_str(),
+                         out_fname.c_str(),
+                         temperature_to_string[c->output_temperature()].c_str(),
+                         io_s.ToString().c_str());
+        break;
+      }
+      ROCKS_LOG_BUFFER(log_buffer,
+                       "[%s] Successfully copying from: %s\n"
+                       " temperature=%s, to=%s, temperature=%s, io_status=%s",
+                       c->column_family_data()->GetName().c_str(),
+                       in_fname.c_str(),
+                       temperature_to_string[in_file->temperature].c_str(),
+                       out_fname.c_str(),
+                       temperature_to_string[c->output_temperature()].c_str(),
+                       io_s.ToString().c_str());
+
+      FileMetaData out_file_metadata{
+          out_file_number,
+          c->output_path_id(),
+          in_file->fd.GetFileSize(),
+          in_file->smallest,
+          in_file->largest,
+          in_file->fd.smallest_seqno,
+          in_file->fd.largest_seqno,
+          false /* marked_for_compact */,
+          c->output_temperature() /* temperature */,
+          in_file->oldest_blob_file_number,
+          in_file->oldest_ancester_time,
+          out_file_creation_time,
+          c->MinInputFileEpochNumber(),
+          dest_writer->GetFileChecksum(),
+          dest_writer->GetFileChecksumFuncName(),
+          in_file->unique_id,
+          in_file->compensated_range_deletion_size,
+          in_file->tail_size,
+          in_file->user_defined_timestamps_persisted};
+
+      out_files.push_back(std::move(out_file_metadata));
+    }
+
+    // Update version set
+    if (status.ok() && io_s.ok()) {
+      // NOTE: ChangeTemperature should only copy one file at one file
+      // hence *c->inputs(0) == out_files.size() == 1 if copy succeeded
+      assert(c->inputs(0)->size() == 1);
+      assert(out_files.size() == 1);
+
+      auto out_file_metadata_it = out_files.begin();
+      for (const auto& in_file : *c->inputs(0)) {
+        if (out_file_metadata_it == out_files.end()) {
+          break;
+        }
+
+        c->edit()->DeleteFile(c->level(), in_file->fd.GetNumber());
+        c->edit()->AddFile(c->level(), *out_file_metadata_it);
+        ++out_file_metadata_it;
+      }
+
+      status = versions_->LogAndApply(
+          c->column_family_data(), read_options, write_options, c->edit(),
+          &mutex_, directories_.GetDbDir(),
+          /*new_descriptor_log=*/false, /*column_family_options=*/nullptr,
+          [&c, &compaction_released](const Status& s) {
+            c->ReleaseCompactionFiles(s);
+            compaction_released = true;
+          });
+    }
+
+    // TODO (mikechuang): Currently skip calling
+    // EventHelper::LogAndNotifyTableFileCreationFinished for the trivial copy.
+    // Since it's a trivial copy we should ideally use the exact TableProperties
+    // from the input file but that will break some existing stress tests. For
+    // now skip the listener call for the FIFO kChangeTemperature trivial copy
+    // move.
+
+    if (io_s.ok()) {
+      io_s = versions_->io_status();
+    }
+
+    InstallSuperVersionAndScheduleWork(
+        c->column_family_data(), job_context->superversion_contexts.data());
+    if (status.ok() && io_s.ok()) {
+      UpdateFIFOCompactionStatus(c);
+    } else {
+      for (const auto& in_file : *c->inputs(0)) {
+        const std::string in_fname =
+            TableFileName(c->immutable_options().cf_paths,
+                          in_file->fd.GetNumber(), in_file->fd.GetPathId());
+        ROCKS_LOG_BUFFER(
+            log_buffer,
+            "[%s] Failed to do trvial copy compaction: %s"
+            " temperature=%s, to temperature=%s, status=%s, io_status=%s",
+            c->column_family_data()->GetName().c_str(), in_fname.c_str(),
+            temperature_to_string[in_file->temperature].c_str(),
+            temperature_to_string[c->output_temperature()].c_str(),
+            status.ToString().c_str(), io_s.ToString().c_str());
+      }
+    }
+    *made_progress = true;
+    TEST_SYNC_POINT_CALLBACK(
+        "DBImpl::BackgroundCompaction:TriviaCopyAfterCompaction",
+        c->column_family_data());
  } else if (!trivial_move_disallowed && c->IsTrivialMove()) {
    TEST_SYNC_POINT("DBImpl::BackgroundCompaction:TrivialMove");
    TEST_SYNC_POINT_CALLBACK("DBImpl::BackgroundCompaction:BeforeCompaction",
@@ -3881,14 +4147,17 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
    ThreadStatusUtil::ResetThreadStatus();
    TEST_SYNC_POINT_CALLBACK("DBImpl::BackgroundCompaction:AfterCompaction",
                             c->column_family_data());
-  } else if (!is_prepicked && c->output_level() > 0 &&
-             c->output_level() ==
+  } else if (!is_prepicked &&
+             Compaction::OutputToNonZeroMaxOutputLevel(
+                 c->output_level(),
                 c->column_family_data()
                     ->current()
                     ->storage_info()
                     ->MaxOutputLevel(
-                         immutable_db_options_.allow_ingest_behind) &&
+                         c->immutable_options().cf_allow_ingest_behind ||
+                         immutable_db_options_.allow_ingest_behind)) &&
             env_->GetBackgroundThreads(Env::Priority::BOTTOM) > 0) {
+    assert(thread_pri == Env::Priority::LOW);
    // Forward compactions involving last level to the bottom pool if it exists,
    // such that compactions unlikely to contribute to write stalls can be
    // delayed or deprioritized.
@@ -3897,7 +4166,23 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
    ca->db = this;
    ca->compaction_pri_ = Env::Priority::BOTTOM;
    ca->prepicked_compaction = new PrepickedCompaction;
-    ca->prepicked_compaction->compaction = c.release();
+
+    // If `universal_reduce_file_locking` is true, we only lock a limited set of
+    // input files by creating an intended compaction to forward to bottom
+    // priority pool and repicking files when bottom priority thread
+    // gets to execute this intended compaction
+    const bool need_repick =
+        c->mutable_cf_options()
+            .compaction_options_universal.reduce_file_locking;
+    if (need_repick) {
+      ca->prepicked_compaction->compaction =
+          CreateIntendedCompactionForwardedToBottomPriorityPool(c.get());
+      c.reset();
+      ca->prepicked_compaction->need_repick = true;
+    } else {
+      ca->prepicked_compaction->compaction = c.release();
+      ca->prepicked_compaction->need_repick = false;
+    }
    ca->prepicked_compaction->manual_compaction_state = nullptr;
    // Transfer requested token, so it doesn't need to do it again.
    ca->prepicked_compaction->task_token = std::move(task_token);
@@ -3912,11 +4197,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
    output_level = c->output_level();
    TEST_SYNC_POINT_CALLBACK("DBImpl::BackgroundCompaction:NonTrivial",
                             &output_level);
-    std::vector<SequenceNumber> snapshot_seqs;
-    SequenceNumber earliest_write_conflict_snapshot;
-    SnapshotChecker* snapshot_checker;
-    GetSnapshotContext(job_context, &snapshot_seqs,
-                       &earliest_write_conflict_snapshot, &snapshot_checker);
+    InitSnapshotContext(job_context);
    assert(is_snapshot_supported_ || snapshots_.empty());

    CompactionJob compaction_job(
@@ -3925,8 +4206,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
        &shutting_down_, log_buffer, directories_.GetDbDir(),
        GetDataDir(c->column_family_data(), c->output_path_id()),
        GetDataDir(c->column_family_data(), 0), stats_, &mutex_,
-        &error_handler_, snapshot_seqs, earliest_write_conflict_snapshot,
-        snapshot_checker, job_context, table_cache_, &event_logger_,
+        &error_handler_, job_context, table_cache_, &event_logger_,
        c->mutable_cf_options().paranoid_file_checks,
        c->mutable_cf_options().report_bg_io_stats, dbname_,
        &compaction_job_stats, thread_pri, io_tracer_,
@@ -3946,8 +4226,15 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
    NotifyOnCompactionBegin(c->column_family_data(), c.get(), status,
                            compaction_job_stats, job_context->job_id);
    mutex_.Unlock();
-    TEST_SYNC_POINT_CALLBACK(
-        "DBImpl::BackgroundCompaction:NonTrivial:BeforeRun", nullptr);
+    if (thread_pri == Env::Priority::LOW) {
+      TEST_SYNC_POINT_CALLBACK(
+          "DBImpl::BackgroundCompaction:NonTrivial:BeforeRun", nullptr);
+    } else {
+      assert(thread_pri == Env::Priority::BOTTOM);
+      TEST_SYNC_POINT(
+          "DBImpl::BackgroundCompaction:NonTrivial:BeforeRunBottomPri");
+    }
+
    // Should handle error?
    compaction_job.Run().PermitUncheckedError();
    TEST_SYNC_POINT("DBImpl::BackgroundCompaction:NonTrivial:AfterRun");
@@ -4041,9 +4328,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
          ->storage_info()
          ->ComputeCompactionScore(c->immutable_options(),
                                   c->mutable_cf_options());
-      if (!cfd->queued_for_compaction()) {
-        AddToCompactionQueue(cfd);
-      }
+      EnqueuePendingCompaction(cfd);
    }
  }
  // this will unref its input_version and column_family_data
@@ -4088,6 +4373,72 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
  return status;
 }

+// Create an intended compaction to forward based on the original picked
+// compaction. It serves two purposes while it is waiting
+// for a bottom-priority thread becomes available to run:
+// - Prevent the last input file (or sorted run if non-L0) from
+// being included in compaction score calculations unnecessarily since the
+// intended compaction is already scheduled to compact it
+// - Allow other input files to be picked by low-priority compactions that can
+// run right away
+//
+// Once a bottom-priority available to run this intended compaction, it will
+// repick files to consider the LSM updates that occurred during the waiting
+// period.
+Compaction* DBImpl::CreateIntendedCompactionForwardedToBottomPriorityPool(
+    Compaction* c) {
+  auto* cfd = c->column_family_data();
+  const auto& io = c->immutable_options();
+  const auto& mo = c->mutable_cf_options();
+  auto* vstorage = c->input_version()->storage_info();
+
+  std::vector<CompactionInputFiles> inputs(1);
+
+  const std::vector<FileMetaData*>* max_intput_level_files = nullptr;
+  int max_intput_level = 0;
+
+  for (size_t i = c->num_input_levels(); i >= 1; --i) {
+    size_t level = i - 1;
+    if (c->num_input_files(level) > 0) {
+      max_intput_level = static_cast<int>(level);
+      max_intput_level_files = c->inputs(level);
+      break;
+    }
+  }
+
+  assert(max_intput_level_files);
+  assert(!max_intput_level_files->empty());
+  inputs[0].level = max_intput_level;
+
+  if (max_intput_level == 0) {
+    // The last input file
+    inputs[0].files.push_back(
+        (*max_intput_level_files)[max_intput_level_files->size() - 1]);
+  } else {
+    // The last input sorted run
+    for (FileMetaData* f : (*max_intput_level_files)) {
+      inputs[0].files.push_back(f);
+    }
+  }
+
+  c->ReleaseCompactionFiles(Status::OK());
+
+  Compaction* intended_compaction =
+      new Compaction(vstorage, io, mo, mutable_db_options_, std::move(inputs),
+                     c->output_level(), c->target_output_file_size(),
+                     c->max_compaction_bytes(), c->output_path_id(),
+                     c->output_compression(), c->output_compression_opts(),
+                     c->output_temperature(), c->max_subcompactions(),
+                     c->grandparents(), std::nullopt /* earliest_snapshot */,
+                     nullptr /* snapshot_checker */, c->compaction_reason());
+
+  cfd->compaction_picker()->RegisterCompaction(intended_compaction);
+  vstorage->ComputeCompactionScore(io, mo);
+  intended_compaction->FinalizeInputInfo(cfd->current());
+
+  return intended_compaction;
+}
+
 bool DBImpl::HasPendingManualCompaction() {
  return (!manual_compaction_dequeue_.empty());
 }
@@ -4176,8 +4527,7 @@ bool DBImpl::MCOverlap(ManualCompactionState* m, ManualCompactionState* m1) {
  return false;
 }

-void DBImpl::UpdateDeletionCompactionStats(
-    const std::unique_ptr<Compaction>& c) {
+void DBImpl::UpdateFIFOCompactionStatus(const std::unique_ptr<Compaction>& c) {
  if (c == nullptr) {
    return;
  }
@@ -4191,6 +4541,9 @@ void DBImpl::UpdateDeletionCompactionStats(
    case CompactionReason::kFIFOTtl:
      RecordTick(stats_, FIFO_TTL_COMPACTIONS);
      break;
+    case CompactionReason::kChangeTemperature:
+      RecordTick(stats_, FIFO_CHANGE_TEMPERATURE_COMPACTIONS);
+      break;
    default:
      assert(false);
      break;
@@ -4307,7 +4660,7 @@ void DBImpl::InstallSuperVersionAndScheduleWork(
  bottommost_files_mark_threshold_ = kMaxSequenceNumber;
  standalone_range_deletion_files_mark_threshold_ = kMaxSequenceNumber;
  for (auto* my_cfd : *versions_->GetColumnFamilySet()) {
-    if (!my_cfd->ioptions().allow_ingest_behind) {
+    if (!my_cfd->AllowIngestBehind()) {
      bottommost_files_mark_threshold_ = std::min(
          bottommost_files_mark_threshold_,
          my_cfd->current()->storage_info()->bottommost_files_mark_threshold());
@@ -4357,31 +4710,33 @@ void DBImpl::SetSnapshotChecker(SnapshotChecker* snapshot_checker) {
  snapshot_checker_.reset(snapshot_checker);
 }

-void DBImpl::GetSnapshotContext(
-    JobContext* job_context, std::vector<SequenceNumber>* snapshot_seqs,
-    SequenceNumber* earliest_write_conflict_snapshot,
-    SnapshotChecker** snapshot_checker_ptr) {
+void DBImpl::InitSnapshotContext(JobContext* job_context) {
  mutex_.AssertHeld();
  assert(job_context != nullptr);
-  assert(snapshot_seqs != nullptr);
-  assert(earliest_write_conflict_snapshot != nullptr);
-  assert(snapshot_checker_ptr != nullptr);
-
-  *snapshot_checker_ptr = snapshot_checker_.get();
-  if (use_custom_gc_ && *snapshot_checker_ptr == nullptr) {
-    *snapshot_checker_ptr = DisableGCSnapshotChecker::Instance();
+  if (job_context->snapshot_context_initialized) {
+    return;
  }
-  if (*snapshot_checker_ptr != nullptr) {
+  SnapshotChecker* snapshot_checker = snapshot_checker_.get();
+  if (use_custom_gc_ && !snapshot_checker) {
+    snapshot_checker = DisableGCSnapshotChecker::Instance();
+  }
+  std::unique_ptr<ManagedSnapshot> managed_snapshot = nullptr;
+  if (snapshot_checker) {
    // If snapshot_checker is used, that means the flush/compaction may
    // contain values not visible to snapshot taken after
    // flush/compaction job starts. Take a snapshot and it will appear
    // in snapshot_seqs and force compaction iterator to consider such
    // snapshots.
-    const Snapshot* job_snapshot =
-        GetSnapshotImpl(false /*write_conflict_boundary*/, false /*lock*/);
-    job_context->job_snapshot.reset(new ManagedSnapshot(this, job_snapshot));
+    const Snapshot* snapshot =
+        GetSnapshotImpl(/*is_write_conflict_boundary=*/false, /*lock=*/false);
+    managed_snapshot.reset(new ManagedSnapshot(this, snapshot));
  }
-  *snapshot_seqs = snapshots_.GetAll(earliest_write_conflict_snapshot);
+  SequenceNumber earliest_write_conflict_snapshot = kMaxSequenceNumber;
+  std::vector<SequenceNumber> snapshot_seqs =
+      snapshots_.GetAll(&earliest_write_conflict_snapshot);
+  job_context->InitSnapshotContext(
+      snapshot_checker, std::move(managed_snapshot),
+      earliest_write_conflict_snapshot, std::move(snapshot_seqs));
 }

 Status DBImpl::WaitForCompact(
@@ -4440,4 +4795,18 @@ Status DBImpl::WaitForCompact(
  }
 }

+bool DBImpl::ShouldPickCompaction(
+    bool is_prepicked, const PrepickedCompaction* prepicked_compaction) {
+  return (!is_prepicked && !compaction_queue_.empty()) ||
+         (is_prepicked && prepicked_compaction->need_repick);
+}
+
+void DBImpl::ResetBottomPriCompactionIntent(ColumnFamilyData* cfd,
+                                            std::unique_ptr<Compaction>& c) {
+  c->ReleaseCompactionFiles(Status::OK());
+  cfd->current()->storage_info()->ComputeCompactionScore(
+      c->immutable_options(), c->mutable_cf_options());
+  c.reset();
+}
+
 }  // namespace ROCKSDB_NAMESPACE
@@ -84,6 +84,7 @@ void DBImpl::TEST_GetFilesMetaData(
 }

 uint64_t DBImpl::TEST_Current_Manifest_FileNo() {
+  InstrumentedMutexLock l(&mutex_);
  return versions_->manifest_file_number();
 }

@@ -224,13 +225,13 @@ void DBImpl::TEST_EndWrite(void* w) {
 }

 size_t DBImpl::TEST_LogsToFreeSize() {
-  InstrumentedMutexLock l(&log_write_mutex_);
-  return logs_to_free_.size();
+  InstrumentedMutexLock l(&wal_write_mutex_);
+  return wals_to_free_.size();
 }

 uint64_t DBImpl::TEST_LogfileNumber() {
  InstrumentedMutexLock l(&mutex_);
-  return logfile_number_;
+  return cur_wal_number_;
 }

 void DBImpl::TEST_GetAllBlockCaches(
@@ -28,7 +28,7 @@ uint64_t DBImpl::MinLogNumberToKeep() {
  return versions_->min_log_number_to_keep();
 }

-uint64_t DBImpl::MinLogNumberToRecycle() { return min_log_number_to_recycle_; }
+uint64_t DBImpl::MinLogNumberToRecycle() { return min_wal_number_to_recycle_; }

 uint64_t DBImpl::MinObsoleteSstNumberToKeep() {
  mutex_.AssertHeld();
@@ -272,77 +272,77 @@ void DBImpl::FindObsoleteFiles(JobContext* job_context, bool force,

  // logs_ is empty when called during recovery, in which case there can't yet
  // be any tracked obsolete logs
-  log_write_mutex_.Lock();
+  wal_write_mutex_.Lock();

-  if (alive_log_files_.empty() || logs_.empty()) {
+  if (alive_wal_files_.empty() || logs_.empty()) {
    mutex_.AssertHeld();
    // We may reach here if the db is DBImplSecondary
-    log_write_mutex_.Unlock();
+    wal_write_mutex_.Unlock();
    return;
  }

  bool mutex_unlocked = false;
-  if (!alive_log_files_.empty() && !logs_.empty()) {
+  if (!alive_wal_files_.empty() && !logs_.empty()) {
    uint64_t min_log_number = job_context->log_number;
-    size_t num_alive_log_files = alive_log_files_.size();
+    size_t num_alive_wal_files = alive_wal_files_.size();
    // find newly obsoleted log files
-    while (alive_log_files_.begin()->number < min_log_number) {
-      auto& earliest = *alive_log_files_.begin();
+    while (alive_wal_files_.begin()->number < min_log_number) {
+      auto& earliest = *alive_wal_files_.begin();
      if (immutable_db_options_.recycle_log_file_num >
-              log_recycle_files_.size() &&
+              wal_recycle_files_.size() &&
          earliest.number >= MinLogNumberToRecycle()) {
        ROCKS_LOG_INFO(immutable_db_options_.info_log,
                       "adding log %" PRIu64 " to recycle list\n",
                       earliest.number);
-        log_recycle_files_.push_back(earliest.number);
+        wal_recycle_files_.push_back(earliest.number);
      } else {
        job_context->log_delete_files.push_back(earliest.number);
      }
      if (job_context->size_log_to_delete == 0) {
-        job_context->prev_total_log_size = total_log_size_;
-        job_context->num_alive_log_files = num_alive_log_files;
+        job_context->prev_wals_total_size = wals_total_size_.LoadRelaxed();
+        job_context->num_alive_wal_files = num_alive_wal_files;
      }
      job_context->size_log_to_delete += earliest.size;
-      total_log_size_ -= earliest.size;
-      alive_log_files_.pop_front();
+      wals_total_size_.FetchSubRelaxed(earliest.size);
+      alive_wal_files_.pop_front();

      // Current log should always stay alive since it can't have
      // number < MinLogNumber().
-      assert(alive_log_files_.size());
+      assert(alive_wal_files_.size());
    }
-    log_write_mutex_.Unlock();
+    wal_write_mutex_.Unlock();
    mutex_.Unlock();
    mutex_unlocked = true;
    TEST_SYNC_POINT_CALLBACK("FindObsoleteFiles::PostMutexUnlock", nullptr);
-    log_write_mutex_.Lock();
+    wal_write_mutex_.Lock();
    while (!logs_.empty() && logs_.front().number < min_log_number) {
      auto& log = logs_.front();
      if (log.IsSyncing()) {
-        log_sync_cv_.Wait();
+        wal_sync_cv_.Wait();
        // logs_ could have changed while we were waiting.
        continue;
      }
      // This WAL file is not live, so it's OK if we never sync the rest of it.
      // If it's already closed, then it's been fully synced. If
      // !background_close_inactive_wals then we need to Close it before
-      // removing from logs_ but not blocking while holding log_write_mutex_.
+      // removing from logs_ but not blocking while holding wal_write_mutex_.
      if (!immutable_db_options_.background_close_inactive_wals &&
          log.writer->file()) {
        // We are taking ownership of and pinning the front entry, so we can
        // expect it to be the same after releasing and re-acquiring the lock
        log.PrepareForSync();
-        log_write_mutex_.Unlock();
+        wal_write_mutex_.Unlock();
        // TODO: maybe check the return value of Close.
        // TODO: plumb Env::IOActivity, Env::IOPriority
        auto s = log.writer->file()->Close({});
        s.PermitUncheckedError();
-        log_write_mutex_.Lock();
+        wal_write_mutex_.Lock();
        log.writer->PublishIfClosed();
        assert(&log == &logs_.front());
        log.FinishSync();
-        log_sync_cv_.SignalAll();
+        wal_sync_cv_.SignalAll();
      }
-      logs_to_free_.push_back(log.ReleaseWriter());
+      wals_to_free_.push_back(log.ReleaseWriter());
      logs_.pop_front();
    }
    // Current log cannot be obsolete.
@@ -350,16 +350,16 @@ void DBImpl::FindObsoleteFiles(JobContext* job_context, bool force,
  }

  // We're just cleaning up for DB::Write().
-  assert(job_context->logs_to_free.empty());
-  job_context->logs_to_free = logs_to_free_;
+  assert(job_context->wals_to_free.empty());
+  job_context->wals_to_free = wals_to_free_;

-  logs_to_free_.clear();
-  log_write_mutex_.Unlock();
+  wals_to_free_.clear();
+  wal_write_mutex_.Unlock();
  if (mutex_unlocked) {
    mutex_.Lock();
  }
-  job_context->log_recycle_files.assign(log_recycle_files_.begin(),
-                                        log_recycle_files_.end());
+  job_context->log_recycle_files.assign(wal_recycle_files_.begin(),
+                                        wal_recycle_files_.end());
 }

 // Delete obsolete files and log status and information of file deletion
@@ -431,7 +431,7 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
                                            state.sst_live.end());
  std::unordered_set<uint64_t> blob_live_set(state.blob_live.begin(),
                                             state.blob_live.end());
-  std::unordered_set<uint64_t> log_recycle_files_set(
+  std::unordered_set<uint64_t> wal_recycle_files_set(
      state.log_recycle_files.begin(), state.log_recycle_files.end());
  std::unordered_set<uint64_t> quarantine_files_set(
      state.files_to_quarantine.begin(), state.files_to_quarantine.end());
@@ -491,13 +491,13 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
      std::unique(candidate_files.begin(), candidate_files.end()),
      candidate_files.end());

-  if (state.prev_total_log_size > 0) {
+  if (state.prev_wals_total_size > 0) {
    ROCKS_LOG_INFO(immutable_db_options_.info_log,
                   "[JOB %d] Try to delete WAL files size %" PRIu64
                   ", prev total WAL file size %" PRIu64
                   ", number of live WAL files %" ROCKSDB_PRIszt ".\n",
                   state.job_id, state.size_log_to_delete,
-                   state.prev_total_log_size, state.num_alive_log_files);
+                   state.prev_wals_total_size, state.num_alive_wal_files);
  }

  std::vector<std::string> old_info_log_files;
@@ -532,7 +532,7 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
  optsfile_num2 = std::min(optsfile_num2, state.min_options_file_number);

  // Close WALs before trying to delete them.
-  for (const auto w : state.logs_to_free) {
+  for (const auto w : state.wals_to_free) {
    // TODO: maybe check the return value of Close.
    // TODO: plumb Env::IOActivity, Env::IOPriority
    auto s = w->Close({});
@@ -559,8 +559,8 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
      case kWalFile:
        keep = ((number >= state.log_number) ||
                (number == state.prev_log_number) ||
-                (log_recycle_files_set.find(number) !=
-                 log_recycle_files_set.end()));
+                (wal_recycle_files_set.find(number) !=
+                 wal_recycle_files_set.end()));
        break;
      case kDescriptorFile:
        // Keep my manifest file, and any newer incarnations'
@@ -70,9 +70,6 @@ Status DBImplFollower::Recover(
    }
    return s;
  }
-  if (immutable_db_options_.paranoid_checks && s.ok()) {
-    s = CheckConsistency();
-  }
  if (s.ok()) {
    default_cf_handle_ = new ColumnFamilyHandleImpl(
        versions_->GetColumnFamilySet()->GetDefault(), this, &mutex_);
@@ -191,12 +191,6 @@ DBOptions SanitizeOptions(const std::string& dbname, const DBOptions& src,
                   "wal_compression is disabled since only zstd is supported");
  }

-  if (!result.paranoid_checks) {
-    result.skip_checking_sst_file_sizes_on_db_open = true;
-    ROCKS_LOG_INFO(result.info_log,
-                   "file size check will be skipped during open.");
-  }
-
  return result;
 }

@@ -605,7 +599,7 @@ Status DBImpl::Recover(
        // allow_ingest_behind does not support Level Compaction,
        // and per_key_placement can have infinite compaction loop for Level
        // Compaction. Adjust to_level here just to be safe.
-        if (cfd->ioptions().allow_ingest_behind ||
+        if (cfd->AllowIngestBehind() ||
            moptions.preclude_last_level_data_seconds > 0) {
          to_level -= 1;
        }
@@ -694,9 +688,6 @@ Status DBImpl::Recover(
    s = MaybeUpdateNextFileNumber(recovery_ctx);
  }

-  if (immutable_db_options_.paranoid_checks && s.ok()) {
-    s = CheckConsistency();
-  }
  if (s.ok() && !read_only) {
    // TODO: share file descriptors (FSDirectory) with SetDirectories above
    std::map<std::string, std::shared_ptr<FSDirectory>> created_dirs;
@@ -1119,7 +1110,7 @@ void DBOpenLogRecordReadReporter::Corruption(size_t bytes, const Status& s,
                 static_cast<int>(bytes), s.ToString().c_str());
  if (status != nullptr && status->ok()) {
    *status = s;
-    corrupted_log_number_ = log_number;
+    corrupted_wal_number_ = log_number;
  }
 }

@@ -1764,8 +1755,12 @@ Status DBImpl::MaybeHandleStopReplayForCorruptionForInconsistency(
        ROCKS_LOG_ERROR(immutable_db_options_.info_log,
                        "Column family inconsistency: SST file contains data"
                        " beyond the point of corruption.");
-        status = Status::Corruption("SST file is ahead of WALs in CF " +
-                                    cfd->GetName());
+        status = Status::Corruption(
+            "Column family inconsistency: SST file contains data"
+            " beyond the point of corruption in CF " +
+            cfd->GetName() +
+            ". WAL recovery stopped at corruption point, but SST files"
+            " contain newer data.");
        return status;
      }
    }
@@ -1902,8 +1897,8 @@ void DBImpl::FinishLogFilesRecovery(int job_id, const Status& status) {
 }

 Status DBImpl::GetLogSizeAndMaybeTruncate(uint64_t wal_number, bool truncate,
-                                          LogFileNumberSize* log_ptr) {
-  LogFileNumberSize log(wal_number);
+                                          WalFileNumberSize* log_ptr) {
+  WalFileNumberSize log(wal_number);
  std::string fname =
      LogFileName(immutable_db_options_.GetWalDir(), wal_number);
  Status s;
@@ -1946,27 +1941,27 @@ Status DBImpl::RestoreAliveLogFiles(const std::vector<uint64_t>& wal_numbers) {
  assert(immutable_db_options_.avoid_flush_during_recovery);
  // Mark these as alive so they'll be considered for deletion later by
  // FindObsoleteFiles()
-  total_log_size_ = 0;
-  log_empty_ = false;
+  wals_total_size_.StoreRelaxed(0);
+  wal_empty_ = false;
  uint64_t min_wal_with_unflushed_data =
      versions_->MinLogNumberWithUnflushedData();
  for (auto wal_number : wal_numbers) {
    if (!allow_2pc() && wal_number < min_wal_with_unflushed_data) {
      // In non-2pc mode, the WAL files not backing unflushed data are not
-      // alive, thus should not be added to the alive_log_files_.
+      // alive, thus should not be added to the alive_wal_files_.
      continue;
    }
    // We preallocate space for wals, but then after a crash and restart, those
    // preallocated space are not needed anymore. It is likely only the last
    // log has such preallocated space, so we only truncate for the last log.
-    LogFileNumberSize log;
+    WalFileNumberSize log;
    s = GetLogSizeAndMaybeTruncate(
        wal_number, /*truncate=*/(wal_number == wal_numbers.back()), &log);
    if (!s.ok()) {
      break;
    }
-    total_log_size_ += log.size;
-    alive_log_files_.push_back(log);
+    wals_total_size_.FetchAddRelaxed(log.size);
+    alive_wal_files_.push_back(log);
  }
  return s;
 }
@@ -2000,6 +1995,9 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
  const size_t ts_sz = ucmp->timestamp_size();
  const bool logical_strip_timestamp =
      ts_sz > 0 && !cfd->ioptions().persist_user_defined_timestamps;
+  // Note that here we treat flush as level 0 compaction in internal stats
+  InternalStats::CompactionStats flush_stats(CompactionReason::kFlush,
+                                             1 /* count */);
  {
    ScopedArenaPtr<InternalIterator> iter(
        logical_strip_timestamp
@@ -2072,19 +2070,20 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
          kMaxSequenceNumber);
      Version* version = cfd->current();
      version->Ref();
-      uint64_t num_input_entries = 0;
-      s = BuildTable(dbname_, versions_.get(), immutable_db_options_, tboptions,
-                     file_options_for_compaction_, cfd->table_cache(),
-                     iter.get(), std::move(range_del_iters), &meta,
-                     &blob_file_additions, snapshot_seqs, earliest_snapshot,
-                     earliest_write_conflict_snapshot, kMaxSequenceNumber,
-                     snapshot_checker, paranoid_file_checks,
-                     cfd->internal_stats(), &io_s, io_tracer_,
-                     BlobFileCreationReason::kRecovery,
-                     nullptr /* seqno_to_time_mapping */, &event_logger_,
-                     job_id, nullptr /* table_properties */, write_hint,
-                     nullptr /*full_history_ts_low*/, &blob_callback_, version,
-                     &num_input_entries);
+      TableProperties temp_table_proerties;
+      s = BuildTable(
+          dbname_, versions_.get(), immutable_db_options_, tboptions,
+          file_options_for_compaction_, cfd->table_cache(), iter.get(),
+          std::move(range_del_iters), &meta, &blob_file_additions,
+          snapshot_seqs, earliest_snapshot, earliest_write_conflict_snapshot,
+          kMaxSequenceNumber, snapshot_checker, paranoid_file_checks,
+          cfd->internal_stats(), &io_s, io_tracer_,
+          BlobFileCreationReason::kRecovery,
+          nullptr /* seqno_to_time_mapping */, &event_logger_, job_id,
+          &temp_table_proerties /* table_properties */, write_hint,
+          nullptr /*full_history_ts_low*/, &blob_callback_, version,
+          nullptr /* memtable_payload_bytes */,
+          nullptr /* memtable_garbage_bytes */, &flush_stats);
      version->Unref();
      LogFlush(immutable_db_options_.info_log);
      ROCKS_LOG_DEBUG(immutable_db_options_.info_log,
@@ -2100,10 +2099,31 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
      }

      uint64_t total_num_entries = mem->NumEntries();
-      if (s.ok() && total_num_entries != num_input_entries) {
+      if (s.ok() && total_num_entries != flush_stats.num_input_records) {
        std::string msg = "Expected " + std::to_string(total_num_entries) +
                          " entries in memtable, but read " +
-                          std::to_string(num_input_entries);
+                          std::to_string(flush_stats.num_input_records);
+        ROCKS_LOG_WARN(immutable_db_options_.info_log,
+                       "[%s] [JOB %d] Level-0 flush during recover: %s",
+                       cfd->GetName().c_str(), job_id, msg.c_str());
+        if (immutable_db_options_.flush_verify_memtable_count) {
+          s = Status::Corruption(msg);
+        }
+      }
+      // Only verify on table with format collects table properties
+      const auto& mutable_cf_options = cfd->GetLatestMutableCFOptions();
+      if (s.ok() &&
+          (mutable_cf_options.table_factory->IsInstanceOf(
+               TableFactory::kBlockBasedTableName()) ||
+           mutable_cf_options.table_factory->IsInstanceOf(
+               TableFactory::kPlainTableName())) &&
+          flush_stats.num_output_records != temp_table_proerties.num_entries) {
+        std::string msg =
+            "Number of keys in flush output SST files does not match "
+            "number of keys added to the table. Expected " +
+            std::to_string(flush_stats.num_output_records) + " but there are " +
+            std::to_string(temp_table_proerties.num_entries) +
+            " in output SST files";
        ROCKS_LOG_WARN(immutable_db_options_.info_log,
                       "[%s] [JOB %d] Level-0 flush during recover: %s",
                       cfd->GetName().c_str(), job_id, msg.c_str());
@@ -2151,25 +2171,25 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
    }
  }

-  InternalStats::CompactionStats stats(CompactionReason::kFlush, 1);
-  stats.micros = immutable_db_options_.clock->NowMicros() - start_micros;
+  flush_stats.micros = immutable_db_options_.clock->NowMicros() - start_micros;

  if (has_output) {
-    stats.bytes_written = meta.fd.GetFileSize();
-    stats.num_output_files = 1;
+    flush_stats.bytes_written = meta.fd.GetFileSize();
+    flush_stats.num_output_files = 1;
  }

  const auto& blobs = edit->GetBlobFileAdditions();
  for (const auto& blob : blobs) {
-    stats.bytes_written_blob += blob.GetTotalBlobBytes();
+    flush_stats.bytes_written_blob += blob.GetTotalBlobBytes();
  }

-  stats.num_output_files_blob = static_cast<int>(blobs.size());
+  flush_stats.num_output_files_blob = static_cast<int>(blobs.size());

-  cfd->internal_stats()->AddCompactionStats(level, Env::Priority::USER, stats);
+  cfd->internal_stats()->AddCompactionStats(level, Env::Priority::USER,
+                                            flush_stats);
  cfd->internal_stats()->AddCFStats(
      InternalStats::BYTES_FLUSHED,
-      stats.bytes_written + stats.bytes_written_blob);
+      flush_stats.bytes_written + flush_stats.bytes_written_blob);
  RecordTick(stats_, COMPACT_WRITE_BYTES, meta.fd.GetFileSize());
  return s;
 }
@@ -2449,18 +2469,18 @@ Status DBImpl::Open(const DBOptions& db_options, const std::string& dbname,
    if (s.ok()) {
      // Prevent log files created by previous instance from being recycled.
      // They might be in alive_log_file_, and might get recycled otherwise.
-      impl->min_log_number_to_recycle_ = new_log_number;
+      impl->min_wal_number_to_recycle_ = new_log_number;
    }
    if (s.ok()) {
-      InstrumentedMutexLock wl(&impl->log_write_mutex_);
-      impl->logfile_number_ = new_log_number;
+      InstrumentedMutexLock wl(&impl->wal_write_mutex_);
+      impl->cur_wal_number_ = new_log_number;
      assert(new_log != nullptr);
      assert(impl->logs_.empty());
      impl->logs_.emplace_back(new_log_number, new_log);
    }

    if (s.ok()) {
-      impl->alive_log_files_.emplace_back(impl->logfile_number_);
+      impl->alive_wal_files_.emplace_back(impl->cur_wal_number_);
      // In WritePrepared there could be gap in sequence numbers. This breaks
      // the trick we use in kPointInTimeRecovery which assumes the first seq in
      // the log right after the corrupted log is one larger than the last seq
@@ -2473,14 +2493,14 @@ Status DBImpl::Open(const DBOptions& db_options, const std::string& dbname,
      if (recovered_seq != kMaxSequenceNumber) {
        WriteBatch empty_batch;
        WriteBatchInternal::SetSequence(&empty_batch, recovered_seq);
-        uint64_t log_used, log_size;
+        uint64_t wal_used, log_size;
        log::Writer* log_writer = impl->logs_.back().writer;
-        LogFileNumberSize& log_file_number_size = impl->alive_log_files_.back();
+        WalFileNumberSize& wal_file_number_size = impl->alive_wal_files_.back();

-        assert(log_writer->get_log_number() == log_file_number_size.number);
+        assert(log_writer->get_log_number() == wal_file_number_size.number);
        impl->mutex_.AssertHeld();
-        s = impl->WriteToWAL(empty_batch, write_options, log_writer, &log_used,
-                             &log_size, log_file_number_size, recovered_seq);
+        s = impl->WriteToWAL(empty_batch, write_options, log_writer, &wal_used,
+                             &log_size, wal_file_number_size, recovered_seq);
        if (s.ok()) {
          // Need to fsync, otherwise it might get lost after a power reset.
          s = impl->FlushWAL(write_options, false);
@@ -185,16 +185,10 @@ Iterator* DBImplReadOnly::NewIterator(const ReadOptions& _read_options,
          ? static_cast<const SnapshotImpl*>(read_options.snapshot)->number_
          : latest_snapshot;
  ReadCallback* read_callback = nullptr;  // No read callback provided.
-  auto db_iter = NewArenaWrappedDbIterator(
-      env_, read_options, cfd->ioptions(), super_version->mutable_cf_options,
-      super_version->current, read_seq,
-      super_version->mutable_cf_options.max_sequential_skip_in_iterations,
-      super_version->version_number, read_callback);
-  auto internal_iter = NewInternalIterator(
-      db_iter->GetReadOptions(), cfd, super_version, db_iter->GetArena(),
-      read_seq, /* allow_unprepared_value */ true, db_iter);
-  db_iter->SetIterUnderDBIter(internal_iter);
-  return db_iter;
+  return NewArenaWrappedDbIterator(
+      env_, read_options, cfh, super_version, read_seq, read_callback, this,
+      /*expose_blob_index=*/false, /*allow_refresh=*/false,
+      /*allow_mark_memtable_for_flush=*/false);
 }

 Status DBImplReadOnly::NewIterators(
@@ -231,36 +225,32 @@ Status DBImplReadOnly::NewIterators(
          ? static_cast<const SnapshotImpl*>(read_options.snapshot)->number_
          : latest_snapshot;

-  autovector<std::tuple<ColumnFamilyData*, SuperVersion*>> cfd_to_sv;
+  autovector<std::tuple<ColumnFamilyHandleImpl*, SuperVersion*>> cfh_to_sv;

  const bool check_read_ts =
      read_options.timestamp && read_options.timestamp->size() > 0;
  for (auto cfh : column_families) {
    auto* cfd = static_cast_with_check<ColumnFamilyHandleImpl>(cfh)->cfd();
    auto* sv = cfd->GetSuperVersion()->Ref();
-    cfd_to_sv.emplace_back(cfd, sv);
+    cfh_to_sv.emplace_back(static_cast_with_check<ColumnFamilyHandleImpl>(cfh),
+                           sv);
    if (check_read_ts) {
      const Status s =
          FailIfReadCollapsedHistory(cfd, sv, *(read_options.timestamp));
      if (!s.ok()) {
-        for (auto prev_entry : cfd_to_sv) {
+        for (auto prev_entry : cfh_to_sv) {
          std::get<1>(prev_entry)->Unref();
        }
        return s;
      }
    }
  }
-  assert(cfd_to_sv.size() == column_families.size());
-  for (auto [cfd, sv] : cfd_to_sv) {
+  assert(cfh_to_sv.size() == column_families.size());
+  for (auto [cfh, sv] : cfh_to_sv) {
    auto* db_iter = NewArenaWrappedDbIterator(
-        env_, read_options, cfd->ioptions(), sv->mutable_cf_options,
-        sv->current, read_seq,
-        sv->mutable_cf_options.max_sequential_skip_in_iterations,
-        sv->version_number, read_callback);
-    auto* internal_iter = NewInternalIterator(
-        db_iter->GetReadOptions(), cfd, sv, db_iter->GetArena(), read_seq,
-        /* allow_unprepared_value */ true, db_iter);
-    db_iter->SetIterUnderDBIter(internal_iter);
+        env_, read_options, cfh, sv, read_seq, read_callback, this,
+        /*expose_blob_index=*/false, /*allow_refresh=*/false,
+        /*allow_mark_memtable_for_flush=*/false);
    iterators->push_back(db_iter);
  }

@@ -49,9 +49,6 @@ Status DBImplSecondary::Recover(
    }
    return s;
  }
-  if (immutable_db_options_.paranoid_checks && s.ok()) {
-    s = CheckConsistency();
-  }
  // Initial max_total_in_memory_state_ before recovery logs.
  max_total_in_memory_state_ = 0;
  for (auto cfd : *versions_->GetColumnFamilySet()) {
@@ -566,17 +563,10 @@ ArenaWrappedDBIter* DBImplSecondary::NewIteratorImpl(
  assert(snapshot == kMaxSequenceNumber);
  snapshot = versions_->LastSequence();
  assert(snapshot != kMaxSequenceNumber);
-  auto db_iter = NewArenaWrappedDbIterator(
-      env_, read_options, cfh->cfd()->ioptions(),
-      super_version->mutable_cf_options, super_version->current, snapshot,
-      super_version->mutable_cf_options.max_sequential_skip_in_iterations,
-      super_version->version_number, read_callback, cfh, expose_blob_index,
-      allow_refresh);
-  auto internal_iter = NewInternalIterator(
-      db_iter->GetReadOptions(), cfh->cfd(), super_version, db_iter->GetArena(),
-      snapshot, /* allow_unprepared_value */ true, db_iter);
-  db_iter->SetIterUnderDBIter(internal_iter);
-  return db_iter;
+  return NewArenaWrappedDbIterator(env_, read_options, cfh, super_version,
+                                   snapshot, read_callback, this,
+                                   expose_blob_index, allow_refresh,
+                                   /*allow_mark_memtable_for_flush=*/false);
 }

 Status DBImplSecondary::NewIterators(
@@ -660,58 +650,15 @@ Status DBImplSecondary::NewIterators(
  return Status::OK();
 }

-Status DBImplSecondary::CheckConsistency() {
-  mutex_.AssertHeld();
-  Status s = DBImpl::CheckConsistency();
-  // If DBImpl::CheckConsistency() which is stricter returns success, then we
-  // do not need to give a second chance.
-  if (s.ok()) {
-    return s;
-  }
-  // It's possible that DBImpl::CheckConssitency() can fail because the primary
-  // may have removed certain files, causing the GetFileSize(name) call to
-  // fail and returning a PathNotFound. In this case, we take a best-effort
-  // approach and just proceed.
-  TEST_SYNC_POINT_CALLBACK(
-      "DBImplSecondary::CheckConsistency:AfterFirstAttempt", &s);
-
-  if (immutable_db_options_.skip_checking_sst_file_sizes_on_db_open) {
-    return Status::OK();
-  }
-
-  std::vector<LiveFileMetaData> metadata;
-  versions_->GetLiveFilesMetaData(&metadata);
-
-  std::string corruption_messages;
-  for (const auto& md : metadata) {
-    // md.name has a leading "/".
-    std::string file_path = md.db_path + md.name;
-
-    uint64_t fsize = 0;
-    s = env_->GetFileSize(file_path, &fsize);
-    if (!s.ok() &&
-        (env_->GetFileSize(Rocks2LevelTableFileName(file_path), &fsize).ok() ||
-         s.IsPathNotFound())) {
-      s = Status::OK();
-    }
-    if (!s.ok()) {
-      corruption_messages +=
-          "Can't access " + md.name + ": " + s.ToString() + "\n";
-    }
-  }
-  return corruption_messages.empty() ? Status::OK()
-                                     : Status::Corruption(corruption_messages);
-}
-
 Status DBImplSecondary::TryCatchUpWithPrimary() {
  assert(versions_.get() != nullptr);
-  assert(manifest_reader_.get() != nullptr);
  Status s;
  // read the manifest and apply new changes to the secondary instance
  std::unordered_set<ColumnFamilyData*> cfds_changed;
  JobContext job_context(0, true /*create_superversion*/);
  {
    InstrumentedMutexLock lock_guard(&mutex_);
+    assert(manifest_reader_.get() != nullptr);
    s = static_cast_with_check<ReactiveVersionSet>(versions_.get())
            ->ReadAndApply(&mutex_, &manifest_reader_,
                           manifest_reader_status_.get(), &cfds_changed,
@@ -735,13 +682,13 @@ Status DBImplSecondary::TryCatchUpWithPrimary() {
    // instance
    if (s.ok()) {
      s = FindAndRecoverLogFiles(&cfds_changed, &job_context);
-    }
-    if (s.IsPathNotFound()) {
-      ROCKS_LOG_INFO(
-          immutable_db_options_.info_log,
-          "Secondary tries to read WAL, but WAL file(s) have already "
-          "been purged by primary.");
-      s = Status::OK();
+      if (s.IsPathNotFound()) {
+        ROCKS_LOG_INFO(
+            immutable_db_options_.info_log,
+            "Secondary tries to read WAL, but WAL file(s) have already "
+            "been purged by primary.");
+        s = Status::OK();
+      }
    }
    if (s.ok()) {
      for (auto cfd : cfds_changed) {
@@ -901,7 +848,6 @@ Status DBImplSecondary::CompactWithoutInstallation(

  VersionStorageInfo* vstorage = version->storage_info();

-  // Use comp_options to reuse some CompactFiles functions
  CompactionOptions comp_options;
  comp_options.compression = kDisableCompressionOption;
  comp_options.output_file_size_limit = MaxFileSizeForLevel(
@@ -916,17 +862,31 @@ Status DBImplSecondary::CompactWithoutInstallation(
    ROCKS_LOG_ERROR(
        immutable_db_options_.info_log,
        "GetCompactionInputsFromFileNumbers() failed - %s.\n DebugString: %s",
-        s.ToString().c_str(), version->DebugString().c_str());
+        s.ToString().c_str(), version->DebugString(/*hex=*/true).c_str());
    return s;
  }

+  const int job_id = next_job_id_.fetch_add(1);
+  JobContext job_context(job_id, true /*create_superversion*/);
+  std::vector<SequenceNumber> snapshots = input.snapshots;
+
+  // TODO - snapshot_checker support in Remote Compaction
+  job_context.InitSnapshotContext(/*checker=*/nullptr,
+                                  /*managed_snapshot=*/nullptr,
+                                  kMaxSequenceNumber, std::move(snapshots));
+
+  // TODO - consider serializing the entire Compaction object and using it as
+  // input instead of recreating it in the remote worker
  std::unique_ptr<Compaction> c;
  assert(cfd->compaction_picker());
-  c.reset(cfd->compaction_picker()->CompactFiles(
+  c.reset(cfd->compaction_picker()->PickCompactionForCompactFiles(
      comp_options, input_files, input.output_level, vstorage,
-      cfd->GetLatestMutableCFOptions(), mutable_db_options_, 0));
+      cfd->GetLatestMutableCFOptions(), mutable_db_options_, 0,
+      /*earliest_snapshot=*/job_context.snapshot_seqs.empty()
+          ? kMaxSequenceNumber
+          : job_context.snapshot_seqs.front(),
+      job_context.snapshot_checker));
  assert(c != nullptr);
-
  c->FinalizeInputInfo(version);

  // Create output directory if it's not existed yet
@@ -939,8 +899,6 @@ Status DBImplSecondary::CompactWithoutInstallation(
  LogBuffer log_buffer(InfoLogLevel::INFO_LEVEL,
                       immutable_db_options_.info_log.get());

-  const int job_id = next_job_id_.fetch_add(1);
-
  // use primary host's db_id for running the compaction, but db_session_id is
  // using the local one, which is to make sure the unique id is unique from
  // the remote compactors. Because the id is generated from db_id,
@@ -951,7 +909,7 @@ Status DBImplSecondary::CompactWithoutInstallation(
      job_id, c.get(), immutable_db_options_, mutable_db_options_,
      file_options_for_compaction_, versions_.get(), &shutting_down_,
      &log_buffer, output_dir.get(), stats_, &mutex_, &error_handler_,
-      input.snapshots, table_cache_, &event_logger_, dbname_, io_tracer_,
+      &job_context, table_cache_, &event_logger_, dbname_, io_tracer_,
      options.canceled ? *options.canceled : kManualCompactionCanceledFalse_,
      input.db_id, db_session_id_, secondary_path_, input, result);

@@ -991,7 +949,7 @@ Status DB::OpenAndCompact(
  }

  // 2. Load the options
-  DBOptions db_options;
+  DBOptions base_db_options;
  ConfigOptions config_options;
  config_options.env = override_options.env;
  config_options.ignore_unknown_options = true;
@@ -1004,13 +962,22 @@ Status DB::OpenAndCompact(
  std::string options_file_name =
      OptionsFileName(name, compaction_input.options_file_number);

-  s = LoadOptionsFromFile(config_options, options_file_name, &db_options,
+  s = LoadOptionsFromFile(config_options, options_file_name, &base_db_options,
                          &all_column_families);
  if (!s.ok()) {
    return s;
  }

-  // 3. Override pointer configurations in DBOptions with
+  // 3. Options to Override
+  // Override serializable configurations from override_options.options_map
+  DBOptions db_options;
+  s = GetDBOptionsFromMap(config_options, base_db_options,
+                          override_options.options_map, &db_options);
+  if (!s.ok()) {
+    return s;
+  }
+
+  // Override options that are directly set as shared ptrs in
  // CompactionServiceOptionsOverride
  db_options.env = override_options.env;
  db_options.file_checksum_gen_factory =
@@ -1021,6 +988,7 @@ Status DB::OpenAndCompact(
  // We will close the DB after the compaction anyway.
  // Open as many files as needed for the compaction.
  db_options.max_open_files = -1;
+  db_options.info_log = override_options.info_log;

  // 4. Filter CFs that are needed for OpenAndCompact()
  // We do not need to open all column families for the remote compaction.
@@ -1030,6 +998,18 @@ Status DB::OpenAndCompact(
  std::vector<ColumnFamilyDescriptor> column_families;
  for (auto& cf : all_column_families) {
    if (cf.name == compaction_input.cf_name) {
+      ColumnFamilyOptions cf_options;
+      // Override serializable configurations from override_options.options_map
+      s = GetColumnFamilyOptionsFromMap(config_options, cf.options,
+                                        override_options.options_map,
+                                        &cf_options);
+      if (!s.ok()) {
+        return s;
+      }
+      cf.options = std::move(cf_options);
+
+      // Override options that are directly set as shared ptrs in
+      // CompactionServiceOptionsOverride
      cf.options.comparator = override_options.comparator;
      cf.options.merge_operator = override_options.merge_operator;
      cf.options.compaction_filter = override_options.compaction_filter;
@@ -1041,6 +1021,7 @@ Status DB::OpenAndCompact(
          override_options.sst_partitioner_factory;
      cf.options.table_properties_collector_factories =
          override_options.table_properties_collector_factories;
+
      column_families.emplace_back(cf);
    } else if (cf.name == kDefaultColumnFamilyName) {
      column_families.emplace_back(cf);
@@ -1057,6 +1038,9 @@ Status DB::OpenAndCompact(
  }
  assert(db);

+  TEST_SYNC_POINT_CALLBACK(
+      "DBImplSecondary::OpenAndCompact::AfterOpenAsSecondary:0", db);
+
  // 6. Find the handle of the Column Family that this will compact
  ColumnFamilyHandle* cfh = nullptr;
  for (auto* handle : handles) {
@@ -248,12 +248,6 @@ class DBImplSecondary : public DBImpl {
  Status MaybeInitLogReader(uint64_t log_number,
                            log::FragmentBufferedReader** log_reader);

-  // Check if all live files exist on file system and that their file sizes
-  // matche to the in-memory records. It is possible that some live files may
-  // have been deleted by the primary. In this case, CheckConsistency() does
-  // not flag the missing file as inconsistency.
-  Status CheckConsistency() override;
-
 #ifndef NDEBUG
  Status TEST_CompactWithoutInstallation(const OpenAndCompactOptions& options,
                                         ColumnFamilyHandle* cfh,
@@ -157,7 +157,7 @@ Status DBImpl::Write(const WriteOptions& write_options, WriteBatch* my_batch) {
  if (s.ok()) {
    s = WriteImpl(write_options, my_batch, /*callback=*/nullptr,
                  /*user_write_cb=*/nullptr,
-                  /*log_used=*/nullptr);
+                  /*wal_used=*/nullptr);
  }
  return s;
 }
@@ -190,11 +190,38 @@ Status DBImpl::WriteWithCallback(const WriteOptions& write_options,
  return s;
 }

-Status DBImpl::IngestWBWI(std::shared_ptr<WriteBatchWithIndex> wbwi,
-                          const WBWIMemTable::SeqnoRange& assigned_seqno,
-                          uint64_t prep_log,
-                          SequenceNumber last_seqno_after_ingest,
-                          bool memtable_updated, bool ignore_missing_cf) {
+Status DBImpl::IngestWriteBatchWithIndex(
+    const WriteOptions& write_options,
+    std::shared_ptr<WriteBatchWithIndex> wbwi) {
+  if (!wbwi) {
+    return Status::InvalidArgument("Batch is nullptr!");
+  }
+  if (!write_options.disableWAL) {
+    return Status::NotSupported(
+        "IngestWriteBatchWithIndex does not support disableWAL=true");
+  }
+  Status s;
+  if (write_options.protection_bytes_per_key > 0) {
+    s = WriteBatchInternal::UpdateProtectionInfo(
+        wbwi->GetWriteBatch(), write_options.protection_bytes_per_key);
+  }
+  if (s.ok()) {
+    WriteBatch dummy_empty_batch;
+    s = WriteImpl(
+        write_options, /*updates=*/&dummy_empty_batch, /*callback=*/nullptr,
+        /*user_write_cb=*/nullptr, /*log_used=*/nullptr, /*log_ref=*/0,
+        /*disable_memtable=*/false, /*seq_used=*/nullptr,
+        /*batch_cnt=*/0, /*pre_release_callback=*/nullptr,
+        /*post_memtable_callback=*/nullptr, /*wbwi=*/wbwi);
+  }
+  return s;
+}
+
+Status DBImpl::IngestWBWIAsMemtable(
+    std::shared_ptr<WriteBatchWithIndex> wbwi,
+    const WBWIMemTable::SeqnoRange& assigned_seqno, uint64_t min_prep_log,
+    SequenceNumber last_seqno_after_ingest, bool memtable_updated,
+    bool ignore_missing_cf) {
  // Keys in new memtable have seqno > last_seqno_after_ingest >= keys in wbwi.
  assert(assigned_seqno.upper_bound <= last_seqno_after_ingest);
  // Keys in the current memtable have seqno <= LastSequence() < keys in wbwi.
@@ -238,12 +265,30 @@ Status DBImpl::IngestWBWI(std::shared_ptr<WriteBatchWithIndex> wbwi,
    wbwi_memtable->AssignSequenceNumbers(assigned_seqno);
    // This is needed to keep the WAL that contains Prepare alive until
    // committed data in this memtable is persisted.
-    wbwi_memtable->SetMinPrepLog(prep_log);
+    wbwi_memtable->SetMinPrepLog(min_prep_log);
    memtables.push_back(wbwi_memtable);
    cfd->Ref();
    cfds.push_back(cfd);
  }

+  autovector<ColumnFamilyData*> cfds_for_atomic_flush;
+  if (immutable_db_options_.atomic_flush) {
+    SelectColumnFamiliesForAtomicFlush(&cfds_for_atomic_flush);
+    for (auto cfd : cfds_for_atomic_flush) {
+      bool found = false;
+      for (auto existing_cfd : cfds) {
+        if (existing_cfd == cfd) {
+          found = true;
+          break;
+        }
+      }
+      if (!found) {
+        cfd->Ref();
+        cfds.push_back(cfd);
+      }
+    }
+  }
+
  // Stop writes to the DB by entering both write threads
  WriteThread::Writer nonmem_w;
  if (two_write_queues_) {
@@ -253,15 +298,16 @@ Status DBImpl::IngestWBWI(std::shared_ptr<WriteBatchWithIndex> wbwi,

  // Switch memtable and add WBWIMemTables
  Status s;
-  for (size_t i = 0; i < memtables.size(); ++i) {
-    assert(!immutable_db_options_.atomic_flush);
-    // NOTE: to support atomic flush, need to call
-    // SelectColumnFamiliesForAtomicFlush()
+  for (size_t i = 0; i < cfds.size(); ++i) {
    WriteContext write_context;
    // TODO: not switch on empty memtable, may need to update metadata
    //   like NextLogNumber(), earliest_seqno and memtable id.
-    s = SwitchMemtable(cfds[i], &write_context, memtables[i],
-                       last_seqno_after_ingest);
+    if (i < memtables.size()) {
+      s = SwitchMemtable(cfds[i], &write_context, memtables[i],
+                         last_seqno_after_ingest);
+    } else {
+      s = SwitchMemtable(cfds[i], &write_context);
+    }
    if (!s.ok()) {
      // SwitchMemtable() can only fail if a new WAL is to be created, this
      // should only happen for the first call to SwitchMemtable(). log will
@@ -301,9 +347,18 @@ Status DBImpl::IngestWBWI(std::shared_ptr<WriteBatchWithIndex> wbwi,
        continue;
      }
      cfd->imm()->FlushRequested();
+      if (!immutable_db_options_.atomic_flush) {
+        FlushRequest flush_req;
+        // TODO: a new flush reason for ingesting memtable
+        GenerateFlushRequest({cfd}, FlushReason::kExternalFileIngestion,
+                             &flush_req);
+        EnqueuePendingFlush(flush_req);
+      }
+    }
+    if (immutable_db_options_.atomic_flush) {
+      AssignAtomicFlushSeq(cfds);
      FlushRequest flush_req;
-      // TODO: a new flush reason for ingesting memtable
-      GenerateFlushRequest({cfd}, FlushReason::kExternalFileIngestion,
+      GenerateFlushRequest(cfds, FlushReason::kExternalFileIngestion,
                           &flush_req);
      EnqueuePendingFlush(flush_req);
    }
@@ -314,13 +369,12 @@ Status DBImpl::IngestWBWI(std::shared_ptr<WriteBatchWithIndex> wbwi,

 Status DBImpl::WriteImpl(const WriteOptions& write_options,
                         WriteBatch* my_batch, WriteCallback* callback,
-                         UserWriteCallback* user_write_cb, uint64_t* log_used,
+                         UserWriteCallback* user_write_cb, uint64_t* wal_used,
                         uint64_t log_ref, bool disable_memtable,
                         uint64_t* seq_used, size_t batch_cnt,
                         PreReleaseCallback* pre_release_callback,
                         PostMemTableCallback* post_memtable_callback,
-                         std::shared_ptr<WriteBatchWithIndex> wbwi,
-                         uint64_t prep_log) {
+                         std::shared_ptr<WriteBatchWithIndex> wbwi) {
  assert(!seq_per_batch_ || batch_cnt != 0);
  assert(my_batch == nullptr || my_batch->Count() == 0 ||
         write_options.protection_bytes_per_key == 0 ||
@@ -409,9 +463,17 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
    return Status::NotSupported(
        "DeleteRange is not compatible with row cache.");
  }
+  // Whether the WBWI is from transaction commit or a direct write
+  // (IngestWriteBatchWithIndex())
+  bool ingest_wbwi_for_commit = false;
  if (wbwi) {
-    assert(prep_log > 0);
-    // Used only in WriteCommittedTxn::CommitInternal() with no `callback`.
+    if (my_batch->HasCommit()) {
+      ingest_wbwi_for_commit = true;
+      assert(log_ref);
+    } else {
+      // Only supports disableWAL for directly ingesting WBWI for now.
+      assert(write_options.disableWAL);
+    }
    assert(!callback);
    if (immutable_db_options_.unordered_write) {
      return Status::NotSupported(
@@ -421,9 +483,9 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
      return Status::NotSupported(
          "Ingesting WriteBatch does not support pipelined_write");
    }
-    if (immutable_db_options_.atomic_flush) {
+    if (!wbwi->GetOverwriteKey()) {
      return Status::NotSupported(
-          "Ingesting WriteBatch does not support atomic_flush");
+          "WriteBatchWithIndex ingestion requires overwrite_key=true");
    }
  }
  // Otherwise IsLatestPersistentState optimization does not make sense
@@ -444,7 +506,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
    // they don't consume sequence.
    return WriteImplWALOnly(
        &nonmem_write_thread_, write_options, my_batch, callback, user_write_cb,
-        log_used, log_ref, seq_used, batch_cnt, pre_release_callback,
+        wal_used, log_ref, seq_used, batch_cnt, pre_release_callback,
        assign_order, kDontPublishLastSeq, disable_memtable);
  }

@@ -458,7 +520,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
    // sequence in in increasing order, iii) call pre_release_callback serially
    Status status = WriteImplWALOnly(
        &write_thread_, write_options, my_batch, callback, user_write_cb,
-        log_used, log_ref, &seq, sub_batch_cnt, pre_release_callback,
+        wal_used, log_ref, &seq, sub_batch_cnt, pre_release_callback,
        kDoAssignOrder, kDoPublishLastSeq, disable_memtable);
    TEST_SYNC_POINT("DBImpl::WriteImpl:UnorderedWriteAfterWriteWAL");
    if (!status.ok()) {
@@ -477,7 +539,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,

  if (immutable_db_options_.enable_pipelined_write) {
    return PipelinedWriteImpl(write_options, my_batch, callback, user_write_cb,
-                              log_used, log_ref, disable_memtable, seq_used);
+                              wal_used, log_ref, disable_memtable, seq_used);
  }

  PERF_TIMER_GUARD(write_pre_and_post_process_time);
@@ -524,16 +586,19 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
          assert(tmp_s.ok());
        }
      }
-      versions_->SetLastSequence(last_sequence);
-      MemTableInsertStatusCheck(w.status);
+      if (w.status.ok()) {  // Don't publish a partial batch write
+        versions_->SetLastSequence(last_sequence);
+      } else {
+        HandleMemTableInsertFailure(w.status);
+      }
      write_thread_.ExitAsBatchGroupFollower(&w);
    }
    assert(w.state == WriteThread::STATE_COMPLETED);
    // STATE_COMPLETED conditional below handles exit
  }
  if (w.state == WriteThread::STATE_COMPLETED) {
-    if (log_used != nullptr) {
-      *log_used = w.log_used;
+    if (wal_used != nullptr) {
+      *wal_used = w.wal_used;
    }
    if (seq_used != nullptr) {
      *seq_used = w.sequence;
@@ -549,7 +614,8 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
  // when it finds suitable, and finish them in the same write batch.
  // This is how a write job could be done by the other writer.
  WriteContext write_context;
-  LogContext log_context(write_options.sync);
+  // FIXME: also check disableWAL like others?
+  WalContext wal_context(write_options.sync);
  WriteThread::WriteGroup write_group;
  bool in_parallel_group = false;
  uint64_t last_sequence = kMaxSequenceNumber;
@@ -563,7 +629,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
    // PreprocessWrite does its own perf timing.
    PERF_TIMER_STOP(write_pre_and_post_process_time);

-    status = PreprocessWrite(write_options, &log_context, &write_context);
+    status = PreprocessWrite(write_options, &wal_context, &write_context);
    if (!two_write_queues_) {
      // Assign it after ::PreprocessWrite since the sequence might advance
      // inside it by WriteRecoverableState
@@ -631,7 +697,13 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
            continue;
          }
          // TODO: maybe handle the tracing status?
-          tracer_->Write(writer->batch).PermitUncheckedError();
+          if (wbwi && !ingest_wbwi_for_commit) {
+            // for transaction write, tracer only needs the commit marker which
+            // is in writer->batch
+            tracer_->Write(wbwi->GetWriteBatch()).PermitUncheckedError();
+          } else {
+            tracer_->Write(writer->batch).PermitUncheckedError();
+          }
        }
      }
    }
@@ -689,22 +761,21 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,

    if (!two_write_queues_) {
      if (status.ok() && !write_options.disableWAL) {
-        assert(log_context.log_file_number_size);
-        LogFileNumberSize& log_file_number_size =
-            *(log_context.log_file_number_size);
+        assert(wal_context.wal_file_number_size);
+        wal_context.prev_size = wal_context.writer->file()->GetFileSize();
        PERF_TIMER_GUARD(write_wal_time);
-        io_s =
-            WriteToWAL(write_group, log_context.writer, log_used,
-                       log_context.need_log_sync, log_context.need_log_dir_sync,
-                       last_sequence + 1, log_file_number_size);
+        io_s = WriteGroupToWAL(write_group, wal_context.writer, wal_used,
+                               wal_context.need_wal_sync,
+                               wal_context.need_wal_dir_sync, last_sequence + 1,
+                               *wal_context.wal_file_number_size);
      }
    } else {
      if (status.ok() && !write_options.disableWAL) {
        PERF_TIMER_GUARD(write_wal_time);
        // LastAllocatedSequence is increased inside WriteToWAL under
        // wal_write_mutex_ to ensure ordered events in WAL
-        io_s = ConcurrentWriteToWAL(write_group, log_used, &last_sequence,
-                                    seq_inc);
+        io_s = ConcurrentWriteGroupToWAL(write_group, wal_used, &last_sequence,
+                                         seq_inc);
      } else {
        // Otherwise we inc seq number for memtable writes
        last_sequence = versions_->FetchAddLastAllocatedSequence(seq_inc);
@@ -716,16 +787,16 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
    last_sequence += seq_inc;
    // Seqno assigned to this write are [current_sequence, last_sequence]

-    if (log_context.need_log_sync) {
+    if (wal_context.need_wal_sync) {
      VersionEdit synced_wals;
-      log_write_mutex_.Lock();
+      wal_write_mutex_.Lock();
      if (status.ok()) {
-        MarkLogsSynced(logfile_number_, log_context.need_log_dir_sync,
+        MarkLogsSynced(cur_wal_number_, wal_context.need_wal_dir_sync,
                       &synced_wals);
      } else {
-        MarkLogsNotSynced(logfile_number_);
+        MarkLogsNotSynced(cur_wal_number_);
      }
-      log_write_mutex_.Unlock();
+      wal_write_mutex_.Unlock();
      if (status.ok() && synced_wals.IsWalAddition()) {
        InstrumentedMutexLock l(&mutex_);
        // TODO: plumb Env::IOActivity, Env::IOPriority
@@ -760,7 +831,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
        writer->sequence = next_sequence;
        if (writer->pre_release_callback) {
          Status ws = writer->pre_release_callback->Callback(
-              writer->sequence, disable_memtable, writer->log_used, index++,
+              writer->sequence, disable_memtable, writer->wal_used, index++,
              pre_release_callback_cnt);
          if (!ws.ok()) {
            status = pre_release_cb_status = ws;
@@ -785,8 +856,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
            write_group, current_sequence, column_family_memtables_.get(),
            &flush_scheduler_, &trim_history_scheduler_,
            write_options.ignore_missing_column_families,
-            0 /*recovery_log_number*/, this, parallel, seq_per_batch_,
-            batch_per_txn_);
+            0 /*recovery_log_number*/, this, seq_per_batch_, batch_per_txn_);
      } else {
        write_group.last_sequence = last_sequence;
        write_thread_.LaunchParallelMemTableWriters(&write_group);
@@ -834,12 +904,13 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
    // handle exit, false means somebody else did
    should_exit_batch_group = write_thread_.CompleteParallelMemTableWriter(&w);
  }
-  if (wbwi) {
-    if (status.ok() && w.status.ok()) {
+  if (wbwi && status.ok() && w.status.ok()) {
+    uint32_t wbwi_count = wbwi->GetWriteBatch()->Count();
+    // skip empty batch case
+    if (wbwi_count) {
      // w.batch contains (potentially empty) commit time batch updates,
      // only ingest wbwi if w.batch is applied to memtable successfully
      uint32_t memtable_update_count = w.batch->Count();
-      uint32_t wbwi_count = wbwi->GetWriteBatch()->Count();
      // Seqno assigned to this write are [last_seq + 1 - seq_inc, last_seq].
      // seq_inc includes w.batch (memtable updates) and wbwi
      // w.batch gets first `memtable_update_count` sequence numbers.
@@ -852,10 +923,12 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
      if (two_write_queues_) {
        assert(ub <= versions_->LastAllocatedSequence());
      }
-      status = IngestWBWI(wbwi, {/*lower_bound=*/lb, /*upper_bound=*/ub},
-                          prep_log, last_sequence,
-                          /*memtable_updated=*/memtable_update_count > 0,
-                          write_options.ignore_missing_column_families);
+      status =
+          IngestWBWIAsMemtable(wbwi, {/*lower_bound=*/lb, /*upper_bound=*/ub},
+                               /*min_prep_log=*/log_ref, last_sequence,
+                               /*memtable_updated=*/memtable_update_count > 0,
+                               write_options.ignore_missing_column_families);
+      RecordTick(stats_, NUMBER_WBWI_INGEST);
    }
  }

@@ -873,9 +946,19 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
      }
      // Note: if we are to resume after non-OK statuses we need to revisit how
      // we react to non-OK statuses here.
-      versions_->SetLastSequence(last_sequence);
+      if (w.status.ok()) {  // Don't publish a partial batch write
+        versions_->SetLastSequence(last_sequence);
+      }
+    }
+    if (!w.status.ok()) {
+      if (wal_context.prev_size < SIZE_MAX) {
+        InstrumentedMutexLock l(&wal_write_mutex_);
+        if (logs_.back().number == wal_context.wal_file_number_size->number) {
+          logs_.back().SetAttemptTruncateSize(wal_context.prev_size);
+        }
+      }
+      HandleMemTableInsertFailure(w.status);
    }
-    MemTableInsertStatusCheck(w.status);
    write_thread_.ExitAsBatchGroupLeader(write_group, status);
  }

@@ -888,7 +971,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
 Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,
                                  WriteBatch* my_batch, WriteCallback* callback,
                                  UserWriteCallback* user_write_cb,
-                                  uint64_t* log_used, uint64_t log_ref,
+                                  uint64_t* wal_used, uint64_t log_ref,
                                  bool disable_memtable, uint64_t* seq_used) {
  PERF_TIMER_GUARD(write_pre_and_post_process_time);
  StopWatch write_sw(immutable_db_options_.clock, stats_, DB_WRITE);
@@ -905,10 +988,10 @@ Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,
    if (w.callback && !w.callback->AllowWriteBatching()) {
      write_thread_.WaitForMemTableWriters();
    }
-    LogContext log_context(!write_options.disableWAL && write_options.sync);
+    WalContext wal_context(!write_options.disableWAL && write_options.sync);
    // PreprocessWrite does its own perf timing.
    PERF_TIMER_STOP(write_pre_and_post_process_time);
-    w.status = PreprocessWrite(write_options, &log_context, &write_context);
+    w.status = PreprocessWrite(write_options, &wal_context, &write_context);
    PERF_TIMER_START(write_pre_and_post_process_time);

    // This can set non-OK status if callback fail.
@@ -977,13 +1060,13 @@ Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,
                          wal_write_group.size - 1);
        RecordTick(stats_, WRITE_DONE_BY_OTHER, wal_write_group.size - 1);
      }
-      assert(log_context.log_file_number_size);
-      LogFileNumberSize& log_file_number_size =
-          *(log_context.log_file_number_size);
-      io_s =
-          WriteToWAL(wal_write_group, log_context.writer, log_used,
-                     log_context.need_log_sync, log_context.need_log_dir_sync,
-                     current_sequence, log_file_number_size);
+      assert(wal_context.wal_file_number_size);
+      WalFileNumberSize& wal_file_number_size =
+          *(wal_context.wal_file_number_size);
+      io_s = WriteGroupToWAL(wal_write_group, wal_context.writer, wal_used,
+                             wal_context.need_wal_sync,
+                             wal_context.need_wal_dir_sync, current_sequence,
+                             wal_file_number_size);
      w.status = io_s;
    }

@@ -995,13 +1078,13 @@ Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,
    }

    VersionEdit synced_wals;
-    if (log_context.need_log_sync) {
-      InstrumentedMutexLock l(&log_write_mutex_);
+    if (wal_context.need_wal_sync) {
+      InstrumentedMutexLock l(&wal_write_mutex_);
      if (w.status.ok()) {
-        MarkLogsSynced(logfile_number_, log_context.need_log_dir_sync,
+        MarkLogsSynced(cur_wal_number_, wal_context.need_wal_dir_sync,
                       &synced_wals);
      } else {
-        MarkLogsNotSynced(logfile_number_);
+        MarkLogsNotSynced(cur_wal_number_);
      }
    }
    if (w.status.ok() && synced_wals.IsWalAddition()) {
@@ -1031,8 +1114,13 @@ Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,
          memtable_write_group, w.sequence, column_family_memtables_.get(),
          &flush_scheduler_, &trim_history_scheduler_,
          write_options.ignore_missing_column_families, 0 /*log_number*/, this,
-          false /*concurrent_memtable_writes*/, seq_per_batch_, batch_per_txn_);
-      versions_->SetLastSequence(memtable_write_group.last_sequence);
+          seq_per_batch_, batch_per_txn_);
+      if (memtable_write_group.status
+              .ok()) {  // Don't publish a partial batch write
+        versions_->SetLastSequence(memtable_write_group.last_sequence);
+      } else {
+        HandleMemTableInsertFailure(memtable_write_group.status);
+      }
      write_thread_.ExitAsMemTableWriter(&w, memtable_write_group);
    }
  } else {
@@ -1061,8 +1149,11 @@ Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,
    PERF_TIMER_START(write_pre_and_post_process_time);

    if (write_thread_.CompleteParallelMemTableWriter(&w)) {
-      MemTableInsertStatusCheck(w.status);
-      versions_->SetLastSequence(w.write_group->last_sequence);
+      if (w.status.ok()) {  // Don't publish a partial batch write
+        versions_->SetLastSequence(w.write_group->last_sequence);
+      } else {
+        HandleMemTableInsertFailure(w.status);
+      }
      write_thread_.ExitAsMemTableWriter(&w, *w.write_group);
    }
  }
@@ -1134,7 +1225,7 @@ Status DBImpl::UnorderedWriteMemtable(const WriteOptions& write_options,
 Status DBImpl::WriteImplWALOnly(
    WriteThread* write_thread, const WriteOptions& write_options,
    WriteBatch* my_batch, WriteCallback* callback,
-    UserWriteCallback* user_write_cb, uint64_t* log_used,
+    UserWriteCallback* user_write_cb, uint64_t* wal_used,
    const uint64_t log_ref, uint64_t* seq_used, const size_t sub_batch_cnt,
    PreReleaseCallback* pre_release_callback, const AssignOrder assign_order,
    const PublishLastSeq publish_last_seq, const bool disable_memtable) {
@@ -1147,8 +1238,8 @@ Status DBImpl::WriteImplWALOnly(
  write_thread->JoinBatchGroup(&w);
  assert(w.state != WriteThread::STATE_PARALLEL_MEMTABLE_WRITER);
  if (w.state == WriteThread::STATE_COMPLETED) {
-    if (log_used != nullptr) {
-      *log_used = w.log_used;
+    if (wal_used != nullptr) {
+      *wal_used = w.wal_used;
    }
    if (seq_used != nullptr) {
      *seq_used = w.sequence;
@@ -1164,10 +1255,10 @@ Status DBImpl::WriteImplWALOnly(

    // TODO(myabandeh): Make preliminary checks thread-safe so we could do them
    // without paying the cost of obtaining the mutex.
-    LogContext log_context;
+    WalContext wal_context;
    WriteContext write_context;
    Status status =
-        PreprocessWrite(write_options, &log_context, &write_context);
+        PreprocessWrite(write_options, &wal_context, &write_context);
    WriteStatusCheckOnLocked(status);

    if (!status.ok()) {
@@ -1264,8 +1355,8 @@ Status DBImpl::WriteImplWALOnly(
  }
  Status status;
  if (!write_options.disableWAL) {
-    IOStatus io_s =
-        ConcurrentWriteToWAL(write_group, log_used, &last_sequence, seq_inc);
+    IOStatus io_s = ConcurrentWriteGroupToWAL(write_group, wal_used,
+                                              &last_sequence, seq_inc);
    status = io_s;
    // last_sequence may not be set if there is an error
    // This error checking and return is moved up to avoid using uninitialized
@@ -1317,7 +1408,7 @@ Status DBImpl::WriteImplWALOnly(
      if (!writer->CallbackFailed() && writer->pre_release_callback) {
        assert(writer->sequence != kMaxSequenceNumber);
        Status ws = writer->pre_release_callback->Callback(
-            writer->sequence, disable_memtable, writer->log_used, index++,
+            writer->sequence, disable_memtable, writer->wal_used, index++,
            pre_release_callback_cnt);
        if (!ws.ok()) {
          status = ws;
@@ -1386,24 +1477,22 @@ void DBImpl::WALIOStatusCheck(const IOStatus& io_status) {
  }
 }

-void DBImpl::MemTableInsertStatusCheck(const Status& status) {
-  // A non-OK status here indicates that the state implied by the
-  // WAL has diverged from the in-memory state.  This could be
-  // because of a corrupt write_batch (very bad), or because the
-  // client specified an invalid column family and didn't specify
-  // ignore_missing_column_families.
-  if (!status.ok()) {
-    mutex_.Lock();
-    assert(!error_handler_.IsBGWorkStopped());
-    error_handler_.SetBGError(status, BackgroundErrorReason::kMemTable);
-    mutex_.Unlock();
-  }
+void DBImpl::HandleMemTableInsertFailure(const Status& status) {
+  assert(!status.ok());
+  // A non-OK status on memtable insert indicates that the state implied by the
+  // WAL has diverged from the in-memory state.  This could be because of a
+  // corrupt write_batch (very bad), or because the client specified an invalid
+  // column family and didn't specify ignore_missing_column_families.
+  mutex_.Lock();
+  assert(!error_handler_.IsBGWorkStopped());
+  error_handler_.SetBGError(status, BackgroundErrorReason::kMemTable);
+  mutex_.Unlock();
 }

 Status DBImpl::PreprocessWrite(const WriteOptions& write_options,
-                               LogContext* log_context,
+                               WalContext* wal_context,
                               WriteContext* write_context) {
-  assert(write_context != nullptr && log_context != nullptr);
+  assert(write_context != nullptr && wal_context != nullptr);
  Status status;

  if (error_handler_.IsDBStopped()) {
@@ -1413,7 +1502,8 @@ Status DBImpl::PreprocessWrite(const WriteOptions& write_options,

  PERF_TIMER_GUARD(write_scheduling_flushes_compactions_time);

-  if (UNLIKELY(status.ok() && total_log_size_ > GetMaxTotalWalSize())) {
+  if (UNLIKELY(status.ok() &&
+               wals_total_size_.LoadRelaxed() > GetMaxTotalWalSize())) {
    assert(versions_);
    InstrumentedMutexLock l(&mutex_);
    const ColumnFamilySet* const column_families =
@@ -1482,17 +1572,17 @@ Status DBImpl::PreprocessWrite(const WriteOptions& write_options,
      WriteBufferManagerStallWrites();
    }
  }
-  InstrumentedMutexLock l(&log_write_mutex_);
-  if (status.ok() && log_context->need_log_sync) {
+  InstrumentedMutexLock l(&wal_write_mutex_);
+  if (status.ok() && wal_context->need_wal_sync) {
    // Wait until the parallel syncs are finished. Any sync process has to sync
    // the front log too so it is enough to check the status of front()
-    // We do a while loop since log_sync_cv_ is signalled when any sync is
+    // We do a while loop since wal_sync_cv_ is signalled when any sync is
    // finished
    // Note: there does not seem to be a reason to wait for parallel sync at
    // this early step but it is not important since parallel sync (SyncWAL) and
-    // need_log_sync are usually not used together.
+    // need_wal_sync are usually not used together.
    while (logs_.front().IsSyncing()) {
-      log_sync_cv_.Wait();
+      wal_sync_cv_.Wait();
    }
    for (auto& log : logs_) {
      // This is just to prevent the logs to be synced by a parallel SyncWAL
@@ -1503,12 +1593,12 @@ Status DBImpl::PreprocessWrite(const WriteOptions& write_options,
      log.PrepareForSync();
    }
  } else {
-    log_context->need_log_sync = false;
+    wal_context->need_wal_sync = false;
  }
-  log_context->writer = logs_.back().writer;
-  log_context->need_log_dir_sync =
-      log_context->need_log_dir_sync && !log_dir_synced_;
-  log_context->log_file_number_size = std::addressof(alive_log_files_.back());
+  wal_context->writer = logs_.back().writer;
+  wal_context->need_wal_dir_sync =
+      wal_context->need_wal_dir_sync && !wal_dir_synced_;
+  wal_context->wal_file_number_size = std::addressof(alive_wal_files_.back());

  return status;
 }
@@ -1559,12 +1649,12 @@ Status DBImpl::MergeBatch(const WriteThread::WriteGroup& write_group,
 }

 // When two_write_queues_ is disabled, this function is called from the only
-// write thread. Otherwise this must be called holding log_write_mutex_.
+// write thread. Otherwise this must be called holding wal_write_mutex_.
 IOStatus DBImpl::WriteToWAL(const WriteBatch& merged_batch,
                            const WriteOptions& write_options,
-                            log::Writer* log_writer, uint64_t* log_used,
+                            log::Writer* log_writer, uint64_t* wal_used,
                            uint64_t* log_size,
-                            LogFileNumberSize& log_file_number_size,
+                            WalFileNumberSize& wal_file_number_size,
                            SequenceNumber sequence) {
  assert(log_size != nullptr);

@@ -1576,7 +1666,7 @@ IOStatus DBImpl::WriteToWAL(const WriteBatch& merged_batch,
  }
  *log_size = log_entry.size();
  // When two_write_queues_ WriteToWAL has to be protected from concurretn calls
-  // from the two queues anyway and log_write_mutex_ is already held. Otherwise
+  // from the two queues anyway and wal_write_mutex_ is already held. Otherwise
  // if manual_wal_flush_ is enabled we need to protect log_writer->AddRecord
  // from possible concurrent calls via the FlushWAL by the application.
  const bool needs_locking = manual_wal_flush_ && !two_write_queues_;
@@ -1584,7 +1674,7 @@ IOStatus DBImpl::WriteToWAL(const WriteBatch& merged_batch,
  // manual_wal_flush_ feature (by UNLIKELY) instead of the more common case
  // when we do not need any locking.
  if (UNLIKELY(needs_locking)) {
-    log_write_mutex_.Lock();
+    wal_write_mutex_.Lock();
  }
  IOStatus io_s = log_writer->MaybeAddUserDefinedTimestampSizeRecord(
      write_options, versions_->GetColumnFamiliesTimestampSizeForRecord());
@@ -1594,23 +1684,24 @@ IOStatus DBImpl::WriteToWAL(const WriteBatch& merged_batch,
  io_s = log_writer->AddRecord(write_options, log_entry, sequence);

  if (UNLIKELY(needs_locking)) {
-    log_write_mutex_.Unlock();
+    wal_write_mutex_.Unlock();
  }
-  if (log_used != nullptr) {
-    *log_used = logfile_number_;
+  if (wal_used != nullptr) {
+    *wal_used = cur_wal_number_;
+    assert(*wal_used == wal_file_number_size.number);
  }
-  total_log_size_ += log_entry.size();
-  log_file_number_size.AddSize(*log_size);
-  log_empty_ = false;
+  wals_total_size_.FetchAddRelaxed(log_entry.size());
+  wal_file_number_size.AddSize(*log_size);
+  wal_empty_ = false;

  return io_s;
 }

-IOStatus DBImpl::WriteToWAL(const WriteThread::WriteGroup& write_group,
-                            log::Writer* log_writer, uint64_t* log_used,
-                            bool need_log_sync, bool need_log_dir_sync,
-                            SequenceNumber sequence,
-                            LogFileNumberSize& log_file_number_size) {
+IOStatus DBImpl::WriteGroupToWAL(const WriteThread::WriteGroup& write_group,
+                                 log::Writer* log_writer, uint64_t* wal_used,
+                                 bool need_wal_sync, bool need_wal_dir_sync,
+                                 SequenceNumber sequence,
+                                 WalFileNumberSize& wal_file_number_size) {
  IOStatus io_s;
  assert(!two_write_queues_);
  assert(!write_group.leader->disable_wal);
@@ -1625,10 +1716,10 @@ IOStatus DBImpl::WriteToWAL(const WriteThread::WriteGroup& write_group,
  }

  if (merged_batch == write_group.leader->batch) {
-    write_group.leader->log_used = logfile_number_;
+    write_group.leader->wal_used = cur_wal_number_;
  } else if (write_with_wal > 1) {
    for (auto writer : write_group) {
-      writer->log_used = logfile_number_;
+      writer->wal_used = cur_wal_number_;
    }
  }

@@ -1640,14 +1731,14 @@ IOStatus DBImpl::WriteToWAL(const WriteThread::WriteGroup& write_group,
  WriteOptions write_options;
  write_options.rate_limiter_priority =
      write_group.leader->rate_limiter_priority;
-  io_s = WriteToWAL(*merged_batch, write_options, log_writer, log_used,
-                    &log_size, log_file_number_size, sequence);
+  io_s = WriteToWAL(*merged_batch, write_options, log_writer, wal_used,
+                    &log_size, wal_file_number_size, sequence);
  if (to_be_cached_state) {
    cached_recoverable_state_ = *to_be_cached_state;
    cached_recoverable_state_empty_ = false;
  }

-  if (io_s.ok() && need_log_sync) {
+  if (io_s.ok() && need_wal_sync) {
    StopWatch sw(immutable_db_options_.clock, stats_, WAL_FILE_SYNC_MICROS);
    // It's safe to access logs_ with unlocked mutex_ here because:
    //  - we've set getting_synced=true for all logs,
@@ -1657,15 +1748,15 @@ IOStatus DBImpl::WriteToWAL(const WriteThread::WriteGroup& write_group,
    //  - as long as other threads don't modify it, it's safe to read
    //    from std::deque from multiple threads concurrently.
    //
-    // Sync operation should work with locked log_write_mutex_, because:
+    // Sync operation should work with locked wal_write_mutex_, because:
    //   when DBOptions.manual_wal_flush_ is set,
    //   FlushWAL function will be invoked by another thread.
-    //   if without locked log_write_mutex_, the log file may get data
+    //   if without locked wal_write_mutex_, the log file may get data
    //   corruption

    const bool needs_locking = manual_wal_flush_ && !two_write_queues_;
    if (UNLIKELY(needs_locking)) {
-      log_write_mutex_.Lock();
+      wal_write_mutex_.Lock();
    }

    if (io_s.ok()) {
@@ -1688,10 +1779,10 @@ IOStatus DBImpl::WriteToWAL(const WriteThread::WriteGroup& write_group,
    }

    if (UNLIKELY(needs_locking)) {
-      log_write_mutex_.Unlock();
+      wal_write_mutex_.Unlock();
    }

-    if (io_s.ok() && need_log_dir_sync) {
+    if (io_s.ok() && need_wal_dir_sync) {
      // We only sync WAL directory the first time WAL syncing is
      // requested, so that in case users never turn on WAL sync,
      // we can avoid the disk I/O in the write code path.
@@ -1706,7 +1797,7 @@ IOStatus DBImpl::WriteToWAL(const WriteThread::WriteGroup& write_group,
  }
  if (io_s.ok()) {
    auto stats = default_cf_internal_stats_;
-    if (need_log_sync) {
+    if (need_wal_sync) {
      stats->AddDBStats(InternalStats::kIntStatsWalFileSynced, 1);
      RecordTick(stats_, WAL_FILE_SYNCED);
    }
@@ -1723,8 +1814,8 @@ IOStatus DBImpl::WriteToWAL(const WriteThread::WriteGroup& write_group,
  return io_s;
 }

-IOStatus DBImpl::ConcurrentWriteToWAL(
-    const WriteThread::WriteGroup& write_group, uint64_t* log_used,
+IOStatus DBImpl::ConcurrentWriteGroupToWAL(
+    const WriteThread::WriteGroup& write_group, uint64_t* wal_used,
    SequenceNumber* last_sequence, size_t seq_inc) {
  IOStatus io_s;

@@ -1741,14 +1832,14 @@ IOStatus DBImpl::ConcurrentWriteToWAL(
    return io_s;
  }

-  // We need to lock log_write_mutex_ since logs_ and alive_log_files might be
+  // We need to lock wal_write_mutex_ since logs_ and alive_wal_files might be
  // pushed back concurrently
-  log_write_mutex_.Lock();
+  wal_write_mutex_.Lock();
  if (merged_batch == write_group.leader->batch) {
-    write_group.leader->log_used = logfile_number_;
+    write_group.leader->wal_used = cur_wal_number_;
  } else if (write_with_wal > 1) {
    for (auto writer : write_group) {
-      writer->log_used = logfile_number_;
+      writer->wal_used = cur_wal_number_;
    }
  }
  *last_sequence = versions_->FetchAddLastAllocatedSequence(seq_inc);
@@ -1756,9 +1847,9 @@ IOStatus DBImpl::ConcurrentWriteToWAL(
  WriteBatchInternal::SetSequence(merged_batch, sequence);

  log::Writer* log_writer = logs_.back().writer;
-  LogFileNumberSize& log_file_number_size = alive_log_files_.back();
+  WalFileNumberSize& wal_file_number_size = alive_wal_files_.back();

-  assert(log_writer->get_log_number() == log_file_number_size.number);
+  assert(log_writer->get_log_number() == wal_file_number_size.number);

  uint64_t log_size;

@@ -1766,13 +1857,13 @@ IOStatus DBImpl::ConcurrentWriteToWAL(
  WriteOptions write_options;
  write_options.rate_limiter_priority =
      write_group.leader->rate_limiter_priority;
-  io_s = WriteToWAL(*merged_batch, write_options, log_writer, log_used,
-                    &log_size, log_file_number_size, sequence);
+  io_s = WriteToWAL(*merged_batch, write_options, log_writer, wal_used,
+                    &log_size, wal_file_number_size, sequence);
  if (to_be_cached_state) {
    cached_recoverable_state_ = *to_be_cached_state;
    cached_recoverable_state_empty_ = false;
  }
-  log_write_mutex_.Unlock();
+  wal_write_mutex_.Unlock();

  if (io_s.ok()) {
    const bool concurrent = true;
@@ -1800,7 +1891,7 @@ Status DBImpl::WriteRecoverableState() {
    bool dont_care_bool;
    SequenceNumber next_seq;
    if (two_write_queues_) {
-      log_write_mutex_.Lock();
+      wal_write_mutex_.Lock();
    }
    SequenceNumber seq;
    if (two_write_queues_) {
@@ -1815,13 +1906,17 @@ Status DBImpl::WriteRecoverableState() {
        0 /*recovery_log_number*/, this, false /* concurrent_memtable_writes */,
        &next_seq, &dont_care_bool, seq_per_batch_);
    auto last_seq = next_seq - 1;
-    if (two_write_queues_) {
-      versions_->FetchAddLastAllocatedSequence(last_seq - seq);
-      versions_->SetLastPublishedSequence(last_seq);
+    if (status.ok()) {  // Don't publish a partial batch write
+      if (two_write_queues_) {
+        versions_->FetchAddLastAllocatedSequence(last_seq - seq);
+        versions_->SetLastPublishedSequence(last_seq);
+      }
+      versions_->SetLastSequence(last_seq);
+    } else {
+      HandleMemTableInsertFailure(status);
    }
-    versions_->SetLastSequence(last_seq);
    if (two_write_queues_) {
-      log_write_mutex_.Unlock();
+      wal_write_mutex_.Unlock();
    }
    if (status.ok() && recoverable_state_pre_release_callback_) {
      const bool DISABLE_MEMTABLE = true;
@@ -1893,7 +1988,10 @@ void DBImpl::AssignAtomicFlushSeq(const autovector<ColumnFamilyData*>& cfds) {
  assert(immutable_db_options_.atomic_flush);
  auto seq = versions_->LastSequence();
  for (auto cfd : cfds) {
-    cfd->imm()->AssignAtomicFlushSeq(seq);
+    // cfd can be nullptr, see ScheduleFlushes()
+    if (cfd) {
+      cfd->imm()->AssignAtomicFlushSeq(seq);
+    }
  }
 }

@@ -1902,11 +2000,11 @@ Status DBImpl::SwitchWAL(WriteContext* write_context) {
  assert(write_context != nullptr);
  Status status;

-  if (alive_log_files_.begin()->getting_flushed) {
+  if (alive_wal_files_.begin()->getting_flushed) {
    return status;
  }

-  auto oldest_alive_log = alive_log_files_.begin()->number;
+  auto oldest_alive_log = alive_wal_files_.begin()->number;
  bool flush_wont_release_oldest_log = false;
  if (allow_2pc()) {
    auto oldest_log_with_uncommitted_prep =
@@ -1936,14 +2034,14 @@ Status DBImpl::SwitchWAL(WriteContext* write_context) {
    // transactions then we cannot flush this log until those transactions are
    // commited.
    unable_to_release_oldest_log_ = false;
-    alive_log_files_.begin()->getting_flushed = true;
+    alive_wal_files_.begin()->getting_flushed = true;
  }

  ROCKS_LOG_INFO(
      immutable_db_options_.info_log,
      "Flushing all column families with data in WAL number %" PRIu64
      ". Total log size is %" PRIu64 " while max_total_wal_size is %" PRIu64,
-      oldest_alive_log, total_log_size_.load(), GetMaxTotalWalSize());
+      oldest_alive_log, wals_total_size_.LoadRelaxed(), GetMaxTotalWalSize());
  // no need to refcount because drop is happening in write thread, so can't
  // happen while we're in the write thread
  autovector<ColumnFamilyData*> cfds;
@@ -2413,21 +2511,21 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context,
  // Do this without holding the dbmutex lock.
  assert(versions_->prev_log_number() == 0);
  if (two_write_queues_) {
-    log_write_mutex_.Lock();
+    wal_write_mutex_.Lock();
  }
-  bool creating_new_log = !log_empty_;
+  bool creating_new_log = !wal_empty_;
  if (two_write_queues_) {
-    log_write_mutex_.Unlock();
+    wal_write_mutex_.Unlock();
  }
  uint64_t recycle_log_number = 0;
  // If file deletion is disabled, don't recycle logs since it'll result in
  // the file getting renamed
  if (creating_new_log && immutable_db_options_.recycle_log_file_num &&
-      !log_recycle_files_.empty() && IsFileDeletionsEnabled()) {
-    recycle_log_number = log_recycle_files_.front();
+      !wal_recycle_files_.empty() && IsFileDeletionsEnabled()) {
+    recycle_log_number = wal_recycle_files_.front();
  }
  uint64_t new_log_number =
-      creating_new_log ? versions_->NewFileNumber() : logfile_number_;
+      creating_new_log ? versions_->NewFileNumber() : cur_wal_number_;
  // For use outside of holding DB mutex
  const MutableCFOptions mutable_cf_options_copy =
      cfd->GetLatestMutableCFOptions();
@@ -2453,14 +2551,14 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context,
  mutex_.Unlock();
  if (creating_new_log) {
    PredecessorWALInfo info;
-    log_write_mutex_.Lock();
+    wal_write_mutex_.Lock();
    if (!logs_.empty()) {
      log::Writer* cur_log_writer = logs_.back().writer;
      info = PredecessorWALInfo(cur_log_writer->get_log_number(),
                                cur_log_writer->file()->GetFileSize(),
                                cur_log_writer->GetLastSeqnoRecorded());
    }
-    log_write_mutex_.Unlock();
+    wal_write_mutex_.Unlock();
    // TODO: Write buffer size passed in should be max of all CF's instead
    // of mutable_cf_options.write_buffer_size.
    io_s = CreateWAL(write_options, new_log_number, recycle_log_number,
@@ -2501,11 +2599,11 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context,
    // concurrent full purges don't delete the file while we're recycling it.
    // To achieve that we hold the old log number in the recyclable list until
    // after it has been renamed.
-    assert(log_recycle_files_.front() == recycle_log_number);
-    log_recycle_files_.pop_front();
+    assert(wal_recycle_files_.front() == recycle_log_number);
+    wal_recycle_files_.pop_front();
  }
  if (s.ok() && creating_new_log) {
-    InstrumentedMutexLock l(&log_write_mutex_);
+    InstrumentedMutexLock l(&wal_write_mutex_);
    assert(new_log != nullptr);
    if (!logs_.empty()) {
      // Alway flush the buffer of the last log before switching to a new one
@@ -2527,11 +2625,11 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context,
      }
    }
    if (s.ok()) {
-      logfile_number_ = new_log_number;
-      log_empty_ = true;
-      log_dir_synced_ = false;
-      logs_.emplace_back(logfile_number_, new_log);
-      alive_log_files_.emplace_back(logfile_number_);
+      cur_wal_number_ = new_log_number;
+      wal_empty_ = true;
+      wal_dir_synced_ = false;
+      logs_.emplace_back(cur_wal_number_, new_log);
+      alive_wal_files_.emplace_back(cur_wal_number_);
    }
  }

@@ -2562,7 +2660,7 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context,
    // obsolete. So we should track the WAL obsoletion event before actually
    // updating the empty CF's log number.
    uint64_t min_wal_number_to_keep =
-        versions_->PreComputeMinLogNumberWithUnflushedData(logfile_number_);
+        versions_->PreComputeMinLogNumberWithUnflushedData(cur_wal_number_);
    if (min_wal_number_to_keep >
        versions_->GetWalSet().GetMinWalNumberToKeep()) {
      // TODO: plumb Env::IOActivity, Env::IOPriority
@@ -2597,7 +2695,7 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context,

      for (auto cf : empty_cfs) {
        if (cf->IsEmpty()) {
-          cf->SetLogNumber(logfile_number_);
+          cf->SetLogNumber(cur_wal_number_);
          // MEMPURGE: No need to change this, because new adds
          // should still receive new sequence numbers.
          cf->mem()->SetCreationSeq(versions_->LastSequence());
@@ -2614,14 +2712,14 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context,
      // advance the log number. no need to persist this in the manifest
      if (cf->IsEmpty()) {
        if (creating_new_log) {
-          cf->SetLogNumber(logfile_number_);
+          cf->SetLogNumber(cur_wal_number_);
        }
        cf->mem()->SetCreationSeq(versions_->LastSequence());
      }
    }
  }

-  cfd->mem()->SetNextLogNumber(logfile_number_);
+  cfd->mem()->SetNextLogNumber(cur_wal_number_);
  assert(new_mem != nullptr);
  cfd->imm()->Add(cfd->mem(), &context->memtables_to_free_);
  if (new_imm) {
@@ -2633,7 +2731,7 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context,
    // we always try to flush all immutable memtable. For atomic flush, these
    // two memtables will be marked eligible for flush in the same call to
    // AssignAtomicFlushSeq().
-    new_imm->SetNextLogNumber(logfile_number_);
+    new_imm->SetNextLogNumber(cur_wal_number_);
    cfd->imm()->Add(new_imm, &context->memtables_to_free_);
  }
  new_mem->Ref();
@@ -7,6 +7,8 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

+#include <iomanip>
+
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
 #include "test_util/testutil.h"
@@ -9,7 +9,6 @@

 #include "db/db_iter.h"

-#include <iostream>
 #include <limits>
 #include <string>

@@ -42,9 +41,8 @@ DBIter::DBIter(Env* _env, const ReadOptions& read_options,
               const MutableCFOptions& mutable_cf_options,
               const Comparator* cmp, InternalIterator* iter,
               const Version* version, SequenceNumber s, bool arena_mode,
-               uint64_t max_sequential_skip_in_iterations,
               ReadCallback* read_callback, ColumnFamilyHandleImpl* cfh,
-               bool expose_blob_index)
+               bool expose_blob_index, ReadOnlyMemTable* active_mem)
    : prefix_extractor_(mutable_cf_options.prefix_extractor.get()),
      env_(_env),
      clock_(ioptions.clock),
@@ -58,11 +56,21 @@ DBIter::DBIter(Env* _env, const ReadOptions& read_options,
      read_callback_(read_callback),
      sequence_(s),
      statistics_(ioptions.stats),
-      max_skip_(max_sequential_skip_in_iterations),
+      max_skip_(mutable_cf_options.max_sequential_skip_in_iterations),
      max_skippable_internal_keys_(read_options.max_skippable_internal_keys),
      num_internal_keys_skipped_(0),
      iterate_lower_bound_(read_options.iterate_lower_bound),
      iterate_upper_bound_(read_options.iterate_upper_bound),
+      cfh_(cfh),
+      timestamp_ub_(read_options.timestamp),
+      timestamp_lb_(read_options.iter_start_ts),
+      timestamp_size_(timestamp_ub_ ? timestamp_ub_->size() : 0),
+      active_mem_(active_mem),
+      memtable_seqno_lb_(kMaxSequenceNumber),
+      memtable_op_scan_flush_trigger_(0),
+      avg_op_scan_flush_trigger_(0),
+      iter_step_since_seek_(1),
+      mem_hidden_op_scanned_since_seek_(0),
      direction_(kForward),
      valid_(false),
      current_entry_is_merged_(false),
@@ -76,11 +84,7 @@ DBIter::DBIter(Env* _env, const ReadOptions& read_options,
      expose_blob_index_(expose_blob_index),
      allow_unprepared_value_(read_options.allow_unprepared_value),
      is_blob_(false),
-      arena_mode_(arena_mode),
-      cfh_(cfh),
-      timestamp_ub_(read_options.timestamp),
-      timestamp_lb_(read_options.iter_start_ts),
-      timestamp_size_(timestamp_ub_ ? timestamp_ub_->size() : 0) {
+      arena_mode_(arena_mode) {
  RecordTick(statistics_, NO_ITERATOR_CREATED);
  if (pin_thru_lifetime_) {
    pinned_iters_mgr_.StartPinning();
@@ -94,6 +98,25 @@ DBIter::DBIter(Env* _env, const ReadOptions& read_options,
  // prefix_seek_opt_in_only should force total_order_seek whereever the caller
  // is duplicating the original ReadOptions
  assert(!ioptions.prefix_seek_opt_in_only || read_options.total_order_seek);
+  if (active_mem_) {
+    // FIXME: GetEarliestSequenceNumber() may return a seqno that is one smaller
+    // than the smallest seqno in the memtable. This violates its comment and
+    // entries with that seqno may not be in the active memtable. Before it's
+    // fixed, we use GetFirstSequenceNumber() for more accurate result.
+    memtable_seqno_lb_ = active_mem_->IsEmpty()
+                             ? active_mem_->GetEarliestSequenceNumber()
+                             : active_mem_->GetFirstSequenceNumber();
+    memtable_op_scan_flush_trigger_ =
+        mutable_cf_options.memtable_op_scan_flush_trigger;
+    if (memtable_op_scan_flush_trigger_) {
+      // avg_op_scan_flush_trigger_ requires memtable_op_scan_flush_trigger_ > 0
+      avg_op_scan_flush_trigger_ =
+          mutable_cf_options.memtable_avg_op_scan_flush_trigger;
+    }
+  } else {
+    // memtable_op_scan_flush_trigger_ and avg_op_scan_flush_trigger_ are
+    // initialized to 0(disabled) as default.
+  }
 }

 Status DBIter::GetProperty(std::string prop_name, std::string* prop) {
@@ -155,6 +178,7 @@ void DBIter::Next() {
  local_stats_.skip_count_ += num_internal_keys_skipped_;
  local_stats_.skip_count_--;
  num_internal_keys_skipped_ = 0;
+  iter_step_since_seek_++;
  bool ok = true;
  if (direction_ == kReverse) {
    is_key_seqnum_zero_ = false;
@@ -369,6 +393,7 @@ bool DBIter::FindNextUserEntryInternal(bool skipping_saved_key,
  // to one.
  bool reseek_done = false;

+  uint64_t mem_hidden_op_scanned = 0;
  do {
    // Will update is_key_seqnum_zero_ as soon as we parsed the current key
    // but we need to save the previous value to be used in the loop.
@@ -425,6 +450,7 @@ bool DBIter::FindNextUserEntryInternal(bool skipping_saved_key,
          CompareKeyForSkip(ikey_.user_key, saved_key_.GetUserKey()) <= 0) {
        num_skipped++;  // skip this entry
        PERF_COUNTER_ADD(internal_key_skipped_count, 1);
+        MarkMemtableForFlushForPerOpTrigger(mem_hidden_op_scanned);
      } else {
        assert(!skipping_saved_key ||
               CompareKeyForSkip(ikey_.user_key, saved_key_.GetUserKey()) > 0);
@@ -446,6 +472,7 @@ bool DBIter::FindNextUserEntryInternal(bool skipping_saved_key,
                                      !iter_.iter()->IsKeyPinned() /* copy */);
              skipping_saved_key = true;
              PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
+              MarkMemtableForFlushForPerOpTrigger(mem_hidden_op_scanned);
            }
            break;
          case kTypeValue:
@@ -484,7 +511,6 @@ bool DBIter::FindNextUserEntryInternal(bool skipping_saved_key,

            valid_ = true;
            return true;
-            break;
          case kTypeMerge:
            if (!PrepareValueInternal()) {
              return false;
@@ -496,7 +522,6 @@ bool DBIter::FindNextUserEntryInternal(bool skipping_saved_key,
            current_entry_is_merged_ = true;
            valid_ = true;
            return MergeValuesNewToOld();  // Go to a different state machine
-            break;
          default:
            valid_ = false;
            status_ = Status::Corruption(
@@ -1097,7 +1122,6 @@ bool DBIter::FindValueForCurrentKey() {
        }
        return true;
      }
-      break;
    case kTypeValue:
    case kTypeValuePreferredSeqno:
      SetValueAndColumnsFromPlain(pinned_value_);
@@ -1224,6 +1248,8 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {

    if (timestamp_lb_ != nullptr) {
      saved_key_.SetInternalKey(ikey);
+    } else {
+      saved_key_.SetUserKey(ikey.user_key);
    }

    valid_ = true;
@@ -1568,6 +1594,7 @@ void DBIter::Seek(const Slice& target) {
  ResetBlobData();
  ResetValueAndColumns();
  ResetInternalKeysSkippedCounter();
+  MarkMemtableForFlushForAvgTrigger();

  // Seek the inner iterator based on the target key.
  {
@@ -1644,6 +1671,7 @@ void DBIter::SeekForPrev(const Slice& target) {
  ResetBlobData();
  ResetValueAndColumns();
  ResetInternalKeysSkippedCounter();
+  MarkMemtableForFlushForAvgTrigger();

  // Seek the inner iterator based on the target key.
  {
@@ -1705,6 +1733,7 @@ void DBIter::SeekToFirst() {
  ResetBlobData();
  ResetValueAndColumns();
  ResetInternalKeysSkippedCounter();
+  MarkMemtableForFlushForAvgTrigger();
  ClearSavedValue();
  is_key_seqnum_zero_ = false;

@@ -1768,6 +1797,7 @@ void DBIter::SeekToLast() {
  ResetBlobData();
  ResetValueAndColumns();
  ResetInternalKeysSkippedCounter();
+  MarkMemtableForFlushForAvgTrigger();
  ClearSavedValue();
  is_key_seqnum_zero_ = false;

@@ -1790,21 +1820,4 @@ void DBIter::SeekToLast() {
        StripTimestampFromUserKey(saved_key_.GetUserKey(), timestamp_size_)));
  }
 }
-
-Iterator* NewDBIterator(Env* env, const ReadOptions& read_options,
-                        const ImmutableOptions& ioptions,
-                        const MutableCFOptions& mutable_cf_options,
-                        const Comparator* user_key_comparator,
-                        InternalIterator* internal_iter, const Version* version,
-                        const SequenceNumber& sequence,
-                        uint64_t max_sequential_skip_in_iterations,
-                        ReadCallback* read_callback,
-                        ColumnFamilyHandleImpl* cfh, bool expose_blob_index) {
-  DBIter* db_iter = new DBIter(
-      env, read_options, ioptions, mutable_cf_options, user_key_comparator,
-      internal_iter, version, sequence, false,
-      max_sequential_skip_in_iterations, read_callback, cfh, expose_blob_index);
-  return db_iter;
-}
-
 }  // namespace ROCKSDB_NAMESPACE
@@ -12,7 +12,6 @@
 #include <string>

 #include "db/db_impl/db_impl.h"
-#include "db/range_del_aggregator.h"
 #include "memory/arena.h"
 #include "options/cf_options.h"
 #include "rocksdb/db.h"
@@ -57,6 +56,34 @@ class Version;
 // numbers, deletion markers, overwrites, etc.
 class DBIter final : public Iterator {
 public:
+  // Return a new DBIter that reads from `internal_iter` at the specified
+  // `sequence` number.
+  //
+  // @param active_mem Pointer to the active memtable that `internal_iter`
+  // is reading from. If not null, the memtable can be marked for flush
+  // according to options mutable_cf_options.memtable_op_scan_flush_trigger
+  // and mutable_cf_options.memtable_avg_op_scan_flush_trigger.
+  // @param arena_mode If true, the DBIter will be allocated from the arena.
+  static DBIter* NewIter(Env* env, const ReadOptions& read_options,
+                         const ImmutableOptions& ioptions,
+                         const MutableCFOptions& mutable_cf_options,
+                         const Comparator* user_key_comparator,
+                         InternalIterator* internal_iter,
+                         const Version* version, const SequenceNumber& sequence,
+                         ReadCallback* read_callback,
+                         ReadOnlyMemTable* active_mem,
+                         ColumnFamilyHandleImpl* cfh = nullptr,
+                         bool expose_blob_index = false,
+                         Arena* arena = nullptr) {
+    void* mem = arena ? arena->AllocateAligned(sizeof(DBIter))
+                      : operator new(sizeof(DBIter));
+    DBIter* db_iter = new (mem)
+        DBIter(env, read_options, ioptions, mutable_cf_options,
+               user_key_comparator, internal_iter, version, sequence, arena,
+               read_callback, cfh, expose_blob_index, active_mem);
+    return db_iter;
+  }
+
  // The following is grossly complicated. TODO: clean it up
  // Which direction is the iterator currently moving?
  // (1) When moving forward:
@@ -113,19 +140,12 @@ class DBIter final : public Iterator {
    uint64_t skip_count_;
  };

-  DBIter(Env* _env, const ReadOptions& read_options,
-         const ImmutableOptions& ioptions,
-         const MutableCFOptions& mutable_cf_options, const Comparator* cmp,
-         InternalIterator* iter, const Version* version, SequenceNumber s,
-         bool arena_mode, uint64_t max_sequential_skip_in_iterations,
-         ReadCallback* read_callback, ColumnFamilyHandleImpl* cfh,
-         bool expose_blob_index);
-
  // No copying allowed
  DBIter(const DBIter&) = delete;
  void operator=(const DBIter&) = delete;

  ~DBIter() override {
+    MarkMemtableForFlushForAvgTrigger();
    ThreadStatus::OperationType cur_op_type =
        ThreadStatusUtil::GetThreadOperation();
    ThreadStatusUtil::SetThreadOperation(
@@ -220,7 +240,26 @@ class DBIter final : public Iterator {

  bool PrepareValue() override;

+  void Prepare(const MultiScanArgs& scan_opts) override {
+    std::optional<MultiScanArgs> new_scan_opts;
+    new_scan_opts.emplace(scan_opts);
+    scan_opts_.swap(new_scan_opts);
+    if (!scan_opts.empty()) {
+      iter_.Prepare(&scan_opts_.value());
+    } else {
+      iter_.Prepare(nullptr);
+    }
+  }
+
 private:
+  DBIter(Env* _env, const ReadOptions& read_options,
+         const ImmutableOptions& ioptions,
+         const MutableCFOptions& mutable_cf_options, const Comparator* cmp,
+         InternalIterator* iter, const Version* version, SequenceNumber s,
+         bool arena_mode, ReadCallback* read_callback,
+         ColumnFamilyHandleImpl* cfh, bool expose_blob_index,
+         ReadOnlyMemTable* active_mem);
+
  class BlobReader {
   public:
    BlobReader(const Version* version, ReadTier read_tier,
@@ -379,6 +418,36 @@ class DBIter final : public Iterator {
    return true;
  }

+  void MarkMemtableForFlushForAvgTrigger() {
+    if (avg_op_scan_flush_trigger_ &&
+        mem_hidden_op_scanned_since_seek_ >= memtable_op_scan_flush_trigger_ &&
+        mem_hidden_op_scanned_since_seek_ >=
+            static_cast<uint64_t>(iter_step_since_seek_) *
+                avg_op_scan_flush_trigger_) {
+      assert(memtable_op_scan_flush_trigger_ > 0);
+      active_mem_->MarkForFlush();
+      avg_op_scan_flush_trigger_ = 0;
+      memtable_op_scan_flush_trigger_ = 0;
+    }
+    iter_step_since_seek_ = 1;
+    mem_hidden_op_scanned_since_seek_ = 0;
+  }
+
+  void MarkMemtableForFlushForPerOpTrigger(uint64_t& mem_hidden_op_scanned) {
+    if (memtable_op_scan_flush_trigger_ &&
+        ikey_.sequence >= memtable_seqno_lb_) {
+      if (++mem_hidden_op_scanned >= memtable_op_scan_flush_trigger_) {
+        active_mem_->MarkForFlush();
+        // Turn off the flush trigger checks.
+        memtable_op_scan_flush_trigger_ = 0;
+        avg_op_scan_flush_trigger_ = 0;
+      }
+      if (avg_op_scan_flush_trigger_) {
+        ++mem_hidden_op_scanned_since_seek_;
+      }
+    }
+  }
+
  const SliceTransform* prefix_extractor_;
  Env* const env_;
  SystemClock* clock_;
@@ -425,6 +494,24 @@ class DBIter final : public Iterator {
  IterKey prefix_;

  Status status_;
+  Slice lazy_blob_index_;
+
+  // List of operands for merge operator.
+  MergeContext merge_context_;
+  LocalStatistics local_stats_;
+  PinnedIteratorsManager pinned_iters_mgr_;
+  ColumnFamilyHandleImpl* cfh_;
+  const Slice* const timestamp_ub_;
+  const Slice* const timestamp_lb_;
+  const size_t timestamp_size_;
+  std::string saved_timestamp_;
+  std::optional<MultiScanArgs> scan_opts_;
+  ReadOnlyMemTable* const active_mem_;
+  SequenceNumber memtable_seqno_lb_;
+  uint32_t memtable_op_scan_flush_trigger_;
+  uint32_t avg_op_scan_flush_trigger_;
+  uint32_t iter_step_since_seek_;
+  uint32_t mem_hidden_op_scanned_since_seek_;
  Direction direction_;
  bool valid_;
  bool current_entry_is_merged_;
@@ -443,29 +530,7 @@ class DBIter final : public Iterator {
  // the stacked BlobDB implementation is used, false otherwise.
  bool expose_blob_index_;
  bool allow_unprepared_value_;
-  Slice lazy_blob_index_;
  bool is_blob_;
  bool arena_mode_;
-  // List of operands for merge operator.
-  MergeContext merge_context_;
-  LocalStatistics local_stats_;
-  PinnedIteratorsManager pinned_iters_mgr_;
-  ColumnFamilyHandleImpl* cfh_;
-  const Slice* const timestamp_ub_;
-  const Slice* const timestamp_lb_;
-  const size_t timestamp_size_;
-  std::string saved_timestamp_;
 };
-
-// Return a new iterator that converts internal keys (yielded by
-// "*internal_iter") that were live at the specified `sequence` number
-// into appropriate user keys.
-Iterator* NewDBIterator(
-    Env* env, const ReadOptions& read_options, const ImmutableOptions& ioptions,
-    const MutableCFOptions& mutable_cf_options,
-    const Comparator* user_key_comparator, InternalIterator* internal_iter,
-    const Version* version, const SequenceNumber& sequence,
-    uint64_t max_sequential_skip_in_iterations, ReadCallback* read_callback,
-    ColumnFamilyHandleImpl* cfh = nullptr, bool expose_blob_index = false);
-
 }  // namespace ROCKSDB_NAMESPACE
@@ -528,12 +528,11 @@ TEST_F(DBIteratorStressTest, StressTest) {
                  internal_iter->target_hidden_fraction =
                      target_hidden_fraction;
                  internal_iter->trace = trace;
-                  db_iter.reset(NewDBIterator(
+                  db_iter.reset(DBIter::NewIter(
                      env_, ropt, ImmutableOptions(options),
                      MutableCFOptions(options), BytewiseComparator(),
-                      internal_iter, nullptr /* version */, sequence,
-                      options.max_sequential_skip_in_iterations,
-                      nullptr /*read_callback*/));
+                      internal_iter, /*version=*/nullptr, sequence,
+                      nullptr /*read_callback*/, /*active_mem=*/nullptr));
                }

                // Do a random operation. It's important to do it on ref_it
@@ -259,11 +259,10 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    internal_iter->Finish();

    ReadOptions ro;
-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -294,11 +293,10 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    internal_iter->Finish();

    ReadOptions ro;
-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -322,11 +320,10 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    ReadOptions ro;
    ro.iterate_upper_bound = &prefix;

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -356,11 +353,10 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    ReadOptions ro;
    ro.iterate_upper_bound = &prefix;

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -393,11 +389,10 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    ReadOptions ro;
    ro.iterate_upper_bound = &prefix;

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToLast();
    ASSERT_TRUE(!db_iter->Valid());
@@ -425,11 +420,10 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    ReadOptions ro;
    ro.iterate_upper_bound = &prefix;

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 7 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    SetPerfLevel(kEnableCount);
    ASSERT_TRUE(GetPerfLevel() == kEnableCount);
@@ -465,11 +459,10 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    ReadOptions ro;
    ro.iterate_upper_bound = &prefix;

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 4 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -492,11 +485,10 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    ReadOptions ro;
    ro.iterate_upper_bound = &prefix;

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToLast();
    ASSERT_TRUE(!db_iter->Valid());
@@ -517,11 +509,10 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    ReadOptions ro;
    ro.iterate_upper_bound = &prefix;

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -554,11 +545,10 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    ReadOptions ro;
    ro.iterate_upper_bound = &prefix;

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 7 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    SetPerfLevel(kEnableCount);
    ASSERT_TRUE(GetPerfLevel() == kEnableCount);
@@ -586,11 +576,10 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    internal_iter->Finish();

    ReadOptions ro;
-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToFirst();
    ASSERT_TRUE(db_iter->Valid());
@@ -631,11 +620,10 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    internal_iter->Finish();

    ReadOptions ro;
-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 2 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "b");
@@ -664,11 +652,10 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    internal_iter->Finish();

    ReadOptions ro;
-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -696,11 +683,10 @@ TEST_F(DBIteratorTest, DBIteratorEmpty) {
    TestIterator* internal_iter = new TestIterator(BytewiseComparator());
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 0 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(!db_iter->Valid());
    ASSERT_OK(db_iter->status());
@@ -710,11 +696,10 @@ TEST_F(DBIteratorTest, DBIteratorEmpty) {
    TestIterator* internal_iter = new TestIterator(BytewiseComparator());
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 0 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToFirst();
    ASSERT_TRUE(!db_iter->Valid());
    ASSERT_OK(db_iter->status());
@@ -735,11 +720,10 @@ TEST_F(DBIteratorTest, DBIteratorUseSkipCountSkips) {
  }
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      2 /* sequence */, options.max_sequential_skip_in_iterations,
-      nullptr /* read_callback */));
+      2 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));
  db_iter->SeekToLast();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -782,11 +766,11 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      internal_iter->Finish();

      options.statistics = ROCKSDB_NAMESPACE::CreateDBStatistics();
-      std::unique_ptr<Iterator> db_iter(NewDBIterator(
+      std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
          env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
          internal_iter, nullptr /* version */, i + 2 /* sequence */,
-          options.max_sequential_skip_in_iterations,
-          nullptr /* read_callback */));
+
+          nullptr /* read_callback */, /*active_mem=*/nullptr));
      db_iter->SeekToLast();
      ASSERT_TRUE(db_iter->Valid());

@@ -820,11 +804,11 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      internal_iter->AddPut("c", "200");
      internal_iter->Finish();

-      std::unique_ptr<Iterator> db_iter(NewDBIterator(
+      std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
          env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
          internal_iter, nullptr /* version */, i + 2 /* sequence */,
-          options.max_sequential_skip_in_iterations,
-          nullptr /* read_callback */));
+
+          nullptr /* read_callback */, /*active_mem=*/nullptr));
      db_iter->SeekToLast();
      ASSERT_TRUE(db_iter->Valid());

@@ -851,11 +835,10 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      internal_iter->AddPut("c", "200");
      internal_iter->Finish();

-      std::unique_ptr<Iterator> db_iter(NewDBIterator(
+      std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
          env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
          internal_iter, nullptr /* version */, 202 /* sequence */,
-          options.max_sequential_skip_in_iterations,
-          nullptr /* read_callback */));
+          nullptr /* read_callback */, /*active_mem=*/nullptr));
      db_iter->SeekToLast();
      ASSERT_TRUE(db_iter->Valid());

@@ -886,11 +869,11 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      }
      internal_iter->AddPut("c", "200");
      internal_iter->Finish();
-      std::unique_ptr<Iterator> db_iter(NewDBIterator(
+      std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
          env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
          internal_iter, nullptr /* version */, i /* sequence */,
-          options.max_sequential_skip_in_iterations,
-          nullptr /* read_callback */));
+
+          nullptr /* read_callback */, /*active_mem=*/nullptr));
      db_iter->SeekToLast();
      ASSERT_TRUE(!db_iter->Valid());
      ASSERT_OK(db_iter->status());
@@ -906,11 +889,10 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
    }
    internal_iter->AddPut("c", "200");
    internal_iter->Finish();
-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 200 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -944,11 +926,11 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      }
      internal_iter->Finish();

-      std::unique_ptr<Iterator> db_iter(NewDBIterator(
+      std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
          env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
          internal_iter, nullptr /* version */, i + 2 /* sequence */,
-          options.max_sequential_skip_in_iterations,
-          nullptr /* read_callback */));
+
+          nullptr /* read_callback */, /*active_mem=*/nullptr));
      db_iter->SeekToLast();
      ASSERT_TRUE(db_iter->Valid());

@@ -981,11 +963,11 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      }
      internal_iter->Finish();

-      std::unique_ptr<Iterator> db_iter(NewDBIterator(
+      std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
          env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
          internal_iter, nullptr /* version */, i + 2 /* sequence */,
-          options.max_sequential_skip_in_iterations,
-          nullptr /* read_callback */));
+
+          nullptr /* read_callback */, /*active_mem=*/nullptr));
      db_iter->SeekToLast();
      ASSERT_TRUE(db_iter->Valid());

@@ -1033,11 +1015,10 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
    internal_iter->Finish();

    ro.max_skippable_internal_keys = 0;
-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToFirst();
    ASSERT_TRUE(db_iter->Valid());
@@ -1081,11 +1062,10 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
    internal_iter->Finish();

    ro.max_skippable_internal_keys = 2;
-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToFirst();
    ASSERT_TRUE(db_iter->Valid());
@@ -1127,11 +1107,10 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
    internal_iter->Finish();

    ro.max_skippable_internal_keys = 2;
-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToFirst();
    ASSERT_TRUE(db_iter->Valid());
@@ -1167,11 +1146,10 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
    internal_iter->Finish();

    ro.max_skippable_internal_keys = 2;
-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToFirst();
    ASSERT_TRUE(db_iter->Valid());
@@ -1204,11 +1182,10 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
    internal_iter->Finish();

    ro.max_skippable_internal_keys = 2;
-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -1236,11 +1213,10 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
    internal_iter->Finish();

    ro.max_skippable_internal_keys = 2;
-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToFirst();
    ASSERT_TRUE(db_iter->Valid());
@@ -1275,11 +1251,10 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
    internal_iter->Finish();

    ro.max_skippable_internal_keys = 2;
-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToFirst();
    ASSERT_TRUE(db_iter->Valid());
@@ -1314,11 +1289,11 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {
      internal_iter->Finish();

      ro.max_skippable_internal_keys = i;
-      std::unique_ptr<Iterator> db_iter(NewDBIterator(
+      std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
          env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
          internal_iter, nullptr /* version */, 2 * i + 1 /* sequence */,
-          options.max_sequential_skip_in_iterations,
-          nullptr /* read_callback */));
+
+          nullptr /* read_callback */, /*active_mem=*/nullptr));

      db_iter->SeekToFirst();
      ASSERT_TRUE(db_iter->Valid());
@@ -1369,11 +1344,10 @@ TEST_F(DBIteratorTest, DBIteratorSkipInternalKeys) {

      options.max_sequential_skip_in_iterations = 1000;
      ro.max_skippable_internal_keys = i;
-      std::unique_ptr<Iterator> db_iter(NewDBIterator(
-          env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
+      std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
+          env_, ro, ioptions, MutableCFOptions(options), BytewiseComparator(),
          internal_iter, nullptr /* version */, 2 * i + 1 /* sequence */,
-          options.max_sequential_skip_in_iterations,
-          nullptr /* read_callback */));
+          nullptr /* read_callback */, /*active_mem=*/nullptr));

      db_iter->SeekToFirst();
      ASSERT_TRUE(db_iter->Valid());
@@ -1412,11 +1386,11 @@ TEST_F(DBIteratorTest, DBIteratorTimedPutBasic) {
  internal_iter->AddTimedPut("d", "3", /*write_unix_time=*/0);
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  options.max_sequential_skip_in_iterations = 1;
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      7 /* sequence */, /*max_sequential_skip_in_iterations*/ 1,
-      nullptr /* read_callback */));
+      7 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1463,11 +1437,10 @@ TEST_F(DBIteratorTest, DBIterator1) {
  internal_iter->AddMerge("b", "2");
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      1 /* sequence */, options.max_sequential_skip_in_iterations,
-      nullptr /* read_callback */));
+      1 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1493,11 +1466,10 @@ TEST_F(DBIteratorTest, DBIterator2) {
  internal_iter->AddMerge("b", "2");
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      0 /* sequence */, options.max_sequential_skip_in_iterations,
-      nullptr /* read_callback */));
+      0 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1519,11 +1491,10 @@ TEST_F(DBIteratorTest, DBIterator3) {
  internal_iter->AddMerge("b", "2");
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      2 /* sequence */, options.max_sequential_skip_in_iterations,
-      nullptr /* read_callback */));
+      2 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1545,11 +1516,10 @@ TEST_F(DBIteratorTest, DBIterator4) {
  internal_iter->AddMerge("b", "2");
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      4 /* sequence */, options.max_sequential_skip_in_iterations,
-      nullptr /* read_callback */));
+      4 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1580,11 +1550,10 @@ TEST_F(DBIteratorTest, DBIterator5) {
    internal_iter->AddMerge("a", "merge_6");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 0 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1605,11 +1574,10 @@ TEST_F(DBIteratorTest, DBIterator5) {
    internal_iter->AddMerge("a", "merge_6");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 1 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1630,11 +1598,10 @@ TEST_F(DBIteratorTest, DBIterator5) {
    internal_iter->AddMerge("a", "merge_6");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 2 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1655,11 +1622,10 @@ TEST_F(DBIteratorTest, DBIterator5) {
    internal_iter->AddMerge("a", "merge_6");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 3 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1680,11 +1646,10 @@ TEST_F(DBIteratorTest, DBIterator5) {
    internal_iter->AddMerge("a", "merge_6");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 4 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1705,11 +1670,10 @@ TEST_F(DBIteratorTest, DBIterator5) {
    internal_iter->AddMerge("a", "merge_6");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 5 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1730,11 +1694,10 @@ TEST_F(DBIteratorTest, DBIterator5) {
    internal_iter->AddMerge("a", "merge_6");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 6 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1753,11 +1716,10 @@ TEST_F(DBIteratorTest, DBIterator5) {
    internal_iter->AddMerge("a", "merge_2");
    internal_iter->AddPut("b", "val_b");
    internal_iter->Finish();
-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 10 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->Seek("b");
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "b");
@@ -1785,11 +1747,10 @@ TEST_F(DBIteratorTest, DBIterator6) {
    internal_iter->AddMerge("a", "merge_6");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 0 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1810,11 +1771,10 @@ TEST_F(DBIteratorTest, DBIterator6) {
    internal_iter->AddMerge("a", "merge_6");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 1 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1835,11 +1795,10 @@ TEST_F(DBIteratorTest, DBIterator6) {
    internal_iter->AddMerge("a", "merge_6");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 2 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1860,11 +1819,10 @@ TEST_F(DBIteratorTest, DBIterator6) {
    internal_iter->AddMerge("a", "merge_6");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 3 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(!db_iter->Valid());
    ASSERT_OK(db_iter->status());
@@ -1881,11 +1839,10 @@ TEST_F(DBIteratorTest, DBIterator6) {
    internal_iter->AddMerge("a", "merge_6");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 4 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1906,11 +1863,10 @@ TEST_F(DBIteratorTest, DBIterator6) {
    internal_iter->AddMerge("a", "merge_6");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 5 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1931,11 +1887,10 @@ TEST_F(DBIteratorTest, DBIterator6) {
    internal_iter->AddMerge("a", "merge_6");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 6 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1976,11 +1931,10 @@ TEST_F(DBIteratorTest, DBIterator7) {
    internal_iter->AddDeletion("c");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 0 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -2013,11 +1967,10 @@ TEST_F(DBIteratorTest, DBIterator7) {
    internal_iter->AddDeletion("c");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 2 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -2056,11 +2009,10 @@ TEST_F(DBIteratorTest, DBIterator7) {
    internal_iter->AddDeletion("c");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 4 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -2099,11 +2051,10 @@ TEST_F(DBIteratorTest, DBIterator7) {
    internal_iter->AddDeletion("c");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 5 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -2147,11 +2098,10 @@ TEST_F(DBIteratorTest, DBIterator7) {
    internal_iter->AddDeletion("c");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 6 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -2196,11 +2146,10 @@ TEST_F(DBIteratorTest, DBIterator7) {
    internal_iter->AddDeletion("c");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 7 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -2239,11 +2188,10 @@ TEST_F(DBIteratorTest, DBIterator7) {
    internal_iter->AddDeletion("c");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 9 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -2288,11 +2236,10 @@ TEST_F(DBIteratorTest, DBIterator7) {
    internal_iter->AddDeletion("c");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 13 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -2338,11 +2285,10 @@ TEST_F(DBIteratorTest, DBIterator7) {
    internal_iter->AddDeletion("c");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+    std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
        env_, ro, ioptions, mutable_cf_options, BytewiseComparator(),
        internal_iter, nullptr /* version */, 14 /* sequence */,
-        options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+        nullptr /* read_callback */, /*active_mem=*/nullptr));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -2371,11 +2317,10 @@ TEST_F(DBIteratorTest, DBIterator8) {
  internal_iter->AddPut("b", "0");
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      10 /* sequence */, options.max_sequential_skip_in_iterations,
-      nullptr /* read_callback */));
+      10 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));
  db_iter->SeekToLast();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "b");
@@ -2403,11 +2348,11 @@ TEST_F(DBIteratorTest, DBIterator9) {
    internal_iter->AddMerge("d", "merge_6");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
-        env_, ro, ImmutableOptions(options), MutableCFOptions(options),
-        BytewiseComparator(), internal_iter, nullptr /* version */,
-        10 /* sequence */, options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+    std::unique_ptr<Iterator> db_iter(
+        DBIter::NewIter(env_, ro, ImmutableOptions(options),
+                        MutableCFOptions(options), BytewiseComparator(),
+                        internal_iter, nullptr /* version */, 10 /* sequence */,
+                        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -2471,11 +2416,10 @@ TEST_F(DBIteratorTest, DBIterator10) {
  internal_iter->AddPut("d", "4");
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      10 /* sequence */, options.max_sequential_skip_in_iterations,
-      nullptr /* read_callback */));
+      10 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));

  db_iter->Seek("c");
  ASSERT_TRUE(db_iter->Valid());
@@ -2512,10 +2456,10 @@ TEST_F(DBIteratorTest, SeekToLastOccurrenceSeq0) {
  internal_iter->AddPut("b", "2");
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      10 /* sequence */, 0 /* force seek */, nullptr /* read_callback */));
+      10 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -2542,11 +2486,10 @@ TEST_F(DBIteratorTest, DBIterator11) {
  internal_iter->AddMerge("b", "2");
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      1 /* sequence */, options.max_sequential_skip_in_iterations,
-      nullptr /* read_callback */));
+      1 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -2571,10 +2514,10 @@ TEST_F(DBIteratorTest, DBIterator12) {
  internal_iter->AddSingleDeletion("b");
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      10 /* sequence */, 0 /* force seek */, nullptr /* read_callback */));
+      10 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));
  db_iter->SeekToLast();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -2610,11 +2553,11 @@ TEST_F(DBIteratorTest, DBIterator13) {
  internal_iter->AddPut(key, "8");
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  options.max_sequential_skip_in_iterations = 3;
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      2 /* sequence */, 3 /* max_sequential_skip_in_iterations */,
-      nullptr /* read_callback */));
+      2 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));
  db_iter->Seek("b");
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), key);
@@ -2640,11 +2583,11 @@ TEST_F(DBIteratorTest, DBIterator14) {
  internal_iter->AddPut("c", "9");
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  options.max_sequential_skip_in_iterations = 1;
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      4 /* sequence */, 1 /* max_sequential_skip_in_iterations */,
-      nullptr /* read_callback */));
+      4 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));
  db_iter->Seek("b");
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "b");
@@ -2680,11 +2623,12 @@ class DBIterWithMergeIterTest : public testing::Test {
    InternalIterator* merge_iter =
        NewMergingIterator(&icomp_, child_iters.data(), 2u);

-    db_iter_.reset(NewDBIterator(
+    options_.max_sequential_skip_in_iterations = 3;
+    db_iter_.reset(DBIter::NewIter(
        env_, ro_, ImmutableOptions(options_), MutableCFOptions(options_),
        BytewiseComparator(), merge_iter, nullptr /* version */,
-        8 /* read data earlier than seqId 8 */,
-        3 /* max iterators before reseek */, nullptr /* read_callback */));
+        8 /* read data earlier than seqId 8 */, nullptr /* read_callback */,
+        /*active_mem=*/nullptr));
  }

  Env* env_;
@@ -3120,11 +3064,10 @@ TEST_F(DBIteratorTest, SeekPrefixTombstones) {
  internal_iter->Finish();

  ro.prefix_same_as_start = true;
-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      10 /* sequence */, options.max_sequential_skip_in_iterations,
-      nullptr /* read_callback */));
+      10 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));

  int skipped_keys = 0;

@@ -3157,11 +3100,11 @@ TEST_F(DBIteratorTest, SeekToFirstLowerBound) {
    Slice lower_bound(lower_bound_str);
    ro.iterate_lower_bound = &lower_bound;
    Options options;
-    std::unique_ptr<Iterator> db_iter(NewDBIterator(
-        env_, ro, ImmutableOptions(options), MutableCFOptions(options),
-        BytewiseComparator(), internal_iter, nullptr /* version */,
-        10 /* sequence */, options.max_sequential_skip_in_iterations,
-        nullptr /* read_callback */));
+    std::unique_ptr<Iterator> db_iter(
+        DBIter::NewIter(env_, ro, ImmutableOptions(options),
+                        MutableCFOptions(options), BytewiseComparator(),
+                        internal_iter, nullptr /* version */, 10 /* sequence */,
+                        nullptr /* read_callback */, /*active_mem=*/nullptr));

    db_iter->SeekToFirst();
    if (i == kNumKeys + 1) {
@@ -3197,11 +3140,10 @@ TEST_F(DBIteratorTest, PrevLowerBound) {
  Slice lower_bound(lower_bound_str);
  ro.iterate_lower_bound = &lower_bound;
  Options options;
-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      10 /* sequence */, options.max_sequential_skip_in_iterations,
-      nullptr /* read_callback */));
+      10 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));

  db_iter->SeekToLast();
  for (int i = kNumKeys; i >= kLowerBound; --i) {
@@ -3226,11 +3168,10 @@ TEST_F(DBIteratorTest, SeekLessLowerBound) {
  Slice lower_bound(lower_bound_str);
  ro.iterate_lower_bound = &lower_bound;
  Options options;
-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ro, ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      10 /* sequence */, options.max_sequential_skip_in_iterations,
-      nullptr /* read_callback */));
+      10 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));

  auto before_lower_bound_str = std::to_string(kLowerBound - 1);
  Slice before_lower_bound(lower_bound_str);
@@ -3252,11 +3193,10 @@ TEST_F(DBIteratorTest, ReverseToForwardWithDisappearingKeys) {
  }
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+  std::unique_ptr<Iterator> db_iter(DBIter::NewIter(
      env_, ReadOptions(), ImmutableOptions(options), MutableCFOptions(options),
      BytewiseComparator(), internal_iter, nullptr /* version */,
-      10 /* sequence */, options.max_sequential_skip_in_iterations,
-      nullptr /* read_callback */));
+      10 /* sequence */, nullptr /* read_callback */, /*active_mem=*/nullptr));

  db_iter->SeekForPrev("a");
  ASSERT_TRUE(db_iter->Valid());
@@ -8,6 +8,8 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

 #include <functional>
+#include <iomanip>
+#include <iostream>

 #include "db/arena_wrapped_db_iter.h"
 #include "db/db_iter.h"
@@ -3824,6 +3826,555 @@ TEST_F(DBIteratorTest, IteratorsConsistentViewExplicitSnapshot) {
  }
 }

+TEST_P(DBIteratorTest, MemtableOpsScanFlushTriggerWithSeek) {
+  // Tests that option memtable_op_scan_flush_trigger works when the limit
+  // is reached during a Seek() operation.
+  const int kTrigger = 10;
+  Random* r = Random::GetTLSInstance();
+
+  for (int trigger : {kTrigger, kTrigger + 1}) {
+    for (bool delete_only : {false, true}) {
+      Options options;
+      options.create_if_missing = true;
+      options.memtable_op_scan_flush_trigger = trigger;
+      options.level_compaction_dynamic_level_bytes = true;
+      DestroyAndReopen(options);
+
+      // Base data that will be covered by a consecutive sequence of tombstones.
+      int kNumKeys = delete_only ? kTrigger : kTrigger / 2;
+      for (int i = 0; i < kNumKeys; ++i) {
+        ASSERT_OK(Put(Key(i), r->RandomString(100)));
+      }
+      ASSERT_OK(Flush());
+      ASSERT_OK(db_->CompactRange({}, nullptr, nullptr));
+      ASSERT_EQ(1, NumTableFilesAtLevel(6));
+
+      if (delete_only) {
+        for (int i = 0; i < kNumKeys; ++i) {
+          ASSERT_OK(SingleDelete(Key(i)));
+        }
+      } else {
+        for (int i = 0; i < kNumKeys; ++i) {
+          ASSERT_OK(Put(Key(i), r->RandomString(100)));
+        }
+        for (int i = 0; i < kNumKeys; ++i) {
+          ASSERT_OK(Delete(Key(i)));
+        }
+      }
+
+      SetPerfLevel(PerfLevel::kEnableCount);
+      get_perf_context()->Reset();
+      ReadOptions ro;
+      std::unique_ptr<Iterator> iter(db_->NewIterator(ro));
+
+      // Seek to the first key, this will scan through all the tombstones and
+      // hidden puts
+      iter->Seek(Key(0));
+      ASSERT_FALSE(
+          iter->Valid());  // All keys are deleted, so iterator is not valid
+      ASSERT_OK(iter->status());
+      ASSERT_EQ(get_perf_context()->next_on_memtable_count, kTrigger);
+
+      // Skipping kNumTrigger memtable entries in a single iterator operation
+      // should mark the memtable for flush.
+      //
+      // At the end of a write, we check and update memtable to request a flush
+      ASSERT_OK(Put(Key(11), "val"));
+      // Before a write, we schedule memtables for flush if requested.
+      ASSERT_OK(Put(Key(12), "val"));
+      ASSERT_OK(db_->WaitForCompact({}));
+
+      if (trigger <= kTrigger) {
+        // Check if memtable was flushed due to scan trigger
+        ASSERT_EQ(1, NumTableFilesAtLevel(0));
+        uint64_t val = 0;
+        ASSERT_TRUE(
+            db_->GetIntProperty("rocksdb.num-deletes-active-mem-table", &val));
+        ASSERT_EQ(0, val);
+      } else {
+        ASSERT_EQ(0, NumTableFilesAtLevel(0));
+        uint64_t val = 0;
+        ASSERT_TRUE(
+            db_->GetIntProperty("rocksdb.num-deletes-active-mem-table", &val));
+        ASSERT_EQ(kNumKeys, val);
+      }
+    }
+  }
+}
+
+TEST_P(DBIteratorTest, MemtableOpsScanFlushTriggerWithNext) {
+  // Tests that option memtable_op_scan_flush_trigger works when the limit
+  // is reached during a Next() operation, and not trigger a flush when
+  // the limit is reached across multiple Next() operations.
+  const int kTrigger = 10;
+  Random* r = Random::GetTLSInstance();
+
+  for (int trigger : {kTrigger, kTrigger + 1}) {
+    for (bool delete_only : {false, true}) {
+      Options options;
+      options.create_if_missing = true;
+      options.memtable_op_scan_flush_trigger = trigger;
+      options.level_compaction_dynamic_level_bytes = true;
+      DestroyAndReopen(options);
+
+      // Base data that will be covered by a consecutive sequence of tombstones.
+      int kNumKeys = delete_only ? kTrigger : kTrigger / 2;
+      for (int i = 0; i <= kNumKeys; ++i) {
+        ASSERT_OK(Put(Key(i), r->RandomString(100)));
+      }
+      ASSERT_OK(Flush());
+      ASSERT_OK(db_->CompactRange({}, nullptr, nullptr));
+      ASSERT_EQ(1, NumTableFilesAtLevel(6));
+
+      ASSERT_OK(Put(Key(0), "val"));
+      if (delete_only) {
+        for (int i = 1; i <= kNumKeys; ++i) {
+          ASSERT_OK(SingleDelete(Key(i)));
+        }
+      } else {
+        for (int i = 1; i <= kNumKeys; ++i) {
+          ASSERT_OK(Put(Key(i), r->RandomString(100)));
+        }
+        for (int i = 1; i <= kNumKeys; ++i) {
+          ASSERT_OK(Delete(Key(i)));
+        }
+      }
+
+      // Total number of tombstones and hidden puts scanned across multiple
+      // Next() operations below will be kTrigger, and it should not trigger a
+      // flush when the limit is kTrigger + 1.
+      ASSERT_OK(Put(Key(kNumKeys + 1), "v1"));
+      ASSERT_OK(Delete(Key(kNumKeys + 2)));
+      ASSERT_OK(Put(Key(kNumKeys + 3), "v3"));
+
+      SetPerfLevel(PerfLevel::kEnableCount);
+      get_perf_context()->Reset();
+      ReadOptions ro;
+      std::unique_ptr<Iterator> iter(db_->NewIterator(ro));
+      iter->Seek(Key(0));
+      ASSERT_TRUE(iter->Valid());
+      ASSERT_EQ(iter->value(), "val");
+      ASSERT_OK(iter->status());
+      ASSERT_EQ(get_perf_context()->next_on_memtable_count, 0);
+      iter->Next();
+      // kTrigger tombstones and invisible puts and 1 for the visible put
+      ASSERT_EQ(get_perf_context()->next_on_memtable_count, kTrigger + 1);
+      iter->Next();
+      ASSERT_EQ(get_perf_context()->next_on_memtable_count, kTrigger + 3);
+
+      // Skipping kNumTrigger memtable entries in a single iterator operation
+      // should mark the memtable for flush.
+      //
+      // At the end of a write, we check and update memtable to request a flush
+      ASSERT_OK(Put(Key(11), "val"));
+      // Before a write, we schedule memtables for flush if requested.
+      ASSERT_OK(Put(Key(12), "val"));
+      ASSERT_OK(db_->WaitForCompact({}));
+
+      if (trigger <= kTrigger) {
+        // Check if memtable was flushed due to scan trigger
+        ASSERT_EQ(1, NumTableFilesAtLevel(0));
+        uint64_t val = 0;
+        ASSERT_TRUE(
+            db_->GetIntProperty("rocksdb.num-deletes-active-mem-table", &val));
+        ASSERT_EQ(0, val);
+      } else {
+        uint64_t val = 0;
+        ASSERT_TRUE(
+            db_->GetIntProperty("rocksdb.num-deletes-active-mem-table", &val));
+        ASSERT_EQ(kNumKeys + 1, val);
+      }
+    }
+  }
+}
+
+TEST_P(DBIteratorTest, AverageMemtableOpsScanFlushTrigger) {
+  // Tests option memtable_avg_op_scan_flush_trigger with
+  // long tombstone sequences.
+  Random* r = Random::GetTLSInstance();
+
+  const int kAvgTrigger = 10;
+  const int kMaxTrigger = 500;
+  Options options;
+  options.create_if_missing = true;
+  options.memtable_op_scan_flush_trigger = kMaxTrigger;
+  options.memtable_avg_op_scan_flush_trigger = kAvgTrigger;
+  options.level_compaction_dynamic_level_bytes = true;
+  DestroyAndReopen(options);
+
+  const int kNumKeys = 1000;
+  // Base data that will be covered by a consecutive sequence of tombstones.
+  for (int i = 0; i < kNumKeys; ++i) {
+    ASSERT_OK(Put(Key(i), r->RandomString(50)));
+  }
+  ASSERT_OK(Flush());
+  ASSERT_OK(db_->CompactRange({}, nullptr, nullptr));
+  ASSERT_EQ(1, NumTableFilesAtLevel(6));
+
+  for (int i = 0; i < kNumKeys; ++i) {
+    // We issue slightly more deletions than kAvgTrigger between visible keys
+    // to ensure avg skipped entries exceed kAvgTrigger.
+    if (i % (kAvgTrigger + 2) != 0) {
+      ASSERT_OK(SingleDelete(Key(i)));
+    }
+  }
+
+  // Each operation, except the first Seek, is expected to see kAvgTrigger + 1
+  // tombstones (from the active memtable) before it finds the next visible key.
+  SetPerfLevel(PerfLevel::kEnableCount);
+  get_perf_context()->Reset();
+  std::unique_ptr<Iterator> iter(db_->NewIterator(ReadOptions()));
+  iter->Seek(Key(1));
+  ASSERT_EQ(get_perf_context()->next_on_memtable_count, kAvgTrigger + 1);
+  iter.reset();
+  // Should not flush since total entries skipped is below
+  // memtable_op_scan_flush_trigger
+  ASSERT_OK(Put(Key(0), "dummy write"));
+  ASSERT_OK(Put(Key(0), "dummy write"));
+  ASSERT_OK(db_->WaitForCompact({}));
+  ASSERT_EQ(0, NumTableFilesAtLevel(0));
+
+  get_perf_context()->Reset();
+  iter.reset(db_->NewIterator(ReadOptions()));
+  int num_ops = 1;
+  uint64_t num_skipped = 0;
+  iter->Seek(Key(0));
+  ASSERT_EQ(iter->key(), Key(0));
+  uint64_t last_memtable_next_count =
+      get_perf_context()->next_on_memtable_count;
+  iter->Next();
+  num_ops++;
+  while (iter->Valid()) {
+    ASSERT_OK(iter->status());
+    uint64_t num_skipped_in_op =
+        get_perf_context()->next_on_memtable_count - last_memtable_next_count;
+    ASSERT_GE(num_skipped_in_op, kAvgTrigger + 1);
+    last_memtable_next_count = get_perf_context()->next_on_memtable_count;
+    num_skipped += num_skipped_in_op;
+    iter->Next();
+    num_ops++;
+  }
+  // During iterator destruction we mark memtable for flush
+  iter.reset();
+
+  // avg trigger
+  ASSERT_GE(num_skipped, kAvgTrigger * num_ops);
+  // memtable_op_scan_flush_trigger
+  ASSERT_GE(num_skipped, kMaxTrigger);
+  // Average hidden entries scanned from memtable per operation is more than
+  // kAvgTrigger and the total skipped is more than
+  // memtable_op_scan_flush_trigger, the current memtable should be marked for
+  // flush. The following two writes will trigger the flush.
+  ASSERT_OK(Put(Key(0), "dummy write"));
+  // Before a write, we schedule memtables for flush if requested.
+  ASSERT_OK(Put(Key(0), "dummy write"));
+  ASSERT_OK(db_->WaitForCompact({}));
+  ASSERT_EQ(1, NumTableFilesAtLevel(0));
+}
+
+TEST_P(DBIteratorTest, AverageMemtableOpsScanFlushTriggerByOverwrites) {
+  // Tests option memtable_avg_op_scan_flush_trigger with overwrites to keys.
+  Random* r = Random::GetTLSInstance();
+
+  const int kAvgTrigger = 25;
+  Options options;
+  options.create_if_missing = true;
+  options.memtable_op_scan_flush_trigger = 250;
+  options.memtable_avg_op_scan_flush_trigger = kAvgTrigger;
+  options.level_compaction_dynamic_level_bytes = true;
+  DestroyAndReopen(options);
+
+  const int kNumKeys = 100;
+  // Base data that will be covered by a consecutive sequence of tombstones.
+  for (int i = 0; i < kNumKeys; ++i) {
+    ASSERT_OK(Put(Key(i), r->RandomString(50)));
+  }
+  ASSERT_OK(Flush());
+  ASSERT_OK(db_->CompactRange({}, nullptr, nullptr));
+  ASSERT_EQ(1, NumTableFilesAtLevel(6));
+
+  // One visible key every 10 keys.
+  // Each non-visible user key has 3 non-visible entries in the active memtable.
+  for (int i = 0; i < kNumKeys; ++i) {
+    if (i % 10 != 0) {
+      ASSERT_OK(Put(Key(i), r->RandomString(50)));
+      ASSERT_OK(Put(Key(i), r->RandomString(50)));
+      ASSERT_OK(Delete(Key(i)));
+    }
+  }
+
+  SetPerfLevel(PerfLevel::kEnableCount);
+  get_perf_context()->Reset();
+  ReadOptions ro;
+  std::unique_ptr<Iterator> iter(db_->NewIterator(ro));
+  iter->Seek(Key(1));
+  ASSERT_GT(get_perf_context()->next_on_memtable_count, kAvgTrigger);
+  // Re-seek to trigger check for flush trigger
+  iter->Seek(Key(1));
+  // Should not flush since total entries skipped is below
+  // memtable_op_scan_flush_trigger
+  ASSERT_FALSE(static_cast<ColumnFamilyHandleImpl*>(db_->DefaultColumnFamily())
+                   ->cfd()
+                   ->mem()
+                   ->IsMarkedForFlush());
+  ASSERT_OK(Put(Key(0), "dummy write"));
+  ASSERT_OK(Put(Key(0), "dummy write"));
+  ASSERT_OK(db_->WaitForCompact({}));
+  ASSERT_EQ(0, NumTableFilesAtLevel(0));
+  get_perf_context()->Reset();
+
+  int num_ops = 1;
+  iter->Seek(Key(1));
+  while (iter->Valid()) {
+    num_ops++;
+    iter->Next();
+  }
+  ASSERT_GT(get_perf_context()->next_on_memtable_count, num_ops * kAvgTrigger);
+
+  // Re-seek should check conditions for marking memtable for flush
+  iter->Seek(Key(80));
+
+  // Average hidden entries scanned from memtable per operation is 2.
+  ASSERT_OK(Put(Key(0), "dummy write"));
+  // Before a write, we schedule memtables for flush if requested.
+  ASSERT_OK(Put(Key(0), "dummy write"));
+  ASSERT_OK(db_->WaitForCompact({}));
+  ASSERT_EQ(1, NumTableFilesAtLevel(0));
+}
+
+class DBMultiScanIteratorTest : public DBTestBase,
+                                public ::testing::WithParamInterface<bool> {
+ public:
+  DBMultiScanIteratorTest()
+      : DBTestBase("db_multi_scan_iterator_test", /*env_do_fsync=*/true) {}
+};
+
+INSTANTIATE_TEST_CASE_P(DBMultiScanIteratorTest, DBMultiScanIteratorTest,
+                        ::testing::Bool());
+
+TEST_P(DBMultiScanIteratorTest, BasicTest) {
+  // Create a file
+  for (int i = 0; i < 100; ++i) {
+    std::stringstream ss;
+    ss << std::setw(2) << std::setfill('0') << i;
+    ASSERT_OK(Put("k" + ss.str(), "val" + ss.str()));
+  }
+  ASSERT_OK(Flush());
+
+  std::vector<std::string> key_ranges({"k03", "k10", "k25", "k50"});
+  ReadOptions ro;
+  ro.fill_cache = GetParam();
+  MultiScanArgs scan_options(BytewiseComparator());
+  scan_options.insert(key_ranges[0], key_ranges[1]);
+  scan_options.insert(key_ranges[2], key_ranges[3]);
+  ColumnFamilyHandle* cfh = dbfull()->DefaultColumnFamily();
+  std::unique_ptr<MultiScan> iter =
+      dbfull()->NewMultiScan(ro, cfh, scan_options);
+  try {
+    int idx = 0;
+    int count = 0;
+    for (auto range : *iter) {
+      for (auto it : range) {
+        ASSERT_GE(it.first.ToString().compare(key_ranges[idx]), 0);
+        ASSERT_LT(it.first.ToString().compare(key_ranges[idx + 1]), 0);
+        count++;
+      }
+      idx += 2;
+    }
+    ASSERT_EQ(count, 32);
+  } catch (MultiScanException& ex) {
+    // Make sure exception contains the status
+    ASSERT_NOK(ex.status());
+    std::cerr << "Iterator returned status " << ex.what();
+    abort();
+  } catch (std::logic_error& ex) {
+    std::cerr << "Iterator returned logic error " << ex.what();
+    abort();
+  }
+  iter.reset();
+
+  // Test the overlapping scan case
+  key_ranges[1] = "k30";
+  scan_options = MultiScanArgs(BytewiseComparator());
+  scan_options.insert(key_ranges[0], key_ranges[1]);
+  scan_options.insert(key_ranges[2], key_ranges[3]);
+
+  iter = dbfull()->NewMultiScan(ro, cfh, scan_options);
+  try {
+    int idx = 0;
+    int count = 0;
+    for (auto range : *iter) {
+      for (auto it : range) {
+        ASSERT_GE(it.first.ToString().compare(key_ranges[idx]), 0);
+        ASSERT_LT(it.first.ToString().compare(key_ranges[idx + 1]), 0);
+        count++;
+      }
+      idx += 2;
+    }
+    ASSERT_EQ(count, 52);
+  } catch (MultiScanException& ex) {
+    // Make sure exception contains the status
+    ASSERT_NOK(ex.status());
+    std::cerr << "Iterator returned status " << ex.what();
+    abort();
+  } catch (std::logic_error& ex) {
+    std::cerr << "Iterator returned logic error " << ex.what();
+    abort();
+  }
+  iter.reset();
+
+  // Test the no limit scan case
+  scan_options = MultiScanArgs(BytewiseComparator());
+  scan_options.insert(key_ranges[0]);
+  scan_options.insert(key_ranges[2]);
+  iter = dbfull()->NewMultiScan(ro, cfh, scan_options);
+  try {
+    int idx = 0;
+    int count = 0;
+    for (auto range : *iter) {
+      for (auto it : range) {
+        ASSERT_GE(it.first.ToString().compare(key_ranges[idx]), 0);
+        if (it.first.ToString().compare(key_ranges[idx + 1]) == 0) {
+          break;
+        }
+        count++;
+      }
+      idx += 2;
+    }
+    ASSERT_EQ(count, 52);
+  } catch (MultiScanException& ex) {
+    // Make sure exception contains the status
+    ASSERT_NOK(ex.status());
+    std::cerr << "Iterator returned status " << ex.what();
+    abort();
+  } catch (std::logic_error& ex) {
+    std::cerr << "Iterator returned logic error " << ex.what();
+    abort();
+  }
+  iter.reset();
+}
+
+TEST_P(DBMultiScanIteratorTest, MixedBoundsTest) {
+  // Create a file
+  for (int i = 0; i < 100; ++i) {
+    std::stringstream ss;
+    ss << std::setw(2) << std::setfill('0') << i;
+    ASSERT_OK(Put("k" + ss.str(), "val" + ss.str()));
+  }
+  ASSERT_OK(Flush());
+
+  std::vector<std::string> key_ranges(
+      {"k03", "k10", "k25", "k50", "k75", "k90"});
+  ReadOptions ro;
+  ro.fill_cache = GetParam();
+  MultiScanArgs scan_options(BytewiseComparator());
+  scan_options.insert(key_ranges[0], key_ranges[1]);
+  scan_options.insert(key_ranges[2]);
+  scan_options.insert(key_ranges[4], key_ranges[5]);
+  ColumnFamilyHandle* cfh = dbfull()->DefaultColumnFamily();
+  std::unique_ptr<MultiScan> iter =
+      dbfull()->NewMultiScan(ro, cfh, scan_options);
+  try {
+    int idx = 0;
+    int count = 0;
+    for (auto range : *iter) {
+      for (auto it : range) {
+        ASSERT_GE(
+            it.first.ToString().compare(
+                scan_options.GetScanRanges()[idx].range.start->ToString()),
+            0);
+        if (scan_options.GetScanRanges()[idx].range.limit) {
+          ASSERT_LT(
+              it.first.ToString().compare(
+                  scan_options.GetScanRanges()[idx].range.limit->ToString()),
+              0);
+        }
+        count++;
+      }
+      idx++;
+    }
+    ASSERT_EQ(count, 97);
+  } catch (MultiScanException& ex) {
+    // Make sure exception contains the status
+    ASSERT_NOK(ex.status());
+    std::cerr << "Iterator returned status " << ex.what();
+    abort();
+  } catch (std::logic_error& ex) {
+    std::cerr << "Iterator returned logic error " << ex.what();
+    abort();
+  }
+  iter.reset();
+  scan_options = MultiScanArgs(BytewiseComparator());
+  scan_options.insert(key_ranges[0]);
+  scan_options.insert(key_ranges[2], key_ranges[3]);
+  scan_options.insert(key_ranges[4]);
+  iter = dbfull()->NewMultiScan(ro, cfh, scan_options);
+  try {
+    int idx = 0;
+    int count = 0;
+    for (auto range : *iter) {
+      for (auto it : range) {
+        ASSERT_GE(
+            it.first.ToString().compare(
+                scan_options.GetScanRanges()[idx].range.start->ToString()),
+            0);
+        if (scan_options.GetScanRanges()[idx].range.limit) {
+          ASSERT_LT(
+              it.first.ToString().compare(
+                  scan_options.GetScanRanges()[idx].range.limit->ToString()),
+              0);
+        }
+        count++;
+      }
+      idx++;
+    }
+    ASSERT_EQ(count, 147);
+  } catch (MultiScanException& ex) {
+    // Make sure exception contains the status
+    ASSERT_NOK(ex.status());
+    std::cerr << "Iterator returned status " << ex.what();
+    abort();
+  } catch (std::logic_error& ex) {
+    std::cerr << "Iterator returned logic error " << ex.what();
+    abort();
+  }
+  iter.reset();
+}
+
+TEST_P(DBMultiScanIteratorTest, RangeAcrossFiles) {
+  auto options = CurrentOptions();
+  options.target_file_size_base = 100 << 10;  // 20KB
+  options.compaction_style = kCompactionStyleUniversal;
+  options.num_levels = 50;
+  options.compression = kNoCompression;
+  DestroyAndReopen(options);
+
+  auto rnd = Random::GetTLSInstance();
+  // Write ~200KB data
+  for (int i = 0; i < 100; ++i) {
+    ASSERT_OK(Put(Key(i), rnd->RandomString(2 << 10)));
+  }
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(db_->CompactRange({}, nullptr, nullptr));
+  ASSERT_EQ(2, NumTableFilesAtLevel(49));
+  std::vector<std::string> key_ranges({Key(10), Key(90)});
+  ReadOptions ro;
+  ro.fill_cache = GetParam();
+  MultiScanArgs scan_options(BytewiseComparator());
+  scan_options.insert(key_ranges[0], key_ranges[1]);
+  ColumnFamilyHandle* cfh = dbfull()->DefaultColumnFamily();
+  std::unique_ptr<MultiScan> iter =
+      dbfull()->NewMultiScan(ro, cfh, scan_options);
+  int i = 10;
+  for (auto range : *iter) {
+    for (auto it : range) {
+      ASSERT_EQ(it.first.ToString(), Key(i));
+      ++i;
+    }
+  }
+  ASSERT_EQ(i, 90);
+}
 }  // namespace ROCKSDB_NAMESPACE

 int main(int argc, char** argv) {
@@ -312,12 +312,12 @@ TEST_P(DbKvChecksumTest, WriteToWALCorrupted) {
    // Corrupted write batch leads to read-only mode, so we have to
    // reopen for every attempt.
    Reopen(options);
-    auto log_size_pre_write = dbfull()->TEST_total_log_size();
+    auto log_size_pre_write = dbfull()->TEST_wals_total_size();

    SyncPoint::GetInstance()->EnableProcessing();
    ASSERT_TRUE(ExecuteWrite(nullptr /* cf_handle */).IsCorruption());
    // Confirm that nothing was written to WAL
-    ASSERT_EQ(log_size_pre_write, dbfull()->TEST_total_log_size());
+    ASSERT_EQ(log_size_pre_write, dbfull()->TEST_wals_total_size());
    ASSERT_TRUE(dbfull()->TEST_GetBGError().IsCorruption());
    SyncPoint::GetInstance()->DisableProcessing();

@@ -350,12 +350,12 @@ TEST_P(DbKvChecksumTest, WriteToWALWithColumnFamilyCorrupted) {
    // Corrupted write batch leads to read-only mode, so we have to
    // reopen for every attempt.
    ReopenWithColumnFamilies({kDefaultColumnFamilyName, "pikachu"}, options);
-    auto log_size_pre_write = dbfull()->TEST_total_log_size();
+    auto log_size_pre_write = dbfull()->TEST_wals_total_size();

    SyncPoint::GetInstance()->EnableProcessing();
    ASSERT_TRUE(ExecuteWrite(nullptr /* cf_handle */).IsCorruption());
    // Confirm that nothing was written to WAL
-    ASSERT_EQ(log_size_pre_write, dbfull()->TEST_total_log_size());
+    ASSERT_EQ(log_size_pre_write, dbfull()->TEST_wals_total_size());
    ASSERT_TRUE(dbfull()->TEST_GetBGError().IsCorruption());
    SyncPoint::GetInstance()->DisableProcessing();

@@ -487,7 +487,7 @@ TEST_P(DbKvChecksumTestMergedBatch, WriteToWALCorrupted) {
    // Reopen DB since it failed WAL write which lead to read-only mode
    Reopen(options);
    SyncPoint::GetInstance()->EnableProcessing();
-    auto log_size_pre_write = dbfull()->TEST_total_log_size();
+    auto log_size_pre_write = dbfull()->TEST_wals_total_size();
    leader_batch_and_status =
        GetWriteBatch(GetCFHandleToUse(nullptr, op_type1_),
                      8 /* protection_bytes_per_key */, op_type1_);
@@ -499,7 +499,7 @@ TEST_P(DbKvChecksumTestMergedBatch, WriteToWALCorrupted) {
    SyncPoint::GetInstance()->ClearCallBack("WriteThread::JoinBatchGroup:Wait");
    ASSERT_EQ(1, leader_count);
    // Nothing should have been written to WAL
-    ASSERT_EQ(log_size_pre_write, dbfull()->TEST_total_log_size());
+    ASSERT_EQ(log_size_pre_write, dbfull()->TEST_wals_total_size());
    ASSERT_TRUE(dbfull()->TEST_GetBGError().IsCorruption());

    corrupt_byte_offset++;
@@ -599,7 +599,7 @@ TEST_P(DbKvChecksumTestMergedBatch, WriteToWALWithColumnFamilyCorrupted) {
    // Reopen DB since it failed WAL write which lead to read-only mode
    ReopenWithColumnFamilies({kDefaultColumnFamilyName, "ramen"}, options);
    SyncPoint::GetInstance()->EnableProcessing();
-    auto log_size_pre_write = dbfull()->TEST_total_log_size();
+    auto log_size_pre_write = dbfull()->TEST_wals_total_size();
    leader_batch_and_status =
        GetWriteBatch(GetCFHandleToUse(handles_[1], op_type1_),
                      8 /* protection_bytes_per_key */, op_type1_);
@@ -612,7 +612,7 @@ TEST_P(DbKvChecksumTestMergedBatch, WriteToWALWithColumnFamilyCorrupted) {

    ASSERT_EQ(1, leader_count);
    // Nothing should have been written to WAL
-    ASSERT_EQ(log_size_pre_write, dbfull()->TEST_total_log_size());
+    ASSERT_EQ(log_size_pre_write, dbfull()->TEST_wals_total_size());
    ASSERT_TRUE(dbfull()->TEST_GetBGError().IsCorruption());

    corrupt_byte_offset++;
@@ -424,6 +424,96 @@ TEST_F(DBMemTableTest, IntegrityChecks) {
    ASSERT_FALSE(iter->Valid());
  }
 }
+
+TEST_F(DBMemTableTest, VectorConcurrentInsert) {
+  Options options;
+  options.create_if_missing = true;
+  options.create_missing_column_families = true;
+  options.allow_concurrent_memtable_write = true;
+  options.memtable_factory.reset(new VectorRepFactory());
+  DestroyAndReopen(options);
+  CreateAndReopenWithCF({"cf1"}, options);
+
+  // Multi-threaded writes
+  {
+    WriteOptions write_options;
+    std::vector<port::Thread> threads;
+    for (int i = 0; i < 10; ++i) {
+      threads.emplace_back([&, i]() {
+        int start = i * 100;
+        int end = start + 100;
+        WriteBatch batch;
+        for (int j = start; j < end; ++j) {
+          ASSERT_OK(
+              batch.Put(handles_[0], Key(j), "value" + std::to_string(j)));
+        }
+        ASSERT_OK(db_->Write(write_options, &batch));
+      });
+    }
+    for (auto& t : threads) {
+      t.join();
+    }
+
+    std::unique_ptr<Iterator> iter(
+        db_->NewIterator(ReadOptions(), handles_[0]));
+    iter->SeekToFirst();
+    for (int i = 0; i < 1000; ++i) {
+      ASSERT_TRUE(iter->Valid());
+      ASSERT_EQ(iter->key().ToString(), Key(i));
+      ASSERT_EQ(iter->value().ToString(), "value" + std::to_string(i));
+      iter->Next();
+    }
+    ASSERT_FALSE(iter->Valid());
+    ASSERT_OK(iter->status());
+  }
+
+  // Multi-threaded writes, multi CF
+  {
+    WriteOptions write_options;
+    std::vector<port::Thread> threads;
+    for (int i = 0; i < 10; ++i) {
+      threads.emplace_back([&, i]() {
+        int start = i * 100;
+        int end = start + 100;
+        WriteBatch batch;
+        for (int j = start; j < end; ++j) {
+          ASSERT_OK(batch.Put(handles_[0], Key(j), "CF0" + std::to_string(j)));
+          ASSERT_OK(batch.Put(handles_[1], Key(j), "CF1" + std::to_string(j)));
+        }
+        ASSERT_OK(db_->Write(write_options, &batch));
+      });
+    }
+
+    for (auto& t : threads) {
+      t.join();
+    }
+
+    std::unique_ptr<Iterator> iter0(
+        db_->NewIterator(ReadOptions(), handles_[0]));
+    std::unique_ptr<Iterator> iter1(
+        db_->NewIterator(ReadOptions(), handles_[1]));
+    iter0->SeekToFirst();
+    iter1->SeekToFirst();
+    for (int i = 0; i < 1000; ++i) {
+      ASSERT_TRUE(iter0->Valid());
+      ASSERT_EQ(iter0->key().ToString(), Key(i));
+      ASSERT_EQ(iter0->value().ToString(), "CF0" + std::to_string(i));
+      iter0->Next();
+
+      ASSERT_TRUE(iter1->Valid());
+      ASSERT_EQ(iter1->key().ToString(), Key(i));
+      ASSERT_EQ(iter1->value().ToString(), "CF1" + std::to_string(i));
+      iter1->Next();
+    }
+    ASSERT_FALSE(iter0->Valid());
+    ASSERT_OK(iter0->status());
+    ASSERT_FALSE(iter1->Valid());
+    ASSERT_OK(iter1->status());
+  }
+
+  ASSERT_OK(Flush(0));
+  ASSERT_OK(Flush(1));
+}
 }  // namespace ROCKSDB_NAMESPACE

 int main(int argc, char** argv) {
@@ -322,31 +322,26 @@ TEST_F(DBOptionsTest, SetWithCustomMemTableFactory) {
  }
  Options options;
  options.create_if_missing = true;
-  // Try with fail_if_options_file_error=false/true to update the options
-  for (bool on_error : {false, true}) {
-    options.fail_if_options_file_error = on_error;
-    options.env = env_;
-    options.disable_auto_compactions = false;
+  options.env = env_;
+  options.disable_auto_compactions = false;

-    options.memtable_factory.reset(new DummySkipListFactory());
-    Reopen(options);
+  options.memtable_factory.reset(new DummySkipListFactory());
+  Reopen(options);

-    ColumnFamilyHandle* cfh = dbfull()->DefaultColumnFamily();
-    ASSERT_OK(
-        dbfull()->SetOptions(cfh, {{"disable_auto_compactions", "true"}}));
-    ColumnFamilyDescriptor cfd;
-    ASSERT_OK(cfh->GetDescriptor(&cfd));
-    ASSERT_STREQ(cfd.options.memtable_factory->Name(),
-                 DummySkipListFactory::kClassName());
-    ColumnFamilyHandle* test = nullptr;
-    ASSERT_OK(dbfull()->CreateColumnFamily(options, "test", &test));
-    ASSERT_OK(test->GetDescriptor(&cfd));
-    ASSERT_STREQ(cfd.options.memtable_factory->Name(),
-                 DummySkipListFactory::kClassName());
+  ColumnFamilyHandle* cfh = dbfull()->DefaultColumnFamily();
+  ASSERT_OK(dbfull()->SetOptions(cfh, {{"disable_auto_compactions", "true"}}));
+  ColumnFamilyDescriptor cfd;
+  ASSERT_OK(cfh->GetDescriptor(&cfd));
+  ASSERT_STREQ(cfd.options.memtable_factory->Name(),
+               DummySkipListFactory::kClassName());
+  ColumnFamilyHandle* test = nullptr;
+  ASSERT_OK(dbfull()->CreateColumnFamily(options, "test", &test));
+  ASSERT_OK(test->GetDescriptor(&cfd));
+  ASSERT_STREQ(cfd.options.memtable_factory->Name(),
+               DummySkipListFactory::kClassName());

-    ASSERT_OK(dbfull()->DropColumnFamily(test));
-    delete test;
-  }
+  ASSERT_OK(dbfull()->DropColumnFamily(test));
+  delete test;
 }

 TEST_F(DBOptionsTest, SetBytesPerSync) {
@@ -377,6 +377,8 @@ TEST_F(DBPropertiesTest, AggregatedTableProperties) {
        NewBloomFilterPolicy(kBloomBitsPerKey, false));
    table_options.block_size = 1024;
    options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+    // The checks assume kTableCount number of files
+    options.disable_auto_compactions = true;

    DestroyAndReopen(options);

@@ -567,7 +569,7 @@ TEST_F(DBPropertiesTest, AggregatedTablePropertiesAtLevel) {
  options.target_file_size_base = 8192;
  options.max_bytes_for_level_base = 10000;
  options.max_bytes_for_level_multiplier = 2;
-  // This ensures there no compaction happening when we call GetProperty().
+  // The checks assume kTableCount number of files
  options.disable_auto_compactions = true;
  options.merge_operator.reset(new TestPutOperator());

@@ -508,6 +508,81 @@ TEST_F(DBSecondaryTest, OpenAsSecondary) {
  verify_db_func("new_foo_value", "new_bar_value");
 }

+TEST_F(DBSecondaryTest, OptionsOverrideTest) {
+  Options options;
+  options.env = env_;
+  options.preserve_internal_time_seconds = 300;
+  options.compaction_readahead_size = 200;
+  options.blob_compaction_readahead_size = 100;
+  Reopen(options);
+
+  for (int i = 0; i < 3; ++i) {
+    ASSERT_OK(Put("foo", "foo_value" + std::to_string(i)));
+    ASSERT_OK(Put("bar", "bar_value" + std::to_string(i)));
+    ASSERT_OK(Flush());
+  }
+
+  CompactionServiceInput input;
+
+  ColumnFamilyMetaData meta;
+  db_->GetColumnFamilyMetaData(&meta);
+  for (auto& file : meta.levels[0].files) {
+    ASSERT_EQ(0, meta.levels[0].level);
+    input.input_files.push_back(file.name);
+  }
+  ASSERT_EQ(input.input_files.size(), 3);
+
+  input.output_level = 1;
+  input.options_file_number = dbfull()->GetVersionSet()->options_file_number();
+  input.cf_name = kDefaultColumnFamilyName;
+  ASSERT_OK(db_->GetDbIdentity(input.db_id));
+
+  ASSERT_EQ(db_->GetOptions().compaction_readahead_size, 200);
+  ASSERT_EQ(db_->GetOptions().blob_compaction_readahead_size, 100);
+
+  Close();
+
+  std::string compaction_input_binary;
+  ASSERT_OK(input.Write(&compaction_input_binary));
+  std::string compaction_result_binary;
+
+  CompactionServiceOptionsOverride override_options;
+  override_options.env = env_;
+  override_options.table_factory.reset(
+      NewBlockBasedTableFactory(BlockBasedTableOptions()));
+
+  ASSERT_OK(
+      StringToMap("compaction_readahead_size=8388608;"
+                  "blob_compaction_readahead_size=4194304;"
+                  "some_invalid_option=ignore_me;"
+                  "env=this_should_not_fail;"
+                  "max_open_files=100;",  // this should be always overriden as
+                                          // -1 in remote compaction
+                  &override_options.options_map));
+
+  bool verified = false;
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "DBImplSecondary::OpenAndCompact::AfterOpenAsSecondary:0",
+      [&](void* arg) {
+        auto secondary_db = static_cast<DB*>(arg);
+        auto secondary_db_options = secondary_db->GetOptions();
+        // DBOption
+        ASSERT_EQ(secondary_db_options.compaction_readahead_size, 8388608);
+        ASSERT_EQ(secondary_db_options.max_open_files, -1);
+        // CFOption
+        ASSERT_EQ(secondary_db_options.blob_compaction_readahead_size, 4194304);
+        verified = true;
+      });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  ASSERT_OK(DB::OpenAndCompact(OpenAndCompactOptions(), dbname_,
+                               secondary_path_, compaction_input_binary,
+                               &compaction_result_binary, override_options));
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+  ASSERT_TRUE(verified);
+}
+
 namespace {
 class TraceFileEnv : public EnvWrapper {
 public:
@@ -530,6 +605,9 @@ class TraceFileEnv : public EnvWrapper {
                  char* scratch) const override {
        return target_->Read(offset, n, result, scratch);
      }
+      Status GetFileSize(uint64_t* file_size) override {
+        return target_->GetFileSize(file_size);
+      }

     private:
      std::unique_ptr<RandomAccessFile> target_;
@@ -1216,46 +1294,6 @@ TEST_F(DBSecondaryTest, CatchUpAfterFlush) {
  ASSERT_OK(iter3->status());
 }

-TEST_F(DBSecondaryTest, CheckConsistencyWhenOpen) {
-  bool called = false;
-  Options options;
-  options.env = env_;
-  options.disable_auto_compactions = true;
-  Reopen(options);
-  SyncPoint::GetInstance()->DisableProcessing();
-  SyncPoint::GetInstance()->ClearAllCallBacks();
-  SyncPoint::GetInstance()->SetCallBack(
-      "DBImplSecondary::CheckConsistency:AfterFirstAttempt", [&](void* arg) {
-        ASSERT_NE(nullptr, arg);
-        called = true;
-        auto* s = static_cast<Status*>(arg);
-        ASSERT_NOK(*s);
-      });
-  SyncPoint::GetInstance()->LoadDependency(
-      {{"DBImpl::CheckConsistency:AfterGetLiveFilesMetaData",
-        "BackgroundCallCompaction:0"},
-       {"DBImpl::BackgroundCallCompaction:PurgedObsoleteFiles",
-        "DBImpl::CheckConsistency:BeforeGetFileSize"}});
-  SyncPoint::GetInstance()->EnableProcessing();
-
-  ASSERT_OK(Put("a", "value0"));
-  ASSERT_OK(Put("c", "value0"));
-  ASSERT_OK(Flush());
-  ASSERT_OK(Put("b", "value1"));
-  ASSERT_OK(Put("d", "value1"));
-  ASSERT_OK(Flush());
-  port::Thread thread([this]() {
-    Options opts;
-    opts.env = env_;
-    opts.max_open_files = -1;
-    OpenSecondary(opts);
-  });
-  ASSERT_OK(dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr));
-  ASSERT_OK(dbfull()->TEST_WaitForCompact());
-  thread.join();
-  ASSERT_TRUE(called);
-}
-
 TEST_F(DBSecondaryTest, StartFromInconsistent) {
  Options options = CurrentOptions();
  DestroyAndReopen(options);
@@ -135,21 +135,6 @@ TEST_F(DBSSTTest, SSTsWithLdbSuffixHandling) {
  Destroy(options);
 }

-// Check that we don't crash when opening DB with
-// DBOptions::skip_checking_sst_file_sizes_on_db_open = true.
-TEST_F(DBSSTTest, SkipCheckingSSTFileSizesOnDBOpen) {
-  ASSERT_OK(Put("pika", "choo"));
-  ASSERT_OK(Flush());
-
-  // Just open the DB with the option set to true and check that we don't crash.
-  Options options;
-  options.env = env_;
-  options.skip_checking_sst_file_sizes_on_db_open = true;
-  Reopen(options);
-
-  ASSERT_EQ("choo", Get("pika"));
-}
-
 TEST_F(DBSSTTest, DontDeleteMovedFile) {
  // This test triggers move compaction and verifies that the file is not
  // deleted when it's part of move compaction
@@ -1748,45 +1733,6 @@ TEST_F(DBSSTTest, GetTotalSstFilesSize) {
  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
 }

-TEST_F(DBSSTTest, OpenDBWithoutGetFileSizeInvocations) {
-  Options options = CurrentOptions();
-  std::unique_ptr<MockEnv> env{MockEnv::Create(Env::Default())};
-  options.env = env.get();
-  options.disable_auto_compactions = true;
-  options.compression = kNoCompression;
-  options.enable_blob_files = true;
-  options.blob_file_size = 32;  // create one blob per file
-  options.skip_checking_sst_file_sizes_on_db_open = true;
-
-  DestroyAndReopen(options);
-  // Generate 5 files in L0
-  for (int i = 0; i < 5; i++) {
-    for (int j = 0; j < 10; j++) {
-      std::string val = "val_file_" + std::to_string(i);
-      ASSERT_OK(Put(Key(j), val));
-    }
-    ASSERT_OK(Flush());
-  }
-  Close();
-
-  bool is_get_file_size_called = false;
-  SyncPoint::GetInstance()->SetCallBack(
-      "MockFileSystem::GetFileSize:CheckFileType", [&](void* arg) {
-        std::string* filename = static_cast<std::string*>(arg);
-        if (filename->find(".blob") != std::string::npos) {
-          is_get_file_size_called = true;
-        }
-      });
-
-  SyncPoint::GetInstance()->EnableProcessing();
-  Reopen(options);
-  ASSERT_FALSE(is_get_file_size_called);
-  SyncPoint::GetInstance()->DisableProcessing();
-  SyncPoint::GetInstance()->ClearAllCallBacks();
-
-  Destroy(options);
-}
-
 TEST_F(DBSSTTest, GetTotalSstFilesSizeVersionsFilesShared) {
  Options options = CurrentOptions();
  options.disable_auto_compactions = true;
@@ -1991,6 +1937,70 @@ TEST_F(DBSSTTest, DBWithSFMForBlobFilesAtomicFlush) {
  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
 }

+TEST_F(DBSSTTest, SstGetFileSizeFails) {
+  // Build an SST file
+  ASSERT_OK(Put("x", "zaphod"));
+  ASSERT_OK(Flush());
+  std::vector<LiveFileMetaData> metadata;
+  db_->GetLiveFilesMetaData(&metadata);
+  ASSERT_EQ(1U, metadata.size());
+  std::string filename = dbname_ + metadata[0].name;
+
+  // Prepare for fault injection
+  std::shared_ptr<FaultInjectionTestFS> fault_fs =
+      std::make_shared<FaultInjectionTestFS>(
+          CurrentOptions().env->GetFileSystem());
+  std::unique_ptr<Env> fault_fs_env(NewCompositeEnv(fault_fs));
+  Options options = CurrentOptions();
+  options.env = fault_fs_env.get();
+  options.paranoid_checks = false;  // don't check file sizes on open
+
+  for (int i = 0; i < 4; i++) {
+    SCOPED_TRACE("Iteration = " + std::to_string(i));
+    fault_fs->SetFailRandomAccessGetFileSizeSst(false);
+    fault_fs->SetFailFilesystemGetFileSizeSst(false);
+    Close();
+
+    if (i == 1) {
+      // Just FSRandomAccessFile::GetFileSize fails, which should be worked
+      // around
+      fault_fs->SetFailRandomAccessGetFileSizeSst(true);
+    } else if (i == 2) {
+      // FileSystem::GetFileSize fails, which should be worked around if
+      // FSRandomAccessFile::GetFileSize is supported
+      fault_fs->SetFailFilesystemGetFileSizeSst(true);
+    } else if (i == 3) {
+      // Both GetFileSize APIs fail with an IOError
+      fault_fs->SetFailRandomAccessGetFileSizeSst(true);
+      fault_fs->SetFailFilesystemGetFileSizeSst(true);
+    }
+
+    ASSERT_OK(TryReopen(options));
+    std::string value;
+    Status get_status = db_->Get({}, "x", &value);
+    if (i < 2) {
+      ASSERT_OK(get_status);
+    } else if (i == 2) {
+      if (encrypted_env_) {
+        // Can't recover because RandomAccessFile::GetFileSize is not supported
+        // on EncryptedEnv
+        // Fail with propagated IOError. (Not Corruption nor NotSupported!)
+        ASSERT_EQ(get_status.code(), Status::Code::kIOError);
+        ASSERT_STREQ(get_status.getState(), "FileSystem::GetFileSize failed");
+      } else {
+        // Never sees the FileSystem::GetFileSize failure
+        ASSERT_OK(get_status);
+      }
+    } else {
+      ASSERT_EQ(i, 3);
+      // Fail with propagated IOError. (Not Corruption nor NotSupported!)
+      ASSERT_EQ(get_status.code(), Status::Code::kIOError);
+      ASSERT_STREQ(get_status.getState(), "FileSystem::GetFileSize failed");
+    }
+  }
+  Close();
+}
+
 }  // namespace ROCKSDB_NAMESPACE

 int main(int argc, char** argv) {
@@ -144,6 +144,121 @@ TEST_F(DBTest, MockEnvTest) {
  delete db;
 }

+TEST_F(DBTest, RequestIdPlumbingTest) {
+  // test that request_id is passed to the filesystem, from
+  // ReadOptions to IODebugContext
+  Options options = CurrentOptions();
+  options.env = env_;
+
+  // Create a mock environment to capture IODebugContext during reads
+  IODebugContext dbgCopy;
+  const std::string* captured_request_id_dbg;
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "RandomAccessFileReader::Read:IODebugContext", [&](void* arg) {
+        IODebugContext* dbg = static_cast<IODebugContext*>(arg);
+        if (dbg == nullptr) {
+          captured_request_id_dbg = nullptr;
+        } else {
+          captured_request_id_dbg = dbg->request_id;
+          // Test IODebugContext assignment operator
+          dbgCopy = *dbg;
+        }
+      });
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
+  ASSERT_OK(Put("k1", "v1"));
+  ASSERT_OK(Flush());
+
+  // test request_id plumbing during a get
+  {
+    const std::string test_request_id = "test_request_id_123";
+    ReadOptions read_opts;
+    read_opts.request_id = &test_request_id;
+    std::string value;
+    ASSERT_OK(db_->Get(read_opts, "k1", &value));
+
+    // Verify the request_id was propagated to the file system
+    ASSERT_NE(captured_request_id_dbg, nullptr);
+    ASSERT_EQ(*captured_request_id_dbg, test_request_id);
+
+    ASSERT_NE(dbgCopy.request_id, nullptr);
+    ASSERT_NE(dbgCopy.request_id, captured_request_id_dbg);
+    ASSERT_EQ(*dbgCopy.request_id, test_request_id);
+  }
+
+  captured_request_id_dbg = nullptr;
+
+  // test request_id plumbing during iterator seek
+  ASSERT_OK(Put("k2", "v2"));
+  ASSERT_OK(Flush());
+  {
+    ReadOptions read_opts;
+    const std::string request_id = "test_request_id_456";
+    read_opts.request_id = &request_id;
+
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_opts));
+    iter->Seek("k2");
+    ASSERT_TRUE(iter->Valid());
+
+    // Verify the request_id was propagated to the file system
+    ASSERT_NE(captured_request_id_dbg, nullptr);
+    ASSERT_EQ(*captured_request_id_dbg, request_id);
+
+    ASSERT_NE(dbgCopy.request_id, nullptr);
+    ASSERT_NE(dbgCopy.request_id, captured_request_id_dbg);
+    ASSERT_EQ(*dbgCopy.request_id, request_id);
+
+    // Test IODebugContext copy constructor
+    IODebugContext dbgCopy2(dbgCopy);
+    ASSERT_NE(dbgCopy2.request_id, nullptr);
+    ASSERT_NE(dbgCopy2.request_id, captured_request_id_dbg);
+    ASSERT_NE(dbgCopy2.request_id, dbgCopy.request_id);
+    ASSERT_EQ(*dbgCopy2.request_id, request_id);
+  }
+
+  // test request_id plumbing during multiget
+  captured_request_id_dbg = nullptr;
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "RandomAccessFileReader::MultiRead:IODebugContext", [&](void* arg) {
+        IODebugContext* dbg = static_cast<IODebugContext*>(arg);
+        if (dbg == nullptr) {
+          captured_request_id_dbg = nullptr;
+        } else {
+          captured_request_id_dbg = dbg->request_id;
+        }
+      });
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
+  ASSERT_OK(Put("k3", "v3"));
+  ASSERT_OK(Put("k4", "v4"));
+  ASSERT_OK(Flush());
+
+  {
+    ReadOptions read_opts;
+    const std::string multiget_request_id = "test_request_id_789";
+    read_opts.request_id = &multiget_request_id;
+
+    std::vector<std::string> values;
+    std::vector<Slice> keys = {Slice("k3"), Slice("k4")};
+
+    values.resize(keys.size());
+
+    std::vector<ColumnFamilyHandle*> cfhs(keys.size(),
+                                          db_->DefaultColumnFamily());
+    db_->MultiGet(read_opts, cfhs, keys, &values);
+
+    ASSERT_NE(captured_request_id_dbg, nullptr);
+    ASSERT_EQ(*captured_request_id_dbg, multiget_request_id);
+  }
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
+}
+
 TEST_F(DBTest, MemEnvTest) {
  std::unique_ptr<Env> env{NewMemEnv(Env::Default())};
  Options options;
@@ -1163,12 +1278,6 @@ class DelayFilterFactory : public CompactionFilterFactory {
 };
 }  // anonymous namespace

-static std::string CompressibleString(Random* rnd, int len) {
-  std::string r;
-  test::CompressibleString(rnd, 0.8, len, &r);
-  return r;
-}
-
 TEST_F(DBTest, FailMoreDbPaths) {
  Options options = CurrentOptions();
  options.db_paths.emplace_back(dbname_, 10000000);
@@ -3351,11 +3460,6 @@ class ModelDB : public DB {
  using DB::NumberLevels;
  int NumberLevels(ColumnFamilyHandle* /*column_family*/) override { return 1; }

-  using DB::MaxMemCompactionLevel;
-  int MaxMemCompactionLevel(ColumnFamilyHandle* /*column_family*/) override {
-    return 1;
-  }
-
  using DB::Level0StopWriteTrigger;
  int Level0StopWriteTrigger(ColumnFamilyHandle* /*column_family*/) override {
    return -1;
@@ -3412,7 +3516,7 @@ class ModelDB : public DB {
  }

  Status GetCurrentWalFile(
-      std::unique_ptr<LogFile>* /*current_log_file*/) override {
+      std::unique_ptr<LogFile>* /*current_wal_file*/) override {
    return Status::OK();
  }

@@ -3451,6 +3555,11 @@ class ModelDB : public DB {
    return Status::OK();
  }

+  Status GetNewestUserDefinedTimestamp(
+      ColumnFamilyHandle* /*cf*/, std::string* /*newest_timestamp*/) override {
+    return Status::OK();
+  }
+
  ColumnFamilyHandle* DefaultColumnFamily() const override { return nullptr; }

 private:
@@ -5292,272 +5401,6 @@ TEST_F(DBTest, FlushOnDestroy) {
  CancelAllBackgroundWork(db_);
 }

-TEST_F(DBTest, DynamicLevelCompressionPerLevel) {
-  if (!Snappy_Supported()) {
-    return;
-  }
-  const int kNKeys = 120;
-  int keys[kNKeys];
-  for (int i = 0; i < kNKeys; i++) {
-    keys[i] = i;
-  }
-
-  Random rnd(301);
-  Options options;
-  options.env = env_;
-  options.create_if_missing = true;
-  options.db_write_buffer_size = 20480;
-  options.write_buffer_size = 20480;
-  options.max_write_buffer_number = 2;
-  options.level0_file_num_compaction_trigger = 2;
-  options.level0_slowdown_writes_trigger = 2;
-  options.level0_stop_writes_trigger = 2;
-  options.target_file_size_base = 20480;
-  options.level_compaction_dynamic_level_bytes = true;
-  options.max_bytes_for_level_base = 102400;
-  options.max_bytes_for_level_multiplier = 4;
-  options.max_background_compactions = 1;
-  options.num_levels = 5;
-  options.statistics = CreateDBStatistics();
-
-  options.compression_per_level.resize(3);
-  // No compression for L0
-  options.compression_per_level[0] = kNoCompression;
-  // No compression for the Ln whre L0 is compacted to
-  options.compression_per_level[1] = kNoCompression;
-  // Snappy compression for Ln+1
-  options.compression_per_level[2] = kSnappyCompression;
-
-  OnFileDeletionListener* listener = new OnFileDeletionListener();
-  options.listeners.emplace_back(listener);
-
-  DestroyAndReopen(options);
-
-  // Insert more than 80K. L4 should be base level. Neither L0 nor L4 should
-  // be compressed, so there shouldn't be any compression.
-  for (int i = 0; i < 20; i++) {
-    ASSERT_OK(Put(Key(keys[i]), CompressibleString(&rnd, 4000)));
-    ASSERT_OK(dbfull()->TEST_WaitForBackgroundWork());
-  }
-  ASSERT_OK(Flush());
-  ASSERT_OK(dbfull()->TEST_WaitForCompact());
-
-  ASSERT_EQ(NumTableFilesAtLevel(1), 0);
-  ASSERT_EQ(NumTableFilesAtLevel(2), 0);
-  ASSERT_EQ(NumTableFilesAtLevel(3), 0);
-  ASSERT_TRUE(NumTableFilesAtLevel(0) > 0 || NumTableFilesAtLevel(4) > 0);
-
-  // Verify there was no compression
-  auto num_block_compressed =
-      options.statistics->getTickerCount(NUMBER_BLOCK_COMPRESSED);
-  ASSERT_EQ(num_block_compressed, 0);
-
-  // Insert 400KB and there will be some files end up in L3. According to the
-  // above compression settings for each level, there will be some compression.
-  ASSERT_OK(options.statistics->Reset());
-  ASSERT_EQ(num_block_compressed, 0);
-  for (int i = 20; i < 120; i++) {
-    ASSERT_OK(Put(Key(keys[i]), CompressibleString(&rnd, 4000)));
-    ASSERT_OK(dbfull()->TEST_WaitForBackgroundWork());
-  }
-  ASSERT_OK(Flush());
-  ASSERT_OK(dbfull()->TEST_WaitForCompact());
-  ASSERT_EQ(NumTableFilesAtLevel(1), 0);
-  ASSERT_EQ(NumTableFilesAtLevel(2), 0);
-  ASSERT_GE(NumTableFilesAtLevel(3), 1);
-  ASSERT_GE(NumTableFilesAtLevel(4), 1);
-
-  // Verify there was compression
-  num_block_compressed =
-      options.statistics->getTickerCount(NUMBER_BLOCK_COMPRESSED);
-  ASSERT_GT(num_block_compressed, 0);
-
-  // Make sure data in files in L3 is not compacted by removing all files
-  // in L4 and calculate number of rows
-  ASSERT_OK(dbfull()->SetOptions({
-      {"disable_auto_compactions", "true"},
-  }));
-  ColumnFamilyMetaData cf_meta;
-  db_->GetColumnFamilyMetaData(&cf_meta);
-
-  // Ensure that L1+ files are non-overlapping and together with L0 encompass
-  // full key range between smallestkey and largestkey from CF file metadata.
-  int largestkey_in_prev_level = -1;
-  int keys_found = 0;
-  for (int level = (int)cf_meta.levels.size() - 1; level >= 0; level--) {
-    int files_in_level = (int)cf_meta.levels[level].files.size();
-    int largestkey_in_prev_file = -1;
-    for (int j = 0; j < files_in_level; j++) {
-      int smallestkey = IdFromKey(cf_meta.levels[level].files[j].smallestkey);
-      int largestkey = IdFromKey(cf_meta.levels[level].files[j].largestkey);
-      int num_entries = (int)cf_meta.levels[level].files[j].num_entries;
-      ASSERT_EQ(num_entries, largestkey - smallestkey + 1);
-      keys_found += num_entries;
-      if (level > 0) {
-        if (j == 0) {
-          ASSERT_GT(smallestkey, largestkey_in_prev_level);
-        }
-        if (j > 0) {
-          ASSERT_GT(smallestkey, largestkey_in_prev_file);
-        }
-        if (j == files_in_level - 1) {
-          largestkey_in_prev_level = largestkey;
-        }
-      }
-      largestkey_in_prev_file = largestkey;
-    }
-  }
-  ASSERT_EQ(keys_found, kNKeys);
-
-  for (const auto& file : cf_meta.levels[4].files) {
-    listener->SetExpectedFileName(dbname_ + file.name);
-    const RangeOpt ranges(file.smallestkey, file.largestkey);
-    // Given verification from above, we're guaranteed that by deleting all the
-    // files in [<smallestkey>, <largestkey>] range, we're effectively deleting
-    // that very single file and nothing more.
-    EXPECT_OK(dbfull()->DeleteFilesInRanges(dbfull()->DefaultColumnFamily(),
-                                            &ranges, true /* include_end */));
-  }
-  listener->VerifyMatchedCount(cf_meta.levels[4].files.size());
-
-  int num_keys = 0;
-  std::unique_ptr<Iterator> iter(db_->NewIterator(ReadOptions()));
-  for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
-    num_keys++;
-  }
-  ASSERT_OK(iter->status());
-
-  ASSERT_EQ(NumTableFilesAtLevel(1), 0);
-  ASSERT_EQ(NumTableFilesAtLevel(2), 0);
-  ASSERT_GE(NumTableFilesAtLevel(3), 1);
-  ASSERT_EQ(NumTableFilesAtLevel(4), 0);
-
-  ASSERT_GT(SizeAtLevel(0) + SizeAtLevel(3), num_keys * 4000U + num_keys * 10U);
-}
-
-TEST_F(DBTest, DynamicLevelCompressionPerLevel2) {
-  if (!Snappy_Supported() || !LZ4_Supported() || !Zlib_Supported()) {
-    return;
-  }
-  const int kNKeys = 500;
-  int keys[kNKeys];
-  for (int i = 0; i < kNKeys; i++) {
-    keys[i] = i;
-  }
-  RandomShuffle(std::begin(keys), std::end(keys));
-
-  Random rnd(301);
-  Options options;
-  options.create_if_missing = true;
-  options.db_write_buffer_size = 6000000;
-  options.write_buffer_size = 600000;
-  options.max_write_buffer_number = 2;
-  options.level0_file_num_compaction_trigger = 2;
-  options.level0_slowdown_writes_trigger = 2;
-  options.level0_stop_writes_trigger = 2;
-  options.soft_pending_compaction_bytes_limit = 1024 * 1024;
-  options.target_file_size_base = 20;
-  options.env = env_;
-  options.level_compaction_dynamic_level_bytes = true;
-  options.max_bytes_for_level_base = 200;
-  options.max_bytes_for_level_multiplier = 8;
-  options.max_background_compactions = 1;
-  options.num_levels = 5;
-  options.compaction_verify_record_count = false;
-  std::shared_ptr<mock::MockTableFactory> mtf(new mock::MockTableFactory);
-  options.table_factory = mtf;
-
-  options.compression_per_level.resize(3);
-  options.compression_per_level[0] = kNoCompression;
-  options.compression_per_level[1] = kLZ4Compression;
-  options.compression_per_level[2] = kZlibCompression;
-
-  DestroyAndReopen(options);
-  // When base level is L4, L4 is LZ4.
-  std::atomic<int> num_zlib(0);
-  std::atomic<int> num_lz4(0);
-  std::atomic<int> num_no(0);
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
-      "LevelCompactionPicker::PickCompaction:Return", [&](void* arg) {
-        Compaction* compaction = static_cast<Compaction*>(arg);
-        if (compaction->output_level() == 4) {
-          ASSERT_TRUE(compaction->output_compression() == kLZ4Compression);
-          num_lz4.fetch_add(1);
-        }
-      });
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
-      "FlushJob::WriteLevel0Table:output_compression", [&](void* arg) {
-        auto* compression = static_cast<CompressionType*>(arg);
-        ASSERT_TRUE(*compression == kNoCompression);
-        num_no.fetch_add(1);
-      });
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
-
-  for (int i = 0; i < 100; i++) {
-    std::string value = rnd.RandomString(200);
-    ASSERT_OK(Put(Key(keys[i]), value));
-    if (i % 25 == 24) {
-      ASSERT_OK(Flush());
-      ASSERT_OK(dbfull()->TEST_WaitForCompact());
-    }
-  }
-
-  ASSERT_OK(Flush());
-  ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());
-  ASSERT_OK(dbfull()->TEST_WaitForCompact());
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
-
-  ASSERT_EQ(NumTableFilesAtLevel(1), 0);
-  ASSERT_EQ(NumTableFilesAtLevel(2), 0);
-  ASSERT_EQ(NumTableFilesAtLevel(3), 0);
-  ASSERT_GT(NumTableFilesAtLevel(4), 0);
-  ASSERT_GT(num_no.load(), 2);
-  ASSERT_GT(num_lz4.load(), 0);
-  int prev_num_files_l4 = NumTableFilesAtLevel(4);
-
-  // After base level turn L4->L3, L3 becomes LZ4 and L4 becomes Zlib
-  num_lz4.store(0);
-  num_no.store(0);
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
-      "LevelCompactionPicker::PickCompaction:Return", [&](void* arg) {
-        Compaction* compaction = static_cast<Compaction*>(arg);
-        if (compaction->output_level() == 4 && compaction->start_level() == 3) {
-          ASSERT_TRUE(compaction->output_compression() == kZlibCompression);
-          num_zlib.fetch_add(1);
-        } else {
-          ASSERT_TRUE(compaction->output_compression() == kLZ4Compression);
-          num_lz4.fetch_add(1);
-        }
-      });
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
-      "FlushJob::WriteLevel0Table:output_compression", [&](void* arg) {
-        auto* compression = static_cast<CompressionType*>(arg);
-        ASSERT_TRUE(*compression == kNoCompression);
-        num_no.fetch_add(1);
-      });
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
-
-  for (int i = 101; i < 500; i++) {
-    std::string value = rnd.RandomString(200);
-    ASSERT_OK(Put(Key(keys[i]), value));
-    if (i % 100 == 99) {
-      ASSERT_OK(Flush());
-      ASSERT_OK(dbfull()->TEST_WaitForCompact());
-    }
-  }
-
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
-  ASSERT_EQ(NumTableFilesAtLevel(1), 0);
-  ASSERT_EQ(NumTableFilesAtLevel(2), 0);
-  ASSERT_GT(NumTableFilesAtLevel(3), 0);
-  ASSERT_GT(NumTableFilesAtLevel(4), prev_num_files_l4);
-  ASSERT_GT(num_no.load(), 2);
-  ASSERT_GT(num_lz4.load(), 0);
-  ASSERT_GT(num_zlib.load(), 0);
-}
-
 TEST_F(DBTest, DynamicCompactionOptions) {
  // minimum write buffer size is enforced at 64KB
  const uint64_t k32KB = 1 << 15;
@@ -6095,8 +5938,8 @@ TEST_F(DBTest, L0L1L2AndUpHitCounter) {
 }

 TEST_F(DBTest, EncodeDecompressedBlockSizeTest) {
-  bool& allow_unsupported_fv =
-      BlockBasedTableFactory::AllowUnsupportedFormatVersion();
+  // Allow testing format_version=1
+  bool& allow_unsupported_fv = TEST_AllowUnsupportedFormatVersion();
  SaveAndRestore guard(&allow_unsupported_fv);
  ASSERT_FALSE(allow_unsupported_fv);

@@ -6419,7 +6262,7 @@ TEST_F(DBTest, TestLogCleanup) {

  for (int i = 0; i < 100000; ++i) {
    ASSERT_OK(Put(Key(i), "val"));
-    // only 2 memtables will be alive, so logs_to_free needs to always be below
+    // only 2 memtables will be alive, so wals_to_free needs to always be below
    // 2
    ASSERT_LT(dbfull()->TEST_LogsToFreeSize(), static_cast<size_t>(3));
  }
@@ -7235,27 +7078,6 @@ TEST_F(DBTest, LastWriteBufferDelay) {
 }
 #endif  // !defined(ROCKSDB_DISABLE_STALL_NOTIFICATION)

-TEST_F(DBTest, FailWhenCompressionNotSupportedTest) {
-  CompressionType compressions[] = {kZlibCompression, kBZip2Compression,
-                                    kLZ4Compression, kLZ4HCCompression,
-                                    kXpressCompression};
-  for (auto comp : compressions) {
-    if (!CompressionTypeSupported(comp)) {
-      // not supported, we should fail the Open()
-      Options options = CurrentOptions();
-      options.compression = comp;
-      ASSERT_TRUE(!TryReopen(options).ok());
-      // Try if CreateColumnFamily also fails
-      options.compression = kNoCompression;
-      ASSERT_OK(TryReopen(options));
-      ColumnFamilyOptions cf_options(options);
-      cf_options.compression = comp;
-      ColumnFamilyHandle* handle;
-      ASSERT_TRUE(!db_->CreateColumnFamily(cf_options, "name", &handle).ok());
-    }
-  }
-}
-
 TEST_F(DBTest, CreateColumnFamilyShouldFailOnIncompatibleOptions) {
  Options options = CurrentOptions();
  options.max_open_files = 100;
@@ -10,7 +10,6 @@
 #include <atomic>
 #include <cstdlib>
 #include <functional>
-#include <iostream>
 #include <memory>

 #include "db/db_test_util.h"
@@ -1186,705 +1185,6 @@ TEST_F(DBTest2, WalFilterTestWithColumnFamilies) {
  ASSERT_EQ(index, keys_cf.size());
 }

-TEST_F(DBTest2, PresetCompressionDict) {
-  // Verifies that compression ratio improves when dictionary is enabled, and
-  // improves even further when the dictionary is trained by ZSTD.
-  const size_t kBlockSizeBytes = 4 << 10;
-  const size_t kL0FileBytes = 128 << 10;
-  const size_t kApproxPerBlockOverheadBytes = 50;
-  const int kNumL0Files = 5;
-
-  Options options;
-  // Make sure to use any custom env that the test is configured with.
-  options.env = CurrentOptions().env;
-  options.allow_concurrent_memtable_write = false;
-  options.arena_block_size = kBlockSizeBytes;
-  options.create_if_missing = true;
-  options.disable_auto_compactions = true;
-  options.level0_file_num_compaction_trigger = kNumL0Files;
-  options.memtable_factory.reset(
-      test::NewSpecialSkipListFactory(kL0FileBytes / kBlockSizeBytes));
-  options.num_levels = 2;
-  options.target_file_size_base = kL0FileBytes;
-  options.target_file_size_multiplier = 2;
-  options.write_buffer_size = kL0FileBytes;
-  BlockBasedTableOptions table_options;
-  table_options.block_size = kBlockSizeBytes;
-  std::vector<CompressionType> compression_types;
-  if (Zlib_Supported()) {
-    compression_types.push_back(kZlibCompression);
-  }
-#if LZ4_VERSION_NUMBER >= 10400  // r124+
-  compression_types.push_back(kLZ4Compression);
-  compression_types.push_back(kLZ4HCCompression);
-#endif  // LZ4_VERSION_NUMBER >= 10400
-  if (ZSTD_Supported()) {
-    compression_types.push_back(kZSTD);
-  }
-
-  enum DictionaryTypes : int {
-    kWithoutDict,
-    kWithDict,
-    kWithZSTDfinalizeDict,
-    kWithZSTDTrainedDict,
-    kDictEnd,
-  };
-
-  for (auto compression_type : compression_types) {
-    options.compression = compression_type;
-    size_t bytes_without_dict = 0;
-    size_t bytes_with_dict = 0;
-    size_t bytes_with_zstd_finalize_dict = 0;
-    size_t bytes_with_zstd_trained_dict = 0;
-    for (int i = kWithoutDict; i < kDictEnd; i++) {
-      // First iteration: compress without preset dictionary
-      // Second iteration: compress with preset dictionary
-      // Third iteration (zstd only): compress with zstd-trained dictionary
-      //
-      // To make sure the compression dictionary has the intended effect, we
-      // verify the compressed size is smaller in successive iterations. Also in
-      // the non-first iterations, verify the data we get out is the same data
-      // we put in.
-      switch (i) {
-        case kWithoutDict:
-          options.compression_opts.max_dict_bytes = 0;
-          options.compression_opts.zstd_max_train_bytes = 0;
-          break;
-        case kWithDict:
-          options.compression_opts.max_dict_bytes = kBlockSizeBytes;
-          options.compression_opts.zstd_max_train_bytes = 0;
-          break;
-        case kWithZSTDfinalizeDict:
-          if (compression_type != kZSTD ||
-              !ZSTD_FinalizeDictionarySupported()) {
-            continue;
-          }
-          options.compression_opts.max_dict_bytes = kBlockSizeBytes;
-          options.compression_opts.zstd_max_train_bytes = kL0FileBytes;
-          options.compression_opts.use_zstd_dict_trainer = false;
-          break;
-        case kWithZSTDTrainedDict:
-          if (compression_type != kZSTD || !ZSTD_TrainDictionarySupported()) {
-            continue;
-          }
-          options.compression_opts.max_dict_bytes = kBlockSizeBytes;
-          options.compression_opts.zstd_max_train_bytes = kL0FileBytes;
-          options.compression_opts.use_zstd_dict_trainer = true;
-          break;
-        default:
-          assert(false);
-      }
-
-      options.statistics = ROCKSDB_NAMESPACE::CreateDBStatistics();
-      options.table_factory.reset(NewBlockBasedTableFactory(table_options));
-      CreateAndReopenWithCF({"pikachu"}, options);
-      Random rnd(301);
-      std::string seq_datas[10];
-      for (int j = 0; j < 10; ++j) {
-        seq_datas[j] =
-            rnd.RandomString(kBlockSizeBytes - kApproxPerBlockOverheadBytes);
-      }
-
-      ASSERT_EQ(0, NumTableFilesAtLevel(0, 1));
-      for (int j = 0; j < kNumL0Files; ++j) {
-        for (size_t k = 0; k < kL0FileBytes / kBlockSizeBytes + 1; ++k) {
-          auto key_num = j * (kL0FileBytes / kBlockSizeBytes) + k;
-          ASSERT_OK(Put(1, Key(static_cast<int>(key_num)),
-                        seq_datas[(key_num / 10) % 10]));
-        }
-        ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable(handles_[1]));
-        ASSERT_EQ(j + 1, NumTableFilesAtLevel(0, 1));
-      }
-      ASSERT_OK(dbfull()->TEST_CompactRange(0, nullptr, nullptr, handles_[1],
-                                            true /* disallow_trivial_move */));
-      ASSERT_EQ(0, NumTableFilesAtLevel(0, 1));
-      ASSERT_GT(NumTableFilesAtLevel(1, 1), 0);
-
-      // Get the live sst files size
-      size_t total_sst_bytes = TotalSize(1);
-      if (i == kWithoutDict) {
-        bytes_without_dict = total_sst_bytes;
-      } else if (i == kWithDict) {
-        bytes_with_dict = total_sst_bytes;
-      } else if (i == kWithZSTDfinalizeDict) {
-        bytes_with_zstd_finalize_dict = total_sst_bytes;
-      } else if (i == kWithZSTDTrainedDict) {
-        bytes_with_zstd_trained_dict = total_sst_bytes;
-      }
-
-      for (size_t j = 0; j < kNumL0Files * (kL0FileBytes / kBlockSizeBytes);
-           j++) {
-        ASSERT_EQ(seq_datas[(j / 10) % 10], Get(1, Key(static_cast<int>(j))));
-      }
-      if (i == kWithDict) {
-        ASSERT_GT(bytes_without_dict, bytes_with_dict);
-      } else if (i == kWithZSTDTrainedDict) {
-        // In zstd compression, it is sometimes possible that using a finalized
-        // dictionary does not get as good a compression ratio as raw content
-        // dictionary. But using a dictionary should always get better
-        // compression ratio than not using one.
-        ASSERT_TRUE(bytes_with_dict > bytes_with_zstd_finalize_dict ||
-                    bytes_without_dict > bytes_with_zstd_finalize_dict);
-      } else if (i == kWithZSTDTrainedDict) {
-        // In zstd compression, it is sometimes possible that using a trained
-        // dictionary does not get as good a compression ratio as without
-        // training.
-        // But using a dictionary (with or without training) should always get
-        // better compression ratio than not using one.
-        ASSERT_TRUE(bytes_with_dict > bytes_with_zstd_trained_dict ||
-                    bytes_without_dict > bytes_with_zstd_trained_dict);
-      }
-
-      DestroyAndReopen(options);
-    }
-  }
-}
-
-TEST_F(DBTest2, PresetCompressionDictLocality) {
-  if (!ZSTD_Supported()) {
-    return;
-  }
-  // Verifies that compression dictionary is generated from local data. The
-  // verification simply checks all output SSTs have different compression
-  // dictionaries. We do not verify effectiveness as that'd likely be flaky in
-  // the future.
-  const int kNumEntriesPerFile = 1 << 10;  // 1KB
-  const int kNumBytesPerEntry = 1 << 10;   // 1KB
-  const int kNumFiles = 4;
-  Options options = CurrentOptions();
-  options.compression = kZSTD;
-  options.compression_opts.max_dict_bytes = 1 << 14;        // 16KB
-  options.compression_opts.zstd_max_train_bytes = 1 << 18;  // 256KB
-  options.statistics = ROCKSDB_NAMESPACE::CreateDBStatistics();
-  options.target_file_size_base = kNumEntriesPerFile * kNumBytesPerEntry;
-  BlockBasedTableOptions table_options;
-  table_options.cache_index_and_filter_blocks = true;
-  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
-  Reopen(options);
-
-  Random rnd(301);
-  for (int i = 0; i < kNumFiles; ++i) {
-    for (int j = 0; j < kNumEntriesPerFile; ++j) {
-      ASSERT_OK(Put(Key(i * kNumEntriesPerFile + j),
-                    rnd.RandomString(kNumBytesPerEntry)));
-    }
-    ASSERT_OK(Flush());
-    MoveFilesToLevel(1);
-    ASSERT_EQ(NumTableFilesAtLevel(1), i + 1);
-  }
-
-  // Store all the dictionaries generated during a full compaction.
-  std::vector<std::string> compression_dicts;
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
-      "BlockBasedTableBuilder::WriteCompressionDictBlock:RawDict",
-      [&](void* arg) {
-        compression_dicts.emplace_back(static_cast<Slice*>(arg)->ToString());
-      });
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
-  CompactRangeOptions compact_range_opts;
-  compact_range_opts.bottommost_level_compaction =
-      BottommostLevelCompaction::kForceOptimized;
-  ASSERT_OK(db_->CompactRange(compact_range_opts, nullptr, nullptr));
-
-  // Dictionary compression should not be so good as to compress four totally
-  // random files into one. If it does then there's probably something wrong
-  // with the test.
-  ASSERT_GT(NumTableFilesAtLevel(1), 1);
-
-  // Furthermore, there should be one compression dictionary generated per file.
-  // And they should all be different from each other.
-  ASSERT_EQ(NumTableFilesAtLevel(1),
-            static_cast<int>(compression_dicts.size()));
-  for (size_t i = 1; i < compression_dicts.size(); ++i) {
-    std::string& a = compression_dicts[i - 1];
-    std::string& b = compression_dicts[i];
-    size_t alen = a.size();
-    size_t blen = b.size();
-    ASSERT_TRUE(alen != blen || memcmp(a.data(), b.data(), alen) != 0);
-  }
-}
-
-class PresetCompressionDictTest
-    : public DBTestBase,
-      public testing::WithParamInterface<std::tuple<CompressionType, bool>> {
- public:
-  PresetCompressionDictTest()
-      : DBTestBase("db_test2", false /* env_do_fsync */),
-        compression_type_(std::get<0>(GetParam())),
-        bottommost_(std::get<1>(GetParam())) {}
-
- protected:
-  const CompressionType compression_type_;
-  const bool bottommost_;
-};
-
-INSTANTIATE_TEST_CASE_P(
-    DBTest2, PresetCompressionDictTest,
-    ::testing::Combine(::testing::ValuesIn(GetSupportedDictCompressions()),
-                       ::testing::Bool()));
-
-TEST_P(PresetCompressionDictTest, Flush) {
-  // Verifies that dictionary is generated and written during flush only when
-  // `ColumnFamilyOptions::compression` enables dictionary. Also verifies the
-  // size of the dictionary is within expectations according to the limit on
-  // buffering set by `CompressionOptions::max_dict_buffer_bytes`.
-  const size_t kValueLen = 256;
-  const size_t kKeysPerFile = 1 << 10;
-  const size_t kDictLen = 16 << 10;
-  const size_t kBlockLen = 4 << 10;
-
-  Options options = CurrentOptions();
-  if (bottommost_) {
-    options.bottommost_compression = compression_type_;
-    options.bottommost_compression_opts.enabled = true;
-    options.bottommost_compression_opts.max_dict_bytes = kDictLen;
-    options.bottommost_compression_opts.max_dict_buffer_bytes = kBlockLen;
-  } else {
-    options.compression = compression_type_;
-    options.compression_opts.max_dict_bytes = kDictLen;
-    options.compression_opts.max_dict_buffer_bytes = kBlockLen;
-  }
-  options.memtable_factory.reset(test::NewSpecialSkipListFactory(kKeysPerFile));
-  options.statistics = CreateDBStatistics();
-  BlockBasedTableOptions bbto;
-  bbto.block_size = kBlockLen;
-  bbto.cache_index_and_filter_blocks = true;
-  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
-  Reopen(options);
-
-  Random rnd(301);
-  for (size_t i = 0; i <= kKeysPerFile; ++i) {
-    ASSERT_OK(Put(Key(static_cast<int>(i)), rnd.RandomString(kValueLen)));
-  }
-  ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());
-
-  // We can use `BLOCK_CACHE_COMPRESSION_DICT_BYTES_INSERT` to detect whether a
-  // compression dictionary exists since dictionaries would be preloaded when
-  // the flush finishes.
-  if (bottommost_) {
-    // Flush is never considered bottommost. This should change in the future
-    // since flushed files may have nothing underneath them, like the one in
-    // this test case.
-    ASSERT_EQ(
-        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_BYTES_INSERT),
-        0);
-  } else {
-    ASSERT_GT(
-        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_BYTES_INSERT),
-        0);
-    // TODO(ajkr): fix the below assertion to work with ZSTD. The expectation on
-    // number of bytes needs to be adjusted in case the cached block is in
-    // ZSTD's digested dictionary format.
-    if (compression_type_ != kZSTD) {
-      // Although we limited buffering to `kBlockLen`, there may be up to two
-      // blocks of data included in the dictionary since we only check limit
-      // after each block is built.
-      ASSERT_LE(TestGetTickerCount(options,
-                                   BLOCK_CACHE_COMPRESSION_DICT_BYTES_INSERT),
-                2 * kBlockLen);
-    }
-  }
-}
-
-TEST_P(PresetCompressionDictTest, CompactNonBottommost) {
-  // Verifies that dictionary is generated and written during compaction to
-  // non-bottommost level only when `ColumnFamilyOptions::compression` enables
-  // dictionary. Also verifies the size of the dictionary is within expectations
-  // according to the limit on buffering set by
-  // `CompressionOptions::max_dict_buffer_bytes`.
-  const size_t kValueLen = 256;
-  const size_t kKeysPerFile = 1 << 10;
-  const size_t kDictLen = 16 << 10;
-  const size_t kBlockLen = 4 << 10;
-
-  Options options = CurrentOptions();
-  if (bottommost_) {
-    options.bottommost_compression = compression_type_;
-    options.bottommost_compression_opts.enabled = true;
-    options.bottommost_compression_opts.max_dict_bytes = kDictLen;
-    options.bottommost_compression_opts.max_dict_buffer_bytes = kBlockLen;
-  } else {
-    options.compression = compression_type_;
-    options.compression_opts.max_dict_bytes = kDictLen;
-    options.compression_opts.max_dict_buffer_bytes = kBlockLen;
-  }
-  options.disable_auto_compactions = true;
-  options.statistics = CreateDBStatistics();
-  BlockBasedTableOptions bbto;
-  bbto.block_size = kBlockLen;
-  bbto.cache_index_and_filter_blocks = true;
-  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
-  Reopen(options);
-
-  Random rnd(301);
-  for (size_t j = 0; j <= kKeysPerFile; ++j) {
-    ASSERT_OK(Put(Key(static_cast<int>(j)), rnd.RandomString(kValueLen)));
-  }
-  ASSERT_OK(Flush());
-  MoveFilesToLevel(2);
-
-  for (int i = 0; i < 2; ++i) {
-    for (size_t j = 0; j <= kKeysPerFile; ++j) {
-      ASSERT_OK(Put(Key(static_cast<int>(j)), rnd.RandomString(kValueLen)));
-    }
-    ASSERT_OK(Flush());
-  }
-  ASSERT_EQ("2,0,1", FilesPerLevel(0));
-
-  uint64_t prev_compression_dict_bytes_inserted =
-      TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_BYTES_INSERT);
-  // This L0->L1 compaction merges the two L0 files into L1. The produced L1
-  // file is not bottommost due to the existing L2 file covering the same key-
-  // range.
-  ASSERT_OK(dbfull()->TEST_CompactRange(0, nullptr, nullptr));
-  ASSERT_EQ("0,1,1", FilesPerLevel(0));
-  // We can use `BLOCK_CACHE_COMPRESSION_DICT_BYTES_INSERT` to detect whether a
-  // compression dictionary exists since dictionaries would be preloaded when
-  // the compaction finishes.
-  if (bottommost_) {
-    ASSERT_EQ(
-        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_BYTES_INSERT),
-        prev_compression_dict_bytes_inserted);
-  } else {
-    ASSERT_GT(
-        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_BYTES_INSERT),
-        prev_compression_dict_bytes_inserted);
-    // TODO(ajkr): fix the below assertion to work with ZSTD. The expectation on
-    // number of bytes needs to be adjusted in case the cached block is in
-    // ZSTD's digested dictionary format.
-    if (compression_type_ != kZSTD) {
-      // Although we limited buffering to `kBlockLen`, there may be up to two
-      // blocks of data included in the dictionary since we only check limit
-      // after each block is built.
-      ASSERT_LE(TestGetTickerCount(options,
-                                   BLOCK_CACHE_COMPRESSION_DICT_BYTES_INSERT),
-                prev_compression_dict_bytes_inserted + 2 * kBlockLen);
-    }
-  }
-}
-
-TEST_P(PresetCompressionDictTest, CompactBottommost) {
-  // Verifies that dictionary is generated and written during compaction to
-  // non-bottommost level only when either `ColumnFamilyOptions::compression` or
-  // `ColumnFamilyOptions::bottommost_compression` enables dictionary. Also
-  // verifies the size of the dictionary is within expectations according to the
-  // limit on buffering set by `CompressionOptions::max_dict_buffer_bytes`.
-  const size_t kValueLen = 256;
-  const size_t kKeysPerFile = 1 << 10;
-  const size_t kDictLen = 16 << 10;
-  const size_t kBlockLen = 4 << 10;
-
-  Options options = CurrentOptions();
-  if (bottommost_) {
-    options.bottommost_compression = compression_type_;
-    options.bottommost_compression_opts.enabled = true;
-    options.bottommost_compression_opts.max_dict_bytes = kDictLen;
-    options.bottommost_compression_opts.max_dict_buffer_bytes = kBlockLen;
-  } else {
-    options.compression = compression_type_;
-    options.compression_opts.max_dict_bytes = kDictLen;
-    options.compression_opts.max_dict_buffer_bytes = kBlockLen;
-  }
-  options.disable_auto_compactions = true;
-  options.statistics = CreateDBStatistics();
-  BlockBasedTableOptions bbto;
-  bbto.block_size = kBlockLen;
-  bbto.cache_index_and_filter_blocks = true;
-  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
-  Reopen(options);
-
-  Random rnd(301);
-  for (int i = 0; i < 2; ++i) {
-    for (size_t j = 0; j <= kKeysPerFile; ++j) {
-      ASSERT_OK(Put(Key(static_cast<int>(j)), rnd.RandomString(kValueLen)));
-    }
-    ASSERT_OK(Flush());
-  }
-  ASSERT_EQ("2", FilesPerLevel(0));
-
-  uint64_t prev_compression_dict_bytes_inserted =
-      TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_BYTES_INSERT);
-  CompactRangeOptions cro;
-  ASSERT_OK(db_->CompactRange(cro, nullptr, nullptr));
-  ASSERT_EQ("0,1", FilesPerLevel(0));
-  ASSERT_GT(
-      TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_BYTES_INSERT),
-      prev_compression_dict_bytes_inserted);
-  // TODO(ajkr): fix the below assertion to work with ZSTD. The expectation on
-  // number of bytes needs to be adjusted in case the cached block is in ZSTD's
-  // digested dictionary format.
-  if (compression_type_ != kZSTD) {
-    // Although we limited buffering to `kBlockLen`, there may be up to two
-    // blocks of data included in the dictionary since we only check limit after
-    // each block is built.
-    ASSERT_LE(
-        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_BYTES_INSERT),
-        prev_compression_dict_bytes_inserted + 2 * kBlockLen);
-  }
-}
-
-class CompactionCompressionListener : public EventListener {
- public:
-  explicit CompactionCompressionListener(Options* db_options)
-      : db_options_(db_options) {}
-
-  void OnCompactionCompleted(DB* db, const CompactionJobInfo& ci) override {
-    // Figure out last level with files
-    int bottommost_level = 0;
-    for (int level = 0; level < db->NumberLevels(); level++) {
-      std::string files_at_level;
-      ASSERT_TRUE(
-          db->GetProperty("rocksdb.num-files-at-level" + std::to_string(level),
-                          &files_at_level));
-      if (files_at_level != "0") {
-        bottommost_level = level;
-      }
-    }
-
-    if (db_options_->bottommost_compression != kDisableCompressionOption &&
-        ci.output_level == bottommost_level) {
-      ASSERT_EQ(ci.compression, db_options_->bottommost_compression);
-    } else if (db_options_->compression_per_level.size() != 0) {
-      ASSERT_EQ(ci.compression,
-                db_options_->compression_per_level[ci.output_level]);
-    } else {
-      ASSERT_EQ(ci.compression, db_options_->compression);
-    }
-    max_level_checked = std::max(max_level_checked, ci.output_level);
-  }
-
-  int max_level_checked = 0;
-  const Options* db_options_;
-};
-
-enum CompressionFailureType {
-  kTestCompressionFail,
-  kTestDecompressionFail,
-  kTestDecompressionCorruption
-};
-
-class CompressionFailuresTest
-    : public DBTest2,
-      public testing::WithParamInterface<std::tuple<
-          CompressionFailureType, CompressionType, uint32_t, uint32_t>> {
- public:
-  CompressionFailuresTest() {
-    std::tie(compression_failure_type_, compression_type_,
-             compression_max_dict_bytes_, compression_parallel_threads_) =
-        GetParam();
-  }
-
-  CompressionFailureType compression_failure_type_ = kTestCompressionFail;
-  CompressionType compression_type_ = kNoCompression;
-  uint32_t compression_max_dict_bytes_ = 0;
-  uint32_t compression_parallel_threads_ = 0;
-};
-
-INSTANTIATE_TEST_CASE_P(
-    DBTest2, CompressionFailuresTest,
-    ::testing::Combine(::testing::Values(kTestCompressionFail,
-                                         kTestDecompressionFail,
-                                         kTestDecompressionCorruption),
-                       ::testing::ValuesIn(GetSupportedCompressions()),
-                       ::testing::Values(0, 10), ::testing::Values(1, 4)));
-
-TEST_P(CompressionFailuresTest, CompressionFailures) {
-  if (compression_type_ == kNoCompression) {
-    return;
-  }
-
-  Options options = CurrentOptions();
-  options.level0_file_num_compaction_trigger = 2;
-  options.max_bytes_for_level_base = 1024;
-  options.max_bytes_for_level_multiplier = 2;
-  options.num_levels = 7;
-  options.max_background_compactions = 1;
-  options.target_file_size_base = 512;
-
-  BlockBasedTableOptions table_options;
-  table_options.block_size = 512;
-  table_options.verify_compression = true;
-  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
-
-  options.compression = compression_type_;
-  options.compression_opts.parallel_threads = compression_parallel_threads_;
-  options.compression_opts.max_dict_bytes = compression_max_dict_bytes_;
-  options.bottommost_compression_opts.parallel_threads =
-      compression_parallel_threads_;
-  options.bottommost_compression_opts.max_dict_bytes =
-      compression_max_dict_bytes_;
-
-  if (compression_failure_type_ == kTestCompressionFail) {
-    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
-        "CompressData:TamperWithReturnValue", [](void* arg) {
-          bool* ret = static_cast<bool*>(arg);
-          *ret = false;
-        });
-  } else if (compression_failure_type_ == kTestDecompressionFail) {
-    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
-        "UncompressBlockData:TamperWithReturnValue", [](void* arg) {
-          Status* ret = static_cast<Status*>(arg);
-          ASSERT_OK(*ret);
-          *ret = Status::Corruption("kTestDecompressionFail");
-        });
-  } else if (compression_failure_type_ == kTestDecompressionCorruption) {
-    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
-        "UncompressBlockData:"
-        "TamperWithDecompressionOutput",
-        [](void* arg) {
-          BlockContents* contents = static_cast<BlockContents*>(arg);
-          // Ensure uncompressed data != original data
-          const size_t len = contents->data.size() + 1;
-          std::unique_ptr<char[]> fake_data(new char[len]());
-          *contents = BlockContents(std::move(fake_data), len);
-        });
-  }
-
-  std::map<std::string, std::string> key_value_written;
-
-  const int kKeySize = 5;
-  const int kValUnitSize = 16;
-  const int kValSize = 256;
-  Random rnd(405);
-
-  Status s = Status::OK();
-
-  DestroyAndReopen(options);
-  // Write 10 random files
-  for (int i = 0; i < 10; i++) {
-    for (int j = 0; j < 5; j++) {
-      std::string key = rnd.RandomString(kKeySize);
-      // Ensure good compression ratio
-      std::string valueUnit = rnd.RandomString(kValUnitSize);
-      std::string value;
-      for (int k = 0; k < kValSize; k += kValUnitSize) {
-        value += valueUnit;
-      }
-      s = Put(key, value);
-      if (compression_failure_type_ == kTestCompressionFail) {
-        key_value_written[key] = value;
-        ASSERT_OK(s);
-      }
-    }
-    s = Flush();
-    if (compression_failure_type_ == kTestCompressionFail) {
-      ASSERT_OK(s);
-    }
-    s = dbfull()->TEST_WaitForCompact();
-    if (compression_failure_type_ == kTestCompressionFail) {
-      ASSERT_OK(s);
-    }
-    if (i == 4) {
-      // Make compression fail at the mid of table building
-      ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
-    }
-  }
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
-
-  if (compression_failure_type_ == kTestCompressionFail) {
-    // Should be kNoCompression, check content consistency
-    std::unique_ptr<Iterator> db_iter(db_->NewIterator(ReadOptions()));
-    for (db_iter->SeekToFirst(); db_iter->Valid(); db_iter->Next()) {
-      std::string key = db_iter->key().ToString();
-      std::string value = db_iter->value().ToString();
-      ASSERT_NE(key_value_written.find(key), key_value_written.end());
-      ASSERT_EQ(key_value_written[key], value);
-      key_value_written.erase(key);
-    }
-    ASSERT_OK(db_iter->status());
-    ASSERT_EQ(0, key_value_written.size());
-  } else if (compression_failure_type_ == kTestDecompressionFail) {
-    ASSERT_EQ(std::string(s.getState()),
-              "Could not decompress: kTestDecompressionFail");
-  } else if (compression_failure_type_ == kTestDecompressionCorruption) {
-    ASSERT_EQ(std::string(s.getState()),
-              "Decompressed block did not match pre-compression block");
-  }
-}
-
-TEST_F(DBTest2, CompressionOptions) {
-  if (!Zlib_Supported() || !Snappy_Supported()) {
-    return;
-  }
-
-  Options options = CurrentOptions();
-  options.level0_file_num_compaction_trigger = 2;
-  options.max_bytes_for_level_base = 100;
-  options.max_bytes_for_level_multiplier = 2;
-  options.num_levels = 7;
-  options.max_background_compactions = 1;
-
-  CompactionCompressionListener* listener =
-      new CompactionCompressionListener(&options);
-  options.listeners.emplace_back(listener);
-
-  const int kKeySize = 5;
-  const int kValSize = 20;
-  Random rnd(301);
-
-  std::vector<uint32_t> compression_parallel_threads = {1, 4};
-
-  std::map<std::string, std::string> key_value_written;
-
-  for (int iter = 0; iter <= 2; iter++) {
-    listener->max_level_checked = 0;
-
-    if (iter == 0) {
-      // Use different compression algorithms for different levels but
-      // always use Zlib for bottommost level
-      options.compression_per_level = {kNoCompression,     kNoCompression,
-                                       kNoCompression,     kSnappyCompression,
-                                       kSnappyCompression, kSnappyCompression,
-                                       kZlibCompression};
-      options.compression = kNoCompression;
-      options.bottommost_compression = kZlibCompression;
-    } else if (iter == 1) {
-      // Use Snappy except for bottommost level use ZLib
-      options.compression_per_level = {};
-      options.compression = kSnappyCompression;
-      options.bottommost_compression = kZlibCompression;
-    } else if (iter == 2) {
-      // Use Snappy everywhere
-      options.compression_per_level = {};
-      options.compression = kSnappyCompression;
-      options.bottommost_compression = kDisableCompressionOption;
-    }
-
-    for (auto num_threads : compression_parallel_threads) {
-      options.compression_opts.parallel_threads = num_threads;
-      options.bottommost_compression_opts.parallel_threads = num_threads;
-
-      DestroyAndReopen(options);
-      // Write 10 random files
-      for (int i = 0; i < 10; i++) {
-        for (int j = 0; j < 5; j++) {
-          std::string key = rnd.RandomString(kKeySize);
-          std::string value = rnd.RandomString(kValSize);
-          key_value_written[key] = value;
-          ASSERT_OK(Put(key, value));
-        }
-        ASSERT_OK(Flush());
-        ASSERT_OK(dbfull()->TEST_WaitForCompact());
-      }
-
-      // Make sure that we wrote enough to check all 7 levels
-      ASSERT_EQ(listener->max_level_checked, 6);
-
-      // Make sure database content is the same as key_value_written
-      std::unique_ptr<Iterator> db_iter(db_->NewIterator(ReadOptions()));
-      for (db_iter->SeekToFirst(); db_iter->Valid(); db_iter->Next()) {
-        std::string key = db_iter->key().ToString();
-        std::string value = db_iter->value().ToString();
-        ASSERT_NE(key_value_written.find(key), key_value_written.end());
-        ASSERT_EQ(key_value_written[key], value);
-        key_value_written.erase(key);
-      }
-      ASSERT_OK(db_iter->status());
-      ASSERT_EQ(0, key_value_written.size());
-    }
-  }
-}
-
 class CompactionStallTestListener : public EventListener {
 public:
  CompactionStallTestListener()
@@ -3010,7 +2310,7 @@ TEST_F(DBTest2, PausingManualCompaction1) {
      "TestCompactFiles:PausingManualCompaction:3", [&](void* arg) {
        auto paused = static_cast<std::atomic<int>*>(arg);
        // CompactFiles() relies on manual_compactions_paused to
-        // determine if thie compaction should be paused or not
+        // determine if this compaction should be paused or not
        ASSERT_EQ(0, paused->load(std::memory_order_acquire));
        paused->fetch_add(1, std::memory_order_release);
      });
@@ -3122,6 +2422,7 @@ TEST_F(DBTest2, PausingManualCompaction3) {
  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();

  dbfull()->DisableManualCompaction();
+
  ASSERT_TRUE(dbfull()
                  ->CompactRange(compact_options, nullptr, nullptr)
                  .IsManualCompactionPaused());
@@ -5421,6 +4722,103 @@ TEST_F(DBTest2, TestCompactFiles) {
  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
 }

+TEST_F(DBTest2, TestCancelCompactFiles) {
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  Options options;
+  options.env = env_;
+  options.num_levels = 2;
+  options.disable_auto_compactions = true;
+  Reopen(options);
+
+  auto* handle = db_->DefaultColumnFamily();
+  ASSERT_EQ(db_->NumberLevels(handle), 2);
+
+  ROCKSDB_NAMESPACE::SstFileWriter sst_file_writer{
+      ROCKSDB_NAMESPACE::EnvOptions(), options};
+
+  // ingest large SST files
+  std::vector<std::string> external_sst_file_names;
+  int key_counter = 0;
+  const int num_keys_per_file = 100000;
+  const int num_files = 10;
+  for (int i = 0; i < num_files; ++i) {
+    std::string file_name =
+        dbname_ + "/test_compact_files" + std::to_string(i) + ".sst_t";
+    external_sst_file_names.push_back(file_name);
+    ASSERT_OK(sst_file_writer.Open(file_name));
+    for (int j = 0; j < num_keys_per_file; ++j) {
+      ASSERT_OK(sst_file_writer.Put(Key(j + num_keys_per_file * key_counter),
+                                    std::to_string(j)));
+    }
+    key_counter += 1;
+    ASSERT_OK(sst_file_writer.Finish());
+  }
+
+  ASSERT_OK(db_->IngestExternalFile(handle, external_sst_file_names,
+                                    IngestExternalFileOptions()));
+  ASSERT_EQ(NumTableFilesAtLevel(1, 0), num_files);
+  std::vector<std::string> files;
+  GetSstFiles(env_, dbname_, &files);
+  ASSERT_EQ(files.size(), num_files);
+
+  // Test that 0 compactions happen - canceled is set to True initially
+  CompactionOptions compaction_options;
+  std::atomic<bool> canceled(true);
+  compaction_options.canceled = &canceled;
+
+  ASSERT_TRUE(db_->CompactFiles(compaction_options, handle, files, 1)
+                  .IsManualCompactionPaused());
+  ASSERT_EQ(NumTableFilesAtLevel(1, 0), num_files);
+
+  // Test cancellation before the check to cancel compaction happens -
+  // compaction should not occur
+  bool disable_compaction = false;
+  compaction_options.canceled->store(false, std::memory_order_release);
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "TestCancelCompactFiles:SuccessfulCompaction", [&](void* arg) {
+        auto paused = static_cast<std::atomic<int>*>(arg);
+        if (disable_compaction) {
+          db_->DisableManualCompaction();
+          ASSERT_EQ(1, paused->load(std::memory_order_acquire));
+        } else {
+          compaction_options.canceled->store(true, std::memory_order_release);
+          ASSERT_EQ(0, paused->load(std::memory_order_acquire));
+        }
+      });
+
+  ASSERT_TRUE(db_->CompactFiles(compaction_options, handle, files, 1)
+                  .IsManualCompactionPaused());
+  ASSERT_EQ(NumTableFilesAtLevel(1, 0), num_files);
+
+  // DisableManualCompaction() should successfully cancel compaction
+  disable_compaction = true;
+  compaction_options.canceled->store(false, std::memory_order_release);
+  ASSERT_TRUE(db_->CompactFiles(compaction_options, handle, files, 1)
+                  .IsManualCompactionPaused());
+  ASSERT_EQ(NumTableFilesAtLevel(1, 0), num_files);
+  // unlike CompactRange, value of compaction_options.canceled will be
+  // unaffected by calling DisableManualCompactions()
+  ASSERT_FALSE(compaction_options.canceled->load());
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
+  db_->EnableManualCompaction();
+
+  // Test cancelation after the check to cancel compaction - compaction should
+  // occur, leaving only 1 file
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "CompactFilesImpl:0", [&](void* /*arg*/) {
+        compaction_options.canceled->store(true, std::memory_order_release);
+      });
+
+  compaction_options.canceled->store(false, std::memory_order_release);
+  ASSERT_OK(db_->CompactFiles(compaction_options, handle, files, 1));
+  ASSERT_EQ(NumTableFilesAtLevel(1, 0), 1);
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
+}
+
 TEST_F(DBTest2, MultiDBParallelOpenTest) {
  const int kNumDbs = 2;
  Options options = CurrentOptions();
@@ -8068,11 +7466,27 @@ TEST_F(DBTest2, GetFileChecksumsFromCurrentManifest_CRC32) {
  FlushOptions fopts;
  fopts.wait = true;
  Random rnd(test::RandomSeed());
+
+  // Write 4 files into the default column family.
  for (int i = 0; i < 4; i++) {
    ASSERT_OK(db->Put(wopts, Key(i), rnd.RandomString(100)));
    ASSERT_OK(db->Flush(fopts));
  }

+  // Create a new column family, write 1 file into it and drop it.
+  ColumnFamilyHandle* cf;
+  ASSERT_OK(
+      db->CreateColumnFamily(ColumnFamilyOptions(), "soon_to_be_deleted", &cf));
+  ASSERT_OK(db->Put(wopts, cf, "some_key", "some_value"));
+  ASSERT_OK(db->Flush(fopts, cf));
+
+  // Drop column family should generate corresponding version edit
+  // in manifest, which we expect to be correctly interpreted by
+  // GetFileChecksumsFromCurrentManifest API after db close.
+  ASSERT_OK(db->DropColumnFamily(cf));
+  delete cf;
+  cf = nullptr;
+
  // Obtain rich files metadata for source of truth.
  std::vector<LiveFileMetaData> live_files;
  db->GetLiveFilesMetaData(&live_files);
@@ -366,11 +366,6 @@ Options DBTestBase::GetOptions(
    table_options.block_cache = NewLRUCache(/* too small */ 1);
  }

-  // Test anticipated new default as much as reasonably possible (and remove
-  // this code when obsolete)
-  assert(!table_options.decouple_partitioned_filters);
-  table_options.decouple_partitioned_filters = true;
-
  bool can_allow_mmap = IsMemoryMappedAccessSupported();
  switch (option_config) {
    case kHashSkipList:
@@ -592,7 +587,6 @@ Options DBTestBase::GetOptions(
      options_override.level_compaction_dynamic_level_bytes;
  options.env = env_;
  options.create_if_missing = true;
-  options.fail_if_options_file_error = true;
  return options;
 }

@@ -1160,16 +1154,18 @@ size_t DBTestBase::CountLiveFiles() {
 }

 int DBTestBase::NumTableFilesAtLevel(int level, int cf) {
-  std::string property;
-  if (cf == 0) {
-    // default cfd
-    EXPECT_TRUE(db_->GetProperty(
-        "rocksdb.num-files-at-level" + std::to_string(level), &property));
-  } else {
-    EXPECT_TRUE(db_->GetProperty(
-        handles_[cf], "rocksdb.num-files-at-level" + std::to_string(level),
-        &property));
+  return NumTableFilesAtLevel(level,
+                              cf ? handles_[cf] : db_->DefaultColumnFamily());
+}
+
+int DBTestBase::NumTableFilesAtLevel(int level, ColumnFamilyHandle* cfh,
+                                     DB* db) {
+  if (!db) {
+    db = db_;
  }
+  std::string property;
+  EXPECT_TRUE(db->GetProperty(
+      cfh, "rocksdb.num-files-at-level" + std::to_string(level), &property));
  return atoi(property.c_str());
 }

@@ -1202,12 +1198,22 @@ int DBTestBase::TotalTableFiles(int cf, int levels) {

 // Return spread of files per level
 std::string DBTestBase::FilesPerLevel(int cf) {
-  int num_levels =
-      (cf == 0) ? db_->NumberLevels() : db_->NumberLevels(handles_[cf]);
+  if (cf == 0) {
+    return FilesPerLevel(db_->DefaultColumnFamily());
+  } else {
+    return FilesPerLevel(handles_[cf]);
+  }
+}
+
+std::string DBTestBase::FilesPerLevel(ColumnFamilyHandle* cfh, DB* db) {
+  if (!db) {
+    db = db_;
+  }
+  int num_levels = db->NumberLevels(cfh);
  std::string result;
  size_t last_non_zero_offset = 0;
  for (int level = 0; level < num_levels; level++) {
-    int f = NumTableFilesAtLevel(level, cf);
+    int f = NumTableFilesAtLevel(level, cfh, db);
    char buf[100];
    snprintf(buf, sizeof(buf), "%s%d", (level ? "," : ""), f);
    result += buf;
@@ -1340,12 +1346,14 @@ void DBTestBase::FillLevels(const std::string& smallest,
 }

 void DBTestBase::MoveFilesToLevel(int level, int cf) {
+  MoveFilesToLevel(level, cf ? handles_[cf] : db_->DefaultColumnFamily());
+}
+
+void DBTestBase::MoveFilesToLevel(int level, ColumnFamilyHandle* column_family,
+                                  DB* db) {
+  DBImpl* db_impl = db ? static_cast<DBImpl*>(db) : dbfull();
  for (int l = 0; l < level; ++l) {
-    if (cf > 0) {
-      EXPECT_OK(dbfull()->TEST_CompactRange(l, nullptr, nullptr, handles_[cf]));
-    } else {
-      EXPECT_OK(dbfull()->TEST_CompactRange(l, nullptr, nullptr));
-    }
+    EXPECT_OK(db_impl->TEST_CompactRange(l, nullptr, nullptr, column_family));
  }
 }

@@ -452,6 +452,10 @@ class SpecialEnv : public EnvWrapper {
        return s;
      }

+      Status GetFileSize(uint64_t* s) override {
+        return target_->GetFileSize(s);
+      }
+
     private:
      std::unique_ptr<RandomAccessFile> target_;
      anon::AtomicCounter* counter_;
@@ -478,6 +482,10 @@ class SpecialEnv : public EnvWrapper {
        return target_->Prefetch(offset, n);
      }

+      Status GetFileSize(uint64_t* s) override {
+        return target_->GetFileSize(s);
+      }
+
     private:
      std::unique_ptr<RandomAccessFile> target_;
      std::atomic<uint64_t>* fail_cnt_;
@@ -1272,6 +1280,9 @@ class DBTestBase : public testing::Test {

  int NumTableFilesAtLevel(int level, int cf = 0);

+  int NumTableFilesAtLevel(int level, ColumnFamilyHandle* column_family,
+                           DB* db = nullptr);
+
  double CompressionRatioAtLevel(int level, int cf = 0);

  int TotalTableFiles(int cf = 0, int levels = -1);
@@ -1281,6 +1292,8 @@ class DBTestBase : public testing::Test {
  // Return spread of files per level
  std::string FilesPerLevel(int cf = 0);

+  std::string FilesPerLevel(ColumnFamilyHandle* cfh, DB* db = nullptr);
+
  size_t CountFiles();

  Status CountFiles(size_t* count);
@@ -1312,6 +1325,9 @@ class DBTestBase : public testing::Test {

  void MoveFilesToLevel(int level, int cf = 0);

+  void MoveFilesToLevel(int level, ColumnFamilyHandle* column_family,
+                        DB* db = nullptr);
+
  void DumpFileCounts(const char* label);

  std::string DumpSSTableList();
@@ -1672,55 +1672,75 @@ TEST_P(DBTestUniversalCompaction, ConcurrentBottomPriLowPriCompactions) {
  }
  const int kNumFilesTrigger = 3;
  Env::Default()->SetBackgroundThreads(1, Env::Priority::BOTTOM);
-  Options options = CurrentOptions();
-  options.compaction_style = kCompactionStyleUniversal;
-  options.max_background_compactions = 2;
-  options.num_levels = num_levels_;
-  options.write_buffer_size = 100 << 10;     // 100KB
-  options.target_file_size_base = 32 << 10;  // 32KB
-  options.level0_file_num_compaction_trigger = kNumFilesTrigger;
-  // Trigger compaction if size amplification exceeds 110%
-  options.compaction_options_universal.max_size_amplification_percent = 110;
-  DestroyAndReopen(options);

-  // Need to get a token to enable compaction parallelism up to
-  // `max_background_compactions` jobs.
-  auto pressure_token =
-      dbfull()->TEST_write_controler().GetCompactionPressureToken();
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
-      {// wait for the full compaction to be picked before adding files intended
-       // for the second one.
-       {"DBImpl::BackgroundCompaction:ForwardToBottomPriPool",
-        "DBTestUniversalCompaction:ConcurrentBottomPriLowPriCompactions:0"},
-       // the full (bottom-pri) compaction waits until a partial (low-pri)
-       // compaction has started to verify they can run in parallel.
-       {"DBImpl::BackgroundCompaction:NonTrivial",
-        "DBImpl::BGWorkBottomCompaction"}});
-  SyncPoint::GetInstance()->EnableProcessing();
+  for (bool universal_reduce_file_locking : {true, false}) {
+    Options options = CurrentOptions();
+    options.compaction_style = kCompactionStyleUniversal;
+    options.compaction_options_universal.reduce_file_locking =
+        universal_reduce_file_locking;
+    options.max_background_compactions = 2;
+    options.num_levels = num_levels_;
+    options.write_buffer_size = 100 << 10;     // 100KB
+    options.target_file_size_base = 32 << 10;  // 32KB
+    options.level0_file_num_compaction_trigger = kNumFilesTrigger;
+    // Trigger compaction if size amplification exceeds 110%
+    options.compaction_options_universal.max_size_amplification_percent = 110;
+    DestroyAndReopen(options);

-  Random rnd(301);
-  for (int i = 0; i < 2; ++i) {
-    for (int num = 0; num < kNumFilesTrigger; num++) {
-      int key_idx = 0;
-      GenerateNewFile(&rnd, &key_idx, true /* no_wait */);
-      // use no_wait above because that one waits for flush and compaction. We
-      // don't want to wait for compaction because the full compaction is
-      // intentionally blocked while more files are flushed.
-      ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());
+    // Need to get a token to enable compaction parallelism up to
+    // `max_background_compactions` jobs.
+    auto pressure_token =
+        dbfull()->TEST_write_controler().GetCompactionPressureToken();
+    if (universal_reduce_file_locking) {
+      ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
+          {// Wait for the full compaction to be repicked before adding files
+           // intended for the second compaction.
+           {"DBImpl::BackgroundCompaction():AfterPickCompactionBottomPri",
+            "DBTestUniversalCompaction:ConcurrentBottomPriLowPriCompactions:0"},
+           // Wait for the second compaction to run before running the full
+           // compaction to verify they can run in parallel
+           {"DBImpl::BackgroundCompaction:NonTrivial:BeforeRun",
+            "DBImpl::BackgroundCompaction:NonTrivial:BeforeRunBottomPri"}});
+    } else {
+      ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
+          {// Wait for the full compaction to be forwarded before adding files
+           // intended for the second compaction.
+           {"DBImpl::BackgroundCompaction:ForwardToBottomPriPool",
+            "DBTestUniversalCompaction:ConcurrentBottomPriLowPriCompactions:0"},
+           // Wait for the second compaction to run before running the full
+           // compaction to verify they can run in parallel
+           {"DBImpl::BackgroundCompaction:NonTrivial:BeforeRun",
+            "DBImpl::BackgroundCompaction:NonTrivial:BeforeRunBottomPri"}});
    }
-    if (i == 0) {
-      TEST_SYNC_POINT(
-          "DBTestUniversalCompaction:ConcurrentBottomPriLowPriCompactions:0");
+
+    SyncPoint::GetInstance()->EnableProcessing();
+
+    Random rnd(301);
+    for (int i = 0; i < 2; ++i) {
+      for (int num = 0; num < kNumFilesTrigger; num++) {
+        int key_idx = 0;
+        GenerateNewFile(&rnd, &key_idx, true /* no_wait */);
+        // use no_wait above because that one waits for flush and compaction. We
+        // don't want to wait for compaction because the full compaction is
+        // intentionally blocked while more files are flushed.
+        ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());
+      }
+      if (i == 0) {
+        TEST_SYNC_POINT(
+            "DBTestUniversalCompaction:ConcurrentBottomPriLowPriCompactions:0");
+      }
    }
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+
+    // First compaction should output to bottom level. Second should output to
+    // L0 since older L0 files pending compaction prevent it from being placed
+    // lower.
+    ASSERT_EQ(NumSortedRuns(), 2);
+    ASSERT_GT(NumTableFilesAtLevel(0), 0);
+    ASSERT_GT(NumTableFilesAtLevel(num_levels_ - 1), 0);
+    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
  }
-  ASSERT_OK(dbfull()->TEST_WaitForCompact());

-  // First compaction should output to bottom level. Second should output to L0
-  // since older L0 files pending compaction prevent it from being placed lower.
-  ASSERT_EQ(NumSortedRuns(), 2);
-  ASSERT_GT(NumTableFilesAtLevel(0), 0);
-  ASSERT_GT(NumTableFilesAtLevel(num_levels_ - 1), 0);
-  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
  Env::Default()->SetBackgroundThreads(0, Env::Priority::BOTTOM);
 }

@@ -2086,46 +2106,79 @@ TEST_F(DBTestUniversalCompaction2, OverlappingL0) {
 }

 TEST_F(DBTestUniversalCompaction2, IngestBehind) {
-  const int kNumKeys = 3000;
-  const int kWindowSize = 100;
-  const int kNumDelsTrigger = 90;
+  for (bool cf_option : {false, true}) {
+    SCOPED_TRACE("cf_option = " + std::to_string(cf_option));
+    const int kNumKeys = 3000;
+    const int kWindowSize = 100;
+    const int kNumDelsTrigger = 90;

-  Options opts = CurrentOptions();
-  opts.table_properties_collector_factories.emplace_back(
-      NewCompactOnDeletionCollectorFactory(kWindowSize, kNumDelsTrigger));
-  opts.compaction_style = kCompactionStyleUniversal;
-  opts.level0_file_num_compaction_trigger = 2;
-  opts.compression = kNoCompression;
-  opts.allow_ingest_behind = true;
-  opts.compaction_options_universal.size_ratio = 10;
-  opts.compaction_options_universal.min_merge_width = 2;
-  opts.compaction_options_universal.max_size_amplification_percent = 200;
-  Reopen(opts);
-
-  // add an L1 file to prevent tombstones from dropping due to obsolescence
-  // during flush
-  int i;
-  for (i = 0; i < 2000; ++i) {
-    ASSERT_OK(Put(Key(i), "val"));
-  }
-  ASSERT_OK(Flush());
-  //  MoveFilesToLevel(6);
-  ASSERT_OK(dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr));
-
-  for (i = 1999; i < kNumKeys; ++i) {
-    if (i >= kNumKeys - kWindowSize &&
-        i < kNumKeys - kWindowSize + kNumDelsTrigger) {
-      ASSERT_OK(Delete(Key(i)));
+    Options opts = CurrentOptions();
+    opts.table_properties_collector_factories.emplace_back(
+        NewCompactOnDeletionCollectorFactory(kWindowSize, kNumDelsTrigger));
+    opts.compaction_style = kCompactionStyleUniversal;
+    opts.level0_file_num_compaction_trigger = 2;
+    opts.compression = kNoCompression;
+    if (cf_option) {
+      opts.cf_allow_ingest_behind = true;
    } else {
+      opts.allow_ingest_behind = true;
+    }
+    opts.compaction_options_universal.size_ratio = 10;
+    opts.compaction_options_universal.min_merge_width = 2;
+    opts.compaction_options_universal.max_size_amplification_percent = 200;
+    Reopen(opts);
+
+    // add an L1 file to prevent tombstones from dropping due to obsolescence
+    // during flush
+    int i;
+    for (i = 0; i < 2000; ++i) {
      ASSERT_OK(Put(Key(i), "val"));
    }
-  }
-  ASSERT_OK(Flush());
+    ASSERT_OK(Flush());
+    //  MoveFilesToLevel(6);
+    ASSERT_OK(dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr));

-  ASSERT_OK(dbfull()->TEST_WaitForCompact());
-  ASSERT_EQ(0, NumTableFilesAtLevel(0));
-  ASSERT_EQ(0, NumTableFilesAtLevel(6));
-  ASSERT_GT(NumTableFilesAtLevel(5), 0);
+    for (i = 1999; i < kNumKeys; ++i) {
+      if (i >= kNumKeys - kWindowSize &&
+          i < kNumKeys - kWindowSize + kNumDelsTrigger) {
+        ASSERT_OK(Delete(Key(i)));
+      } else {
+        ASSERT_OK(Put(Key(i), "val"));
+      }
+    }
+    ASSERT_OK(Flush());
+
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    ASSERT_EQ(0, NumTableFilesAtLevel(0));
+    ASSERT_EQ(0, NumTableFilesAtLevel(6));
+    ASSERT_GT(NumTableFilesAtLevel(5), 0);
+
+    if (cf_option) {
+      // Test that another CF does not allow ingest behind
+      ColumnFamilyHandle* new_cfh;
+      Options new_cf_option;
+      new_cf_option.compaction_style = kCompactionStyleUniversal;
+      new_cf_option.num_levels = 7;
+      // CreateColumnFamilies({"new_cf"}, new_cf_option);
+      ASSERT_OK(db_->CreateColumnFamily(new_cf_option, "new_cf", &new_cfh));
+      // handles_.push_back(new_cfh);
+      for (i = 0; i < 10; ++i) {
+        // ASSERT_OK(Put(1, Key(i), "val"));
+        ASSERT_OK(db_->Put(WriteOptions(), new_cfh, Key(i), "val"));
+      }
+      ASSERT_OK(
+          db_->CompactRange(CompactRangeOptions(), new_cfh, nullptr, nullptr));
+      // This CF can use the last leve.
+      std::string property;
+      EXPECT_TRUE(db_->GetProperty(
+          new_cfh, "rocksdb.num-files-at-level" + std::to_string(6),
+          &property));
+      ASSERT_EQ(1, atoi(property.c_str()));
+
+      ASSERT_OK(db_->DropColumnFamily(new_cfh));
+      ASSERT_OK(db_->DestroyColumnFamilyHandle(new_cfh));
+    }
+  }
 }

 TEST_F(DBTestUniversalCompaction2, PeriodicCompactionDefault) {
@@ -3024,13 +3024,13 @@ TEST_F(DBWALTest, GetCompressedWalsAfterSync) {
  options.wal_compression = kZSTD;
  DestroyAndReopen(options);

-  // Write something to memtable and WAL so that log_empty_ will be false after
+  // Write something to memtable and WAL so that wal_empty_ will be false after
  // next DB::Open().
  ASSERT_OK(Put("a", "v"));

  Reopen(options);

-  // New WAL is created, thanks to !log_empty_.
+  // New WAL is created, thanks to !wal_empty_.
  ASSERT_OK(dbfull()->TEST_SwitchWAL());

  ASSERT_OK(Put("b", "v"));
@@ -19,6 +19,13 @@
 #include "utilities/merge_operators/string_append/stringappend2.h"

 namespace ROCKSDB_NAMESPACE {
+namespace {
+std::string EncodeAsUint64(uint64_t v) {
+  std::string dst;
+  PutFixed64(&dst, v);
+  return dst;
+}
+}  // namespace
 class DBBasicTestWithTimestamp : public DBBasicTestWithTimestampBase {
 public:
  DBBasicTestWithTimestamp()
@@ -1480,13 +1487,24 @@ TEST_F(DBBasicTestWithTimestamp, ReseekToUserKeyBeforeSavedKey) {
  Close();
 }

-TEST_F(DBBasicTestWithTimestamp,
-       FIXME_ReverseIterationWithBlobAndUnpreparedValue) {
+class ReverseIterationWithUnpreparedBlobTest
+    : public DBBasicTestWithTimestampBase,
+      public testing::WithParamInterface<std::tuple<bool, uint64_t>> {
+ public:
+  ReverseIterationWithUnpreparedBlobTest()
+      : DBBasicTestWithTimestampBase(
+            "db_basic_test_with_timestamp_reverse_with_unprepare") {}
+};
+INSTANTIATE_TEST_CASE_P(ReverseIterationWithUnpreparedBlobTest,
+                        ReverseIterationWithUnpreparedBlobTest,
+                        ::testing::Combine(::testing::Values(true, false),
+                                           ::testing::Values(0, 2)));
+TEST_P(ReverseIterationWithUnpreparedBlobTest, Basic) {
  Options options = CurrentOptions();
  options.create_if_missing = true;
  options.env = env_;
  options.enable_blob_files = true;
-  options.max_sequential_skip_in_iterations = 0;
+  options.max_sequential_skip_in_iterations = std::get<1>(GetParam());

  const size_t kTimestampSize = Timestamp(0, 0).size();
  TestComparator test_cmp(kTimestampSize);
@@ -1501,7 +1519,7 @@ TEST_F(DBBasicTestWithTimestamp,
  for (uint64_t key = 0; key <= kMaxKey; ++key) {
    for (size_t i = 0; i < write_timestamps.size(); ++i) {
      ASSERT_OK(db_->Put(WriteOptions(), Key1(key), write_timestamps[i],
-                         "value" + std::to_string(i)));
+                         Key1(key) + "value" + std::to_string(i)));
    }
  }

@@ -1513,17 +1531,28 @@ TEST_F(DBBasicTestWithTimestamp,

    ReadOptions read_opts;
    read_opts.timestamp = &read_timestamp;
-    read_opts.allow_unprepared_value = true;
+    read_opts.allow_unprepared_value = std::get<0>(GetParam());

    std::unique_ptr<Iterator> it(db_->NewIterator(read_opts));

    it->SeekForPrev(Key1(kMaxKey));
-    ASSERT_TRUE(it->Valid());
-    ASSERT_OK(it->status());
+    uint64_t key = kMaxKey;
+    int count = 0;
+    while (it->Valid()) {
+      ASSERT_OK(it->status());

-    // FIXME: PrepareValue() should succeed and status() should remain OK
-    ASSERT_FALSE(it->PrepareValue());
-    ASSERT_TRUE(it->status().IsCorruption());
+      ASSERT_TRUE(it->PrepareValue());
+      ASSERT_TRUE(it->Valid());
+      ASSERT_OK(it->status());
+      ASSERT_EQ(it->key(), Key1(key));
+      ASSERT_EQ(it->timestamp(), Timestamp(3, 0));
+      ASSERT_EQ(it->value(), Key1(key) + "value" + std::to_string(1));
+      key--;
+      count++;
+      it->Prev();
+    }
+    ASSERT_OK(it->status());
+    ASSERT_EQ(kMaxKey + 1, count);
  }

  Close();
@@ -2371,7 +2400,6 @@ class DataVisibilityTest : public DBBasicTestWithTimestampBase {
    }
  }
 };
-constexpr int DataVisibilityTest::kTestDataSize;

 // Application specifies timestamp but not snapshot.
 //           reader              writer
@@ -3746,17 +3774,42 @@ INSTANTIATE_TEST_CASE_P(
        test::UserDefinedTimestampTestMode::kStripUserDefinedTimestamp,
        test::UserDefinedTimestampTestMode::kNormal));

-TEST_F(DBBasicTestWithTimestamp, EnableDisableUDT) {
+// Test params:
+// 1) whether to flush before close
+class EnableDisableUDTTest : public DBBasicTestWithTimestampBase,
+                             public testing::WithParamInterface<bool> {
+ public:
+  EnableDisableUDTTest()
+      : DBBasicTestWithTimestampBase("/enable_disable_udt") {}
+};
+
+INSTANTIATE_TEST_CASE_P(EnableDisableUDTTest, EnableDisableUDTTest,
+                        ::testing::Values(true, false));
+
+TEST_P(EnableDisableUDTTest, Basic) {
  Options options = CurrentOptions();
+  // Un-flushed data before close will involve a WAL replay on DB reopen.
+  bool flush_before_close = GetParam();
  options.env = env_;
-  // Create a column family without user-defined timestamps.
  options.comparator = BytewiseComparator();
  options.persist_user_defined_timestamps = true;
  DestroyAndReopen(options);

+  ReadOptions ropts;
+  std::string read_ts;
+  std::string value;
+  std::string key_ts;
+
  // Create one SST file, its user keys have no user-defined timestamps.
-  ASSERT_OK(db_->Put(WriteOptions(), "foo", "val1"));
-  ASSERT_OK(Flush(0));
+  ASSERT_OK(db_->Put(WriteOptions(), "foo", "val0"));
+  ASSERT_OK(db_->Put(WriteOptions(), "bar", "val0"));
+  ASSERT_OK(db_->DeleteRange(WriteOptions(), "bar", "baz"));
+  ASSERT_OK(db_->Get(ReadOptions(), "foo", &value));
+  ASSERT_EQ("val0", value);
+  ASSERT_TRUE(db_->Get(ReadOptions(), "bar", &value).IsNotFound());
+  if (flush_before_close) {
+    ASSERT_OK(Flush(0));
+  }
  Close();

  // Reopen the existing column family and enable user-defined timestamps
@@ -3765,47 +3818,63 @@ TEST_F(DBBasicTestWithTimestamp, EnableDisableUDT) {
  options.persist_user_defined_timestamps = false;
  options.allow_concurrent_memtable_write = false;
  Reopen(options);
-
-  std::string value;
-  ASSERT_TRUE(db_->Get(ReadOptions(), "foo", &value).IsInvalidArgument());
-  std::string read_ts;
-  PutFixed64(&read_ts, 0);
-  ReadOptions ropts;
+  // Read data from previous session before and after compaction.
+  read_ts = EncodeAsUint64(1);
  Slice read_ts_slice = read_ts;
  ropts.timestamp = &read_ts_slice;
-  std::string key_ts;
-  // Entries in pre-existing SST files are treated as if they have minimum
-  // user-defined timestamps.
-  ASSERT_OK(db_->Get(ropts, "foo", &value, &key_ts));
-  ASSERT_EQ("val1", value);
-  ASSERT_EQ(read_ts, key_ts);
+  for (int i = 0; i < 2; i++) {
+    ASSERT_TRUE(db_->Get(ReadOptions(), "foo", &value).IsInvalidArgument());
+    // Entries in pre-existing SST files are treated as if they have minimum
+    // user-defined timestamps.
+    ASSERT_OK(db_->Get(ropts, "foo", &value, &key_ts));
+    ASSERT_EQ("val0", value);
+    ASSERT_EQ(EncodeAsUint64(0), key_ts);
+    ASSERT_TRUE(db_->Get(ropts, "bar", &value, &key_ts).IsNotFound());
+    ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  }

  // Do timestamped read / write.
-  std::string write_ts;
-  PutFixed64(&write_ts, 1);
-  ASSERT_OK(db_->Put(WriteOptions(), "foo", write_ts, "val2"));
-  read_ts.clear();
-  PutFixed64(&read_ts, 1);
+  ASSERT_OK(db_->Put(WriteOptions(), "foo", EncodeAsUint64(1), "val1"));
+  ASSERT_OK(db_->Put(WriteOptions(), "bar", EncodeAsUint64(1), "val1"));
+  ASSERT_OK(db_->DeleteRange(WriteOptions(), "bar", "baz", EncodeAsUint64(2)));
  ASSERT_OK(db_->Get(ropts, "foo", &value, &key_ts));
-  ASSERT_EQ("val2", value);
-  ASSERT_EQ(write_ts, key_ts);
+  ASSERT_EQ("val1", value);
+  ASSERT_EQ(EncodeAsUint64(1), key_ts);
+  ASSERT_OK(db_->Get(ropts, "bar", &value, &key_ts));
+  ASSERT_EQ("val1", value);
+  ASSERT_EQ(EncodeAsUint64(1), key_ts);
+  read_ts = EncodeAsUint64(2);
+  ASSERT_TRUE(db_->Get(ropts, "bar", &value, &key_ts).IsNotFound());
  // The user keys in this SST file don't have user-defined timestamps either,
  // because `persist_user_defined_timestamps` flag is set to false.
-  ASSERT_OK(Flush(0));
+  if (flush_before_close) {
+    ASSERT_OK(Flush(0));
+  }
  Close();

  // Reopen the existing column family while disabling user-defined timestamps.
  options.comparator = BytewiseComparator();
  Reopen(options);

-  ASSERT_TRUE(db_->Get(ropts, "foo", &value).IsInvalidArgument());
-  ASSERT_OK(db_->Get(ReadOptions(), "foo", &value));
-  ASSERT_EQ("val2", value);
+  // Read data from previous session before and after compaction.
+  for (int i = 0; i < 2; i++) {
+    ASSERT_TRUE(db_->Get(ropts, "foo", &value).IsInvalidArgument());
+    ASSERT_OK(db_->Get(ReadOptions(), "foo", &value));
+    ASSERT_EQ("val1", value);
+    ASSERT_TRUE(db_->Get(ReadOptions(), "bar", &value).IsNotFound());
+    ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  }

  // Continue to write / read the column family without user-defined timestamps.
-  ASSERT_OK(db_->Put(WriteOptions(), "foo", "val3"));
+  ASSERT_OK(db_->Put(WriteOptions(), "foo", "val2"));
+  ASSERT_OK(db_->Put(WriteOptions(), "bar", "val2"));
+  ASSERT_OK(db_->DeleteRange(WriteOptions(), "bar", "baz"));
  ASSERT_OK(db_->Get(ReadOptions(), "foo", &value));
-  ASSERT_EQ("val3", value);
+  ASSERT_EQ("val2", value);
+  ASSERT_TRUE(db_->Get(ReadOptions(), "bar", &value).IsNotFound());
+  if (flush_before_close) {
+    ASSERT_OK(Flush(0));
+  }
  Close();
 }

@@ -4844,6 +4913,117 @@ TEST_F(DBBasicTestWithTimestamp, TimestampFilterTableReadOnGet) {
  Close();
 }

+class GetNewestUserDefinedTimestampTest : public DBBasicTestWithTimestampBase {
+ public:
+  explicit GetNewestUserDefinedTimestampTest()
+      : DBBasicTestWithTimestampBase("get_newest_udt_test") {}
+};
+
+TEST_F(GetNewestUserDefinedTimestampTest, Basic) {
+  std::string newest_timestamp;
+  // UDT disabled, get InvalidArgument.
+  ASSERT_TRUE(db_->GetNewestUserDefinedTimestamp(nullptr, &newest_timestamp)
+                  .IsInvalidArgument());
+  Options options = CurrentOptions();
+  options.env = env_;
+  options.create_if_missing = true;
+  options.max_write_buffer_number = 5;
+  options.min_write_buffer_number_to_merge = 4;
+  options.comparator = test::BytewiseComparatorWithU64TsWrapper();
+
+  DestroyAndReopen(options);
+  // UDT persisted, get NotSupported.
+  ASSERT_TRUE(db_->GetNewestUserDefinedTimestamp(nullptr, &newest_timestamp)
+                  .IsNotSupported());
+
+  options.persist_user_defined_timestamps = false;
+  options.allow_concurrent_memtable_write = false;
+
+  DestroyAndReopen(options);
+  ASSERT_TRUE(
+      db_->GetNewestUserDefinedTimestamp(nullptr, nullptr).IsInvalidArgument());
+
+  ColumnFamilyHandleImpl* cfh = static_cast_with_check<ColumnFamilyHandleImpl>(
+      db_->DefaultColumnFamily());
+  ColumnFamilyData* cfd = cfh->cfd();
+  // The column family hasn't seen any user defined timestamp
+  ASSERT_OK(db_->GetNewestUserDefinedTimestamp(nullptr, &newest_timestamp));
+  ASSERT_TRUE(newest_timestamp.empty());
+
+  ASSERT_OK(db_->Put(WriteOptions(), Key(1), EncodeAsUint64(1), "val1"));
+  // Testing get newest timestamp from mutable memtable.
+  ASSERT_OK(db_->GetNewestUserDefinedTimestamp(nullptr, &newest_timestamp));
+  ASSERT_EQ(EncodeAsUint64(1), newest_timestamp);
+
+  ASSERT_OK(db_->Put(WriteOptions(), Key(1), EncodeAsUint64(2), "val2"));
+  ASSERT_OK(dbfull()->TEST_SwitchMemtable(cfd));
+  // Testing get the newest timestamp from immutable memtable because the
+  // mutable one is empty.
+  ASSERT_OK(db_->GetNewestUserDefinedTimestamp(nullptr, &newest_timestamp));
+  ASSERT_EQ(EncodeAsUint64(2), newest_timestamp);
+
+  ASSERT_OK(db_->Put(WriteOptions(), Key(1), EncodeAsUint64(3), "val3"));
+  ASSERT_OK(db_->Put(WriteOptions(), Key(1), EncodeAsUint64(4), "val4"));
+  ASSERT_OK(dbfull()->TEST_SwitchMemtable(cfd));
+  // Testing get the newest timestamp from the more recent immutable memtable
+  // when there are multiple immutable memtables.
+  ASSERT_OK(db_->GetNewestUserDefinedTimestamp(nullptr, &newest_timestamp));
+  ASSERT_EQ(EncodeAsUint64(4), newest_timestamp);
+
+  ASSERT_OK(db_->Put(WriteOptions(), Key(1), EncodeAsUint64(5), "val5"));
+  // Testing get newest timestamp from mutable memtable when it has data, in the
+  // presence of immutable memtables.
+  ASSERT_OK(db_->GetNewestUserDefinedTimestamp(nullptr, &newest_timestamp));
+  ASSERT_EQ(EncodeAsUint64(5), newest_timestamp);
+
+  ASSERT_OK(Flush());
+  // After flushing and all the user defined timestamp are flushed. User defined
+  // timestamp info for SST files is available from MANIFEST.
+  ASSERT_OK(db_->GetNewestUserDefinedTimestamp(nullptr, &newest_timestamp));
+  ASSERT_EQ(EncodeAsUint64(5), newest_timestamp);
+
+  Reopen(options);
+  // Similar after flush, when there is no memtables, but some SST files,
+  // if MANIFEST records the upperbound of flushed timestamps because timestamps
+  // are not persisted in SST files, this info can be found.
+  ASSERT_OK(db_->GetNewestUserDefinedTimestamp(nullptr, &newest_timestamp));
+  ASSERT_EQ(EncodeAsUint64(5), newest_timestamp);
+
+  Close();
+}
+
+TEST_F(GetNewestUserDefinedTimestampTest, ConcurrentWrites) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.comparator = test::BytewiseComparatorWithU64TsWrapper();
+  options.persist_user_defined_timestamps = false;
+  options.allow_concurrent_memtable_write = false;
+
+  DestroyAndReopen(options);
+
+  std::vector<std::thread> threads;
+  threads.reserve(10);
+  std::atomic<uint64_t> current_ts{0};
+  for (int i = 0; i < 10; i++) {
+    threads.emplace_back([this, i, &current_ts]() {
+      if (i % 2 == 0) {
+        std::string newest_timestamp;
+        ASSERT_OK(
+            db_->GetNewestUserDefinedTimestamp(nullptr, &newest_timestamp));
+      } else {
+        uint64_t write_ts = current_ts.fetch_add(1);
+        ASSERT_OK(db_->Put(WriteOptions(), Key(1), EncodeAsUint64(write_ts),
+                           "val" + std::to_string(i)));
+      }
+    });
+  }
+
+  for (auto& t : threads) {
+    t.join();
+  }
+  Close();
+}
+
 }  // namespace ROCKSDB_NAMESPACE

 int main(int argc, char** argv) {
@@ -987,7 +987,7 @@ TEST_P(DBWriteTest, RecycleLogToggleTest) {

  options.recycle_log_file_num = 1;
  Reopen(options);
-  // 1.log is added to alive_log_files_
+  // 1.log is added to alive_wal_files_
  ASSERT_OK(Put(Key(2), "val1"));
  ASSERT_OK(Flush());
  // 1.log should be deleted and not recycled, since it
@@ -1000,6 +1000,80 @@ TEST_P(DBWriteTest, RecycleLogToggleTest) {
  ASSERT_EQ(Get(Key(1)), "val2");
 }

+TEST_P(DBWriteTest, IngestWriteBatchWithIndex) {
+  if (GetParam() == kPipelinedWrite) {
+    return;
+  }
+
+  Options options = GetOptions();
+  options.disable_auto_compactions = true;
+  Reopen(options);
+  Options cf_options = GetOptions();
+  cf_options.merge_operator = MergeOperators::CreateStringAppendOperator();
+  CreateColumnFamilies({"cf1", "cf2"}, cf_options);
+  ReopenWithColumnFamilies({"default", "cf1", "cf2"},
+                           {options, cf_options, cf_options});
+
+  // default cf
+  auto wbwi1 = std::make_shared<WriteBatchWithIndex>(options.comparator, 0,
+                                                     /*overwrite_key=*/true);
+  ASSERT_OK(wbwi1->Put("key1", "value1"));
+  ASSERT_OK(wbwi1->Put("key2", "value2"));
+  if (GetParam() == kPipelinedWrite) {
+    ASSERT_TRUE(db_->IngestWriteBatchWithIndex({}, wbwi1).IsNotSupported());
+    return;
+  }
+  // Test disableWAL=false
+  ASSERT_TRUE(db_->IngestWriteBatchWithIndex({}, wbwi1).IsNotSupported());
+
+  WriteOptions wo;
+  wo.disableWAL = true;
+  ASSERT_OK(db_->IngestWriteBatchWithIndex(wo, wbwi1));
+  ASSERT_EQ("value1", Get("key1"));
+  ASSERT_EQ("value2", Get("key2"));
+
+  // Test with overwrites
+  auto wbwi = std::make_shared<WriteBatchWithIndex>(options.comparator, 0,
+                                                    /*overwrite_key=*/true);
+  ASSERT_OK(wbwi->Put("key2", "value3"));
+  ASSERT_OK(wbwi->Delete("key1"));  // Delete an existing key
+  ASSERT_OK(db_->IngestWriteBatchWithIndex(wo, wbwi));
+  ASSERT_EQ("NOT_FOUND", Get("key1"));
+  ASSERT_EQ("value3", Get("key2"));
+
+  auto wbwi2 = std::make_shared<WriteBatchWithIndex>(options.comparator, 0,
+                                                     /*overwrite_key=*/true);
+  ASSERT_OK(wbwi2->Put(handles_[1], "cf1_key1", "cf1_value1"));
+  ASSERT_OK(wbwi2->Delete(handles_[1], "cf1_key2"));
+  // Test ingestion with column family
+  ASSERT_OK(db_->IngestWriteBatchWithIndex(wo, wbwi2));
+  ASSERT_EQ("cf1_value1", Get(1, "cf1_key1"));
+  ASSERT_EQ("NOT_FOUND", Get(1, "cf1_key2"));
+
+  auto wbwi3 = std::make_shared<WriteBatchWithIndex>(options.comparator, 0,
+                                                     /*overwrite_key=*/true);
+  ASSERT_OK(wbwi3->Merge(handles_[2], "cf2_key1", "cf2_value1"));
+  ASSERT_OK(wbwi3->Merge(handles_[2], "cf2_key1", "cf2_value2"));
+  // Test ingestion with merge operations
+  ASSERT_OK(db_->IngestWriteBatchWithIndex(wo, wbwi3));
+  ASSERT_EQ("cf2_value1,cf2_value2", Get(2, "cf2_key1"));
+
+  // Test with overwrite_key = false
+  auto wbwi_no_overwrite = std::make_shared<WriteBatchWithIndex>(
+      options.comparator, 0, /*overwrite_key=*/false);
+  ASSERT_OK(wbwi_no_overwrite->Put("key1", "value1"));
+  Status s = db_->IngestWriteBatchWithIndex(wo, wbwi_no_overwrite);
+  ASSERT_TRUE(s.IsNotSupported());
+
+  auto empty_wbwi = std::make_shared<WriteBatchWithIndex>(
+      options.comparator, 0, /*overwrite_key=*/true);
+  ASSERT_OK(db_->IngestWriteBatchWithIndex(wo, empty_wbwi));
+
+  DestroyAndReopen(options);
+  // Should fail when trying to ingest to non-existent column family
+  ASSERT_NOK(db_->IngestWriteBatchWithIndex(wo, wbwi2));
+}
+
 INSTANTIATE_TEST_CASE_P(DBWriteTestInstance, DBWriteTest,
                        testing::Values(DBTestBase::kDefault,
                                        DBTestBase::kConcurrentWALWrites,
@@ -275,9 +275,6 @@ void ErrorHandler::HandleKnownErrors(const Status& bg_err,
    return;
  }

-  ROCKS_LOG_INFO(db_options_.info_log,
-                 "ErrorHandler: Set regular background error\n");
-
  bool paranoid = db_options_.paranoid_checks;
  Status::Severity sev = Status::Severity::kFatalError;
  Status new_bg_err;
@@ -335,12 +332,21 @@ void ErrorHandler::HandleKnownErrors(const Status& bg_err,
    if (!s.ok() && (s.severity() > bg_error_.severity())) {
      bg_error_ = s;
    } else {
+      ROCKS_LOG_INFO(db_options_.info_log,
+                     "ErrorHandler: Hit less severe background error\n");
+
      // This error is less severe than previously encountered error. Don't
      // take any further action
      return;
    }
  }

+  bool stop = bg_error_.severity() >= Status::Severity::kHardError;
+  ROCKS_LOG_INFO(
+      db_options_.info_log,
+      "ErrorHandler: Set regular background error, auto_recovery=%d, stop=%d\n",
+      int{auto_recovery}, int{stop});
+
  recover_context_ = context;
  if (auto_recovery) {
    recovery_in_prog_ = true;
@@ -351,7 +357,7 @@ void ErrorHandler::HandleKnownErrors(const Status& bg_err,
      RecoverFromNoSpace();
    }
  }
-  if (bg_error_.severity() >= Status::Severity::kHardError) {
+  if (stop) {
    is_db_stopped_.store(true, std::memory_order_release);
  }
 }
@@ -77,7 +77,12 @@ void EventHelpers::LogAndNotifyTableFileCreationFinished(
    TableFileCreationReason reason, const Status& s,
    const std::string& file_checksum,
    const std::string& file_checksum_func_name) {
-  if (s.ok() && event_logger) {
+  if (!event_logger && listeners.empty()) {
+    s.PermitUncheckedError();
+    return;
+  }
+
+  if (event_logger) {
    JSONWriter jwriter;
    AppendCurrentTime(&jwriter);
    jwriter << "cf_name" << cf_name << "job" << job_id << "event"
@@ -165,6 +170,8 @@ void EventHelpers::LogAndNotifyTableFileCreationFinished(
      jwriter << "oldest_blob_file_number" << oldest_blob_file_number;
    }

+    jwriter << "status" << s.ToString();
+
    jwriter.EndObject();

    event_logger->Log(jwriter);
@@ -195,18 +202,22 @@ void EventHelpers::LogAndNotifyTableFileDeletion(
    const std::string& file_path, const Status& status,
    const std::string& dbname,
    const std::vector<std::shared_ptr<EventListener>>& listeners) {
-  JSONWriter jwriter;
-  AppendCurrentTime(&jwriter);
-
-  jwriter << "job" << job_id << "event" << "table_file_deletion"
-          << "file_number" << file_number;
-  if (!status.ok()) {
-    jwriter << "status" << status.ToString();
+  if (!event_logger && listeners.empty()) {
+    status.PermitUncheckedError();
+    return;
  }

-  jwriter.EndObject();
+  if (event_logger) {
+    JSONWriter jwriter;
+    AppendCurrentTime(&jwriter);

-  event_logger->Log(jwriter);
+    jwriter << "job" << job_id << "event" << "table_file_deletion"
+            << "file_number" << file_number << "status" << status.ToString();
+
+    jwriter.EndObject();
+
+    event_logger->Log(jwriter);
+  }

  if (listeners.empty()) {
    return;
@@ -274,7 +285,12 @@ void EventHelpers::LogAndNotifyBlobFileCreationFinished(
    const std::string& file_checksum,
    const std::string& file_checksum_func_name, uint64_t total_blob_count,
    uint64_t total_blob_bytes) {
-  if (s.ok() && event_logger) {
+  if (!event_logger && listeners.empty()) {
+    s.PermitUncheckedError();
+    return;
+  }
+
+  if (event_logger) {
    JSONWriter jwriter;
    AppendCurrentTime(&jwriter);
    jwriter << "cf_name" << cf_name << "job" << job_id << "event"
@@ -305,15 +321,17 @@ void EventHelpers::LogAndNotifyBlobFileDeletion(
    const std::vector<std::shared_ptr<EventListener>>& listeners, int job_id,
    uint64_t file_number, const std::string& file_path, const Status& status,
    const std::string& dbname) {
+  if (!event_logger && listeners.empty()) {
+    status.PermitUncheckedError();
+    return;
+  }
+
  if (event_logger) {
    JSONWriter jwriter;
    AppendCurrentTime(&jwriter);

    jwriter << "job" << job_id << "event" << "blob_file_deletion"
-            << "file_number" << file_number;
-    if (!status.ok()) {
-      jwriter << "status" << status.ToString();
-    }
+            << "file_number" << file_number << "status" << status.ToString();

    jwriter.EndObject();
    event_logger->Log(jwriter);
@@ -88,11 +88,12 @@ Status GetFileChecksumsFromCurrentManifest(FileSystem* fs,

  // Read all records from the manifest file...
  uint64_t manifest_file_size = std::numeric_limits<uint64_t>::max();
-  FileChecksumRetriever retriever(read_options, manifest_file_size,
-                                  *checksum_list);
+  FileChecksumRetriever retriever(read_options, manifest_file_size);
  retriever.Iterate(reader, &s);
-
-  return retriever.status();
+  if (!retriever.status().ok()) {
+    return retriever.status();
+  }
+  return retriever.FetchFileChecksumList(*checksum_list);
 }

 Status UpdateManifestForFilesState(
@@ -16,6 +16,7 @@
 #include "test_util/testharness.h"
 #include "test_util/testutil.h"
 #include "util/defer.h"
+#include "util/file_checksum_helper.h"
 #include "util/random.h"
 #include "utilities/fault_injection_env.h"

@@ -260,55 +261,6 @@ TEST_F(ExternalSSTFileBasicTest, Basic) {
  s = sst_file_writer.DeleteRange(Key(100), Key(200));
  ASSERT_NOK(s) << s.ToString();

-  DestroyAndReopen(options);
-
-  SyncPoint::GetInstance()->LoadDependency({
-      {"DBImpl::IngestExternalFile:AfterIncIngestFileCounter",
-       "ExternalSSTFileBasicTest.LiveWriteStart"},
-      {"WriteThread::JoinBatchGroup:Wait",
-       "DBImpl::IngestExternalFile:AfterIncIngestFileCounter:2"},
-  });
-  SyncPoint::GetInstance()->EnableProcessing();
-  PerfContext* write_thread_perf_context;
-  std::thread write_thread([&] {
-    TEST_SYNC_POINT("ExternalSSTFileBasicTest.LiveWriteStart");
-    SetPerfLevel(kEnableWait);
-    write_thread_perf_context = get_perf_context();
-    write_thread_perf_context->Reset();
-    ASSERT_OK(db_->Put(WriteOptions(), "bar", "v2"));
-    ASSERT_GT(write_thread_perf_context->write_thread_wait_nanos, 0);
-    // Test sync points were used to make sure this live write enter write
-    // thread after the file ingestion entered write thread. So by the time this
-    // live write finishes, the latest seqno is 1 means file ingestion used
-    // seqno 0.
-    ASSERT_EQ(db_->GetLatestSequenceNumber(), 1U);
-  });
-
-  // Add file using file path
-  SetPerfLevel(kEnableTimeExceptForMutex);
-  PerfContext* perf_ctx = get_perf_context();
-  perf_ctx->Reset();
-  s = DeprecatedAddFile({file1});
-  ASSERT_GT(perf_context.file_ingestion_nanos, 0);
-  ASSERT_GT(perf_context.file_ingestion_blocking_live_writes_nanos, 0);
-  ASSERT_OK(s) << s.ToString();
-  for (int k = 0; k < 100; k++) {
-    ASSERT_EQ(Get(Key(k)), Key(k) + "_val");
-  }
-
-  write_thread.join();
-  SyncPoint::GetInstance()->DisableProcessing();
-
-  // Re-ingest the file just to check the perf context not enabled at and below
-  // kEnableWait.
-  SetPerfLevel(kEnableWait);
-  perf_ctx->Reset();
-  IngestExternalFileOptions opts;
-  opts.allow_global_seqno = true;
-  opts.allow_blocking_flush = true;
-  ASSERT_OK(db_->IngestExternalFile({file1}, opts));
-  ASSERT_EQ(perf_context.file_ingestion_nanos, 0);
-  ASSERT_EQ(perf_context.file_ingestion_blocking_live_writes_nanos, 0);
  DestroyAndRecreateExternalSSTFilesDir();
 }

@@ -395,7 +347,8 @@ class ChecksumVerifyHelper {

  Status GetSingleFileChecksumAndFuncName(
      const std::string& file_path, std::string* file_checksum,
-      std::string* file_checksum_func_name) {
+      std::string* file_checksum_func_name,
+      const std::string& requested_func_name = {}) {
    Status s;
    EnvOptions soptions;
    std::unique_ptr<SequentialFile> file_reader;
@@ -413,6 +366,8 @@ class ChecksumVerifyHelper {
      return Status::OK();
    } else {
      FileChecksumGenContext gen_context;
+      gen_context.file_name = file_path;
+      gen_context.requested_checksum_func_name = requested_func_name;
      std::unique_ptr<FileChecksumGenerator> file_checksum_gen =
          file_checksum_gen_factory->CreateFileChecksumGenerator(gen_context);
      *file_checksum_func_name = file_checksum_gen->Name();
@@ -488,10 +443,50 @@ TEST_F(ExternalSSTFileBasicTest, BasicWithFileChecksumCrc32c) {
  DestroyAndRecreateExternalSSTFilesDir();
 }

+namespace {
+class VariousFileChecksumGenerator : public FileChecksumGenCrc32c {
+ public:
+  explicit VariousFileChecksumGenerator(const std::string& name)
+      : FileChecksumGenCrc32c({}), name_(name) {}
+
+  const char* Name() const override { return name_.c_str(); }
+
+  std::string GetChecksum() const override {
+    return FileChecksumGenCrc32c::GetChecksum() + "_" + name_;
+  }
+
+ private:
+  const std::string name_;
+};
+
+class VariousFileChecksumGenFactory : public FileChecksumGenFactory {
+ public:
+  std::unique_ptr<FileChecksumGenerator> CreateFileChecksumGenerator(
+      const FileChecksumGenContext& context) override {
+    static RelaxedAtomic<int> counter{0};
+    if (Slice(context.requested_checksum_func_name).starts_with("Various")) {
+      return std::make_unique<VariousFileChecksumGenerator>(
+          context.requested_checksum_func_name);
+    } else if (context.requested_checksum_func_name.empty()) {
+      // Lacking a specific request, use a different function name for each
+      // result.
+      return std::make_unique<VariousFileChecksumGenerator>(
+          "Various" + std::to_string(counter.FetchAddRelaxed(1)));
+    } else {
+      return nullptr;
+    }
+  }
+
+  static const char* kClassName() { return "VariousFileChecksumGenFactory"; }
+  const char* Name() const override { return kClassName(); }
+};
+}  // namespace
+
 TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
  Options old_options = CurrentOptions();
  Options options = CurrentOptions();
-  options.file_checksum_gen_factory = GetFileChecksumGenCrc32cFactory();
+  options.file_checksum_gen_factory =
+      std::make_shared<VariousFileChecksumGenFactory>();
  const ImmutableCFOptions ioptions(options);
  ChecksumVerifyHelper checksum_helper(options);

@@ -512,7 +507,8 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
  ASSERT_EQ(file1_info.largest_key, Key(1099));
  std::string file_checksum1, file_checksum_func_name1;
  ASSERT_OK(checksum_helper.GetSingleFileChecksumAndFuncName(
-      file1, &file_checksum1, &file_checksum_func_name1));
+      file1, &file_checksum1, &file_checksum_func_name1,
+      file1_info.file_checksum_func_name));
  ASSERT_EQ(file1_info.file_checksum, file_checksum1);
  ASSERT_EQ(file1_info.file_checksum_func_name, file_checksum_func_name1);

@@ -531,7 +527,8 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
  ASSERT_EQ(file2_info.largest_key, Key(1299));
  std::string file_checksum2, file_checksum_func_name2;
  ASSERT_OK(checksum_helper.GetSingleFileChecksumAndFuncName(
-      file2, &file_checksum2, &file_checksum_func_name2));
+      file2, &file_checksum2, &file_checksum_func_name2,
+      file2_info.file_checksum_func_name));
  ASSERT_EQ(file2_info.file_checksum, file_checksum2);
  ASSERT_EQ(file2_info.file_checksum_func_name, file_checksum_func_name2);

@@ -550,7 +547,8 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
  ASSERT_EQ(file3_info.largest_key, Key(1499));
  std::string file_checksum3, file_checksum_func_name3;
  ASSERT_OK(checksum_helper.GetSingleFileChecksumAndFuncName(
-      file3, &file_checksum3, &file_checksum_func_name3));
+      file3, &file_checksum3, &file_checksum_func_name3,
+      file3_info.file_checksum_func_name));
  ASSERT_EQ(file3_info.file_checksum, file_checksum3);
  ASSERT_EQ(file3_info.file_checksum_func_name, file_checksum_func_name3);

@@ -569,7 +567,8 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
  ASSERT_EQ(file4_info.largest_key, Key(1799));
  std::string file_checksum4, file_checksum_func_name4;
  ASSERT_OK(checksum_helper.GetSingleFileChecksumAndFuncName(
-      file4, &file_checksum4, &file_checksum_func_name4));
+      file4, &file_checksum4, &file_checksum_func_name4,
+      file4_info.file_checksum_func_name));
  ASSERT_EQ(file4_info.file_checksum, file_checksum4);
  ASSERT_EQ(file4_info.file_checksum_func_name, file_checksum_func_name4);

@@ -588,7 +587,8 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
  ASSERT_EQ(file5_info.largest_key, Key(1999));
  std::string file_checksum5, file_checksum_func_name5;
  ASSERT_OK(checksum_helper.GetSingleFileChecksumAndFuncName(
-      file5, &file_checksum5, &file_checksum_func_name5));
+      file5, &file_checksum5, &file_checksum_func_name5,
+      file5_info.file_checksum_func_name));
  ASSERT_EQ(file5_info.file_checksum, file_checksum5);
  ASSERT_EQ(file5_info.file_checksum_func_name, file_checksum_func_name5);

@@ -607,7 +607,8 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
  ASSERT_EQ(file6_info.largest_key, Key(2199));
  std::string file_checksum6, file_checksum_func_name6;
  ASSERT_OK(checksum_helper.GetSingleFileChecksumAndFuncName(
-      file6, &file_checksum6, &file_checksum_func_name6));
+      file6, &file_checksum6, &file_checksum_func_name6,
+      file6_info.file_checksum_func_name));
  ASSERT_EQ(file6_info.file_checksum, file_checksum6);
  ASSERT_EQ(file6_info.file_checksum_func_name, file_checksum_func_name6);

@@ -677,18 +678,23 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
  }
  ASSERT_OK(env_->FileExists(file2));

-  // Enable verify_file_checksum option
-  // No checksum information is provided, generate it when ingesting
-  std::vector<std::string> checksum, checksum_func;
-  s = AddFileWithFileChecksum({file3}, checksum, checksum_func, true, false,
-                              false, false);
+  // Enable verify_file_checksum option. No checksum information is provided,
+  // so it is generated when ingesting. The configured checksum factory will
+  // use a different function than before.
+  s = AddFileWithFileChecksum({file3}, {}, {}, true, false, false, false);
  ASSERT_OK(s) << s.ToString();
  std::vector<LiveFileMetaData> live_files2;
  dbfull()->GetLiveFilesMetaData(&live_files2);
  for (const auto& f : live_files2) {
    if (set1.find(f.name) == set1.end()) {
-      ASSERT_EQ(f.file_checksum, file_checksum3);
-      ASSERT_EQ(f.file_checksum_func_name, file_checksum_func_name3);
+      // Recomputed checksum, different function
+      EXPECT_NE(f.file_checksum_func_name, file_checksum_func_name3);
+      std::string cur_checksum3, cur_checksum_func_name3;
+      ASSERT_OK(checksum_helper.GetSingleFileChecksumAndFuncName(
+          dbname_ + f.name, &cur_checksum3, &cur_checksum_func_name3,
+          f.file_checksum_func_name));
+      EXPECT_EQ(f.file_checksum, cur_checksum3);
+      EXPECT_EQ(f.file_checksum_func_name, cur_checksum_func_name3);
      set1.insert(f.name);
    }
  }
@@ -702,8 +708,9 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
  ASSERT_NOK(s) << s.ToString();

  // Does not enable verify_file_checksum options
-  // Checksum function name matches, store the checksum being ingested.
-  s = AddFileWithFileChecksum({file4}, {"asd"}, {file_checksum_func_name4},
+  // Checksum function name is recognized, so store the checksum being ingested.
+  std::string file_checksum_func_name4alt = "VariousABCD";
+  s = AddFileWithFileChecksum({file4}, {"asd"}, {file_checksum_func_name4alt},
                              false, false, false, false);
  ASSERT_OK(s) << s.ToString();
  std::vector<LiveFileMetaData> live_files3;
@@ -712,7 +719,7 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
    if (set1.find(f.name) == set1.end()) {
      ASSERT_FALSE(f.file_checksum == file_checksum4);
      ASSERT_EQ(f.file_checksum, "asd");
-      ASSERT_EQ(f.file_checksum_func_name, file_checksum_func_name4);
+      ASSERT_EQ(f.file_checksum_func_name, file_checksum_func_name4alt);
      set1.insert(f.name);
    }
  }
@@ -721,7 +728,8 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {

  // enable verify_file_checksum options, DB enable checksum, and enable
  // write_global_seq. So the checksum stored is different from the one
-  // ingested due to the sequence number changes.
+  // ingested due to the sequence number changes. The checksum function name
+  // may also change since the checksum is recomputed.
  s = AddFileWithFileChecksum({file5}, {file_checksum5},
                              {file_checksum_func_name5}, true, false, false,
                              true);
@@ -730,11 +738,14 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
  dbfull()->GetLiveFilesMetaData(&live_files4);
  for (const auto& f : live_files4) {
    if (set1.find(f.name) == set1.end()) {
+      // Recomputed checksum, different function
+      EXPECT_NE(f.file_checksum_func_name, file_checksum_func_name5);
      std::string cur_checksum5, cur_checksum_func_name5;
      ASSERT_OK(checksum_helper.GetSingleFileChecksumAndFuncName(
-          dbname_ + f.name, &cur_checksum5, &cur_checksum_func_name5));
-      ASSERT_EQ(f.file_checksum, cur_checksum5);
-      ASSERT_EQ(f.file_checksum_func_name, file_checksum_func_name5);
+          dbname_ + f.name, &cur_checksum5, &cur_checksum_func_name5,
+          f.file_checksum_func_name));
+      EXPECT_EQ(f.file_checksum, cur_checksum5);
+      EXPECT_EQ(f.file_checksum_func_name, cur_checksum_func_name5);
      set1.insert(f.name);
    }
  }
@@ -742,18 +753,22 @@ TEST_F(ExternalSSTFileBasicTest, IngestFileWithFileChecksum) {
  ASSERT_OK(env_->FileExists(file5));

  // Does not enable verify_file_checksum options and also the ingested file
-  // checksum information is empty. DB will generate and store the checksum
-  // in Manifest.
-  std::vector<std::string> files_c6, files_name6;
-  s = AddFileWithFileChecksum({file6}, files_c6, files_name6, false, false,
-                              false, false);
+  // checksum information is empty. DB will generate and store file checksum
+  // in Manifest, which could be different from the previous invocation.
+  s = AddFileWithFileChecksum({file6}, {}, {}, false, false, false, false);
  ASSERT_OK(s) << s.ToString();
  std::vector<LiveFileMetaData> live_files6;
  dbfull()->GetLiveFilesMetaData(&live_files6);
  for (const auto& f : live_files6) {
    if (set1.find(f.name) == set1.end()) {
-      ASSERT_EQ(f.file_checksum, file_checksum6);
-      ASSERT_EQ(f.file_checksum_func_name, file_checksum_func_name6);
+      // Recomputed checksum, different function
+      EXPECT_NE(f.file_checksum_func_name, file_checksum_func_name6);
+      std::string cur_checksum6, cur_checksum_func_name6;
+      ASSERT_OK(checksum_helper.GetSingleFileChecksumAndFuncName(
+          dbname_ + f.name, &cur_checksum6, &cur_checksum_func_name6,
+          f.file_checksum_func_name));
+      EXPECT_EQ(f.file_checksum, cur_checksum6);
+      EXPECT_EQ(f.file_checksum_func_name, cur_checksum_func_name6);
      set1.insert(f.name);
    }
  }
@@ -2552,7 +2567,14 @@ TEST_F(ExternalSSTFileBasicTest, IngestWithTemperature) {
    options.default_write_temperature = Temperature::kHot;
    SstFileWriter sst_file_writer(EnvOptions(), options);
    options.level0_file_num_compaction_trigger = 2;
-    options.allow_ingest_behind = (mode == "ingest_behind");
+    bool cf_option = Random::GetTLSInstance()->OneIn(2);
+    SCOPED_TRACE(std::string("Use ") + (cf_option ? "CF" : "DB") +
+                 " option for ingest behind");
+    if (cf_option) {
+      options.cf_allow_ingest_behind = (mode == "ingest_behind");
+    } else {
+      options.allow_ingest_behind = (mode == "ingest_behind");
+    }
    Reopen(options);
    Defer destroyer([&]() { Destroy(options); });

@@ -2693,6 +2715,81 @@ TEST_F(ExternalSSTFileBasicTest, IngestWithTemperature) {
  }
 }

+// This tests an internal user's exact usage and expectation of the
+// IngestExternalFiles APIs to bulk load and replace files.
+TEST_F(ExternalSSTFileBasicTest,
+       AtomicReplaceColumnFamilyWithIngestedVersionKey) {
+  Options options = GetDefaultOptions();
+  options.create_if_missing = true;
+  options.compaction_style = CompactionStyle::kCompactionStyleUniversal;
+  options.num_levels = 7;
+  options.disallow_memtable_writes = false;
+
+  DestroyAndReopen(options);
+  SstFileWriter sst_file_writer(EnvOptions(), options);
+  std::string data_file_original = sst_files_dir_ + "data_original";
+  ASSERT_OK(sst_file_writer.Open(data_file_original));
+  ASSERT_OK(sst_file_writer.Put("ukey1", "uval1_orig"));
+  ASSERT_OK(sst_file_writer.Put("ukey2", "uval2_orig"));
+  ASSERT_OK(sst_file_writer.Finish());
+  ASSERT_OK(db_->IngestExternalFile(db_->DefaultColumnFamily(),
+                                    {data_file_original},
+                                    IngestExternalFileOptions()));
+
+  ASSERT_OK(Put("data_version", "v_original"));
+  ASSERT_OK(Flush());
+  std::string value;
+  ASSERT_OK(db_->Get(ReadOptions(), "data_version", &value));
+  ASSERT_EQ(value, "v_original");
+  ASSERT_OK(db_->Get(ReadOptions(), "ukey1", &value));
+  ASSERT_EQ(value, "uval1_orig");
+  ASSERT_OK(db_->Get(ReadOptions(), "ukey2", &value));
+  ASSERT_EQ(value, "uval2_orig");
+  // Set up a 1) data version key file on L0, and 2) a user data file on L6
+  // to test the initial transitioning to use `atomic_replace_range`.
+  ASSERT_EQ("1,0,0,0,0,0,1", FilesPerLevel());
+
+  // Test multiple cycles of replacing by atomically ingest a data file and a
+  // version key file while replace the whole range in the column family.
+  for (int i = 0; i < 10; i++) {
+    std::string version_file_path =
+        sst_files_dir_ + "version" + std::to_string(i);
+    ASSERT_OK(sst_file_writer.Open(version_file_path));
+    ASSERT_OK(sst_file_writer.Put("data_version", "v" + std::to_string(i)));
+    ASSERT_OK(sst_file_writer.Finish());
+
+    std::string file_path = sst_files_dir_ + std::to_string(i);
+    ASSERT_OK(sst_file_writer.Open(file_path));
+    ASSERT_OK(sst_file_writer.Put("ukey1", "uval1" + std::to_string(i)));
+    ASSERT_OK(sst_file_writer.Put("ukey2", "uval2" + std::to_string(i)));
+    ASSERT_OK(sst_file_writer.Finish());
+
+    IngestExternalFileArg arg;
+    arg.column_family = db_->DefaultColumnFamily();
+    arg.external_files = {version_file_path, file_path};
+    arg.atomic_replace_range = {{nullptr, nullptr}};
+    // Test both fail_if_not_bottomost_level: true and false
+    arg.options.fail_if_not_bottommost_level = i % 2 == 0;
+    arg.options.snapshot_consistency = false;
+    // Ingest 1) a new data version file and 2) a new user data file while erase
+    // the whole column family
+    Status s = db_->IngestExternalFiles({arg});
+    ASSERT_OK(s);
+
+    // Check ingestion result and the expected LSM shape:
+    // Two files on L6, 1) a data version file 2) a user data file.
+    ASSERT_OK(db_->Get(ReadOptions(), "ukey1", &value));
+    ASSERT_EQ(value, "uval1" + std::to_string(i));
+    ASSERT_OK(db_->Get(ReadOptions(), "ukey2", &value));
+    ASSERT_EQ(value, "uval2" + std::to_string(i));
+    ASSERT_OK(db_->Get(ReadOptions(), "data_version", &value));
+    ASSERT_EQ(value, "v" + std::to_string(i));
+    ASSERT_EQ("0,0,0,0,0,0,2", FilesPerLevel());
+  }
+
+  Close();
+}
+
 TEST_F(ExternalSSTFileBasicTest, FailIfNotBottommostLevelAndDisallowMemtable) {
  for (bool disallow_memtable : {false, true}) {
    Options options = GetDefaultOptions();
@@ -122,24 +122,28 @@ Status ExternalSstFileIngestionJob::Prepare(
    }
  }

-  if (ingestion_options_.ingest_behind && files_overlap_) {
-    return Status::NotSupported(
-        "Files with overlapping ranges cannot be ingested with ingestion "
-        "behind mode.");
-  }
+  if (files_overlap_) {
+    if (ingestion_options_.ingest_behind) {
+      return Status::NotSupported(
+          "Files with overlapping ranges cannot be ingested with ingestion "
+          "behind mode.");
+    }

-  // Overlapping files need at least two different sequence numbers. If settings
-  // disables global seqno, ingestion will fail anyway, so fail fast in prepare.
-  if (!ingestion_options_.allow_global_seqno && files_overlap_) {
-    return Status::InvalidArgument(
-        "Global seqno is required, but disabled (because external files key "
-        "range overlaps).");
-  }
+    // Overlapping files need at least two different sequence numbers. If
+    // settings disables global seqno, ingestion will fail anyway, so fail
+    // fast in prepare.
+    if (!ingestion_options_.allow_global_seqno &&
+        !ingestion_options_.allow_db_generated_files) {
+      return Status::InvalidArgument(
+          "Global seqno is required, but disabled (because external files key "
+          "range overlaps).");
+    }

-  if (ucmp_->timestamp_size() > 0 && files_overlap_) {
-    return Status::NotSupported(
-        "Files with overlapping ranges cannot be ingested to column "
-        "family with user-defined timestamp enabled.");
+    if (ucmp_->timestamp_size() > 0) {
+      return Status::NotSupported(
+          "Files with overlapping ranges cannot be ingested to column "
+          "family with user-defined timestamp enabled.");
+    }
  }

  // Copy/Move external files into DB
@@ -156,6 +160,14 @@ Status ExternalSstFileIngestionJob::Prepare(
        // It is unsafe to assume application had sync the file and file
        // directory before ingest the file. For integrity of RocksDB we need
        // to sync the file.
+
+        // TODO(xingbo), We should in general be moving away from production
+        // uses of ReuseWritableFile (except explicitly for WAL recycling),
+        // ReopenWritableFile, and NewRandomRWFile. We should create a
+        // FileSystem::SyncFile/FsyncFile API that by default does the
+        // re-open+sync+close combo but can (a) be reused easily, and (b) be
+        // overridden to do that more cleanly, e.g. in EncryptedEnv.
+        // https://github.com/facebook/rocksdb/issues/13741
        std::unique_ptr<FSWritableFile> file_to_sync;
        Status s = fs_->ReopenWritableFile(path_inside_db, env_options_,
                                           &file_to_sync, nullptr);
@@ -260,10 +272,6 @@ Status ExternalSstFileIngestionJob::Prepare(
    } else {
      need_generate_file_checksum_ = true;
    }
-    FileChecksumGenContext gen_context;
-    std::unique_ptr<FileChecksumGenerator> file_checksum_gen =
-        db_options_.file_checksum_gen_factory->CreateFileChecksumGenerator(
-            gen_context);
    std::vector<std::string> generated_checksums;
    std::vector<std::string> generated_checksum_func_names;
    // Step 1: generate the checksum for ingested sst file.
@@ -271,7 +279,9 @@ Status ExternalSstFileIngestionJob::Prepare(
      for (size_t i = 0; i < files_to_ingest_.size(); i++) {
        std::string generated_checksum;
        std::string generated_checksum_func_name;
-        std::string requested_checksum_func_name;
+        std::string requested_checksum_func_name =
+            i < files_checksum_func_names.size() ? files_checksum_func_names[i]
+                                                 : "";
        // TODO: rate limit file reads for checksum calculation during file
        // ingestion.
        // TODO: plumb Env::IOActivity
@@ -314,40 +324,50 @@ Status ExternalSstFileIngestionJob::Prepare(
            if (files_checksum_func_names[i] !=
                generated_checksum_func_names[i]) {
              status = Status::InvalidArgument(
-                  "Checksum function name does not match with the checksum "
-                  "function name of this DB");
-              ROCKS_LOG_WARN(
-                  db_options_.info_log,
-                  "Sst file checksum verification of file: %s failed: %s",
-                  external_files_paths[i].c_str(), status.ToString().c_str());
+                  "DB file checksum gen factory " +
+                  std::string(db_options_.file_checksum_gen_factory->Name()) +
+                  " generated checksum function name " +
+                  generated_checksum_func_names[i] + " for file " +
+                  external_files_paths[i] +
+                  " which does not match requested/provided " +
+                  files_checksum_func_names[i]);
              break;
            }
            if (files_checksums[i] != generated_checksums[i]) {
              status = Status::Corruption(
-                  "Ingested checksum does not match with the generated "
-                  "checksum");
-              ROCKS_LOG_WARN(
-                  db_options_.info_log,
-                  "Sst file checksum verification of file: %s failed: %s",
-                  files_to_ingest_[i].internal_file_path.c_str(),
-                  status.ToString().c_str());
+                  "Checksum verification mismatch for ingestion file " +
+                  external_files_paths[i] + " using function " +
+                  generated_checksum_func_names[i] + ". Expected: " +
+                  Slice(files_checksums[i]).ToString(/*hex=*/true) +
+                  " Computed: " +
+                  Slice(generated_checksums[i]).ToString(/*hex=*/true));
              break;
            }
          }
        } else {
-          // If verify_file_checksum is not enabled, we only verify the
-          // checksum function name. If it does not match, fail the ingestion.
-          // If matches, we trust the ingested checksum information and store
-          // in the Manifest.
+          // If verify_file_checksum is not enabled, we only verify the factory
+          // recognizes the checksum function name. If it does not match, fail
+          // the ingestion. If matches, we trust the ingested checksum
+          // information and store in the Manifest.
          for (size_t i = 0; i < files_to_ingest_.size(); i++) {
-            if (files_checksum_func_names[i] != file_checksum_gen->Name()) {
+            FileChecksumGenContext gen_context;
+            gen_context.file_name = files_to_ingest_[i].internal_file_path;
+            gen_context.requested_checksum_func_name =
+                files_checksum_func_names[i];
+            auto file_checksum_gen =
+                db_options_.file_checksum_gen_factory
+                    ->CreateFileChecksumGenerator(gen_context);
+
+            if (file_checksum_gen == nullptr ||
+                files_checksum_func_names[i] != file_checksum_gen->Name()) {
              status = Status::InvalidArgument(
-                  "Checksum function name does not match with the checksum "
-                  "function name of this DB");
-              ROCKS_LOG_WARN(
-                  db_options_.info_log,
-                  "Sst file checksum verification of file: %s failed: %s",
-                  external_files_paths[i].c_str(), status.ToString().c_str());
+                  "Checksum function name " + files_checksum_func_names[i] +
+                  " for file " + external_files_paths[i] +
+                  " not recognized by DB checksum gen factory" +
+                  db_options_.file_checksum_gen_factory->Name() +
+                  (file_checksum_gen ? (" Returned function " +
+                                        std::string(file_checksum_gen->Name()))
+                                     : ""));
              break;
            }
            files_to_ingest_[i].file_checksum = files_checksums[i];
@@ -362,12 +382,11 @@ Status ExternalSstFileIngestionJob::Prepare(
        status = Status::InvalidArgument(
            "The checksum information of ingested sst files are nonempty and "
            "the size of checksums or the size of the checksum function "
-            "names "
-            "does not match with the number of ingested sst files");
-        ROCKS_LOG_WARN(
-            db_options_.info_log,
-            "The ingested sst files checksum information is incomplete: %s",
-            status.ToString().c_str());
+            "names does not match with the number of ingested sst files");
+      }
+      if (!status.ok()) {
+        ROCKS_LOG_WARN(db_options_.info_log, "Ingestion failed: %s",
+                       status.ToString().c_str());
      }
    }
  }
@@ -525,6 +544,8 @@ Status ExternalSstFileIngestionJob::Run() {

  // Find levels to ingest into
  std::optional<int> prev_batch_uppermost_level;
+  // batches at the front of file_batches_to_ingest_ contains older updates and
+  // are placed in smaller levels.
  for (auto& batch : file_batches_to_ingest_) {
    int batch_uppermost_level = 0;
    status = AssignLevelsForOneBatch(batch, super_version, force_global_seqno,
@@ -575,6 +596,15 @@ Status ExternalSstFileIngestionJob::AssignLevelsForOneBatch(
    if (!status.ok()) {
      return status;
    }
+
+    // If any ingested file overlaps with the DB, it will fail here.
+    if (ingestion_options_.allow_db_generated_files && assigned_seqno != 0) {
+      return Status::InvalidArgument(
+          "An ingested file overlaps with existing data in the DB and has been "
+          "assigned a non-zero sequence number, which is not allowed when "
+          "'allow_db_generated_files' is enabled.");
+    }
+
    if (smallest_parsed.sequence == 0 && assigned_seqno != 0) {
      UpdateInternalKey(file->smallest_internal_key.rep(), assigned_seqno,
                        smallest_parsed.type);
@@ -593,8 +623,8 @@ Status ExternalSstFileIngestionJob::AssignLevelsForOneBatch(
    assert(assigned_seqno == 0 || assigned_seqno == *last_seqno + 1);
    if (assigned_seqno > *last_seqno) {
      *last_seqno = assigned_seqno;
-      ++consumed_seqno_count_;
    }
+    max_assigned_seqno_ = std::max(max_assigned_seqno_, assigned_seqno);

    status = GenerateChecksumForIngestedFile(file);
    if (!status.ok()) {
@@ -610,30 +640,31 @@ Status ExternalSstFileIngestionJob::AssignLevelsForOneBatch(
      current_time = oldest_ancester_time =
          static_cast<uint64_t>(temp_current_time);
    }
-    uint64_t tail_size = 0;
-    bool contain_no_data_blocks = file->table_properties.num_entries > 0 &&
-                                  (file->table_properties.num_entries ==
-                                   file->table_properties.num_range_deletions);
-    if (file->table_properties.tail_start_offset > 0 ||
-        contain_no_data_blocks) {
-      uint64_t file_size = file->fd.GetFileSize();
-      assert(file->table_properties.tail_start_offset <= file_size);
-      tail_size = file_size - file->table_properties.tail_start_offset;
-    }
+    uint64_t tail_size = FileMetaData::CalculateTailSize(
+        file->fd.GetFileSize(), file->table_properties);

    bool marked_for_compaction =
        file->table_properties.num_range_deletions == 1 &&
        (file->table_properties.num_entries ==
         file->table_properties.num_range_deletions);
+    SequenceNumber smallest_seqno = file->assigned_seqno;
+    SequenceNumber largest_seqno = file->assigned_seqno;
+    if (ingestion_options_.allow_db_generated_files) {
+      assert(file->assigned_seqno == 0);
+      assert(file->smallest_seqno != kMaxSequenceNumber);
+      assert(file->largest_seqno != kMaxSequenceNumber);
+      smallest_seqno = file->smallest_seqno;
+      largest_seqno = file->largest_seqno;
+      max_assigned_seqno_ = std::max(max_assigned_seqno_, file->largest_seqno);
+    }
    FileMetaData f_metadata(
        file->fd.GetNumber(), file->fd.GetPathId(), file->fd.GetFileSize(),
-        file->smallest_internal_key, file->largest_internal_key,
-        file->assigned_seqno, file->assigned_seqno, false,
-        file->file_temperature, kInvalidBlobFileNumber, oldest_ancester_time,
-        current_time,
+        file->smallest_internal_key, file->largest_internal_key, smallest_seqno,
+        largest_seqno, false, file->file_temperature, kInvalidBlobFileNumber,
+        oldest_ancester_time, current_time,
        ingestion_options_.ingest_behind
            ? kReservedEpochNumberForFileIngestedBehind
-            : cfd_->NewEpochNumber(),
+            : cfd_->NewEpochNumber(),  // orders files ingested to L0
        file->file_checksum, file->file_checksum_func_name, file->unique_id, 0,
        tail_size, file->user_defined_timestamps_persisted);
    f_metadata.temperature = file->file_temperature;
@@ -690,10 +721,9 @@ void ExternalSstFileIngestionJob::CreateEquivalentFileIngestingCompactions() {
        0 /* max_subcompaction, not applicable */,
        {} /* grandparents, not applicable */,
        std::nullopt /* earliest_snapshot */, nullptr /* snapshot_checker */,
-        false /* is manual */, "" /* trim_ts */, -1 /* score, not applicable */,
-        false /* is deletion compaction, not applicable */,
-        files_overlap_ /* l0_files_might_overlap, not applicable */,
-        CompactionReason::kExternalSstIngestion));
+        CompactionReason::kExternalSstIngestion, "" /* trim_ts */,
+        -1 /* score, not applicable */,
+        files_overlap_ /* l0_files_might_overlap, not applicable */));
  }
 }

@@ -781,7 +811,6 @@ void ExternalSstFileIngestionJob::Cleanup(const Status& status) {
    // We failed to add the files to the database
    // remove all the files we copied
    DeleteInternalFiles();
-    consumed_seqno_count_ = 0;
    files_overlap_ = false;
  } else if (status.ok() && ingestion_options_.move_files) {
    // The files were moved and added successfully, remove original file links
@@ -842,7 +871,8 @@ Status ExternalSstFileIngestionJob::ResetTableReader(
      ro,
      TableReaderOptions(
          cfd_->ioptions(), sv->mutable_cf_options.prefix_extractor,
-          env_options_, cfd_->internal_comparator(),
+          sv->mutable_cf_options.compression_manager.get(), env_options_,
+          cfd_->internal_comparator(),
          sv->mutable_cf_options.block_protection_bytes_per_key,
          /*skip_filters*/ false, /*immortal*/ false,
          /*force_direct_prefetch*/ false, /*level*/ -1,
@@ -994,6 +1024,32 @@ Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
    return status;
  }

+  const bool allow_data_in_errors = db_options_.allow_data_in_errors;
+  ParsedInternalKey key;
+  if (ingestion_options_.allow_db_generated_files) {
+    // We are ingesting a DB generated SST file for which we don't reassign
+    // sequence numbers. We need its smallest sequence number and largest
+    // sequence number for FileMetaData.
+    Status seqno_status = GetSeqnoBoundaryForFile(
+        table_reader.get(), sv, file_to_ingest, allow_data_in_errors);
+
+    if (!seqno_status.ok()) {
+      return seqno_status;
+    }
+    assert(file_to_ingest->smallest_seqno <= file_to_ingest->largest_seqno);
+    assert(file_to_ingest->largest_seqno < kMaxSequenceNumber);
+  } else {
+    SequenceNumber largest_seqno =
+        table_reader.get()->GetTableProperties()->key_largest_seqno;
+    // UINT64_MAX means unknown and the file is generated before table property
+    // `key_largest_seqno` is introduced.
+    if (largest_seqno != UINT64_MAX && largest_seqno > 0) {
+      return Status::Corruption(
+          "External file has non zero largest sequence number " +
+          std::to_string(largest_seqno));
+    }
+  }
+
  if (ingestion_options_.verify_checksums_before_ingest) {
    // If customized readahead size is needed, we can pass a user option
    // all the way to here. Right now we just rely on the default readahead
@@ -1009,7 +1065,6 @@ Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
    }
  }

-  ParsedInternalKey key;
  // TODO: plumb Env::IOActivity, Env::IOPriority
  ReadOptions ro;
  ro.fill_cache = ingestion_options_.fill_cache;
@@ -1018,7 +1073,6 @@ Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
      /*skip_filters=*/false, TableReaderCaller::kExternalSSTIngestion));

  // Get first (smallest) and last (largest) key from file.
-  bool allow_data_in_errors = db_options_.allow_data_in_errors;
  iter->SeekToFirst();
  if (iter->Valid()) {
    Status pik_status =
@@ -1027,7 +1081,7 @@ Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
      return Status::Corruption("Corrupted key in external file. ",
                                pik_status.getState());
    }
-    if (key.sequence != 0) {
+    if (key.sequence != 0 && !ingestion_options_.allow_db_generated_files) {
      return Status::Corruption("External file has non zero sequence number");
    }
    file_to_ingest->smallest_internal_key.SetFrom(key);
@@ -1064,41 +1118,13 @@ Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
      return Status::Corruption("Corrupted key in external file. ",
                                pik_status.getState());
    }
-    if (key.sequence != 0) {
+    if (key.sequence != 0 && !ingestion_options_.allow_db_generated_files) {
      return Status::Corruption("External file has non zero sequence number");
    }
    file_to_ingest->largest_internal_key.SetFrom(key);
  } else if (!iter->status().ok()) {
    return iter->status();
  }
-  SequenceNumber largest_seqno =
-      table_reader.get()->GetTableProperties()->key_largest_seqno;
-  // UINT64_MAX means unknown and the file is generated before table property
-  // `key_largest_seqno` is introduced.
-  if (largest_seqno != UINT64_MAX && largest_seqno > 0) {
-    return Status::Corruption(
-        "External file has non zero largest sequence number " +
-        std::to_string(largest_seqno));
-  }
-  if (ingestion_options_.allow_db_generated_files &&
-      largest_seqno == UINT64_MAX) {
-    // Need to verify that all keys have seqno zero.
-    for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
-      Status pik_status =
-          ParseInternalKey(iter->key(), &key, allow_data_in_errors);
-      if (!pik_status.ok()) {
-        return Status::Corruption("Corrupted key in external file. ",
-                                  pik_status.getState());
-      }
-      if (key.sequence != 0) {
-        return Status::NotSupported(
-            "External file has a key with non zero sequence number.");
-      }
-    }
-    if (!iter->status().ok()) {
-      return iter->status();
-    }
-  }

  std::unique_ptr<InternalIterator> range_del_iter(
      table_reader->NewRangeTombstoneIterator(ro));
@@ -1113,7 +1139,7 @@ Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
        return Status::Corruption("Corrupted key in external file. ",
                                  pik_status.getState());
      }
-      if (key.sequence != 0) {
+      if (key.sequence != 0 && !ingestion_options_.allow_db_generated_files) {
        return Status::Corruption(
            "External file has a range deletion with non zero sequence "
            "number.");
@@ -1161,12 +1187,14 @@ Status ExternalSstFileIngestionJob::AssignLevelAndSeqnoForIngestedFile(
  const size_t ts_sz = ucmp_->timestamp_size();
  assert(!prev_batch_uppermost_level.has_value() ||
         prev_batch_uppermost_level.value() < cfd_->NumberLevels());
-  bool must_assign_to_l0 = prev_batch_uppermost_level.has_value() &&
-                           prev_batch_uppermost_level.value() == 0;
-  if (force_global_seqno || files_overlap_ ||
-      compaction_style == kCompactionStyleFIFO || must_assign_to_l0) {
+  bool must_assign_to_l0 = (prev_batch_uppermost_level.has_value() &&
+                            prev_batch_uppermost_level.value() == 0) ||
+                           compaction_style == kCompactionStyleFIFO;
+
+  if (force_global_seqno || (!ingestion_options_.allow_db_generated_files &&
+                             (files_overlap_ || must_assign_to_l0))) {
    *assigned_seqno = last_seqno + 1;
-    if (compaction_style == kCompactionStyleFIFO || must_assign_to_l0) {
+    if (must_assign_to_l0) {
      assert(ts_sz == 0);
      file_to_ingest->picked_level = 0;
      if (ingestion_options_.fail_if_not_bottommost_level &&
@@ -1187,16 +1215,26 @@ Status ExternalSstFileIngestionJob::AssignLevelAndSeqnoForIngestedFile(
  ro.total_order_seek = true;
  int target_level = 0;
  auto* vstorage = cfd_->current()->storage_info();
-  assert(!must_assign_to_l0);
-  int exclusive_end_level = prev_batch_uppermost_level.has_value()
-                                ? prev_batch_uppermost_level.value()
-                                : cfd_->NumberLevels();
+  assert(!must_assign_to_l0 || ingestion_options_.allow_db_generated_files);
+  int assigned_level_exclusive_end = cfd_->NumberLevels();
+  if (must_assign_to_l0) {
+    assigned_level_exclusive_end = 0;
+  } else if (prev_batch_uppermost_level.has_value()) {
+    assigned_level_exclusive_end = prev_batch_uppermost_level.value();
+  }

-  for (int lvl = 0; lvl < exclusive_end_level; lvl++) {
+  // When ingesting db generated files, we require that ingested files do not
+  // overlap with any file in the DB. So we need to check all levels.
+  int overlap_checking_exclusive_end =
+      ingestion_options_.allow_db_generated_files
+          ? cfd_->NumberLevels()
+          : assigned_level_exclusive_end;
+  for (int lvl = 0; lvl < overlap_checking_exclusive_end; lvl++) {
    if (lvl > 0 && lvl < vstorage->base_level()) {
      continue;
    }
-    if (atomic_replace_range_.has_value()) {
+    if (lvl < assigned_level_exclusive_end &&
+        atomic_replace_range_.has_value()) {
      target_level = lvl;
      continue;
    }
@@ -1227,7 +1265,8 @@ Status ExternalSstFileIngestionJob::AssignLevelAndSeqnoForIngestedFile(

    // We don't overlap with any keys in this level, but we still need to check
    // if our file can fit in it
-    if (IngestedFileFitInLevel(file_to_ingest, lvl)) {
+    if (lvl < assigned_level_exclusive_end &&
+        IngestedFileFitInLevel(file_to_ingest, lvl)) {
      target_level = lvl;
    }
  }
@@ -1236,8 +1275,9 @@ Status ExternalSstFileIngestionJob::AssignLevelAndSeqnoForIngestedFile(
      target_level < cfd_->NumberLevels() - 1) {
    status = Status::TryAgain(
        "Files cannot be ingested to Lmax. Please make sure key range of Lmax "
-        "and ongoing compaction's output to Lmax"
-        "does not overlap with files to ingest.");
+        "and ongoing compaction's output to Lmax does not overlap with files "
+        "to ingest. Input files overlapping with each other can cause some "
+        "file to be assigned to non Lmax level.");
    return status;
  }

@@ -1258,11 +1298,6 @@ Status ExternalSstFileIngestionJob::AssignLevelAndSeqnoForIngestedFile(
    }
  }

-  if (ingestion_options_.allow_db_generated_files && *assigned_seqno != 0) {
-    return Status::InvalidArgument(
-        "An ingested file is assigned to a non-zero sequence number, which is "
-        "incompatible with ingestion option allow_db_generated_files.");
-  }
  return status;
 }

@@ -1279,13 +1314,13 @@ Status ExternalSstFileIngestionJob::CheckLevelForIngestedBehindFile(
        "at the last level!");
  }

-  // Second, check if despite allow_ingest_behind=true we still have 0 seqnums
-  // at some upper level
+  // Second, check if despite cf_allow_ingest_behind=true we still have 0
+  // seqnums at some upper level
  for (int lvl = 0; lvl < cfd_->NumberLevels() - 1; lvl++) {
    for (auto file : vstorage->LevelFiles(lvl)) {
      if (file->fd.smallest_seqno == 0) {
        return Status::InvalidArgument(
-            "Can't ingest_behind file as despite allow_ingest_behind=true "
+            "Can't ingest_behind file as despite cf_allow_ingest_behind=true "
            "there are files with 0 seqno in database at upper levels!");
      }
    }
@@ -1297,8 +1332,12 @@ Status ExternalSstFileIngestionJob::CheckLevelForIngestedBehindFile(

 Status ExternalSstFileIngestionJob::AssignGlobalSeqnoForIngestedFile(
    IngestedFileInfo* file_to_ingest, SequenceNumber seqno) {
+  if (ingestion_options_.allow_db_generated_files) {
+    assert(seqno == 0);
+    assert(file_to_ingest->original_seqno == 0);
+  }
  if (file_to_ingest->original_seqno == seqno) {
-    // This file already have the correct global seqno
+    // This file already has the correct global seqno.
    return Status::OK();
  } else if (!ingestion_options_.allow_global_seqno) {
    return Status::InvalidArgument("Global seqno is required, but disabled");
@@ -1412,4 +1451,86 @@ Status ExternalSstFileIngestionJob::SyncIngestedFile(TWritableFile* file) {
  }
 }

+Status ExternalSstFileIngestionJob::GetSeqnoBoundaryForFile(
+    TableReader* table_reader, SuperVersion* sv,
+    IngestedFileInfo* file_to_ingest, bool allow_data_in_errors) {
+  const bool has_largest_seqno =
+      table_reader->GetTableProperties()->HasKeyLargestSeqno();
+  SequenceNumber largest_seqno =
+      table_reader->GetTableProperties()->key_largest_seqno;
+  if (has_largest_seqno && largest_seqno == 0) {
+    file_to_ingest->largest_seqno = 0;
+    file_to_ingest->smallest_seqno = 0;
+    return Status::OK();
+  }
+  // The following file scan is only executed when ingesting files with
+  // non-zero seqno.
+  // TODO: record smallest_seqno in table properties to avoid the
+  // file scan here.
+  SequenceNumber smallest_seqno = kMaxSequenceNumber;
+
+  SequenceNumber largest_seqno_from_iter = 0;
+  ReadOptions ro;
+  ro.fill_cache = ingestion_options_.fill_cache;
+  std::unique_ptr<InternalIterator> iter(table_reader->NewIterator(
+      ro, sv->mutable_cf_options.prefix_extractor.get(), /*arena=*/nullptr,
+      /*skip_filters=*/false, TableReaderCaller::kExternalSSTIngestion));
+  ParsedInternalKey key;
+  iter->SeekToFirst();
+  while (iter->Valid()) {
+    Status pik_status =
+        ParseInternalKey(iter->key(), &key, allow_data_in_errors);
+    if (!pik_status.ok()) {
+      return Status::Corruption("Corrupted key in external file. ",
+                                pik_status.getState());
+    }
+    smallest_seqno = std::min(smallest_seqno, key.sequence);
+    largest_seqno_from_iter = std::max(largest_seqno_from_iter, key.sequence);
+    iter->Next();
+  }
+  if (!iter->status().ok()) {
+    return iter->status();
+  }
+
+  if (table_reader->GetTableProperties()->num_range_deletions > 0) {
+    std::unique_ptr<InternalIterator> range_del_iter(
+        table_reader->NewRangeTombstoneIterator(ro));
+    if (range_del_iter != nullptr) {
+      for (range_del_iter->SeekToFirst(); range_del_iter->Valid();
+           range_del_iter->Next()) {
+        Status pik_status =
+            ParseInternalKey(range_del_iter->key(), &key, allow_data_in_errors);
+        if (!pik_status.ok()) {
+          return Status::Corruption("Corrupted key in external file. ",
+                                    pik_status.getState());
+        }
+        smallest_seqno = std::min(smallest_seqno, key.sequence);
+        largest_seqno_from_iter =
+            std::max(largest_seqno_from_iter, key.sequence);
+      }
+      if (!range_del_iter->status().ok()) {
+        return range_del_iter->status();
+      }
+    }
+  }
+
+  file_to_ingest->smallest_seqno = smallest_seqno;
+  if (!has_largest_seqno) {
+    file_to_ingest->largest_seqno = largest_seqno_from_iter;
+  } else {
+    assert(largest_seqno == largest_seqno_from_iter);
+    file_to_ingest->largest_seqno = largest_seqno;
+  }
+
+  if (file_to_ingest->largest_seqno == kMaxSequenceNumber) {
+    return Status::InvalidArgument(
+        "Unknown smallest seqno for db generated file.");
+  }
+  if (file_to_ingest->smallest_seqno == kMaxSequenceNumber) {
+    return Status::InvalidArgument(
+        "Unknown largest seqno for db generated file.");
+  }
+  return Status::OK();
+}
+
 }  // namespace ROCKSDB_NAMESPACE
@@ -180,6 +180,9 @@ struct IngestedFileInfo : public KeyRangeInfo {
  // the user key's format in the external file matches the column family's
  // setting.
  bool user_defined_timestamps_persisted = true;
+
+  SequenceNumber largest_seqno = kMaxSequenceNumber;
+  SequenceNumber smallest_seqno = kMaxSequenceNumber;
 };

 // A batch of files.
@@ -230,7 +233,7 @@ class ExternalSstFileIngestionJob {
        directories_(directories),
        event_logger_(event_logger),
        job_start_time_(clock_->NowMicros()),
-        consumed_seqno_count_(0),
+        max_assigned_seqno_(0),
        io_tracer_(io_tracer) {
    assert(directories != nullptr);
    assert(cfd_);
@@ -287,8 +290,16 @@ class ExternalSstFileIngestionJob {
    return files_to_ingest_;
  }

-  // How many sequence numbers did we consume as part of the ingestion job?
-  int ConsumedSequenceNumbersCount() const { return consumed_seqno_count_; }
+  // Return the maximum assigned sequence number for all files in this job.
+  // When allow_db_generated_files = false, we may assign global sequence
+  // numbers to ingested files. The global sequence numbers are sequence numbers
+  // following versions_->LastSequence().
+  // When allow_db_generated_files = true, we ingest files that already have
+  // sequence numbers assigned. max_assigned_seqno_ will be the max sequence
+  // number among ingested files.
+  SequenceNumber MaxAssignedSequenceNumber() const {
+    return max_assigned_seqno_;
+  }

 private:
  Status ResetTableReader(const std::string& external_file,
@@ -349,7 +360,7 @@ class ExternalSstFileIngestionJob {
      std::optional<int> prev_batch_uppermost_level);

  // File that we want to ingest behind always goes to the lowest level;
-  // we just check that it fits in the level, that DB allows ingest_behind,
+  // we just check that it fits in the level, that the CF allows ingest_behind,
  // and that we don't have 0 seqnums at the upper levels.
  // REQUIRES: Mutex held
  Status CheckLevelForIngestedBehindFile(IngestedFileInfo* file_to_ingest);
@@ -369,6 +380,13 @@ class ExternalSstFileIngestionJob {
  template <typename TWritableFile>
  Status SyncIngestedFile(TWritableFile* file);

+  // Helper function to obtain the smallest and largest sequence number from a
+  // file. When OK is returned, file_to_ingest->smallest_seqno and
+  // file_to_ingest->largest_seqno will be updated.
+  Status GetSeqnoBoundaryForFile(TableReader* table_reader, SuperVersion* sv,
+                                 IngestedFileInfo* file_to_ingest,
+                                 bool allow_data_in_errors);
+
  // Create equivalent `Compaction` objects to this file ingestion job
  // , which will be used to check range conflict with other ongoing
  // compactions.
@@ -395,7 +413,7 @@ class ExternalSstFileIngestionJob {
  EventLogger* event_logger_;
  VersionEdit edit_;
  uint64_t job_start_time_;
-  int consumed_seqno_count_;
+  SequenceNumber max_assigned_seqno_;
  // Set in ExternalSstFileIngestionJob::Prepare(), if true all files are
  // ingested in L0
  bool files_overlap_{false};
@@ -7,6 +7,7 @@

 #include <functional>
 #include <memory>
+#include <sstream>

 #include "db/db_test_util.h"
 #include "db/dbformat.h"
@@ -2417,102 +2418,130 @@ TEST_F(ExternalSSTFileTest, SnapshotInconsistencyBug) {
 }

 TEST_P(ExternalSSTFileTest, IngestBehind) {
-  Options options = CurrentOptions();
-  options.compaction_style = kCompactionStyleUniversal;
-  options.num_levels = 3;
-  options.disable_auto_compactions = false;
-  DestroyAndReopen(options);
-  std::vector<std::pair<std::string, std::string>> file_data;
-  std::map<std::string, std::string> true_data;
+  for (bool cf_option : {false, true}) {
+    SCOPED_TRACE("cf_option = " + std::to_string(cf_option));
+    Options options = CurrentOptions();
+    options.compaction_style = kCompactionStyleUniversal;
+    options.num_levels = 3;
+    options.disable_auto_compactions = false;
+    DestroyAndReopen(options);
+    std::vector<std::pair<std::string, std::string>> file_data;
+    std::map<std::string, std::string> true_data;

-  // Insert 100 -> 200 into the memtable
-  for (int i = 100; i <= 200; i++) {
-    ASSERT_OK(Put(Key(i), "memtable"));
-  }
-
-  // Insert 100 -> 200 using IngestExternalFile
-  file_data.clear();
-  for (int i = 0; i <= 20; i++) {
-    file_data.emplace_back(Key(i), "ingest_behind");
-    true_data[Key(i)] = "ingest_behind";
-  }
-
-  bool allow_global_seqno = true;
-  bool ingest_behind = true;
-  bool write_global_seqno = std::get<0>(GetParam());
-  bool verify_checksums_before_ingest = std::get<1>(GetParam());
-
-  // Can't ingest behind since allow_ingest_behind isn't set to true
-  ASSERT_NOK(GenerateAndAddExternalFile(
-      options, file_data, -1, allow_global_seqno, write_global_seqno,
-      verify_checksums_before_ingest, ingest_behind, false /*sort_data*/,
-      &true_data));
-
-  options.allow_ingest_behind = true;
-  // check that we still can open the DB, as num_levels should be
-  // sanitized to 3
-  options.num_levels = 2;
-  DestroyAndReopen(options);
-
-  options.num_levels = 3;
-  DestroyAndReopen(options);
-  true_data.clear();
-  // Insert 100 -> 200 into the memtable
-  for (int i = 100; i <= 200; i++) {
-    ASSERT_OK(Put(Key(i), "memtable"));
-    true_data[Key(i)] = "memtable";
-  }
-  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
-  // Universal picker should go at second from the bottom level
-  ASSERT_EQ("0,1", FilesPerLevel());
-  ASSERT_OK(GenerateAndAddExternalFile(
-      options, file_data, -1, allow_global_seqno, write_global_seqno,
-      verify_checksums_before_ingest, true /*ingest_behind*/,
-      false /*sort_data*/, &true_data));
-  ASSERT_EQ("0,1,1", FilesPerLevel());
-  // this time ingest should fail as the file doesn't fit to the bottom level
-  ASSERT_NOK(GenerateAndAddExternalFile(
-      options, file_data, -1, allow_global_seqno, write_global_seqno,
-      verify_checksums_before_ingest, true /*ingest_behind*/,
-      false /*sort_data*/, &true_data));
-  ASSERT_EQ("0,1,1", FilesPerLevel());
-  std::vector<std::vector<FileMetaData>> level_to_files;
-  dbfull()->TEST_GetFilesMetaData(db_->DefaultColumnFamily(), &level_to_files);
-  uint64_t ingested_file_number = level_to_files[2][0].fd.GetNumber();
-  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
-  // Last level should not be compacted
-  ASSERT_EQ("0,1,1", FilesPerLevel());
-  dbfull()->TEST_GetFilesMetaData(db_->DefaultColumnFamily(), &level_to_files);
-  ASSERT_EQ(ingested_file_number, level_to_files[2][0].fd.GetNumber());
-  size_t kcnt = 0;
-  VerifyDBFromMap(true_data, &kcnt, false);
-
-  // Auto-compaction should not include the last level.
-  // Trigger compaction if size amplification exceeds 110%.
-  options.compaction_options_universal.max_size_amplification_percent = 110;
-  options.level0_file_num_compaction_trigger = 4;
-  ASSERT_OK(TryReopen(options));
-  Random rnd(301);
-  for (int i = 0; i < 4; ++i) {
-    for (int j = 0; j < 10; j++) {
-      true_data[Key(j)] = rnd.RandomString(1000);
-      ASSERT_OK(Put(Key(j), true_data[Key(j)]));
+    // Insert 100 -> 200 into the memtable
+    for (int i = 100; i <= 200; i++) {
+      ASSERT_OK(Put(Key(i), "memtable"));
    }
-    ASSERT_OK(Flush());
-  }
-  ASSERT_OK(dbfull()->TEST_WaitForCompact());
-  dbfull()->TEST_GetFilesMetaData(db_->DefaultColumnFamily(), &level_to_files);
-  ASSERT_EQ(1, level_to_files[2].size());
-  ASSERT_EQ(ingested_file_number, level_to_files[2][0].fd.GetNumber());

-  // Turning off the option allows DB to compact ingested files.
-  options.allow_ingest_behind = false;
-  ASSERT_OK(TryReopen(options));
-  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
-  dbfull()->TEST_GetFilesMetaData(db_->DefaultColumnFamily(), &level_to_files);
-  ASSERT_EQ(1, level_to_files[2].size());
-  ASSERT_NE(ingested_file_number, level_to_files[2][0].fd.GetNumber());
-  VerifyDBFromMap(true_data, &kcnt, false);
+    // Insert 100 -> 200 using IngestExternalFile
+    file_data.clear();
+    for (int i = 0; i <= 20; i++) {
+      file_data.emplace_back(Key(i), "ingest_behind");
+      true_data[Key(i)] = "ingest_behind";
+    }
+
+    bool allow_global_seqno = true;
+    bool ingest_behind = true;
+    bool write_global_seqno = std::get<0>(GetParam());
+    bool verify_checksums_before_ingest = std::get<1>(GetParam());
+
+    // Can't ingest behind since allow_ingest_behind isn't set to true
+    ASSERT_NOK(GenerateAndAddExternalFile(
+        options, file_data, -1, allow_global_seqno, write_global_seqno,
+        verify_checksums_before_ingest, ingest_behind, false /*sort_data*/,
+        &true_data));
+
+    if (cf_option) {
+      options.cf_allow_ingest_behind = true;
+    } else {
+      options.allow_ingest_behind = true;
+    }
+    // check that we still can open the DB, as num_levels should be
+    // sanitized to 3
+    options.num_levels = 2;
+    DestroyAndReopen(options);
+
+    options.num_levels = 3;
+    DestroyAndReopen(options);
+    true_data.clear();
+    // Insert 100 -> 200 into the memtable
+    for (int i = 100; i <= 200; i++) {
+      ASSERT_OK(Put(Key(i), "memtable"));
+      true_data[Key(i)] = "memtable";
+    }
+    ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+    // Universal picker should go at second from the bottom level
+    ASSERT_EQ("0,1", FilesPerLevel());
+    ASSERT_OK(GenerateAndAddExternalFile(
+        options, file_data, -1, allow_global_seqno, write_global_seqno,
+        verify_checksums_before_ingest, true /*ingest_behind*/,
+        false /*sort_data*/, &true_data));
+    ASSERT_EQ("0,1,1", FilesPerLevel());
+    // this time ingest should fail as the file doesn't fit to the bottom level
+    ASSERT_NOK(GenerateAndAddExternalFile(
+        options, file_data, -1, allow_global_seqno, write_global_seqno,
+        verify_checksums_before_ingest, true /*ingest_behind*/,
+        false /*sort_data*/, &true_data));
+    ASSERT_EQ("0,1,1", FilesPerLevel());
+    std::vector<std::vector<FileMetaData>> level_to_files;
+    dbfull()->TEST_GetFilesMetaData(db_->DefaultColumnFamily(),
+                                    &level_to_files);
+    uint64_t ingested_file_number = level_to_files[2][0].fd.GetNumber();
+    ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+    // Last level should not be compacted
+    ASSERT_EQ("0,1,1", FilesPerLevel());
+    dbfull()->TEST_GetFilesMetaData(db_->DefaultColumnFamily(),
+                                    &level_to_files);
+    ASSERT_EQ(ingested_file_number, level_to_files[2][0].fd.GetNumber());
+    size_t kcnt = 0;
+    VerifyDBFromMap(true_data, &kcnt, false);
+
+    // Auto-compaction should not include the last level.
+    // Trigger compaction if size amplification exceeds 110%.
+    options.compaction_options_universal.max_size_amplification_percent = 110;
+    options.level0_file_num_compaction_trigger = 4;
+    ASSERT_OK(TryReopen(options));
+    Random rnd(301);
+    for (int i = 0; i < 4; ++i) {
+      for (int j = 0; j < 10; j++) {
+        true_data[Key(j)] = rnd.RandomString(1000);
+        ASSERT_OK(Put(Key(j), true_data[Key(j)]));
+      }
+      ASSERT_OK(Flush());
+    }
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    dbfull()->TEST_GetFilesMetaData(db_->DefaultColumnFamily(),
+                                    &level_to_files);
+    ASSERT_EQ(1, level_to_files[2].size());
+    ASSERT_EQ(ingested_file_number, level_to_files[2][0].fd.GetNumber());
+
+    // Turning off the option allows DB to compact ingested files.
+    if (cf_option) {
+      // Test that another CF does not allow ingest behind
+      ColumnFamilyHandle* new_cfh;
+      Options new_cf_option;
+      ASSERT_OK(db_->CreateColumnFamily(new_cf_option, "new_cf", &new_cfh));
+      ASSERT_TRUE(GenerateAndAddExternalFile(
+                      new_cf_option, file_data, -1, allow_global_seqno,
+                      write_global_seqno, verify_checksums_before_ingest,
+                      true /*ingest_behind*/, false /*sort_data*/, nullptr,
+                      /*cfh=*/new_cfh)
+                      .IsInvalidArgument());
+      ASSERT_OK(db_->DropColumnFamily(new_cfh));
+      ASSERT_OK(db_->DestroyColumnFamilyHandle(new_cfh));
+
+      options.cf_allow_ingest_behind = false;
+    } else {
+      options.allow_ingest_behind = false;
+    }
+    ASSERT_OK(TryReopen(options));
+    ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+    dbfull()->TEST_GetFilesMetaData(db_->DefaultColumnFamily(),
+                                    &level_to_files);
+    ASSERT_EQ(1, level_to_files[2].size());
+    ASSERT_NE(ingested_file_number, level_to_files[2][0].fd.GetNumber());
+    VerifyDBFromMap(true_data, &kcnt, false);
+  }
 }

 TEST_F(ExternalSSTFileTest, SkipBloomFilter) {
@@ -3514,19 +3543,26 @@ TEST_F(ExternalSSTFileWithTimestampTest, SanityCheck) {
  // overlapping key ranges.
  ASSERT_TRUE(IngestExternalUDTFile({file1, file2}).IsNotSupported());

-  options.allow_ingest_behind = true;
-  DestroyAndReopen(options);
-  IngestExternalFileOptions opts;
+  for (bool cf_option : {false, true}) {
+    SCOPED_TRACE("cf_option = " + std::to_string(cf_option));
+    if (cf_option) {
+      options.cf_allow_ingest_behind = true;
+    } else {
+      options.allow_ingest_behind = true;
+    }
+    DestroyAndReopen(options);
+    IngestExternalFileOptions opts;

-  // TODO(yuzhangyu): support ingestion behind for user-defined timestamps?
-  // Ingesting external files with user-defined timestamps requires searching
-  // through the whole lsm tree to make sure there is no key range overlap with
-  // the db. Ingestion behind currently is doing a simply placing it at the
-  // bottom level step without a search, so we don't allow it either.
-  opts.ingest_behind = true;
-  ASSERT_TRUE(db_->IngestExternalFile({file1}, opts).IsNotSupported());
+    // TODO(yuzhangyu): support ingestion behind for user-defined timestamps?
+    // Ingesting external files with user-defined timestamps requires searching
+    // through the whole lsm tree to make sure there is no key range overlap
+    // with the db. Ingestion behind currently is doing a simply placing it at
+    // the bottom level step without a search, so we don't allow it either.
+    opts.ingest_behind = true;
+    ASSERT_TRUE(db_->IngestExternalFile({file1}, opts).IsNotSupported());

-  DestroyAndRecreateExternalSSTFilesDir();
+    DestroyAndRecreateExternalSSTFilesDir();
+  }
 }

 TEST_F(ExternalSSTFileWithTimestampTest, UDTSettingsCompatibilityCheck) {
@@ -3818,99 +3854,32 @@ TEST_P(IngestDBGeneratedFileTest, FailureCase) {
      ASSERT_OK(Put(1, Key(k), "cf1_" + Key(k)));
    }
    ASSERT_OK(Flush(/*cf=*/1));
-    {
-      // Verify that largest key of the file has non-zero seqno.
-      std::vector<std::vector<FileMetaData>> metadata;
-      dbfull()->TEST_GetFilesMetaData(handles_[1], &metadata, nullptr);
-      const FileMetaData& file = metadata[0][0];
-      ValueType vtype;
-      SequenceNumber seq;
-      UnPackSequenceAndType(ExtractInternalKeyFooter(file.largest.Encode()),
-                            &seq, &vtype);
-      ASSERT_GE(seq, 0);
-    }
-    std::vector<LiveFileMetaData> live_meta;
-    db_->GetLiveFilesMetaData(&live_meta);
-    ASSERT_EQ(live_meta.size(), 1);
-    std::vector<std::string> to_ingest_files;
-    to_ingest_files.emplace_back(live_meta[0].directory + "/" +
-                                 live_meta[0].relative_filename);
-    // Ingesting a file whose boundary key has non-zero seqno.
-    Status s = db_->IngestExternalFile(to_ingest_files, ingest_opts);
-    // This error msg is from checking seqno of boundary keys.
-    ASSERT_TRUE(
-        s.ToString().find("External file has non zero sequence number") !=
-        std::string::npos);
-    ASSERT_NOK(s);
-
-    {
-      // Only non-boundary key with non-zero seqno.
-      const Snapshot* snapshot = db_->GetSnapshot();
-      ASSERT_OK(Put(1, Key(70), "cf1_" + Key(70)));
-      ASSERT_OK(Flush(1));
-      CompactRangeOptions cro;
-      cro.bottommost_level_compaction =
-          BottommostLevelCompaction::kForceOptimized;
-      ASSERT_OK(db_->CompactRange(cro, handles_[1], nullptr, nullptr));
-
-      // Verify that only the non-boundary key of the file has non-zero seqno.
-      std::vector<std::vector<FileMetaData>> metadata;
-      // File may be at different level for different options.
-      dbfull()->TEST_GetFilesMetaData(handles_[1], &metadata, nullptr);
-      bool found_file = false;
-      for (const auto& level : metadata) {
-        if (level.empty()) {
-          continue;
-        }
-        ASSERT_FALSE(found_file);
-        found_file = true;
-        ASSERT_EQ(1, level.size());
-        const FileMetaData& file = level[0];
-        ValueType vtype;
-        SequenceNumber seq;
-        UnPackSequenceAndType(ExtractInternalKeyFooter(file.largest.Encode()),
-                              &seq, &vtype);
-        ASSERT_EQ(seq, 0);
-        UnPackSequenceAndType(ExtractInternalKeyFooter(file.smallest.Encode()),
-                              &seq, &vtype);
-        ASSERT_EQ(seq, 0);
-        ASSERT_GT(file.fd.largest_seqno, 0);
-      }
-      ASSERT_TRUE(found_file);
-      live_meta.clear();
-      db_->GetLiveFilesMetaData(&live_meta);
-      ASSERT_EQ(live_meta.size(), 1);
-      to_ingest_files[0] =
-          live_meta[0].directory + "/" + live_meta[0].relative_filename;
-      s = db_->IngestExternalFile(to_ingest_files, ingest_opts);
-      ASSERT_NOK(s);
-      // This error msg is from checking largest seqno in table property.
-      ASSERT_TRUE(s.ToString().find("non zero largest sequence number") !=
-                  std::string::npos);
-      db_->ReleaseSnapshot(snapshot);
-    }

+    Status s;
    CompactRangeOptions cro;
    cro.bottommost_level_compaction =
        BottommostLevelCompaction::kForceOptimized;
    ASSERT_OK(db_->CompactRange(cro, handles_[1], nullptr, nullptr));
-    live_meta.clear();
+
+    std::vector<LiveFileMetaData> live_meta;
+    std::vector<std::string> to_ingest_files;
    db_->GetLiveFilesMetaData(&live_meta);
    ASSERT_EQ(live_meta.size(), 1);
+    ASSERT_EQ(live_meta[0].column_family_name, "toto");
    ASSERT_EQ(0, live_meta[0].largest_seqno);
-    to_ingest_files[0] =
-        live_meta[0].directory + "/" + live_meta[0].relative_filename;
+    to_ingest_files.emplace_back(live_meta[0].directory + "/" +
+                                 live_meta[0].relative_filename);

+    // Ingesting a DB generated file with allow_db_generated_files = false
    ingest_opts.allow_db_generated_files = false;
-    // Ingesting a DB genrate file with allow_db_generated_files = false;
    s = db_->IngestExternalFile(to_ingest_files, ingest_opts);
    ASSERT_TRUE(s.ToString().find("External file version not found") !=
                std::string::npos);
    ASSERT_NOK(s);

    const std::string err =
-        "An ingested file is assigned to a non-zero sequence number, which is "
-        "incompatible with ingestion option allow_db_generated_files";
+        "An ingested file overlaps with existing data in the DB and has been "
+        "assigned a non-zero sequence number";
    ingest_opts.allow_db_generated_files = true;
    s = db_->IngestExternalFile(to_ingest_files, ingest_opts);
    ASSERT_TRUE(s.ToString().find(err) != std::string::npos);
@@ -4111,6 +4080,440 @@ TEST_P(IngestDBGeneratedFileTest2, NotOverlapWithDB) {
    }
  } while (ChangeOptions(kSkipPlainTable | kSkipFIFOCompaction));
 }
+
+TEST_P(IngestDBGeneratedFileTest2, NonZeroSeqno) {
+  // Test ingestion of DB-generated SST files that contain non-zero sequence
+  // numbers.
+  IngestExternalFileOptions ingest_opts;
+  ingest_opts.allow_db_generated_files = true;
+  // This only works since we are ingesting without snapshot
+  // Failure case will be tested below.
+  ingest_opts.snapshot_consistency = std::get<0>(GetParam());
+  ingest_opts.allow_global_seqno = std::get<1>(GetParam());
+  ingest_opts.allow_blocking_flush = std::get<2>(GetParam());
+  ingest_opts.fail_if_not_bottommost_level = std::get<3>(GetParam());
+  ingest_opts.link_files = std::get<4>(GetParam());
+  Random* rnd = Random::GetTLSInstance();
+
+  do {
+    SCOPED_TRACE("option_config_ = " + std::to_string(option_config_));
+
+    Options options = CurrentOptions();
+    options.allow_concurrent_memtable_write =
+        false;  // Required for VectorRepFactory
+    CreateAndReopenWithCF({"non_overlap", "overlap"}, options);
+
+    ColumnFamilyHandle* non_overlap_cf = handles_[1];
+    ColumnFamilyHandle* overlap_cf = handles_[2];
+
+    std::vector<std::string> expected_values;
+    expected_values.resize(100);
+    WriteOptions wo;
+    // Setup target CF with non-overlapping base data Key1 and Key99
+    // Will ingest keys [1, 98] below.
+    expected_values[0] = rnd->RandomString(100);
+    ASSERT_OK(db_->Put(wo, non_overlap_cf, Key(0), expected_values[0]));
+    ASSERT_OK(db_->Flush({}, non_overlap_cf));
+    expected_values[99] = rnd->RandomString(100);
+    ASSERT_OK(db_->Put(wo, non_overlap_cf, Key(99), expected_values[99]));
+
+    // Set up overlapping cf
+    ASSERT_OK(db_->Put(wo, overlap_cf, Key(50), rnd->RandomString(100)));
+
+    // Create temp CF/DB
+    Options temp_cf_opts;
+    ColumnFamilyHandle* temp_cfh = nullptr;
+    DB* from_db = nullptr;
+    std::string temp_db_name;
+    // Using a separate DB also validates that latest sequence number
+    // of target db is updated after ingestion (to the max sequence number
+    // in ingested files).
+    const bool use_temp_db = rnd->OneIn(2);
+    SCOPED_TRACE("use_temp_db: " + std::to_string(use_temp_db));
+
+    std::vector<std::string> sst_file_paths;
+    // optional L5: files in key range [70, 98]
+    // L6: files in key range [1, 79]
+    temp_cf_opts.target_file_size_base =
+        4 << 10;  // Small files to create multiple SSTs
+    temp_cf_opts.num_levels = 7;
+    temp_cf_opts.disable_auto_compactions = true;  // Manually set up LSM
+    temp_cf_opts.env = options.env;
+
+    if (use_temp_db) {
+      temp_cf_opts.create_if_missing = true;
+      temp_db_name = dbname_ + "/temp_db_" + std::to_string(rnd->Next());
+      ASSERT_OK(DB::Open(temp_cf_opts, temp_db_name, &from_db));
+      temp_cfh = from_db->DefaultColumnFamily();
+    } else {
+      from_db = db_;
+      ASSERT_OK(
+          from_db->CreateColumnFamily(temp_cf_opts, "temp_cf", &temp_cfh));
+    }
+
+    // Use snapshot to ensure non-zero sequence numbers after compaction
+    const Snapshot* snapshot = from_db->GetSnapshot();
+
+    for (int k = 1; k < 99; ++k) {
+      expected_values[k] = rnd->RandomString(500);
+      ASSERT_OK(from_db->Put(wo, temp_cfh, Key(k), expected_values[k]));
+    }
+    ASSERT_OK(from_db->Flush({}, temp_cfh));
+    CompactRangeOptions cro;
+    cro.bottommost_level_compaction =
+        BottommostLevelCompaction::kForceOptimized;
+    ASSERT_OK(from_db->CompactRange(cro, temp_cfh, nullptr, nullptr));
+
+    ASSERT_GT(NumTableFilesAtLevel(6, temp_cfh, from_db), 1);
+
+    const bool multi_level_ingestion = rnd->OneIn(2);
+    SCOPED_TRACE("Multi-level ingestion: " +
+                 std::to_string(multi_level_ingestion));
+    if (multi_level_ingestion) {
+      for (int k = 80; k < 99; ++k) {
+        expected_values[k] = rnd->RandomString(500);
+        ASSERT_OK(from_db->Put(wo, temp_cfh, Key(k), expected_values[k]));
+      }
+      ASSERT_OK(from_db->Flush({}, temp_cfh));
+
+      // Do some overwrites, and overlap with previous L0 to avoid trivial move
+      for (int k = 70; k < 82; ++k) {
+        expected_values[k] = rnd->RandomString(500);
+        ASSERT_OK(from_db->Put(wo, temp_cfh, Key(k), expected_values[k]));
+      }
+      ASSERT_OK(from_db->Flush({}, temp_cfh));
+
+      if (rnd->OneIn(2)) {
+        MoveFilesToLevel(5, temp_cfh, from_db);
+        ASSERT_GT(NumTableFilesAtLevel(5, temp_cfh, from_db), 0);
+      }
+      ASSERT_GT(NumTableFilesAtLevel(6, temp_cfh, from_db), 0);
+    }
+    SCOPED_TRACE("LSM of from_db " + FilesPerLevel(temp_cfh, from_db));
+
+    ColumnFamilyMetaData cf_meta;
+    from_db->GetColumnFamilyMetaData(temp_cfh, &cf_meta);
+
+    // Iterate in reverse since IngestExternalFiles expect files to be ordered
+    // from old to new
+    for (auto level_meta = cf_meta.levels.rbegin();
+         level_meta != cf_meta.levels.rend(); ++level_meta) {
+      // L0 files need to be added in reverse order.
+      for (auto file_meta = level_meta->files.rbegin();
+           file_meta != level_meta->files.rend(); ++file_meta) {
+        // Validate that files contain non-zero sequence numbers
+        ASSERT_GT(file_meta->smallest_seqno, 0);
+        ASSERT_GE(file_meta->largest_seqno, file_meta->smallest_seqno);
+        sst_file_paths.emplace_back(file_meta->directory + "/" +
+                                    file_meta->relative_filename);
+      }
+    }
+    from_db->ReleaseSnapshot(snapshot);
+
+    Status s;
+    // Perform ingestion and validate results
+    if (multi_level_ingestion && options.num_levels > 1) {
+      // fail_if_bottommost requres ingesting all files into the last level,
+      // so it fails if we are assiging files to multiple levels.
+      ingest_opts.fail_if_not_bottommost_level = true;
+      s = db_->IngestExternalFile(non_overlap_cf, sst_file_paths, ingest_opts);
+      ASSERT_NOK(s);
+      ASSERT_TRUE(s.ToString().find("Files cannot be ingested to Lmax") !=
+                  std::string::npos);
+      ingest_opts.fail_if_not_bottommost_level = false;
+    }
+    if (ingest_opts.snapshot_consistency) {
+      // snapshot_consisteny requires global sequence number assignment to
+      // ingested files if there is any live snapshot.
+      snapshot = db_->GetSnapshot();
+      s = db_->IngestExternalFile(non_overlap_cf, sst_file_paths, ingest_opts);
+      ASSERT_NOK(s);
+      ASSERT_TRUE(s.ToString().find(
+          "An ingested file overlaps with existing data in the DB and has been "
+          "assigned a non-zero sequence number"));
+      db_->ReleaseSnapshot(snapshot);
+    }
+    ASSERT_OK(
+        db_->IngestExternalFile(non_overlap_cf, sst_file_paths, ingest_opts));
+
+    // Validate ingested data.
+    ReadOptions ro;
+    std::string val;
+    for (int k = 0; k < 100; ++k) {
+      s = db_->Get(ro, handles_[1], Key(k), &val);
+      ASSERT_OK(s) << "Should find ingested key " << Key(k);
+      ASSERT_EQ(val, expected_values[k]) << "key: " << Key(k);
+    }
+
+    // Overlap with data in the CF
+    if (ingest_opts.allow_blocking_flush) {
+      s = db_->IngestExternalFile(overlap_cf, sst_file_paths, ingest_opts);
+
+      ASSERT_NOK(s);
+      ASSERT_TRUE(s.ToString().find("An ingested file overlaps with existing "
+                                    "data in the DB and has been "
+                                    "assigned a non-zero sequence number") !=
+                  std::string::npos)
+          << s.ToString();
+    }
+
+    // Cleanup
+    // FIXME: Without this, the test triggers some data race between dropping
+    // CF and background compaction.
+    ASSERT_OK(db_->WaitForCompact({}));
+    if (use_temp_db) {
+      ASSERT_OK(from_db->Close());
+      delete from_db;
+      ASSERT_OK(DestroyDB(temp_db_name, temp_cf_opts));
+    } else {
+      ASSERT_OK(db_->DropColumnFamily(temp_cfh));
+      ASSERT_OK(db_->DestroyColumnFamilyHandle(temp_cfh));
+    }
+  } while (ChangeOptions(kSkipPlainTable | kSkipFIFOCompaction));
+}
+
+std::string GenSecondaryKey(const std::string& pk, const std::string& val) {
+  return "index_" + val + "_" + pk;
+};
+
+TEST_P(IngestDBGeneratedFileTest2, ZeroAndNonZeroSeqno) {
+  // Test ingestion of SST files with zero and with non-zero sequence numbers.
+  // Generate data using a temp CF and a temp DB:
+  // 1. Temp CF with cf_allow_ingest_behind enabled to preserve non-zero seqno.
+  // 2. Temp DB with everything compacted to have zero seqno.
+  // Then ingest both types of files together into a target CF.
+  // This mimics a user case where temp DB contains data read from a
+  // snapshot while temp CF contains live writes after a snapshot is taken.
+  IngestExternalFileOptions ingest_opts;
+  ingest_opts.allow_db_generated_files = true;
+  ingest_opts.snapshot_consistency = std::get<0>(GetParam());
+  ingest_opts.allow_global_seqno = std::get<1>(GetParam());
+  ingest_opts.allow_blocking_flush = std::get<2>(GetParam());
+  ingest_opts.fail_if_not_bottommost_level = std::get<3>(GetParam());
+  ingest_opts.link_files = std::get<4>(GetParam());
+
+  Random* rnd = Random::GetTLSInstance();
+
+  do {
+    SCOPED_TRACE("option_config_ = " + std::to_string(option_config_));
+    Options options = CurrentOptions();
+    options.allow_concurrent_memtable_write = false;
+    // Force more flushes/compactions and more files to be generated
+    options.target_file_size_base = 1 << 10;     // 1KB
+    options.max_bytes_for_level_base = 2 << 10;  // 2KB
+    options.max_bytes_for_level_multiplier = 2;
+    options.level0_file_num_compaction_trigger = 2;
+    options.level_compaction_dynamic_level_bytes = true;
+    DestroyAndReopen(options);
+    CreateAndReopenWithCF({"target_cf"}, options);
+    auto* target_cfh = handles_[1];
+
+    Options live_write_cf_opts = options;
+    live_write_cf_opts.memtable_factory.reset(new VectorRepFactory());
+    live_write_cf_opts.compaction_style = kCompactionStyleUniversal;
+    live_write_cf_opts.cf_allow_ingest_behind = true;
+    live_write_cf_opts.num_levels = 50;
+    ColumnFamilyHandle* live_write_cfh;
+    ASSERT_OK(db_->CreateColumnFamily(live_write_cf_opts, "live_write_cf",
+                                      &live_write_cfh));
+
+    // Expected value and key
+    std::map<std::string, std::string> expected;
+    std::unordered_set<std::string> deleted;
+    std::stringstream debug_info;
+
+    // Setup base data in target CF, will ingest keys with different prefixes
+    // so they don't overlap with the base data.
+    WriteOptions wo;
+    for (int k = 0; k < 100; ++k) {
+      int random_val = rnd->Uniform(20);
+      expected[Key(k)] = std::to_string(random_val);
+      ASSERT_OK(db_->Put(wo, target_cfh, Key(k), expected[Key(k)]));
+
+      // Force flush every 20 keys to create multiple SST files
+      if (rnd->OneIn(20)) {
+        ASSERT_OK(db_->Flush({}, target_cfh));
+        debug_info << "Flush after " << k
+                   << ", LSM state: " << FilesPerLevel(target_cfh) << "\n";
+      }
+    }
+
+    // Temp DB for snapshot data
+    Options temp_db_opts;
+    temp_db_opts.create_if_missing = true;
+    temp_db_opts.target_file_size_base = 1 << 10;
+    temp_db_opts.write_buffer_size = 1 << 10;
+    temp_db_opts.memtable_factory.reset(new VectorRepFactory());
+    temp_db_opts.allow_concurrent_memtable_write = false;
+    temp_db_opts.compaction_style = kCompactionStyleUniversal;
+    temp_db_opts.env = env_;
+    temp_db_opts.num_levels = 7;
+
+    std::string temp_db_name =
+        dbname_ + "/temp_db_" + std::to_string(rnd->Next());
+    DB* temp_db = nullptr;
+    ASSERT_OK(DB::Open(temp_db_opts, temp_db_name, &temp_db));
+
+    const Snapshot* snapshot = db_->GetSnapshot();
+    ReadOptions ro;
+    ro.snapshot = snapshot;
+    ro.total_order_seek = true;
+    std::unique_ptr<Iterator> iter{db_->NewIterator(ro, target_cfh)};
+    // transform data read from snapshot and write to temp DB
+    // Varying the number of files in temp DB.
+    const int kValSize = rnd->Uniform(200);
+    for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
+      std::string key = iter->key().ToString();
+      std::string value = iter->value().ToString();
+      std::string sk = GenSecondaryKey(key, value);
+      // Usually value is empty, here we use a larger value to generate
+      // multiple SST files in temp_db.
+      std::string sk_val = rnd->RandomString(kValSize);
+      ASSERT_OK(temp_db->Put(wo, sk, sk_val));
+      expected[sk] = sk_val;
+      debug_info << "Snapshot data: " << sk << " -> \n";
+    }
+    ASSERT_OK(iter->status());
+
+    // Do some live writes into target CF and live write CF.
+    for (int i = 0; i < 10; ++i) {
+      WriteBatch wb;
+      for (int j = 0; j < 5; ++j) {
+        std::string key = Key(rnd->Uniform(100));
+        std::string old_val = expected[key];
+        // Value range is 0-19, allow some PK to have the same value.
+        int random_val = rnd->Uniform(20);
+        std::string new_val = std::to_string(random_val);
+        std::string old_index_key = GenSecondaryKey(key, old_val);
+        std::string new_index_key = GenSecondaryKey(key, new_val);
+        ASSERT_OK(wb.SingleDelete(live_write_cfh, old_index_key));
+        std::string sk_val = rnd->RandomString(kValSize);
+        ASSERT_OK(wb.Put(live_write_cfh, new_index_key, sk_val));
+        ASSERT_OK(wb.Put(target_cfh, key, new_val));
+        expected[key] = new_val;
+        expected.erase(old_index_key);
+        expected[new_index_key] = sk_val;
+        deleted.insert(old_index_key);
+        deleted.erase(new_index_key);
+
+        debug_info << "Live write: SD " << old_index_key << "\n";
+        debug_info << "Live write: " << key << " -> " << new_val << "\n";
+        debug_info << "Live write: " << new_index_key << " -> \n";
+      }
+      ASSERT_OK(db_->Write(wo, &wb));
+      if (rnd->OneIn(3)) {
+        debug_info << "Flush after " << i << " live writes\n";
+        ASSERT_OK(db_->Flush({}, live_write_cfh));
+      }
+    }
+    iter.reset();
+    db_->ReleaseSnapshot(snapshot);
+
+    // Compact temp_db to ensure zero sequence numbers
+    CompactRangeOptions cro;
+    cro.bottommost_level_compaction = BottommostLevelCompaction::kForce;
+    ASSERT_OK(temp_db->CompactRange(cro, nullptr, nullptr));
+    SCOPED_TRACE("Temp DB LSM: " +
+                 FilesPerLevel(temp_db->DefaultColumnFamily(), temp_db));
+
+    // Base data from snapshot
+    std::vector<std::string> sst_file_paths_zero_seqno;
+
+    // Collect SST file paths with zero sequence numbers
+    ASSERT_OK(temp_db->DisableFileDeletions());
+    ColumnFamilyMetaData cf_meta_temp_db;
+    temp_db->GetColumnFamilyMetaData(&cf_meta_temp_db);
+    for (const auto& level_meta : cf_meta_temp_db.levels) {
+      if (level_meta.level == 6) {
+        for (const auto& file_meta : level_meta.files) {
+          // Verify files have zero sequence numbers
+          ASSERT_EQ(0, file_meta.largest_seqno)
+              << "File " << file_meta.relative_filename
+              << " should have zero sequence number\n"
+              << debug_info.str();
+          sst_file_paths_zero_seqno.emplace_back(file_meta.directory + "/" +
+                                                 file_meta.relative_filename);
+        }
+      } else {
+        // All files should be in L6
+        ASSERT_EQ(0, level_meta.files.size()) << debug_info.str();
+      }
+    }
+
+    // Flush remaining catch up writes in memtable
+    ASSERT_OK(db_->Flush({}, live_write_cfh));
+    SCOPED_TRACE("LSM of live write cfh " + FilesPerLevel(live_write_cfh));
+    // Collect SST file paths with non-zero sequence numbers
+    ColumnFamilyMetaData live_write_cf_meta;
+    ASSERT_OK(db_->DisableFileDeletions());
+    db_->GetColumnFamilyMetaData(live_write_cfh, &live_write_cf_meta);
+
+    // Live writes after snapshot
+    std::vector<std::string> sst_file_paths_nonzero_seqno;
+    for (auto level_meta = live_write_cf_meta.levels.rbegin();
+         level_meta != live_write_cf_meta.levels.rend(); ++level_meta) {
+      // Reverse order is important for L0, where recent updates are ordered
+      // first
+      for (auto file_meta = level_meta->files.rbegin();
+           file_meta != level_meta->files.rend(); ++file_meta) {
+        sst_file_paths_nonzero_seqno.emplace_back(file_meta->directory + "/" +
+                                                  file_meta->relative_filename);
+        ASSERT_GT(file_meta->smallest_seqno, 0) << debug_info.str();
+      }
+      if (level_meta->level == 49) {
+        // Ingest behind does not compact to the last level
+        ASSERT_EQ(level_meta->files.size(), 0) << debug_info.str();
+      }
+    }
+
+    ASSERT_GT(sst_file_paths_zero_seqno.size(), 0) << debug_info.str();
+    ASSERT_GT(sst_file_paths_nonzero_seqno.size(), 0) << debug_info.str();
+
+    // Combine all SST file paths.
+    // File ingestion takes files from old to new.
+    std::vector<std::string> all_sst_files;
+    all_sst_files.insert(all_sst_files.end(), sst_file_paths_zero_seqno.begin(),
+                         sst_file_paths_zero_seqno.end());
+    all_sst_files.insert(all_sst_files.end(),
+                         sst_file_paths_nonzero_seqno.begin(),
+                         sst_file_paths_nonzero_seqno.end());
+    if (ingest_opts.fail_if_not_bottommost_level && options.num_levels > 1) {
+      // overlapping files will be ingested into different levels, including non
+      // Lmax
+      Status s =
+          db_->IngestExternalFile(target_cfh, all_sst_files, ingest_opts);
+      ASSERT_NOK(s);
+      ASSERT_TRUE(s.ToString().find("Files cannot be ingested to Lmax") !=
+                  std::string::npos);
+    } else {
+      ASSERT_OK(
+          db_->IngestExternalFile(target_cfh, all_sst_files, ingest_opts));
+
+      debug_info << "Zero seqno files: " << sst_file_paths_zero_seqno.size()
+                 << "\nNon-zero seqno files: "
+                 << sst_file_paths_nonzero_seqno.size() << "\n";
+
+      SCOPED_TRACE("Debug info:\n" + debug_info.str());
+      VerifyDBFromMap(expected, nullptr, false, nullptr, target_cfh, &deleted);
+    }
+
+    // clean up
+    ASSERT_OK(db_->EnableFileDeletions());
+    ASSERT_OK(temp_db->EnableFileDeletions());
+
+    // FIXME: Without this, the test triggers some data race between dropping
+    // CF and background compaction.
+    ASSERT_OK(db_->WaitForCompact({}));
+
+    ASSERT_OK(db_->DropColumnFamily(live_write_cfh));
+    ASSERT_OK(db_->DestroyColumnFamilyHandle(live_write_cfh));
+
+    ASSERT_OK(temp_db->Close());
+    delete temp_db;
+    ASSERT_OK(DestroyDB(temp_db_name, temp_db_opts));
+  } while (ChangeOptions(kSkipPlainTable | kSkipFIFOCompaction));
+}
+
 }  // namespace ROCKSDB_NAMESPACE

 int main(int argc, char** argv) {
--- a/Show More
+++ b/Show More