set ignore_unknown_options when parsing options (#13443)

Summary:
In case the primary host has a new option added which isn't available in the remote worker yet, the remote compaction currently fails. In most cases, these new options are not relevant to the remote compaction and the worker should be able to move on by ignoring it.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/13443

Test Plan: Verified internally in Meta Infra.

Reviewed By: anand1976

Differential Revision: D70744359

Pulled By: jaykorean

fbshipit-source-id: eb6a388c2358a7f8089f2e35a378b7017b9e03f3

.circleci/config.yml

@@ -1,905 +0,0 @@
-version: 2.1
-
-orbs:
-  win: circleci/windows@5.0.0
-
-commands:
-  install-cmake-on-macos:
-    steps:
-      - run:
-          name: Install cmake on macos
-          command: |
-            HOMEBREW_NO_AUTO_UPDATE=1 brew install cmake
-
-  install-jdk8-on-macos:
-    steps:
-      - run:
-          name: Install JDK 8 on macos
-          command: |
-            HOMEBREW_NO_AUTO_UPDATE=1 brew tap bell-sw/liberica
-            HOMEBREW_NO_AUTO_UPDATE=1 brew install --cask liberica-jdk8
-
-  increase-max-open-files-on-macos:
-    steps:
-      - run:
-          name: Increase max open files
-          command: |
-            sudo sysctl -w kern.maxfiles=1048576
-            sudo sysctl -w kern.maxfilesperproc=1048576
-            sudo launchctl limit maxfiles 1048576
-
-  pre-steps:
-    steps:
-      - checkout
-      - run:
-          name: Setup Environment Variables
-          command: |
-            echo "export GTEST_THROW_ON_FAILURE=0" >> $BASH_ENV
-            echo "export GTEST_OUTPUT=\"xml:/tmp/test-results/\"" >> $BASH_ENV
-            echo "export SKIP_FORMAT_BUCK_CHECKS=1" >> $BASH_ENV
-            echo "export GTEST_COLOR=1" >> $BASH_ENV
-            echo "export CTEST_OUTPUT_ON_FAILURE=1" >> $BASH_ENV
-            echo "export CTEST_TEST_TIMEOUT=300" >> $BASH_ENV
-            echo "export ZLIB_DOWNLOAD_BASE=https://rocksdb-deps.s3.us-west-2.amazonaws.com/pkgs/zlib" >> $BASH_ENV
-            echo "export BZIP2_DOWNLOAD_BASE=https://rocksdb-deps.s3.us-west-2.amazonaws.com/pkgs/bzip2" >> $BASH_ENV
-            echo "export SNAPPY_DOWNLOAD_BASE=https://rocksdb-deps.s3.us-west-2.amazonaws.com/pkgs/snappy" >> $BASH_ENV
-            echo "export LZ4_DOWNLOAD_BASE=https://rocksdb-deps.s3.us-west-2.amazonaws.com/pkgs/lz4" >> $BASH_ENV
-            echo "export ZSTD_DOWNLOAD_BASE=https://rocksdb-deps.s3.us-west-2.amazonaws.com/pkgs/zstd" >> $BASH_ENV
-
-  windows-build-steps:
-    steps:
-      - checkout
-      - run:
-          name: "Install thirdparty dependencies"
-          command: |
-            echo "Installing CMake..."
-            choco install cmake --installargs 'ADD_CMAKE_TO_PATH=System' -y
-            choco install liberica8jdk -y
-            mkdir $Env:THIRDPARTY_HOME
-            cd $Env:THIRDPARTY_HOME
-            echo "Building Snappy dependency..."
-            curl https://github.com/google/snappy/archive/refs/tags/1.1.8.zip -O snappy-1.1.8.zip
-            unzip -q snappy-1.1.8.zip
-            cd snappy-1.1.8
-            mkdir build
-            cd build
-            & $Env:CMAKE_BIN -G "$Env:CMAKE_GENERATOR" ..
-            msbuild.exe Snappy.sln -maxCpuCount -property:Configuration=Debug -property:Platform=x64
-      - run:
-          name: "Build RocksDB"
-          command: |
-            $env:Path = $env:JAVA_HOME + ";" + $env:Path
-            mkdir build
-            cd build
-            & $Env:CMAKE_BIN -G "$Env:CMAKE_GENERATOR" -DCMAKE_BUILD_TYPE=Debug -DOPTDBG=1 -DPORTABLE="$Env:CMAKE_PORTABLE" -DSNAPPY=1 -DJNI=1 ..
-            cd ..
-            echo "Building with VS version: $Env:CMAKE_GENERATOR"
-            msbuild.exe build/rocksdb.sln -maxCpuCount -property:Configuration=Debug -property:Platform=x64
-      - run:
-          name: "Test RocksDB"
-          shell: powershell.exe
-          command: |
-            build_tools\run_ci_db_test.ps1 -SuiteRun arena_test,db_basic_test,db_test,db_test2,db_merge_operand_test,bloom_test,c_test,coding_test,crc32c_test,dynamic_bloom_test,env_basic_test,env_test,hash_test,random_test -Concurrency 16
-      - run:
-          name: "Test RocksJava"
-          command: |
-            cd build\java
-            & $Env:CTEST_BIN -C Debug -j 16
-  pre-steps-macos:
-      steps:
-        - pre-steps
-
-  post-steps:
-    steps:
-      - store_test_results: # store test result if there's any
-          path: /tmp/test-results
-      - store_artifacts: # store LOG for debugging if there's any
-          path: LOG
-      - run: # on fail, compress Test Logs for diagnosing the issue
-          name: Compress Test Logs
-          command: tar -cvzf t.tar.gz t
-          when: on_fail
-      - store_artifacts: # on fail, store Test Logs for diagnosing the issue
-          path: t.tar.gz
-          destination: test_logs
-          when: on_fail
-      - run: # store core dumps if there's any
-          command: |
-            mkdir -p /tmp/core_dumps
-            cp core.* /tmp/core_dumps
-          when: on_fail
-      - store_artifacts:
-          path: /tmp/core_dumps
-          when: on_fail
-
-  post-pmd-steps:
-    steps:
-      - store_artifacts:
-          path: /home/circleci/project/java/target/pmd.xml
-          when: on_fail
-      - store_artifacts:
-          path: /home/circleci/project/java/target/site
-          when: on_fail
-
-  upgrade-cmake:
-    steps:
-      - run:
-          name: Upgrade cmake
-          command: |
-            sudo apt remove --purge cmake
-            sudo snap install cmake --classic
-
-  install-gflags:
-    steps:
-      - run:
-          name: Install gflags
-          command: |
-            sudo apt-get update -y && sudo apt-get install -y libgflags-dev
-
-  install-gflags-on-macos:
-    steps:
-      - run:
-          name: Install gflags on macos
-          command: |
-            HOMEBREW_NO_AUTO_UPDATE=1 brew install gflags
-
-  install-maven:
-    steps:
-      - run:
-          name: Install Maven
-          command: |
-            wget --no-check-certificate https://dlcdn.apache.org/maven/maven-3/3.9.6/binaries/apache-maven-3.9.6-bin.tar.gz
-            tar zxf apache-maven-3.9.6-bin.tar.gz
-            echo "export M2_HOME=$(pwd)/apache-maven-3.9.6" >> $BASH_ENV
-            echo 'export PATH=$M2_HOME/bin:$PATH' >> $BASH_ENV
-
-  setup-folly:
-    steps:
-      - run:
-          name: Checkout folly sources
-          command: |
-            make checkout_folly
-
-  build-folly:
-    steps:
-      - run:
-          name: Build folly and dependencies
-          command: |
-            make build_folly
-
-  build-for-benchmarks:
-    steps:
-      - pre-steps
-      - run:
-          name: "Linux build for benchmarks"
-          command: #sized for the resource-class rocksdb-benchmark-sys1
-            make V=1 J=8 -j8 release
-
-  perform-benchmarks:
-    steps:
-      - run:
-          name: "Test low-variance benchmarks"
-          command: ./tools/benchmark_ci.py --db_dir /tmp/rocksdb-benchmark-datadir --output_dir /tmp/benchmark-results --num_keys 20000000
-          environment:
-            LD_LIBRARY_PATH: /usr/local/lib
-            # How long to run parts of the test(s)
-            DURATION_RO: 300
-            DURATION_RW: 500
-            # Keep threads within physical capacity of server (much lower than default)
-            NUM_THREADS: 1
-            MAX_BACKGROUND_JOBS: 4
-            # Don't run a couple of "optional" initial tests
-            CI_TESTS_ONLY: "true"
-            # Reduce configured size of levels to ensure more levels in the leveled compaction LSM tree
-            WRITE_BUFFER_SIZE_MB: 16
-            TARGET_FILE_SIZE_BASE_MB: 16
-            MAX_BYTES_FOR_LEVEL_BASE_MB: 64
-            # The benchmark host has 32GB memory
-            # The following values are tailored to work with that
-            # Note, tests may not exercise the targeted issues if the memory is increased on new test hosts.
-            COMPRESSION_TYPE: "none"
-            CACHE_INDEX_AND_FILTER_BLOCKS: 1
-            MIN_LEVEL_TO_COMPRESS: 3
-            CACHE_SIZE_MB: 10240
-            MB_WRITE_PER_SEC: 2
-
-  post-benchmarks:
-    steps:
-      - store_artifacts: # store the benchmark output
-          path: /tmp/benchmark-results
-          destination: test_logs
-      - run:
-          name: Send benchmark report to visualisation
-          command: |
-            set +e
-            set +o pipefail
-            ./build_tools/benchmark_log_tool.py --tsvfile /tmp/benchmark-results/report.tsv --esdocument https://search-rocksdb-bench-k2izhptfeap2hjfxteolsgsynm.us-west-2.es.amazonaws.com/bench_test3_rix/_doc
-            true
-
-executors:
-  linux-docker:
-    docker:
-      # The image configuration is build_tools/ubuntu20_image/Dockerfile
-      # To update and build the image:
-      #  $ cd build_tools/ubuntu20_image
-      #  $ docker build -t zjay437/rocksdb:0.5 .
-      #  $ docker push zjay437/rocksdb:0.5
-      # `zjay437` is the account name for zjay@meta.com which readwrite token is shared internally. To login:
-      #  $ docker login --username zjay437
-      # Or please feel free to change it to your docker hub account for hosting the image, meta employee should already have the account and able to login with SSO.
-      # To avoid impacting the existing CI runs, please bump the version every time creating a new image
-      # to run the CI image environment locally:
-      #  $ docker run --cap-add=SYS_PTRACE --security-opt seccomp=unconfined -it zjay437/rocksdb:0.5 bash
-      # option `--cap-add=SYS_PTRACE --security-opt seccomp=unconfined` is used to enable gdb to attach an existing process
-      - image: zjay437/rocksdb:0.6
-  linux-java-docker:
-    docker:
-      # This is the Docker Image used for building RocksJava releases, see: https://github.com/evolvedbinary/docker-rocksjava
-      - image: evolvedbinary/rocksjava:centos6_x64-be
-
-jobs:
-  build-macos:
-    macos:
-      xcode: 14.3.1
-    resource_class: macos.m1.medium.gen1
-    environment:
-      ROCKSDB_DISABLE_JEMALLOC: 1 # jemalloc cause env_test hang, disable it for now
-    steps:
-      - increase-max-open-files-on-macos
-      - install-gflags-on-macos
-      - pre-steps-macos
-      - run: ulimit -S -n `ulimit -H -n` && make V=1 J=16 -j16 all
-      - post-steps
-
-  build-macos-cmake:
-    macos:
-      xcode: 14.3.1
-    resource_class: macos.m1.medium.gen1
-    parameters:
-      run_even_tests:
-        description: run even or odd tests, used to split tests to 2 groups
-        type: boolean
-        default: true
-    steps:
-      - increase-max-open-files-on-macos
-      - install-cmake-on-macos
-      - install-gflags-on-macos
-      - pre-steps-macos
-      - run:
-          name: "cmake generate project file"
-          command: ulimit -S -n `ulimit -H -n` && mkdir build && cd build && cmake -DWITH_GFLAGS=1 ..
-      - run:
-          name: "Build tests"
-          command: cd build && make V=1 -j16
-      - when:
-          condition: << parameters.run_even_tests >>
-          steps:
-            - run:
-                name: "Run even tests"
-                command: ulimit -S -n `ulimit -H -n` && cd build && ctest -j16 -I 0,,2
-      - when:
-          condition:
-            not: << parameters.run_even_tests >>
-          steps:
-            - run:
-                name: "Run odd tests"
-                command: ulimit -S -n `ulimit -H -n` && cd build && ctest -j16 -I 1,,2
-      - post-steps
-
-  build-linux:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - run: make V=1 J=32 -j32 check
-      - post-steps
-
-  build-linux-encrypted_env-no_compression:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - 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
-      - post-steps
-
-  build-linux-static_lib-alt_namespace-status_checked:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - run: ASSERT_STATUS_CHECKED=1 TEST_UINT128_COMPAT=1 ROCKSDB_MODIFY_NPHASH=1 LIB_MODE=static OPT="-DROCKSDB_NAMESPACE=alternative_rocksdb_ns" make V=1 -j24 check
-      - post-steps
-
-  build-linux-release:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - checkout # check out the code in the project directory
-      - run: make V=1 -j32 LIB_MODE=shared release
-      - run: ls librocksdb.so # ensure shared lib built
-      - run: ./db_stress --version # ensure with gflags
-      - run: make clean
-      - run: make V=1 -j32 release
-      - run: ls librocksdb.a # ensure static lib built
-      - run: ./db_stress --version # ensure with gflags
-      - run: make clean
-      - run: apt-get remove -y libgflags-dev
-      - run: make V=1 -j32 LIB_MODE=shared release
-      - run: ls librocksdb.so # ensure shared lib built
-      - run: if ./db_stress --version; then false; else true; fi # ensure without gflags
-      - run: make clean
-      - run: make V=1 -j32 release
-      - run: ls librocksdb.a # ensure static lib built
-      - run: if ./db_stress --version; then false; else true; fi # ensure without gflags
-      - post-steps
-
-  build-linux-release-rtti:
-    executor: linux-docker
-    resource_class: xlarge
-    steps:
-      - checkout # check out the code in the project directory
-      - run: USE_RTTI=1 DEBUG_LEVEL=0 make V=1 -j16 static_lib tools db_bench
-      - run: ./db_stress --version # ensure with gflags
-      - 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 # ensure without gflags
-
-  build-linux-clang-no_test_run:
-    executor: linux-docker
-    resource_class: xlarge
-    steps:
-      - checkout # check out the code in the project directory
-      - run: CC=clang CXX=clang++ USE_CLANG=1 PORTABLE=1 make V=1 -j16 all
-      - post-steps
-
-  build-linux-clang10-asan:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - 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 # aligned new doesn't work for reason we haven't figured out
-      - post-steps
-
-  build-linux-clang10-mini-tsan:
-    executor: linux-docker
-    resource_class: 2xlarge+
-    steps:
-      - pre-steps
-      - run: COMPILE_WITH_TSAN=1 CC=clang-13 CXX=clang++-13 ROCKSDB_DISABLE_ALIGNED_NEW=1 USE_CLANG=1 make V=1 -j32 check
-      - post-steps
-
-  build-linux-clang10-ubsan:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - run: COMPILE_WITH_UBSAN=1 OPT="-fsanitize-blacklist=.circleci/ubsan_suppression_list.txt" CC=clang-10 CXX=clang++-10 ROCKSDB_DISABLE_ALIGNED_NEW=1 USE_CLANG=1 make V=1 -j32 ubsan_check # aligned new doesn't work for reason we haven't figured out
-      - post-steps
-
-  build-linux-valgrind:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - run: PORTABLE=1 make V=1 -j32 valgrind_test
-      - post-steps
-
-  build-linux-clang10-clang-analyze:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - run: CC=clang-10 CXX=clang++-10 ROCKSDB_DISABLE_ALIGNED_NEW=1 CLANG_ANALYZER="/usr/bin/clang++-10" CLANG_SCAN_BUILD=scan-build-10 USE_CLANG=1 make V=1 -j32 analyze # aligned new doesn't work for reason we haven't figured out. For unknown, reason passing "clang++-10" as CLANG_ANALYZER doesn't work, and we need a full path.
-      - post-steps
-      - run:
-          name: "compress test report"
-          command: tar -cvzf scan_build_report.tar.gz scan_build_report
-          when: on_fail
-      - store_artifacts:
-          path: scan_build_report.tar.gz
-          destination: scan_build_report
-          when: on_fail
-
-  build-linux-runner:
-    machine: true
-    resource_class: facebook/rocksdb-benchmark-sys1
-    steps:
-      - pre-steps
-      - run:
-          name: "Checked Linux build (Runner)"
-          command: make V=1 J=8 -j8 check
-          environment:
-            LD_LIBRARY_PATH: /usr/local/lib
-      - post-steps
-
-  build-linux-cmake-with-folly:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - setup-folly
-      - build-folly
-      - run: (mkdir build && cd build && cmake -DUSE_FOLLY=1 -DWITH_GFLAGS=1 -DROCKSDB_BUILD_SHARED=0 .. && make V=1 -j20 && ctest -j20)
-      - post-steps
-
-  build-linux-cmake-with-folly-lite-no-test:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - setup-folly
-      - run: (mkdir build && cd build && cmake -DUSE_FOLLY_LITE=1 -DWITH_GFLAGS=1 .. && make V=1 -j20)
-      - post-steps
-
-  build-linux-cmake-with-benchmark:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - run: mkdir build && cd build && cmake -DWITH_GFLAGS=1 -DWITH_BENCHMARK=1 .. && make V=1 -j20 && ctest -j20
-      - post-steps
-
-  build-linux-unity-and-headers:
-    docker: # executor type
-      - image: gcc:latest
-    environment:
-      EXTRA_CXXFLAGS: -mno-avx512f # Warnings-as-error in avx512fintrin.h, would be used on newer hardware
-    resource_class: large
-    steps:
-      - checkout # check out the code in the project directory
-      - run: apt-get update -y && apt-get install -y libgflags-dev
-      - run:
-          name: "Unity build"
-          command: make V=1 -j8 unity_test
-          no_output_timeout: 20m
-      - run: make V=1 -j8 -k check-headers # could be moved to a different build
-      - post-steps
-
-  build-linux-gcc-7-with-folly:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - setup-folly
-      - build-folly
-      - run: USE_FOLLY=1 LIB_MODE=static CC=gcc-7 CXX=g++-7 V=1 make -j32 check # TODO: LIB_MODE only to work around unresolved linker failures
-      - post-steps
-
-  build-linux-gcc-7-with-folly-lite-no-test:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - setup-folly
-      - run: USE_FOLLY_LITE=1 CC=gcc-7 CXX=g++-7 V=1 make -j32 all
-      - post-steps
-
-  build-linux-gcc-8-no_test_run:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - run: CC=gcc-8 CXX=g++-8 V=1 make -j32 all
-      - post-steps
-
-  build-linux-cmake-with-folly-coroutines:
-    executor: linux-docker
-    resource_class: 2xlarge
-    environment:
-      CC: gcc-10
-      CXX: g++-10
-    steps:
-      - pre-steps
-      - setup-folly
-      - build-folly
-      - run: (mkdir build && cd build && cmake -DUSE_COROUTINES=1 -DWITH_GFLAGS=1 -DROCKSDB_BUILD_SHARED=0 .. && make V=1 -j20 && ctest -j20)
-      - post-steps
-
-  build-linux-gcc-10-cxx20-no_test_run:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - run: CC=gcc-10 CXX=g++-10 V=1 ROCKSDB_CXX_STANDARD=c++20 make -j32 all
-      - post-steps
-
-  build-linux-gcc-11-no_test_run:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - run: LIB_MODE=static CC=gcc-11 CXX=g++-11 V=1 make -j32 all microbench # TODO: LIB_MODE only to work around unresolved linker failures
-      - post-steps
-
-  build-linux-clang-13-no_test_run:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - run: CC=clang-13 CXX=clang++-13 USE_CLANG=1 make -j32 all microbench
-      - post-steps
-
-  # Ensure ASAN+UBSAN with folly, and full testsuite with clang 13
-  build-linux-clang-13-asan-ubsan-with-folly:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - setup-folly
-      - build-folly
-      - run: CC=clang-13 CXX=clang++-13 LIB_MODE=static USE_CLANG=1 USE_FOLLY=1 COMPILE_WITH_UBSAN=1 COMPILE_WITH_ASAN=1 make -j32 check # TODO: LIB_MODE only to work around unresolved linker failures
-      - post-steps
-
-  # This job is only to make sure the microbench tests are able to run, the benchmark result is not meaningful as the CI host is changing.
-  build-linux-run-microbench:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - run: DEBUG_LEVEL=0 make -j32 run_microbench
-      - post-steps
-
-  build-linux-mini-crashtest:
-    executor: linux-docker
-    resource_class: large
-    steps:
-      - pre-steps
-      - 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
-      - post-steps
-
-  build-linux-crashtest-tiered-storage-bb:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - run:
-          name: "run crashtest"
-          command: ulimit -S -n `ulimit -H -n` && make V=1 -j32 CRASH_TEST_EXT_ARGS='--duration=10800' blackbox_crash_test_with_tiered_storage
-          no_output_timeout: 100m
-      - post-steps
-
-  build-linux-crashtest-tiered-storage-wb:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - run:
-          name: "run crashtest"
-          command: ulimit -S -n `ulimit -H -n` && make V=1 -j32 CRASH_TEST_EXT_ARGS='--duration=10800' whitebox_crash_test_with_tiered_storage
-          no_output_timeout: 100m
-      - post-steps
-
-  build-windows-vs2022-avx2:
-    executor:
-      name: win/server-2022
-      size: 2xlarge
-    environment:
-      THIRDPARTY_HOME: C:/Users/circleci/thirdparty
-      CMAKE_HOME: C:/Program Files/CMake
-      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: C:/Users/circleci/thirdparty/snappy-1.1.8
-      SNAPPY_INCLUDE: C:/Users/circleci/thirdparty/snappy-1.1.8;C:/Users/circleci/thirdparty/snappy-1.1.8/build
-      SNAPPY_LIB_DEBUG: C:/Users/circleci/thirdparty/snappy-1.1.8/build/Debug/snappy.lib
-      CMAKE_GENERATOR: Visual Studio 17 2022
-      CMAKE_PORTABLE: AVX2
-    steps:
-      - windows-build-steps
-
-  build-windows-vs2022:
-    executor:
-      name: win/server-2022
-      size: 2xlarge
-    environment:
-      THIRDPARTY_HOME: C:/Users/circleci/thirdparty
-      CMAKE_HOME: C:/Program Files/CMake
-      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: C:/Users/circleci/thirdparty/snappy-1.1.8
-      SNAPPY_INCLUDE: C:/Users/circleci/thirdparty/snappy-1.1.8;C:/Users/circleci/thirdparty/snappy-1.1.8/build
-      SNAPPY_LIB_DEBUG: C:/Users/circleci/thirdparty/snappy-1.1.8/build/Debug/snappy.lib
-      CMAKE_GENERATOR: Visual Studio 17 2022
-      CMAKE_PORTABLE: 1
-    steps:
-      - windows-build-steps
-
-  build-windows-vs2019:
-    executor:
-      name: win/server-2019
-      size: 2xlarge
-    environment:
-      THIRDPARTY_HOME: C:/Users/circleci/thirdparty
-      CMAKE_HOME: C:/Program Files/CMake
-      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: C:/Users/circleci/thirdparty/snappy-1.1.8
-      SNAPPY_INCLUDE: C:/Users/circleci/thirdparty/snappy-1.1.8;C:/Users/circleci/thirdparty/snappy-1.1.8/build
-      SNAPPY_LIB_DEBUG: C:/Users/circleci/thirdparty/snappy-1.1.8/build/Debug/snappy.lib
-      CMAKE_GENERATOR: Visual Studio 16 2019
-      CMAKE_PORTABLE: 1
-    steps:
-      - windows-build-steps
-
-  build-linux-java:
-    executor: linux-java-docker
-    resource_class: large
-    steps:
-      - pre-steps
-      - run:
-          name: "Set Java Environment"
-          command: |
-            echo "JAVA_HOME=${JAVA_HOME}"
-            echo 'export PATH=$JAVA_HOME/bin:$PATH' >> $BASH_ENV
-            which java && java -version
-            which javac && javac -version
-      - run:
-          name: "Test RocksDBJava"
-          command: scl enable devtoolset-7 'make V=1 J=8 -j8 jtest'
-      - post-steps
-
-  build-linux-java-pmd:
-    executor: linux-java-docker
-    resource_class: large
-    steps:
-      - pre-steps
-      - run:
-          name: "Set Java Environment"
-          command: |
-            echo "JAVA_HOME=${JAVA_HOME}"
-            echo 'export PATH=$JAVA_HOME/bin:$PATH' >> $BASH_ENV
-            which java && java -version
-            which javac && javac -version
-      - install-maven
-      - run:
-          name: "PMD RocksDBJava"
-          command: scl enable devtoolset-7 'make V=1 J=8 -j8 jpmd'
-      - post-pmd-steps
-
-  build-linux-java-static:
-    executor: linux-java-docker
-    resource_class: large
-    steps:
-      - pre-steps
-      - run:
-          name: "Set Java Environment"
-          command: |
-            echo "JAVA_HOME=${JAVA_HOME}"
-            echo 'export PATH=$JAVA_HOME/bin:$PATH' >> $BASH_ENV
-            which java && java -version
-            which javac && javac -version
-      - run:
-          name: "Build RocksDBJava Static Library"
-          command: scl enable devtoolset-7 'make V=1 J=8 -j8 rocksdbjavastatic'
-      - post-steps
-
-  build-macos-java:
-    macos:
-      xcode: 14.3.1
-    resource_class: macos.m1.medium.gen1
-    environment:
-      JAVA_HOME: /Library/Java/JavaVirtualMachines/liberica-jdk-8.jdk/Contents/Home
-      ROCKSDB_DISABLE_JEMALLOC: 1 # jemalloc causes java 8 crash
-    steps:
-      - increase-max-open-files-on-macos
-      - install-gflags-on-macos
-      - install-jdk8-on-macos
-      - pre-steps-macos
-      - run:
-          name: "Set Java Environment"
-          command: |
-            echo "JAVA_HOME=${JAVA_HOME}"
-            echo 'export PATH=$JAVA_HOME/bin:$PATH' >> $BASH_ENV
-            which java && java -version
-            which javac && javac -version
-      - run:
-          name: "Test RocksDBJava"
-          command: make V=1 J=16 -j16 jtest
-          no_output_timeout: 20m
-      - post-steps
-
-  build-macos-java-static:
-    macos:
-      xcode: 14.3.1
-    resource_class: macos.m1.medium.gen1
-    environment:
-      JAVA_HOME: /Library/Java/JavaVirtualMachines/liberica-jdk-8.jdk/Contents/Home
-    steps:
-      - increase-max-open-files-on-macos
-      - install-gflags-on-macos
-      - install-cmake-on-macos
-      - install-jdk8-on-macos
-      - pre-steps-macos
-      - run:
-          name: "Set Java Environment"
-          command: |
-            echo "JAVA_HOME=${JAVA_HOME}"
-            echo 'export PATH=$JAVA_HOME/bin:$PATH' >> $BASH_ENV
-            which java && java -version
-            which javac && javac -version
-      - run:
-          name: "Build RocksDBJava x86 and ARM Static Libraries"
-          command: make V=1 J=16 -j16 rocksdbjavastaticosx
-          no_output_timeout: 20m
-      - post-steps
-
-  build-macos-java-static-universal:
-    macos:
-      xcode: 14.3.1
-    resource_class: macos.m1.medium.gen1
-    environment:
-      JAVA_HOME: /Library/Java/JavaVirtualMachines/liberica-jdk-8.jdk/Contents/Home
-    steps:
-      - increase-max-open-files-on-macos
-      - install-gflags-on-macos
-      - install-cmake-on-macos
-      - install-jdk8-on-macos
-      - pre-steps-macos
-      - run:
-          name: "Set Java Environment"
-          command: |
-            echo "JAVA_HOME=${JAVA_HOME}"
-            echo 'export PATH=$JAVA_HOME/bin:$PATH' >> $BASH_ENV
-            which java && java -version
-            which javac && javac -version
-      - run:
-          name: "Build RocksDBJava Universal Binary Static Library"
-          command: make V=1 J=16 -j16 rocksdbjavastaticosx_ub
-          no_output_timeout: 20m
-      - post-steps
-
-  build-examples:
-    executor: linux-docker
-    resource_class: large
-    steps:
-      - pre-steps
-      - run:
-          name: "Build examples"
-          command: |
-            make V=1 -j4 static_lib && cd examples && make V=1 -j4
-      - post-steps
-
-  build-cmake-mingw:
-    executor: linux-docker
-    resource_class: large
-    steps:
-      - pre-steps
-      - run: update-alternatives --set x86_64-w64-mingw32-g++ /usr/bin/x86_64-w64-mingw32-g++-posix
-      - run:
-          name: "Build cmake-mingw"
-          command: |
-            export PATH=$JAVA_HOME/bin:$PATH
-            echo "JAVA_HOME=${JAVA_HOME}"
-            which java && java -version
-            which javac && javac -version
-            mkdir build && cd build && cmake -DJNI=1 -DWITH_GFLAGS=OFF .. -DCMAKE_C_COMPILER=x86_64-w64-mingw32-gcc -DCMAKE_CXX_COMPILER=x86_64-w64-mingw32-g++ -DCMAKE_SYSTEM_NAME=Windows && make -j4 rocksdb rocksdbjni
-      - post-steps
-
-  build-linux-non-shm:
-    executor: linux-docker
-    resource_class: 2xlarge
-    environment:
-      TEST_TMPDIR: /tmp/rocksdb_test_tmp
-    steps:
-      - pre-steps
-      - run: make V=1 -j32 check
-      - post-steps
-
-  build-linux-arm-test-full:
-    machine:
-      image: ubuntu-2004:202111-02
-    resource_class: arm.large
-    steps:
-      - pre-steps
-      - install-gflags
-      - run: make V=1 J=4 -j4 check
-      - post-steps
-
-  build-linux-arm:
-    machine:
-      image: ubuntu-2004:202111-02
-    resource_class: arm.large
-    steps:
-      - pre-steps
-      - install-gflags
-      - run: ROCKSDBTESTS_PLATFORM_DEPENDENT=only make V=1 J=4 -j4 all_but_some_tests check_some
-      - post-steps
-
-  build-linux-arm-cmake-no_test_run:
-    machine:
-      image: ubuntu-2004:202111-02
-    resource_class: arm.large
-    environment:
-      JAVA_HOME: /usr/lib/jvm/java-8-openjdk-arm64
-    steps:
-      - pre-steps
-      - install-gflags
-      - run:
-          name: "Set Java Environment"
-          command: |
-            echo "JAVA_HOME=${JAVA_HOME}"
-            echo 'export PATH=$JAVA_HOME/bin:$PATH' >> $BASH_ENV
-            which java && java -version
-            which javac && javac -version
-      - run:
-          name: "Build with cmake"
-          command: |
-            mkdir build
-            cd build
-            cmake -DCMAKE_BUILD_TYPE=Release -DWITH_TESTS=0 -DWITH_GFLAGS=1 -DWITH_BENCHMARK_TOOLS=0 -DWITH_TOOLS=0 -DWITH_CORE_TOOLS=1 ..
-            make -j4
-      - run:
-          name: "Build Java with cmake"
-          command: |
-            rm -rf build
-            mkdir build
-            cd build
-            cmake -DJNI=1 -DCMAKE_BUILD_TYPE=Release -DWITH_GFLAGS=1 ..
-            make -j4 rocksdb rocksdbjni
-      - post-steps
-
-  build-format-compatible:
-    executor: linux-docker
-    resource_class: 2xlarge
-    steps:
-      - pre-steps
-      - run:
-          name: "test"
-          command: |
-            export TEST_TMPDIR=/dev/shm/rocksdb
-            rm -rf /dev/shm/rocksdb
-            mkdir /dev/shm/rocksdb
-            tools/check_format_compatible.sh
-      - post-steps
-
-  build-fuzzers:
-    executor: linux-docker
-    resource_class: large
-    steps:
-      - pre-steps
-      - run:
-          name: "Build rocksdb lib"
-          command: CC=clang-13 CXX=clang++-13 USE_CLANG=1 make -j4 static_lib
-      - run:
-          name: "Build fuzzers"
-          command: cd fuzz && make sst_file_writer_fuzzer db_fuzzer db_map_fuzzer
-      - post-steps
-
-  benchmark-linux: #use a private Circle CI runner (resource_class) to run the job
-    machine: true
-    resource_class: facebook/rocksdb-benchmark-sys1
-    steps:
-      - build-for-benchmarks
-      - perform-benchmarks
-      - post-benchmarks
-
-workflows: # Only jobs that haven't been successfully migrated to GitHub Actions
-  version: 2
-  jobs-linux-arm:
-    jobs:
-      - build-linux-arm
-  benchmark-linux:
-    triggers:
-      - schedule:
-          cron: "0 * * * *"
-          filters:
-            branches:
-              only:
-                - main
-    jobs:
-      - benchmark-linux
-  nightly:
-    triggers:
-      - schedule:
-          cron: "0 9 * * *"
-          filters:
-            branches:
-              only:
-                - main
-    jobs:
-      - build-linux-arm-test-full

.circleci/ubsan_suppression_list.txt

@@ -1,6 +0,0 @@
-# Supress UBSAN warnings related to stl_tree.h, e.g.
-# UndefinedBehaviorSanitizer: undefined-behavior /usr/bin/../lib/gcc/x86_64-linux-gnu/5.4.0/../../../../include/c++/5.4.0/bits/stl_tree.h:1505:43 in 
-# /usr/bin/../lib/gcc/x86_64-linux-gnu/5.4.0/../../../../include/c++/5.4.0/bits/stl_tree.h:1505:43:
-# runtime error: upcast of address 0x000001fa8820 with insufficient space for an object of type
-# 'std::_Rb_tree_node<std::pair<const std::__cxx11::basic_string<char>, rocksdb::(anonymous namespace)::LockHoldingInfo> >'
-src:*bits/stl_tree.h

.github/workflows/pr-jobs.yml

@@ -31,7 +31,7 @@ jobs:
   # ======================== Fast Initial Checks ====================== #
   check-format-and-targets:
     if: ${{ github.repository_owner == 'facebook' }}
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-24.04
     steps:
     - uses: actions/checkout@v4.1.0
       with:
@@ -45,7 +45,7 @@ jobs:
     - name: Install argparse
       run: pip install argparse
     - name: Download clang-format-diff.py
-      run: wget https://raw.githubusercontent.com/llvm/llvm-project/release/12.x/clang/tools/clang-format/clang-format-diff.py
+      run: wget https://rocksdb-deps.s3.us-west-2.amazonaws.com/llvm/llvm-project/release/12.x/clang/tools/clang-format/clang-format-diff.py
     - name: Check format
       run: VERBOSE_CHECK=1 make check-format
     - name: Compare buckify output
@@ -390,9 +390,9 @@ jobs:
     steps:
     - uses: actions/checkout@v4.1.0
     - uses: "./.github/actions/pre-steps"
-    - run: COMPILE_WITH_UBSAN=1 OPT="-fsanitize-blacklist=.circleci/ubsan_suppression_list.txt" CC=clang-10 CXX=clang++-10 ROCKSDB_DISABLE_ALIGNED_NEW=1 USE_CLANG=1 make V=1 -j32 ubsan_check
+    - 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
     - uses: "./.github/actions/post-steps"
-  build-linux-clang10-mini-tsan:
+  build-linux-clang13-mini-tsan:
     if: ${{ github.repository_owner == 'facebook' }}
     runs-on:
       labels: 32-core-ubuntu

.gitignore

@@ -91,6 +91,7 @@ buckifier/__pycache__
 .arcconfig
 
 compile_commands.json
+meta_gen_cpp_compile_commands.json
 clang-format-diff.py
 .py3/
 

BUCK

@@ -83,6 +83,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "db/log_writer.cc",
         "db/logs_with_prep_tracker.cc",
         "db/malloc_stats.cc",
+        "db/manifest_ops.cc",
         "db/memtable.cc",
         "db/memtable_list.cc",
         "db/merge_helper.cc",
@@ -213,6 +214,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "table/cuckoo/cuckoo_table_builder.cc",
         "table/cuckoo/cuckoo_table_factory.cc",
         "table/cuckoo/cuckoo_table_reader.cc",
+        "table/external_table.cc",
         "table/format.cc",
         "table/get_context.cc",
         "table/iterator.cc",
@@ -316,6 +318,8 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "utilities/persistent_cache/block_cache_tier_metadata.cc",
         "utilities/persistent_cache/persistent_cache_tier.cc",
         "utilities/persistent_cache/volatile_tier_impl.cc",
+        "utilities/secondary_index/secondary_index_iterator.cc",
+        "utilities/secondary_index/simple_secondary_index.cc",
         "utilities/simulator_cache/cache_simulator.cc",
         "utilities/simulator_cache/sim_cache.cc",
         "utilities/table_properties_collectors/compact_for_tiering_collector.cc",
@@ -367,6 +371,11 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
 
 cpp_library_wrapper(name="rocksdb_whole_archive_lib", srcs=[], deps=[":rocksdb_lib"], headers=[], link_whole=True, extra_test_libs=False)
 
+cpp_library_wrapper(name="rocksdb_with_faiss_lib", srcs=["utilities/secondary_index/faiss_ivf_index.cc"], deps=[
+        "//faiss:faiss",
+        ":rocksdb_lib",
+    ], headers=[], link_whole=False, extra_test_libs=False)
+
 cpp_library_wrapper(name="rocksdb_test_lib", srcs=[
         "db/db_test_util.cc",
         "db/db_with_timestamp_test_util.cc",
@@ -381,11 +390,26 @@ cpp_library_wrapper(name="rocksdb_test_lib", srcs=[
         "utilities/cassandra/test_utils.cc",
     ], deps=[":rocksdb_lib"], headers=[], link_whole=False, extra_test_libs=True)
 
+cpp_library_wrapper(name="rocksdb_with_faiss_test_lib", srcs=[
+        "db/db_test_util.cc",
+        "db/db_with_timestamp_test_util.cc",
+        "table/mock_table.cc",
+        "test_util/mock_time_env.cc",
+        "test_util/secondary_cache_test_util.cc",
+        "test_util/testharness.cc",
+        "test_util/testutil.cc",
+        "tools/block_cache_analyzer/block_cache_trace_analyzer.cc",
+        "tools/trace_analyzer_tool.cc",
+        "utilities/agg_merge/test_agg_merge.cc",
+        "utilities/cassandra/test_utils.cc",
+    ], deps=[":rocksdb_with_faiss_lib"], headers=[], link_whole=False, extra_test_libs=True)
+
 cpp_library_wrapper(name="rocksdb_tools_lib", srcs=[
         "test_util/testutil.cc",
         "tools/block_cache_analyzer/block_cache_trace_analyzer.cc",
         "tools/db_bench_tool.cc",
         "tools/simulated_hybrid_file_system.cc",
+        "tools/tool_hooks.cc",
         "tools/trace_analyzer_tool.cc",
     ], deps=[":rocksdb_lib"], headers=[], link_whole=False, extra_test_libs=False)
 
@@ -5077,6 +5101,12 @@ cpp_unittest_wrapper(name="external_sst_file_test",
             extra_compiler_flags=[])
 
 
+cpp_unittest_wrapper(name="faiss_ivf_index_test",
+            srcs=["utilities/secondary_index/faiss_ivf_index_test.cc"],
+            deps=[":rocksdb_with_faiss_test_lib"],
+            extra_compiler_flags=[])
+
+
 cpp_unittest_wrapper(name="fault_injection_test",
             srcs=["db/fault_injection_test.cc"],
             deps=[":rocksdb_test_lib"],

CMakeLists.txt

@@ -716,6 +716,7 @@ set(SOURCES
         db/log_reader.cc
         db/log_writer.cc
         db/malloc_stats.cc
+        db/manifest_ops.cc
         db/memtable.cc
         db/memtable_list.cc
         db/merge_helper.cc
@@ -834,6 +835,7 @@ set(SOURCES
         table/cuckoo/cuckoo_table_builder.cc
         table/cuckoo/cuckoo_table_factory.cc
         table/cuckoo/cuckoo_table_reader.cc
+        table/external_table.cc
         table/format.cc
         table/get_context.cc
         table/iterator.cc
@@ -938,6 +940,8 @@ set(SOURCES
         utilities/persistent_cache/block_cache_tier_metadata.cc
         utilities/persistent_cache/persistent_cache_tier.cc
         utilities/persistent_cache/volatile_tier_impl.cc
+        utilities/secondary_index/secondary_index_iterator.cc
+        utilities/secondary_index/simple_secondary_index.cc
         utilities/simulator_cache/cache_simulator.cc
         utilities/simulator_cache/sim_cache.cc
         utilities/table_properties_collectors/compact_for_tiering_collector.cc
@@ -1570,6 +1574,7 @@ if(WITH_BENCHMARK_TOOLS)
   add_executable(db_bench${ARTIFACT_SUFFIX}
     tools/simulated_hybrid_file_system.cc
     tools/db_bench.cc
+    tools/tool_hooks.cc
     tools/db_bench_tool.cc)
   target_link_libraries(db_bench${ARTIFACT_SUFFIX}
     ${ROCKSDB_LIB} ${THIRDPARTY_LIBS})

HISTORY.md

@@ -1,6 +1,96 @@
 # Rocksdb Change Log
 > NOTE: Entries for next release do not go here. Follow instructions in `unreleased_history/README.txt`
 
+## 10.0.1 (03/05/2025)
+### Public API Changes
+* Add an unordered map of name/value pairs, ReadOptions::property_bag, to pass opaque options through to an external table when creating an Iterator.
+* Introduced CompactionServiceJobStatus::kAborted to allow handling aborted scenario in Schedule(), Wait() or OnInstallation() APIs in Remote Compactions.
+* Added a column family option disallow_memtable_writes to safely fail any attempts to write to a non-default column family. This can be used for column families that are ingest only.
+
+## 10.0.0 (02/21/2025)
+### New Features
+* Introduced new `auto_refresh_iterator_with_snapshot` opt-in knob that (when enabled) will periodically release obsolete memory and storage resources for as long as the iterator is making progress and its supplied `read_options.snapshot` was initialized with non-nullptr value.
+* Added the ability to plug-in a custom table reader implementation. See include/rocksdb/external_table_reader.h for more details.
+* Experimental feature: RocksDB now supports FAISS inverted file based indices via the secondary indexing framework. Applications can use FAISS secondary indices to automatically quantize embeddings and perform K-nearest-neighbors similarity searches. See `FaissIVFIndex` and `SecondaryIndex` for more details. Note: the FAISS integration currently requires using the BUCK build.
+* Add new DB property `num_running_compaction_sorted_runs` that tracks the number of sorted runs being processed by currently running compactions
+* Experimental feature: added support for simple secondary indices that index the specified column as-is. See `SimpleSecondaryIndex` and `SecondaryIndex` for more details.
+* Added new `TransactionDBOptions::txn_commit_bypass_memtable_threshold`, which enables optimized transaction commit (see `TransactionOptions::commit_bypass_memtable`) when the transaction size exceeds a configured threshold.
+
+### Public API Changes
+* Updated the query API of the experimental secondary indexing feature by removing the earlier `SecondaryIndex::NewIterator` virtual and adding a `SecondaryIndexIterator` class that can be utilized by applications to find the primary keys for a given search target.
+* Added back the ability to leverage the primary key when building secondary index entries. This involved changes to the signatures of `SecondaryIndex::GetSecondary{KeyPrefix,Value}` as well as the addition of a new method `SecondaryIndex::FinalizeSecondaryKeyPrefix`. See the API comments for more details.
+* Minimum supported version of ZSTD is now 1.4.0, for code simplification. Obsolete `CompressionType` `kZSTDNotFinalCompression` is also removed.
+
+### Behavior Changes
+* `VerifyBackup` in `verify_with_checksum`=`true` mode will now evaluate checksums in parallel. As a result, unlike in case of original implementation, the API won't bail out on a very first corruption / mismatch and instead will iterate over all the backup files logging success / _degree_of_failure_ for each.
+* Reversed the order of updates to the same key in WriteBatchWithIndex. This means if there are multiple updates to the same key, the most recent update is ordered first. This affects the output of WBWIIterator. When WriteBatchWithIndex is created with `overwrite_key=true`, this affects the output only if Merge is used (#13387).
+* Added support for Merge operations in transactions using option `TransactionOptions::commit_bypass_memtable`.
+
+### Bug Fixes
+* Fixed GetMergeOperands() API in ReadOnlyDB and SecondaryDB
+* Fix a bug in `GetMergeOperands()` that can return incorrect status (MergeInProgress) and incorrect number of merge operands. This can happen when `GetMergeOperandsOptions::continue_cb` is set, both active and immutable memtables have merge operands and the callback stops the look up at the immutable memtable.
+
+## 9.11.0 (01/17/2025)
+### New Features
+* Introduce CancelAwaitingJobs() in CompactionService interface which will allow users to implement cancellation of running remote compactions from the primary instance
+* Experimental feature: RocksDB now supports defining secondary indices, which are automatically maintained by the storage engine. Secondary indices provide a new customization point: applications can provide their own by implementing the new `SecondaryIndex` interface. See the `SecondaryIndex` API comments for more details. Note: this feature is currently only available in conjunction with write-committed pessimistic transactions, and `Merge` is not yet supported.
+* Provide a new option `track_and_verify_wals` to track and verify various information about WAL during WAL recovery. This is intended to be a better replacement to `track_and_verify_wals_in_manifest`.
+
+### Public API Changes
+* Add `io_buffer_size` to BackupEngineOptions to enable optimal configuration of IO size
+* Clean up all the references to `random_access_max_buffer_size`, related rules and all the clients wrappers. This option has been officially deprecated in 5.4.0.
+* Add `file_ingestion_nanos` and `file_ingestion_blocking_live_writes_nanos` in PerfContext to observe file ingestions
+* Offer new DB::Open and variants that use `std::unique_ptr<DB>*` output parameters and deprecate the old versions that use `DB**` output parameters.
+* The DB::DeleteFile API is officially deprecated.
+
+### Behavior Changes
+* For leveled compaction, manual compaction (CompactRange()) will be more strict about keeping compaction size under `max_compaction_bytes`. This prevents overly large compactions in some cases (#13306).
+* Experimental tiering options `preclude_last_level_data_seconds` and `preserve_internal_time_seconds` are now mutable with `SetOptions()`. Some changes to handling of these features along with long-lived snapshots and range deletes made this possible.
+
+### Bug Fixes
+* Fix a longstanding major bug with SetOptions() in which setting changes can be quietly reverted.
+
+## 9.10.0 (12/12/2024)
+### New Features
+* Introduce `TransactionOptions::commit_bypass_memtable` to enable transaction commit to bypass memtable insertions. This can be beneficial for transactions with many operations, as it reduces commit time that is mostly spent on memtable insertion.
+
+### Public API Changes
+* Deprecated Remote Compaction APIs (StartV2, WaitForCompleteV2) are completely removed from the codebase
+
+### Behavior Changes
+* DB::KeyMayExist() now follows its function comment, which means `value` parameter can be null, and it will be set only if `value_found` is passed in.
+
+### Bug Fixes
+* Fix the issue where compaction incorrectly drops a key when there is a snapshot with a sequence number of zero.
+* Honor ConfigOptions.ignore_unknown_options in ParseStruct()
+
+### Performance Improvements
+* Enable reuse of file system allocated buffer for synchronous prefetching.
+* In buffered IO mode, try to align writes on power of 2 if checksum handoff is not enabled for the file type being written.
+
+## 9.9.0 (11/18/2024)
+### New Features
+* Multi-Column-Family-Iterator (CoalescingIterator/AttributeGroupIterator) is no longer marked as experimental
+* Adds a new table property "rocksdb.newest.key.time" which records the unix timestamp of the newest key. Uses this table property for FIFO TTL and temperature change compaction.
+
+### Public API Changes
+* Added a new API `Transaction::GetAttributeGroupIterator` that can be used to create a multi-column-family attribute group iterator over the specified column families, including the data from both the transaction and the underlying database. This API is currently supported for optimistic and write-committed pessimistic transactions.
+* Added a new API `Transaction::GetCoalescingIterator` that can be used to create a multi-column-family coalescing iterator over the specified column families, including the data from both the transaction and the underlying database. This API is currently supported for optimistic and write-committed pessimistic transactions.
+
+### Behavior Changes
+* `BaseDeltaIterator` now honors the read option `allow_unprepared_value`.
+
+### Bug Fixes
+* `BaseDeltaIterator` now calls `PrepareValue` on the base iterator in case it has been created with the `allow_unprepared_value` read option set. Earlier, such base iterators could lead to incorrect values being exposed from `BaseDeltaIterator`.
+* Fix a leak of obsolete blob files left open until DB::Close(). This bug was introduced in version 9.4.0.
+* Fix missing cases of corruption retry during DB open and read API processing.
+* Fix a bug for transaction db with 2pc where an old WAL may be retained longer than needed (#13127).
+* Fix leaks of some open SST files (until `DB::Close()`) that are written but never become live due to various failures. (We now have a check for such leaks with no outstanding issues.)
+* Fix a bug for replaying WALs for WriteCommitted transaction DB when its user-defined timestamps setting is toggled on/off between DB sessions.
+
+### Performance Improvements
+* Fix regression in issue #12038 due to `Options::compaction_readahead_size` greater than `max_sectors_kb` (i.e, largest I/O size that the OS issues to a block device defined in linux)
+
 ## 9.8.0 (10/25/2024)
 ### New Features
 * All non-`block_cache` options in `BlockBasedTableOptions` are now mutable with `DB::SetOptions()`. See also Bug Fixes below.

LANGUAGE-BINDINGS.md

@@ -2,7 +2,8 @@ This is the list of all known third-party language bindings for RocksDB. If some
 
 * Java - https://github.com/facebook/rocksdb/tree/main/java
 * Python
-    * http://python-rocksdb.readthedocs.io/en/latest/
+    * https://github.com/rocksdict/RocksDict
+    * http://python-rocksdb.readthedocs.io/en/latest/ (unmaintained) 
     * http://pyrocksdb.readthedocs.org/en/latest/ (unmaintained)
 * Perl - https://metacpan.org/pod/RocksDB
 * Node.js - https://npmjs.org/package/rocksdb

Makefile

@@ -659,7 +659,7 @@ ifneq ($(filter check-headers, $(MAKECMDGOALS)),)
 # TODO: add/support JNI headers
 	DEV_HEADER_DIRS := $(sort include/ $(dir $(ALL_SOURCES)))
 # Some headers like in port/ are platform-specific
-	DEV_HEADERS_TO_CHECK := $(shell $(FIND) $(DEV_HEADER_DIRS) -type f -name '*.h' | grep -E -v 'port/|plugin/|lua/|range_tree/')
+	DEV_HEADERS_TO_CHECK := $(shell $(FIND) $(DEV_HEADER_DIRS) -type f -name '*.h' | grep -E -v 'port/|plugin/|lua/|range_tree/|secondary_index/')
 	PUBLIC_HEADERS_TO_CHECK := $(shell $(FIND) include/ -type f -name '*.h' | grep -E -v 'lua/')
 else
 	DEV_HEADERS_TO_CHECK :=
@@ -2489,7 +2489,7 @@ 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 03041f014b6e6ebb6119ffae8b7a37308f52e913
+	cd third-party/folly && git reset --hard f1c1542fb071e9d88181279ec24043400222aef1
 	@# 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

USERS.md

@@ -38,12 +38,11 @@ Snowflake [uses](https://www.snowflake.com/blog/how-foundationdb-powers-snowflak
 The Bing search engine from Microsoft uses RocksDB as the storage engine for its web data platform: https://blogs.bing.com/Engineering-Blog/october-2021/RocksDB-in-Microsoft-Bing
 
 ## LinkedIn
-Two different use cases at Linkedin are using RocksDB as a storage engine:
+1. [Venice](https://venicedb.org/) is a derived data platform using RocksDB as its storage engine. It is LinkedIn's ML feature store, powering thousands of recommender use cases, including the Feed, Video recommendations, and People You May Know.
+2. LinkedIn's follow feed for storing user's activities. Check out the blog post: https://engineering.linkedin.com/blog/2016/03/followfeed--linkedin-s-feed-made-faster-and-smarter
+3. Apache Samza, open source framework for stream processing.
 
-1. LinkedIn's follow feed for storing user's activities. Check out the blog post: https://engineering.linkedin.com/blog/2016/03/followfeed--linkedin-s-feed-made-faster-and-smarter
-2. Apache Samza, open source framework for stream processing
-
-Learn more about those use cases in a Tech Talk by Ankit Gupta and Naveen Somasundaram: http://www.youtube.com/watch?v=plqVp_OnSzg
+Learn more about LinkedIn's follow feed and Apache Samza in a Tech Talk by Ankit Gupta and Naveen Somasundaram: http://www.youtube.com/watch?v=plqVp_OnSzg
 
 ## Yahoo
 Yahoo is using RocksDB as a storage engine for their biggest distributed data store Sherpa. Learn more about it here: http://yahooeng.tumblr.com/post/120730204806/sherpa-scales-new-heights

buckifier/buckify_rocksdb.py

@@ -14,10 +14,10 @@ from targets_builder import TARGETSBuilder, LiteralValue
 
 from util import ColorString
 
-# This script generates TARGETS file for Buck.
+# This script generates BUCK file for Buck.
 # Buck is a build tool specifying dependencies among different build targets.
 # User can pass extra dependencies as a JSON object via command line, and this
-# script can include these dependencies in the generate TARGETS file.
+# script can include these dependencies in the generate BUCK file.
 # Usage:
 # $python3 buckifier/buckify_rocksdb.py
 # (This generates a TARGET file without user-specified dependency for unit
@@ -28,7 +28,7 @@ from util import ColorString
 #                      "extra_compiler_flags": ["-DFOO_BAR", "-Os"]
 #                  }
 #         }'
-# (Generated TARGETS file has test_dep and mock1 as dependencies for RocksDB
+# (Generated BUCK file has test_dep and mock1 as dependencies for RocksDB
 # unit tests, and will use the extra_compiler_flags to compile the unit test
 # source.)
 
@@ -114,9 +114,9 @@ def get_dependencies():
     return deps_map
 
 
-# Prepare TARGETS file for buck
-def generate_targets(repo_path, deps_map):
-    print(ColorString.info("Generating TARGETS"))
+# Prepare BUCK file for buck
+def generate_buck(repo_path, deps_map):
+    print(ColorString.info("Generating BUCK"))
     # parsed src.mk file
     src_mk = parse_src_mk(repo_path)
     # get all .cc files
@@ -133,10 +133,10 @@ def generate_targets(repo_path, deps_map):
         # in how the file was generated.
         extra_argv = " '{}'".format(" ".join(sys.argv[1].split()))
 
-    TARGETS = TARGETSBuilder("%s/TARGETS" % repo_path, extra_argv)
+    BUCK = TARGETSBuilder("%s/BUCK" % repo_path, extra_argv)
 
     # rocksdb_lib
-    TARGETS.add_library(
+    BUCK.add_library(
         "rocksdb_lib",
         src_mk["LIB_SOURCES"] +
         # always add range_tree, it's only excluded on ppc64, which we don't use internally
@@ -152,7 +152,7 @@ def generate_targets(repo_path, deps_map):
         headers=LiteralValue("glob([\"**/*.h\"])")
     )
     # rocksdb_whole_archive_lib
-    TARGETS.add_library(
+    BUCK.add_library(
         "rocksdb_whole_archive_lib",
         [],
         deps=[
@@ -161,8 +161,17 @@ def generate_targets(repo_path, deps_map):
         extra_external_deps="",
         link_whole=True,
     )
+    # rocksdb_with_faiss_lib
+    BUCK.add_library(
+        "rocksdb_with_faiss_lib",
+        src_mk.get("WITH_FAISS_LIB_SOURCES", []),
+        deps=[
+            "//faiss:faiss",
+            ":rocksdb_lib",
+        ],
+    )
     # rocksdb_test_lib
-    TARGETS.add_library(
+    BUCK.add_library(
         "rocksdb_test_lib",
         src_mk.get("MOCK_LIB_SOURCES", [])
         + src_mk.get("TEST_LIB_SOURCES", [])
@@ -171,8 +180,20 @@ def generate_targets(repo_path, deps_map):
         [":rocksdb_lib"],
         extra_test_libs=True,
     )
+    # rocksdb_with_faiss_test_lib
+    BUCK.add_library(
+        "rocksdb_with_faiss_test_lib",
+        src_mk.get("MOCK_LIB_SOURCES", [])
+        + src_mk.get("TEST_LIB_SOURCES", [])
+        + src_mk.get("EXP_LIB_SOURCES", [])
+        + src_mk.get("ANALYZER_LIB_SOURCES", []),
+        deps=[
+            ":rocksdb_with_faiss_lib",
+        ],
+        extra_test_libs=True,
+    )
     # rocksdb_tools_lib
-    TARGETS.add_library(
+    BUCK.add_library(
         "rocksdb_tools_lib",
         src_mk.get("BENCH_LIB_SOURCES", [])
         + src_mk.get("ANALYZER_LIB_SOURCES", [])
@@ -180,51 +201,51 @@ def generate_targets(repo_path, deps_map):
         [":rocksdb_lib"],
     )
     # rocksdb_cache_bench_tools_lib
-    TARGETS.add_library(
+    BUCK.add_library(
         "rocksdb_cache_bench_tools_lib",
         src_mk.get("CACHE_BENCH_LIB_SOURCES", []),
         [":rocksdb_lib"],
     )
     # rocksdb_stress_lib
-    TARGETS.add_rocksdb_library(
+    BUCK.add_rocksdb_library(
         "rocksdb_stress_lib",
         src_mk.get("ANALYZER_LIB_SOURCES", [])
         + src_mk.get("STRESS_LIB_SOURCES", [])
         + ["test_util/testutil.cc"],
     )
     # ldb binary
-    TARGETS.add_binary(
+    BUCK.add_binary(
         "ldb", ["tools/ldb.cc"], [":rocksdb_tools_lib"]
     )
     # db_stress binary
-    TARGETS.add_binary(
+    BUCK.add_binary(
         "db_stress", ["db_stress_tool/db_stress.cc"], [":rocksdb_stress_lib"]
     )
     # db_bench binary
-    TARGETS.add_binary(
+    BUCK.add_binary(
         "db_bench", ["tools/db_bench.cc"], [":rocksdb_tools_lib"]
     )
     # cache_bench binary
-    TARGETS.add_binary(
+    BUCK.add_binary(
         "cache_bench", ["cache/cache_bench.cc"], [":rocksdb_cache_bench_tools_lib"]
     )
     # bench binaries
     for src in src_mk.get("MICROBENCH_SOURCES", []):
         name = src.rsplit("/", 1)[1].split(".")[0] if "/" in src else src.split(".")[0]
-        TARGETS.add_binary(name, [src], [], extra_bench_libs=True)
+        BUCK.add_binary(name, [src], [], extra_bench_libs=True)
     print(f"Extra dependencies:\n{json.dumps(deps_map)}")
 
     # Dictionary test executable name -> relative source file path
     test_source_map = {}
 
-    # c_test.c is added through TARGETS.add_c_test(). If there
+    # c_test.c is added through BUCK.add_c_test(). If there
     # are more than one .c test file, we need to extend
-    # TARGETS.add_c_test() to include other C tests too.
+    # BUCK.add_c_test() to include other C tests too.
     for test_src in src_mk.get("TEST_MAIN_SOURCES_C", []):
         if test_src != "db/c_test.c":
             print("Don't know how to deal with " + test_src)
             return False
-    TARGETS.add_c_test()
+    BUCK.add_c_test()
 
     try:
         with open(f"{repo_path}/buckifier/bench.json") as json_file:
@@ -239,7 +260,7 @@ def generate_targets(repo_path, deps_map):
                         for metric in overloaded_metric_list:
                             if not isinstance(metric, dict):
                                 clean_benchmarks[binary][benchmark].append(metric)
-                TARGETS.add_fancy_bench_config(
+                BUCK.add_fancy_bench_config(
                     config_dict["name"],
                     clean_benchmarks,
                     False,
@@ -261,7 +282,7 @@ def generate_targets(repo_path, deps_map):
                             if not isinstance(metric, dict):
                                 clean_benchmarks[binary][benchmark].append(metric)
             for config_dict in slow_fancy_bench_config_list:
-                TARGETS.add_fancy_bench_config(
+                BUCK.add_fancy_bench_config(
                     config_dict["name"] + "_slow",
                     clean_benchmarks,
                     True,
@@ -274,15 +295,20 @@ def generate_targets(repo_path, deps_map):
     except Exception:
         pass
 
-    TARGETS.add_test_header()
+    BUCK.add_test_header()
 
     for test_src in src_mk.get("TEST_MAIN_SOURCES", []):
         test = test_src.split(".c")[0].strip().split("/")[-1].strip()
-        test_source_map[test] = test_src
+        test_source_map[test] = (test_src, False)
         print("" + test + " " + test_src)
 
+    for test_src in src_mk.get("WITH_FAISS_TEST_MAIN_SOURCES", []):
+        test = test_src.split(".c")[0].strip().split("/")[-1].strip()
+        test_source_map[test] = (test_src, True)
+        print("" + test + " " + test_src + " [FAISS]")
+
     for target_alias, deps in deps_map.items():
-        for test, test_src in sorted(test_source_map.items()):
+        for test, (test_src, with_faiss) in sorted(test_source_map.items()):
             if len(test) == 0:
                 print(ColorString.warning("Failed to get test name for %s" % test_src))
                 continue
@@ -291,31 +317,39 @@ def generate_targets(repo_path, deps_map):
 
             if test in _EXPORTED_TEST_LIBS:
                 test_library = "%s_lib" % test_target_name
-                TARGETS.add_library(
+                BUCK.add_library(
                     test_library,
                     [test_src],
                     deps=[":rocksdb_test_lib"],
                     extra_test_libs=True,
                 )
-                TARGETS.register_test(
+                BUCK.register_test(
                     test_target_name,
                     test_src,
                     deps=json.dumps(deps["extra_deps"] + [":" + test_library]),
                     extra_compiler_flags=json.dumps(deps["extra_compiler_flags"]),
                 )
             else:
-                TARGETS.register_test(
-                    test_target_name,
-                    test_src,
-                    deps=json.dumps(deps["extra_deps"] + [":rocksdb_test_lib"]),
-                    extra_compiler_flags=json.dumps(deps["extra_compiler_flags"]),
-                )
-    TARGETS.export_file("tools/db_crashtest.py")
+                if with_faiss:
+                    BUCK.register_test(
+                        test_target_name,
+                        test_src,
+                        deps=json.dumps(deps["extra_deps"] + [":rocksdb_with_faiss_test_lib"]),
+                        extra_compiler_flags=json.dumps(deps["extra_compiler_flags"]),
+                    )
+                else:
+                    BUCK.register_test(
+                        test_target_name,
+                        test_src,
+                        deps=json.dumps(deps["extra_deps"] + [":rocksdb_test_lib"]),
+                        extra_compiler_flags=json.dumps(deps["extra_compiler_flags"]),
+                    )
+    BUCK.export_file("tools/db_crashtest.py")
 
-    print(ColorString.info("Generated TARGETS Summary:"))
-    print(ColorString.info("- %d libs" % TARGETS.total_lib))
-    print(ColorString.info("- %d binarys" % TARGETS.total_bin))
-    print(ColorString.info("- %d tests" % TARGETS.total_test))
+    print(ColorString.info("Generated BUCK Summary:"))
+    print(ColorString.info("- %d libs" % BUCK.total_lib))
+    print(ColorString.info("- %d binarys" % BUCK.total_bin))
+    print(ColorString.info("- %d tests" % BUCK.total_test))
     return True
 
 
@@ -335,10 +369,10 @@ def exit_with_error(msg):
 
 def main():
     deps_map = get_dependencies()
-    # Generate TARGETS file for buck
-    ok = generate_targets(get_rocksdb_path(), deps_map)
+    # Generate BUCK file for buck
+    ok = generate_buck(get_rocksdb_path(), deps_map)
     if not ok:
-        exit_with_error("Failed to generate TARGETS files")
+        exit_with_error("Failed to generate BUCK files")
 
 
 if __name__ == "__main__":

buckifier/check_buck_targets.sh

@@ -1,32 +1,44 @@
 #!/usr/bin/env bash
 # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
-# If clang_format_diff.py command is not specfied, we assume we are able to
-# access directly without any path.
 
-TGT_DIFF=`git diff TARGETS | head -n 1`
+if [[ ! -f "BUCK" ]]
+then
+  echo "BUCK file is missing!"
+  echo "Please do not remove / rename BUCK file in your commit(s)."
+  exit 1
+fi
+
+TGT_DIFF=`git diff BUCK | head -n 1`
 
 if [ ! -z "$TGT_DIFF" ]
 then
-  echo "TARGETS file has uncommitted changes. Skip this check."
+  echo "BUCK file has uncommitted changes. Skip this check."
   exit 0
 fi
 
-echo Backup original TARGETS file.
+echo Backup original BUCK file.
 
-cp TARGETS TARGETS.bkp
+cp BUCK BUCK.bkp
 
 ${PYTHON:-python3} buckifier/buckify_rocksdb.py
 
-TGT_DIFF=`git diff TARGETS | head -n 1`
+if [[ ! -f "BUCK" ]]
+then
+  echo "BUCK file went missing after running buckifier/buckify_rocksdb.py!"
+  echo "Please do not remove the BUCK file."
+  exit 1
+fi
+
+TGT_DIFF=`git diff BUCK | head -n 1`
 
 if [ -z "$TGT_DIFF" ]
 then
-  mv TARGETS.bkp TARGETS
+  mv BUCK.bkp BUCK
   exit 0
 else
-  echo "Please run '${PYTHON:-python3} buckifier/buckify_rocksdb.py' to update TARGETS file."
-  echo "Do not manually update TARGETS file."
+  echo "Please run '${PYTHON:-python3} buckifier/buckify_rocksdb.py' to update BUCK file."
+  echo "Do not manually update BUCK file."
   ${PYTHON:-python3} --version
-  mv TARGETS.bkp TARGETS
+  mv BUCK.bkp BUCK
   exit 1
 fi

build_tools/build_detect_platform

@@ -360,9 +360,13 @@ EOF
     fi
 
     if ! test $ROCKSDB_DISABLE_ZSTD; then
-        # Test whether zstd library is installed
+        # Test whether zstd library is installed with minimum version
+        # (Keep in sync with compression.h)
         $CXX $PLATFORM_CXXFLAGS $COMMON_FLAGS -x c++ - -o /dev/null 2>/dev/null  <<EOF
           #include <zstd.h>
+          #if ZSTD_VERSION_NUMBER < 10400
+          #error "ZSTD support requires version >= 1.4.0 (libzstd-devel)"
+          #endif  // ZSTD_VERSION_NUMBER
           int main() {}
 EOF
         if [ "$?" = 0 ]; then

build_tools/fbcode_config_platform010.sh

@@ -59,7 +59,7 @@ fi
 if ! test $ROCKSDB_DISABLE_ZSTD; then
   ZSTD_INCLUDE=" -I $ZSTD_BASE/include/"
   ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd${MAYBE_PIC}.a"
-  CFLAGS+=" -DZSTD"
+  CFLAGS+=" -DZSTD -DZSTD_STATIC_LINKING_ONLY"
 fi
 
 # location of gflags headers and libraries

cache/cache.cc

@@ -118,7 +118,6 @@ Status SecondaryCache::CreateFromString(
       sec_cache = NewCompressedSecondaryCache(sec_cache_opts);
     }
 
-
     if (status.ok()) {
       result->swap(sec_cache);
     }

cache/cache_bench_tool.cc

@@ -191,8 +191,7 @@ namespace {
 class SharedState {
  public:
   explicit SharedState(CacheBench* cache_bench)
-      : cv_(&mu_),
-        cache_bench_(cache_bench) {}
+      : cv_(&mu_), cache_bench_(cache_bench) {}
 
   ~SharedState() = default;
 

cache/compressed_secondary_cache_test.cc

@@ -845,7 +845,6 @@ TEST_P(CompressedSecondaryCacheTestWithCompressionParam,
   BasicTestHelper(sec_cache, sec_cache_is_compressed_);
 }
 
-
 TEST_P(CompressedSecondaryCacheTestWithCompressionParam, FailsTest) {
   FailsTest(sec_cache_is_compressed_);
 }

cache/lru_cache.cc

@@ -93,9 +93,8 @@ void LRUHandleTable::Resize() {
 
   uint32_t old_length = uint32_t{1} << length_bits_;
   int new_length_bits = length_bits_ + 1;
-  std::unique_ptr<LRUHandle* []> new_list {
-    new LRUHandle* [size_t{1} << new_length_bits] {}
-  };
+  std::unique_ptr<LRUHandle*[]> new_list{
+      new LRUHandle* [size_t{1} << new_length_bits] {}};
   [[maybe_unused]] uint32_t count = 0;
   for (uint32_t i = 0; i < old_length; i++) {
     LRUHandle* h = list_[i];

cache/lru_cache_test.cc

@@ -546,7 +546,7 @@ TYPED_TEST(ClockCacheTest, Limits) {
     // verify usage tracking on detached entries.)
     {
       size_t n = kCapacity * 5 + 1;
-      std::unique_ptr<HandleImpl* []> ha { new HandleImpl* [n] {} };
+      std::unique_ptr<HandleImpl*[]> ha{new HandleImpl* [n] {}};
       Status s;
       for (size_t i = 0; i < n && s.ok(); ++i) {
         hkey[1] = i;
@@ -713,7 +713,7 @@ TYPED_TEST(ClockCacheTest, ClockEvictionEffortCapTest) {
       // evictable entries are present beyond the cache capacity, despite
       // being evictable.
       constexpr size_t kCount = kCapacity - 1;
-      std::unique_ptr<HandleImpl* []> ha { new HandleImpl* [kCount] {} };
+      std::unique_ptr<HandleImpl*[]> ha{new HandleImpl* [kCount] {}};
       for (size_t i = 0; i < 2 * kCount; ++i) {
         UniqueId64x2 hkey = this->CheapHash(i);
         ASSERT_OK(shard.Insert(

cache/secondary_cache.cc

@@ -7,6 +7,4 @@
 
 #include "cache/cache_entry_roles.h"
 
-namespace ROCKSDB_NAMESPACE {
-
-}  // namespace ROCKSDB_NAMESPACE
+namespace ROCKSDB_NAMESPACE {}  // namespace ROCKSDB_NAMESPACE

cache/tiered_secondary_cache_test.cc

@@ -951,12 +951,20 @@ TEST_F(DBTieredSecondaryCacheTest, FSBufferTest) {
                            IODebugContext* dbg) override {
           for (size_t i = 0; i < num_reqs; ++i) {
             FSReadRequest& req = reqs[i];
-            FSAllocationPtr buffer(new char[req.len], [](void* ptr) {
-              delete[] static_cast<char*>(ptr);
-            });
-            req.fs_scratch = std::move(buffer);
+            // See https://github.com/facebook/rocksdb/pull/13195 for why we
+            // want to set up our test implementation for FSAllocationPtr this
+            // way.
+            char* internalData = new char[req.len];
             req.status = Read(req.offset, req.len, options, &req.result,
-                              static_cast<char*>(req.fs_scratch.get()), dbg);
+                              internalData, dbg);
+
+            Slice* internalSlice = new Slice(internalData, req.len);
+            FSAllocationPtr internalPtr(internalSlice, [](void* ptr) {
+              delete[] static_cast<const char*>(
+                  static_cast<Slice*>(ptr)->data_);
+              delete static_cast<Slice*>(ptr);
+            });
+            req.fs_scratch = std::move(internalPtr);
           }
           return IOStatus::OK();
         }

db/arena_wrapped_db_iter.cc

@@ -64,8 +64,118 @@ void ArenaWrappedDBIter::Init(
   memtable_range_tombstone_iter_ = nullptr;
 }
 
+void ArenaWrappedDBIter::MaybeAutoRefresh(bool is_seek,
+                                          DBIter::Direction direction) {
+  if (cfh_ != nullptr && read_options_.snapshot != nullptr && allow_refresh_ &&
+      read_options_.auto_refresh_iterator_with_snapshot) {
+    // The intent here is to capture the superversion number change
+    // reasonably soon from the time it actually happened. As such,
+    // we're fine with weaker synchronization / ordering guarantees
+    // provided by relaxed atomic (in favor of less CPU / mem overhead).
+    uint64_t cur_sv_number = cfh_->cfd()->GetSuperVersionNumberRelaxed();
+    if ((sv_number_ != cur_sv_number) && status().ok()) {
+      // Changing iterators' direction is pretty heavy-weight operation and
+      // could have unintended consequences when it comes to prefix seek.
+      // Therefore, we need an efficient implementation that does not duplicate
+      // the effort by doing things like double seek(forprev).
+      //
+      // Auto refresh can be triggered on the following groups of operations:
+      //
+      //  1. [Seek]: Seek(), SeekForPrev()
+      //  2. [Non-Seek]: Next(), Prev()
+      //
+      // In case of 'Seek' group, procedure is fairly straightforward as we'll
+      // simply call refresh and then invoke the operation on intended target.
+      //
+      // In case of 'Non-Seek' group, we'll first advance the cursor by invoking
+      // intended user operation (Next() or Prev()), capture the target key T,
+      // refresh the iterator and then reconcile the refreshed iterator by
+      // explicitly calling [Seek(T) or SeekForPrev(T)]. Below is an example
+      // flow for Next(), but same principle applies to Prev():
+      //
+      //
+      //          T0: Before the operation     T1: Execute Next()
+      //                         |                      |
+      //                         |         -------------
+      //                         |        |   * capture the key (T)
+      //           DBIter(SV#A)  |        |
+      //          --------------\ /------\ /---------
+      //  SV #A  |     ... ->  [ X ] -> [ T ] -> ... |
+      //          -----------------------------------
+      //                                / |
+      //                               /  |
+      //                              /  T2: Refresh iterator
+      //                             /
+      //           DBIter(SV#A')    /
+      //          ----------------------------------
+      //  SV #A' |       ... ->  [ T ] -> ...       |
+      //          ----------------/ \---------------
+      //                           |
+      //                            ---- T3: Seek(T)
+      //
+      bool valid = false;
+      std::string key;
+      if (!is_seek && db_iter_->Valid()) {
+        // The key() Slice is valid until the iterator state changes.
+        // Given that refresh is heavy-weight operation it itself,
+        // we should copy the target key upfront to avoid reading bad value.
+        valid = true;
+        key = db_iter_->key().ToString();
+      }
+
+      // It's perfectly fine to unref the corresponding superversion
+      // as we rely on pinning behavior of snapshot for consistency.
+      DoRefresh(read_options_.snapshot, cur_sv_number);
+
+      if (!is_seek && valid) {  // Reconcile new iterator after Next() / Prev()
+        if (direction == DBIter::kForward) {
+          db_iter_->Seek(key);
+        } else {
+          assert(direction == DBIter::kReverse);
+          db_iter_->SeekForPrev(key);
+        }
+      }
+    }
+  }
+}
+
 Status ArenaWrappedDBIter::Refresh() { return Refresh(nullptr); }
 
+void ArenaWrappedDBIter::DoRefresh(const Snapshot* snapshot,
+                                   [[maybe_unused]] uint64_t sv_number) {
+  Env* env = db_iter_->env();
+
+  // NOTE:
+  //
+  // Errors like file deletion (as a part of SV cleanup in ~DBIter) will be
+  // present in the error log, but won't be reflected in the iterator status.
+  // This is by design as we expect compaction to clean up those obsolete files
+  // eventually.
+  db_iter_->~DBIter();
+
+  arena_.~Arena();
+  new (&arena_) Arena();
+
+  auto cfd = cfh_->cfd();
+  auto db_impl = cfh_->db();
+
+  SuperVersion* sv = cfd->GetReferencedSuperVersion(db_impl);
+  assert(sv->version_number >= sv_number);
+  SequenceNumber read_seq = GetSeqNum(db_impl, snapshot);
+  if (read_callback_) {
+    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_);
+
+  InternalIterator* internal_iter = db_impl->NewInternalIterator(
+      read_options_, cfd, sv, &arena_, read_seq,
+      /* allow_unprepared_value */ true, /* db_iter */ this);
+  SetIterUnderDBIter(internal_iter);
+}
+
 Status ArenaWrappedDBIter::Refresh(const Snapshot* snapshot) {
   if (cfh_ == nullptr || !allow_refresh_) {
     return Status::NotSupported("Creating renew iterator is not allowed.");
@@ -85,32 +195,9 @@ Status ArenaWrappedDBIter::Refresh(const Snapshot* snapshot) {
   TEST_SYNC_POINT("ArenaWrappedDBIter::Refresh:1");
   TEST_SYNC_POINT("ArenaWrappedDBIter::Refresh:2");
 
-  auto reinit_internal_iter = [&]() {
-    Env* env = db_iter_->env();
-    db_iter_->~DBIter();
-    arena_.~Arena();
-    new (&arena_) Arena();
-
-    SuperVersion* sv = cfd->GetReferencedSuperVersion(db_impl);
-    assert(sv->version_number >= cur_sv_number);
-    SequenceNumber read_seq = GetSeqNum(db_impl, snapshot);
-    if (read_callback_) {
-      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_);
-
-    InternalIterator* internal_iter = db_impl->NewInternalIterator(
-        read_options_, cfd, sv, &arena_, read_seq,
-        /* allow_unprepared_value */ true, /* db_iter */ this);
-    SetIterUnderDBIter(internal_iter);
-  };
   while (true) {
     if (sv_number_ != cur_sv_number) {
-      reinit_internal_iter();
+      DoRefresh(snapshot, cur_sv_number);
       break;
     } else {
       SequenceNumber read_seq = GetSeqNum(db_impl, snapshot);
@@ -138,7 +225,7 @@ Status ArenaWrappedDBIter::Refresh(const Snapshot* snapshot) {
             db_impl->ReturnAndCleanupSuperVersion(cfd, sv);
             // The memtable under DBIter did not have range tombstone before
             // refresh.
-            reinit_internal_iter();
+            DoRefresh(snapshot, cur_sv_number);
             break;
           } else {
             *memtable_range_tombstone_iter_ =

db/arena_wrapped_db_iter.h

@@ -65,12 +65,26 @@ class ArenaWrappedDBIter : public Iterator {
   void SeekToLast() override { db_iter_->SeekToLast(); }
   // 'target' does not contain timestamp, even if user timestamp feature is
   // enabled.
-  void Seek(const Slice& target) override { db_iter_->Seek(target); }
+  void Seek(const Slice& target) override {
+    MaybeAutoRefresh(true /* is_seek */, DBIter::kForward);
+    db_iter_->Seek(target);
+  }
+
   void SeekForPrev(const Slice& target) override {
+    MaybeAutoRefresh(true /* is_seek */, DBIter::kReverse);
     db_iter_->SeekForPrev(target);
   }
-  void Next() override { db_iter_->Next(); }
-  void Prev() override { db_iter_->Prev(); }
+
+  void Next() override {
+    db_iter_->Next();
+    MaybeAutoRefresh(false /* is_seek */, DBIter::kForward);
+  }
+
+  void Prev() override {
+    db_iter_->Prev();
+    MaybeAutoRefresh(false /* is_seek */, DBIter::kReverse);
+  }
+
   Slice key() const override { return db_iter_->key(); }
   Slice value() const override { return db_iter_->value(); }
   const WideColumns& columns() const override { return db_iter_->columns(); }
@@ -103,6 +117,9 @@ class ArenaWrappedDBIter : public Iterator {
   }
 
  private:
+  void DoRefresh(const Snapshot* snapshot, uint64_t sv_number);
+  void MaybeAutoRefresh(bool is_seek, DBIter::Direction direction);
+
   DBIter* db_iter_ = nullptr;
   Arena arena_;
   uint64_t sv_number_;

db/blob/blob_source.cc

@@ -82,8 +82,8 @@ Status BlobSource::PutBlobIntoCache(
   assert(cached_blob->IsEmpty());
 
   TypedHandle* cache_handle = nullptr;
-  const Status s = InsertEntryIntoCache(cache_key, blob->get(),
-                                        &cache_handle, Cache::Priority::BOTTOM);
+  const Status s = InsertEntryIntoCache(cache_key, blob->get(), &cache_handle,
+                                        Cache::Priority::BOTTOM);
   if (s.ok()) {
     blob->release();
 

db/c.cc

@@ -231,7 +231,7 @@ struct rocksdb_livefiles_t {
 };
 struct rocksdb_column_family_handle_t {
   ColumnFamilyHandle* rep;
-  bool immortal;  /* only true for default cf */
+  bool immortal; /* only true for default cf */
 };
 struct rocksdb_column_family_metadata_t {
   ColumnFamilyMetaData rep;
@@ -1858,10 +1858,6 @@ extern ROCKSDB_LIBRARY_API void rocksdb_approximate_sizes_cf_with_flags(
   delete[] ranges;
 }
 
-void rocksdb_delete_file(rocksdb_t* db, const char* name) {
-  db->rep->DeleteFile(name);
-}
-
 const rocksdb_livefiles_t* rocksdb_livefiles(rocksdb_t* db) {
   rocksdb_livefiles_t* result = new rocksdb_livefiles_t;
   db->rep->GetLiveFilesMetaData(&result->rep);
@@ -5000,6 +4996,35 @@ int rocksdb_compactoptions_get_target_level(rocksdb_compactoptions_t* opt) {
   return opt->rep.target_level;
 }
 
+void rocksdb_compactoptions_set_target_path_id(rocksdb_compactoptions_t* opt,
+                                               int n) {
+  opt->rep.target_path_id = n;
+}
+
+int rocksdb_compactoptions_get_target_path_id(rocksdb_compactoptions_t* opt) {
+  return opt->rep.target_path_id;
+}
+
+void rocksdb_compactoptions_set_allow_write_stall(rocksdb_compactoptions_t* opt,
+                                                  unsigned char v) {
+  opt->rep.allow_write_stall = v;
+}
+
+unsigned char rocksdb_compactoptions_get_allow_write_stall(
+    rocksdb_compactoptions_t* opt) {
+  return opt->rep.allow_write_stall;
+}
+
+void rocksdb_compactoptions_set_max_subcompactions(
+    rocksdb_compactoptions_t* opt, int n) {
+  opt->rep.max_subcompactions = n;
+}
+
+int rocksdb_compactoptions_get_max_subcompactions(
+    rocksdb_compactoptions_t* opt) {
+  return opt->rep.max_subcompactions;
+}
+
 void rocksdb_compactoptions_set_full_history_ts_low(
     rocksdb_compactoptions_t* opt, char* ts, size_t tslen) {
   if (ts == nullptr) {

db/c_test.c

@@ -2938,6 +2938,15 @@ int main(int argc, char** argv) {
     rocksdb_compactoptions_set_target_level(co, 1);
     CheckCondition(1 == rocksdb_compactoptions_get_target_level(co));
 
+    rocksdb_compactoptions_set_target_path_id(co, 1);
+    CheckCondition(1 == rocksdb_compactoptions_get_target_path_id(co));
+
+    rocksdb_compactoptions_set_allow_write_stall(co, 1);
+    CheckCondition(1 == rocksdb_compactoptions_get_allow_write_stall(co));
+
+    rocksdb_compactoptions_set_max_subcompactions(co, 1);
+    CheckCondition(1 == rocksdb_compactoptions_get_max_subcompactions(co));
+
     rocksdb_compactoptions_destroy(co);
   }
 

db/column_family.cc

@@ -53,7 +53,7 @@ ColumnFamilyHandleImpl::ColumnFamilyHandleImpl(
 
 ColumnFamilyHandleImpl::~ColumnFamilyHandleImpl() {
   if (cfd_ != nullptr) {
-    for (auto& listener : cfd_->ioptions()->listeners) {
+    for (auto& listener : cfd_->ioptions().listeners) {
       listener->OnColumnFamilyHandleDeletionStarted(this);
     }
     // Job id == 0 means that this is not our background process, but rather
@@ -168,7 +168,8 @@ Status CheckConcurrentWritesSupported(const ColumnFamilyOptions& cf_options) {
   }
   if (!cf_options.memtable_factory->IsInsertConcurrentlySupported()) {
     return Status::InvalidArgument(
-        "Memtable doesn't concurrent writes (allow_concurrent_memtable_write)");
+        "Memtable doesn't allow concurrent writes "
+        "(allow_concurrent_memtable_write)");
   }
   return Status::OK();
 }
@@ -239,6 +240,10 @@ ColumnFamilyOptions SanitizeOptions(const ImmutableDBOptions& db_options,
 
     result.min_write_buffer_number_to_merge = 1;
   }
+  if (result.disallow_memtable_writes) {
+    // A simple memtable that enforces MarkReadOnly (unlike skip list)
+    result.memtable_factory = std::make_shared<VectorRepFactory>();
+  }
 
   if (result.num_levels < 1) {
     result.num_levels = 1;
@@ -593,7 +598,7 @@ ColumnFamilyData::ColumnFamilyData(
                                       block_cache_tracer, io_tracer,
                                       db_session_id));
     blob_file_cache_.reset(
-        new BlobFileCache(_table_cache, ioptions(), soptions(), id_,
+        new BlobFileCache(_table_cache, &ioptions(), soptions(), id_,
                           internal_stats_->GetBlobFileReadHist(), io_tracer));
     blob_source_.reset(new BlobSource(ioptions_, mutable_cf_options_, db_id,
                                       db_session_id, blob_file_cache_.get()));
@@ -968,7 +973,7 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
     auto write_stall_condition_and_cause = GetWriteStallConditionAndCause(
         imm()->NumNotFlushed(), vstorage->l0_delay_trigger_count(),
         vstorage->estimated_compaction_needed_bytes(), mutable_cf_options,
-        *ioptions());
+        ioptions());
     write_stall_condition = write_stall_condition_and_cause.first;
     auto write_stall_cause = write_stall_condition_and_cause.second;
 
@@ -1160,12 +1165,13 @@ MemTable* ColumnFamilyData::ConstructNewMemtable(
                       write_buffer_manager_, earliest_seq, id_);
 }
 
-void ColumnFamilyData::CreateNewMemtable(
-    const MutableCFOptions& mutable_cf_options, SequenceNumber earliest_seq) {
+void ColumnFamilyData::CreateNewMemtable(SequenceNumber earliest_seq) {
   if (mem_ != nullptr) {
     delete mem_->Unref();
   }
-  SetMemtable(ConstructNewMemtable(mutable_cf_options, earliest_seq));
+  // NOTE: db mutex must be locked for SetMemtable, so safe for
+  // GetLatestMutableCFOptions
+  SetMemtable(ConstructNewMemtable(GetLatestMutableCFOptions(), earliest_seq));
   mem_->Ref();
 }
 
@@ -1208,7 +1214,7 @@ Status ColumnFamilyData::RangesOverlapWithMemtables(
   MergeIteratorBuilder merge_iter_builder(&internal_comparator_, &arena);
   merge_iter_builder.AddIterator(super_version->mem->NewIterator(
       read_opts, /*seqno_to_time_mapping=*/nullptr, &arena,
-      /*prefix_extractor=*/nullptr));
+      /*prefix_extractor=*/nullptr, /*for_flush=*/false));
   super_version->imm->AddIterators(read_opts, /*seqno_to_time_mapping=*/nullptr,
                                    /*prefix_extractor=*/nullptr,
                                    &merge_iter_builder,
@@ -1384,7 +1390,7 @@ void ColumnFamilyData::InstallSuperVersion(
         new_superversion->write_stall_condition) {
       sv_context->PushWriteStallNotification(
           old_superversion->write_stall_condition,
-          new_superversion->write_stall_condition, GetName(), ioptions());
+          new_superversion->write_stall_condition, GetName(), &ioptions());
     }
     if (old_superversion->Unref()) {
       old_superversion->Cleanup();
@@ -1844,10 +1850,7 @@ const ImmutableOptions& GetImmutableOptions(ColumnFamilyHandle* column_family) {
   const ColumnFamilyData* const cfd = handle->cfd();
   assert(cfd);
 
-  const ImmutableOptions* ioptions = cfd->ioptions();
-  assert(ioptions);
-
-  return *ioptions;
+  return cfd->ioptions();
 }
 
 }  // namespace ROCKSDB_NAMESPACE

db/column_family.h

@@ -210,6 +210,7 @@ struct SuperVersion {
   ReadOnlyMemTable* mem;
   MemTableListVersion* imm;
   Version* current;
+  // TODO: do we really need this in addition to what's in current Version?
   MutableCFOptions mutable_cf_options;
   // Version number of the current SuperVersion
   uint64_t version_number;
@@ -221,7 +222,8 @@ struct SuperVersion {
   // enable UDT feature, this is an empty string.
   std::string full_history_ts_low;
 
-  // A shared copy of the DB's seqno to time mapping.
+  // An immutable snapshot of the DB's seqno to time mapping, usually shared
+  // between SuperVersions.
   std::shared_ptr<const SeqnoToTimeMapping> seqno_to_time_mapping{nullptr};
 
   // should be called outside the mutex
@@ -342,17 +344,17 @@ class ColumnFamilyData {
 
   // thread-safe
   const FileOptions* soptions() const;
-  const ImmutableOptions* ioptions() const { return &ioptions_; }
+  const ImmutableOptions& ioptions() const { return ioptions_; }
   // REQUIRES: DB mutex held
   // This returns the MutableCFOptions used by current SuperVersion
   // You should use this API to reference MutableCFOptions most of the time.
-  const MutableCFOptions* GetCurrentMutableCFOptions() const {
-    return &(super_version_->mutable_cf_options);
+  const MutableCFOptions& GetCurrentMutableCFOptions() const {
+    return super_version_->mutable_cf_options;
   }
   // REQUIRES: DB mutex held
   // This returns the latest MutableCFOptions, which may be not in effect yet.
-  const MutableCFOptions* GetLatestMutableCFOptions() const {
-    return &mutable_cf_options_;
+  const MutableCFOptions& GetLatestMutableCFOptions() const {
+    return mutable_cf_options_;
   }
 
   // REQUIRES: DB mutex held
@@ -379,27 +381,35 @@ class ColumnFamilyData {
     return mem()->GetFirstSequenceNumber() == 0 && imm()->NumNotFlushed() == 0;
   }
 
-  Version* current() { return current_; }
   Version* dummy_versions() { return dummy_versions_; }
-  void SetCurrent(Version* _current);
+  Version* current() { return current_; }  // REQUIRE: DB mutex held
+  void SetCurrent(Version* _current);      // REQUIRE: DB mutex held
   uint64_t GetNumLiveVersions() const;    // REQUIRE: DB mutex held
   uint64_t GetTotalSstFilesSize() const;  // REQUIRE: DB mutex held
   uint64_t GetLiveSstFilesSize() const;   // REQUIRE: DB mutex held
   uint64_t GetTotalBlobFileSize() const;  // REQUIRE: DB mutex held
   // REQUIRE: DB mutex held
   void SetMemtable(MemTable* new_mem) {
-    new_mem->SetID(++last_memtable_id_);
+    AssignMemtableID(new_mem);
     mem_ = new_mem;
+    if (ioptions_.disallow_memtable_writes) {
+      mem_->MarkImmutable();
+    }
+  }
+
+  void AssignMemtableID(ReadOnlyMemTable* new_imm) {
+    new_imm->SetID(++last_memtable_id_);
   }
 
   // calculate the oldest log needed for the durability of this column family
   uint64_t OldestLogToKeep();
 
   // See Memtable constructor for explanation of earliest_seq param.
+  // `mutable_cf_options` might need to be a saved copy if calling this without
+  // holding the DB mutex.
   MemTable* ConstructNewMemtable(const MutableCFOptions& mutable_cf_options,
                                  SequenceNumber earliest_seq);
-  void CreateNewMemtable(const MutableCFOptions& mutable_cf_options,
-                         SequenceNumber earliest_seq);
+  void CreateNewMemtable(SequenceNumber earliest_seq);
 
   TableCache* table_cache() const { return table_cache_.get(); }
   BlobFileCache* blob_file_cache() const { return blob_file_cache_.get(); }
@@ -477,6 +487,9 @@ class ColumnFamilyData {
   uint64_t GetSuperVersionNumber() const {
     return super_version_number_.load();
   }
+  uint64_t GetSuperVersionNumberRelaxed() const {
+    return super_version_number_.load(std::memory_order_relaxed);
+  }
   // will return a pointer to SuperVersion* if previous SuperVersion
   // if its reference count is zero and needs deletion or nullptr if not
   // As argument takes a pointer to allocated SuperVersion to enable

db/compact_files_test.cc

@@ -3,7 +3,6 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).
 
-
 #include <mutex>
 #include <string>
 #include <thread>

db/compaction/compaction.cc

@@ -332,11 +332,11 @@ Compaction::Compaction(
               : (_blob_garbage_collection_policy ==
                          BlobGarbageCollectionPolicy::kDisable
                      ? false
-                     : mutable_cf_options()->enable_blob_garbage_collection)),
+                     : mutable_cf_options().enable_blob_garbage_collection)),
       blob_garbage_collection_age_cutoff_(
           _blob_garbage_collection_age_cutoff < 0 ||
                   _blob_garbage_collection_age_cutoff > 1
-              ? mutable_cf_options()->blob_garbage_collection_age_cutoff
+              ? mutable_cf_options().blob_garbage_collection_age_cutoff
               : _blob_garbage_collection_age_cutoff),
       penultimate_level_(
           // For simplicity, we don't support the concept of "penultimate level"
@@ -410,6 +410,7 @@ Compaction::Compaction(
 
 void Compaction::PopulatePenultimateLevelOutputRange() {
   if (!SupportsPerKeyPlacement()) {
+    assert(keep_in_last_level_through_seqno_ == kMaxSequenceNumber);
     return;
   }
 
@@ -452,6 +453,27 @@ void Compaction::PopulatePenultimateLevelOutputRange() {
   GetBoundaryInternalKeys(input_vstorage_, inputs_,
                           &penultimate_level_smallest_,
                           &penultimate_level_largest_, exclude_level);
+
+  if (penultimate_output_range_type_ !=
+      PenultimateOutputRangeType::kFullRange) {
+    // If not full range in penultimate level, must keep everything already
+    // in the last level there, because moving it back up might cause
+    // overlap/placement issues that are difficult to resolve properly in the
+    // presence of range deletes
+    SequenceNumber max_last_level_seqno = 0;
+    for (const auto& input_lvl : inputs_) {
+      if (input_lvl.level == output_level_) {
+        for (const auto& file : input_lvl.files) {
+          max_last_level_seqno =
+              std::max(max_last_level_seqno, file->fd.largest_seqno);
+        }
+      }
+    }
+
+    keep_in_last_level_through_seqno_ = max_last_level_seqno;
+  } else {
+    keep_in_last_level_through_seqno_ = 0;
+  }
 }
 
 Compaction::~Compaction() {
@@ -494,22 +516,29 @@ bool Compaction::OverlapPenultimateLevelOutputRange(
 }
 
 // key includes timestamp if user-defined timestamp is enabled.
-bool Compaction::WithinPenultimateLevelOutputRange(
-    const ParsedInternalKey& ikey) const {
-  if (!SupportsPerKeyPlacement()) {
-    return false;
-  }
+void Compaction::TEST_AssertWithinPenultimateLevelOutputRange(
+    const Slice& user_key, bool expect_failure) const {
+#ifdef NDEBUG
+  (void)user_key;
+  (void)expect_failure;
+#else
+  assert(SupportsPerKeyPlacement());
 
-  if (penultimate_level_smallest_.size() == 0 ||
-      penultimate_level_largest_.size() == 0) {
-    return false;
-  }
+  assert(penultimate_level_smallest_.size() > 0);
+  assert(penultimate_level_largest_.size() > 0);
 
-  const InternalKeyComparator* icmp = input_vstorage_->InternalComparator();
+  auto* cmp = input_vstorage_->user_comparator();
 
   // op_type of a key can change during compaction, e.g. Merge -> Put.
-  return icmp->CompareKeySeq(ikey, penultimate_level_smallest_.Encode()) >= 0 &&
-         icmp->CompareKeySeq(ikey, penultimate_level_largest_.Encode()) <= 0;
+  if (!(cmp->Compare(user_key, penultimate_level_smallest_.user_key()) >= 0)) {
+    assert(expect_failure);
+  } else if (!(cmp->Compare(user_key, penultimate_level_largest_.user_key()) <=
+               0)) {
+    assert(expect_failure);
+  } else {
+    assert(!expect_failure);
+  }
+#endif
 }
 
 bool Compaction::InputCompressionMatchesOutput() const {
@@ -563,7 +592,7 @@ bool Compaction::IsTrivialMove() const {
   // input files are non overlapping
   if ((mutable_cf_options_.compaction_options_universal.allow_trivial_move) &&
       (output_level_ != 0) &&
-      (cfd_->ioptions()->compaction_style == kCompactionStyleUniversal)) {
+      (cfd_->ioptions().compaction_style == kCompactionStyleUniversal)) {
     return is_trivial_move_;
   }
 
@@ -621,7 +650,7 @@ bool Compaction::KeyNotExistsBeyondOutputLevel(
   if (bottommost_level_) {
     return true;
   } else if (output_level_ != 0 &&
-             cfd_->ioptions()->compaction_style == kCompactionStyleLevel) {
+             cfd_->ioptions().compaction_style == kCompactionStyleLevel) {
     // 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++) {
@@ -662,7 +691,7 @@ bool Compaction::KeyRangeNotExistsBeyondOutputLevel(
   if (bottommost_level_) {
     return true /* does not overlap */;
   } else if (output_level_ != 0 &&
-             cfd_->ioptions()->compaction_style == kCompactionStyleLevel) {
+             cfd_->ioptions().compaction_style == kCompactionStyleLevel) {
     const Comparator* user_cmp = cfd_->user_comparator();
     for (int lvl = output_level_ + 1; lvl < number_levels_; lvl++) {
       const std::vector<FileMetaData*>& files =
@@ -838,12 +867,12 @@ uint64_t Compaction::OutputFilePreallocationSize() const {
 }
 
 std::unique_ptr<CompactionFilter> Compaction::CreateCompactionFilter() const {
-  if (!cfd_->ioptions()->compaction_filter_factory) {
+  if (!cfd_->ioptions().compaction_filter_factory) {
     return nullptr;
   }
 
   if (!cfd_->ioptions()
-           ->compaction_filter_factory->ShouldFilterTableFileCreation(
+           .compaction_filter_factory->ShouldFilterTableFileCreation(
                TableFileCreationReason::kCompaction)) {
     return nullptr;
   }
@@ -862,7 +891,7 @@ std::unique_ptr<CompactionFilter> Compaction::CreateCompactionFilter() const {
         "for compaction.");
   }
 
-  return cfd_->ioptions()->compaction_filter_factory->CreateCompactionFilter(
+  return cfd_->ioptions().compaction_filter_factory->CreateCompactionFilter(
       context);
 }
 
@@ -896,8 +925,8 @@ bool Compaction::ShouldFormSubcompactions() const {
 
   // Round-Robin pri under leveled compaction allows subcompactions by default
   // and the number of subcompactions can be larger than max_subcompactions_
-  if (cfd_->ioptions()->compaction_pri == kRoundRobin &&
-      cfd_->ioptions()->compaction_style == kCompactionStyleLevel) {
+  if (cfd_->ioptions().compaction_pri == kRoundRobin &&
+      cfd_->ioptions().compaction_style == kCompactionStyleLevel) {
     return output_level_ > 0;
   }
 
@@ -905,9 +934,9 @@ bool Compaction::ShouldFormSubcompactions() const {
     return false;
   }
 
-  if (cfd_->ioptions()->compaction_style == kCompactionStyleLevel) {
+  if (cfd_->ioptions().compaction_style == kCompactionStyleLevel) {
     return (start_level_ == 0 || is_manual_compaction_) && output_level_ > 0;
-  } else if (cfd_->ioptions()->compaction_style == kCompactionStyleUniversal) {
+  } else if (cfd_->ioptions().compaction_style == kCompactionStyleUniversal) {
     return number_levels_ > 1 && output_level_ > 0;
   } else {
     return false;

db/compaction/compaction.h

@@ -288,14 +288,14 @@ class Compaction {
 
   // Return the ImmutableOptions that should be used throughout the compaction
   // procedure
-  const ImmutableOptions* immutable_options() const {
-    return &immutable_options_;
+  const ImmutableOptions& immutable_options() const {
+    return immutable_options_;
   }
 
   // Return the MutableCFOptions that should be used throughout the compaction
   // procedure
-  const MutableCFOptions* mutable_cf_options() const {
-    return &mutable_cf_options_;
+  const MutableCFOptions& mutable_cf_options() const {
+    return mutable_cf_options_;
   }
 
   // Returns the size in bytes that the output file should be preallocated to.
@@ -388,12 +388,12 @@ class Compaction {
   bool OverlapPenultimateLevelOutputRange(const Slice& smallest_key,
                                           const Slice& largest_key) const;
 
-  // Return true if the key is within penultimate level output range for
-  // per_key_placement feature, which is safe to place the key to the
-  // penultimate level. different compaction strategy has different rules.
-  // If per_key_placement is not supported, always return false.
-  //  key includes timestamp if user-defined timestamp is enabled.
-  bool WithinPenultimateLevelOutputRange(const ParsedInternalKey& ikey) const;
+  // For testing purposes, check that a key is within penultimate level
+  // output range for per_key_placement feature, which is safe to place the key
+  // to the penultimate level. Different compaction strategies have different
+  // rules. `user_key` includes timestamp if user-defined timestamp is enabled.
+  void TEST_AssertWithinPenultimateLevelOutputRange(
+      const Slice& user_key, bool expect_failure = false) const;
 
   CompactionReason compaction_reason() const { return compaction_reason_; }
 
@@ -456,11 +456,17 @@ class Compaction {
       const ImmutableOptions& immutable_options, const int start_level,
       const int 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.
+  SequenceNumber GetKeepInLastLevelThroughSeqno() const {
+    return keep_in_last_level_through_seqno_;
+  }
+
   // mark (or clear) all files that are being compacted
   void MarkFilesBeingCompacted(bool being_compacted) const;
 
  private:
-
   Status InitInputTableProperties();
 
   // get the smallest and largest key present in files to be compacted
@@ -606,6 +612,8 @@ class Compaction {
   // Blob garbage collection age cutoff.
   double blob_garbage_collection_age_cutoff_;
 
+  SequenceNumber keep_in_last_level_through_seqno_ = kMaxSequenceNumber;
+
   // only set when per_key_placement feature is enabled, -1 (kInvalidLevel)
   // means not supported.
   const int penultimate_level_;

db/compaction/compaction_iterator.cc

@@ -38,19 +38,16 @@ CompactionIterator::CompactionIterator(
     const std::atomic<bool>* shutting_down,
     const std::shared_ptr<Logger> info_log,
     const std::string* full_history_ts_low,
-    const SequenceNumber preserve_time_min_seqno,
-    const SequenceNumber preclude_last_level_min_seqno)
+    std::optional<SequenceNumber> preserve_seqno_min)
     : 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,
           manual_compaction_canceled,
-          std::unique_ptr<CompactionProxy>(
-              compaction ? new RealCompaction(compaction) : nullptr),
+          compaction ? std::make_unique<RealCompaction>(compaction) : nullptr,
           must_count_input_entries, compaction_filter, shutting_down, info_log,
-          full_history_ts_low, preserve_time_min_seqno,
-          preclude_last_level_min_seqno) {}
+          full_history_ts_low, preserve_seqno_min) {}
 
 CompactionIterator::CompactionIterator(
     InternalIterator* input, const Comparator* cmp, MergeHelper* merge_helper,
@@ -68,8 +65,7 @@ CompactionIterator::CompactionIterator(
     const std::atomic<bool>* shutting_down,
     const std::shared_ptr<Logger> info_log,
     const std::string* full_history_ts_low,
-    const SequenceNumber preserve_time_min_seqno,
-    const SequenceNumber preclude_last_level_min_seqno)
+    std::optional<SequenceNumber> preserve_seqno_min)
     : input_(input, cmp, must_count_input_entries),
       cmp_(cmp),
       merge_helper_(merge_helper),
@@ -110,10 +106,9 @@ CompactionIterator::CompactionIterator(
       current_key_committed_(false),
       cmp_with_history_ts_low_(0),
       level_(compaction_ == nullptr ? 0 : compaction_->level()),
-      preserve_time_min_seqno_(preserve_time_min_seqno),
-      preclude_last_level_min_seqno_(preclude_last_level_min_seqno) {
+      preserve_seqno_after_(preserve_seqno_min.value_or(earliest_snapshot)) {
   assert(snapshots_ != nullptr);
-  assert(preserve_time_min_seqno_ <= preclude_last_level_min_seqno_);
+  assert(preserve_seqno_after_ <= earliest_snapshot_);
 
   if (compaction_ != nullptr) {
     level_ptrs_ = std::vector<size_t>(compaction_->number_levels(), 0);
@@ -872,8 +867,8 @@ void CompactionIterator::NextFromInput() {
       if (Valid()) {
         at_next_ = true;
       }
-    } else if (last_snapshot == current_user_key_snapshot_ ||
-               (last_snapshot > 0 &&
+    } else if (last_sequence != kMaxSequenceNumber &&
+               (last_snapshot == current_user_key_snapshot_ ||
                 last_snapshot < current_user_key_snapshot_)) {
       // If the earliest snapshot is which this key is visible in
       // is the same as the visibility of a previous instance of the
@@ -1018,7 +1013,6 @@ void CompactionIterator::NextFromInput() {
         } else {
           if (ikey_.sequence != 0) {
             iter_stats_.num_timed_put_swap_preferred_seqno++;
-            saved_seq_for_penul_check_ = ikey_.sequence;
             ikey_.sequence = preferred_seqno;
           }
           ikey_.type = kTypeValue;
@@ -1259,71 +1253,6 @@ void CompactionIterator::GarbageCollectBlobIfNeeded() {
   }
 }
 
-void CompactionIterator::DecideOutputLevel() {
-  assert(compaction_->SupportsPerKeyPlacement());
-  output_to_penultimate_level_ = false;
-  // if the key is newer than the cutoff sequence or within the earliest
-  // snapshot, it should output to the penultimate level.
-  if (ikey_.sequence > preclude_last_level_min_seqno_ ||
-      ikey_.sequence > earliest_snapshot_) {
-    output_to_penultimate_level_ = true;
-  }
-
-#ifndef NDEBUG
-  // Could be overridden by unittest
-  PerKeyPlacementContext context(level_, ikey_.user_key, value_, ikey_.sequence,
-                                 output_to_penultimate_level_);
-  TEST_SYNC_POINT_CALLBACK("CompactionIterator::PrepareOutput.context",
-                           &context);
-  if (ikey_.sequence > earliest_snapshot_) {
-    output_to_penultimate_level_ = true;
-  }
-#endif  // NDEBUG
-
-  // saved_seq_for_penul_check_ is populated in `NextFromInput` when the
-  // entry's sequence number is non zero and validity context for output this
-  // entry is kSwapPreferredSeqno for use in `DecideOutputLevel`. It should be
-  // cleared out here unconditionally. Otherwise, it may end up getting consumed
-  // incorrectly by a different entry.
-  SequenceNumber seq_for_range_check =
-      (saved_seq_for_penul_check_.has_value() &&
-       saved_seq_for_penul_check_.value() != kMaxSequenceNumber)
-          ? saved_seq_for_penul_check_.value()
-          : ikey_.sequence;
-  saved_seq_for_penul_check_ = std::nullopt;
-  ParsedInternalKey ikey_for_range_check = ikey_;
-  if (seq_for_range_check != ikey_.sequence) {
-    ikey_for_range_check.sequence = seq_for_range_check;
-  }
-  if (output_to_penultimate_level_) {
-    // If it's decided to output to the penultimate level, but unsafe to do so,
-    // still output to the last level. For example, moving the data from a lower
-    // level to a higher level outside of the higher-level input key range is
-    // considered unsafe, because the key may conflict with higher-level SSTs
-    // not from this compaction.
-    // TODO: add statistic for declined output_to_penultimate_level
-    bool safe_to_penultimate_level =
-        compaction_->WithinPenultimateLevelOutputRange(ikey_for_range_check);
-    if (!safe_to_penultimate_level) {
-      output_to_penultimate_level_ = false;
-      // It could happen when disable/enable `last_level_temperature` while
-      // holding a snapshot. When `last_level_temperature` is not set
-      // (==kUnknown), the data newer than any snapshot is pushed to the last
-      // level, but when the per_key_placement feature is enabled on the fly,
-      // the data later than the snapshot has to be moved to the penultimate
-      // level, which may or may not be safe. So the user needs to make sure all
-      // snapshot is released before enabling `last_level_temperature` feature
-      // We will migrate the feature to `last_level_temperature` and maybe make
-      // it not dynamically changeable.
-      if (seq_for_range_check > earliest_snapshot_) {
-        status_ = Status::Corruption(
-            "Unsafe to store Seq later than snapshot in the last level if "
-            "per_key_placement is enabled");
-      }
-    }
-  }
-}
-
 void CompactionIterator::PrepareOutput() {
   if (Valid()) {
     if (LIKELY(!is_range_del_)) {
@@ -1332,13 +1261,6 @@ void CompactionIterator::PrepareOutput() {
       } else if (ikey_.type == kTypeBlobIndex) {
         GarbageCollectBlobIfNeeded();
       }
-
-      // For range del sentinel, we don't use it to cut files for bottommost
-      // compaction. So it should not make a difference which output level we
-      // decide.
-      if (compaction_ != nullptr && compaction_->SupportsPerKeyPlacement()) {
-        DecideOutputLevel();
-      }
     }
 
     // Zeroing out the sequence number leads to better compression.
@@ -1356,8 +1278,7 @@ void CompactionIterator::PrepareOutput() {
         !compaction_->allow_ingest_behind() && bottommost_level_ &&
         DefinitelyInSnapshot(ikey_.sequence, earliest_snapshot_) &&
         ikey_.type != kTypeMerge && current_key_committed_ &&
-        !output_to_penultimate_level_ &&
-        ikey_.sequence < preserve_time_min_seqno_ && !is_range_del_) {
+        ikey_.sequence <= preserve_seqno_after_ && !is_range_del_) {
       if (ikey_.type == kTypeDeletion ||
           (ikey_.type == kTypeSingleDeletion && timestamp_size_ == 0)) {
         ROCKS_LOG_FATAL(

db/compaction/compaction_iterator.h

@@ -117,10 +117,6 @@ class CompactionIterator {
     virtual const Compaction* real_compaction() const = 0;
 
     virtual bool SupportsPerKeyPlacement() const = 0;
-
-    // `key` includes timestamp if user-defined timestamp is enabled.
-    virtual bool WithinPenultimateLevelOutputRange(
-        const ParsedInternalKey&) const = 0;
   };
 
   class RealCompaction : public CompactionProxy {
@@ -128,8 +124,6 @@ class CompactionIterator {
     explicit RealCompaction(const Compaction* compaction)
         : compaction_(compaction) {
       assert(compaction_);
-      assert(compaction_->immutable_options());
-      assert(compaction_->mutable_cf_options());
     }
 
     int level() const override { return compaction_->level(); }
@@ -151,11 +145,11 @@ class CompactionIterator {
     }
 
     bool allow_ingest_behind() const override {
-      return compaction_->immutable_options()->allow_ingest_behind;
+      return compaction_->immutable_options().allow_ingest_behind;
     }
 
     bool allow_mmap_reads() const override {
-      return compaction_->immutable_options()->allow_mmap_reads;
+      return compaction_->immutable_options().allow_mmap_reads;
     }
 
     bool enable_blob_garbage_collection() const override {
@@ -167,7 +161,7 @@ class CompactionIterator {
     }
 
     uint64_t blob_compaction_readahead_size() const override {
-      return compaction_->mutable_cf_options()->blob_compaction_readahead_size;
+      return compaction_->mutable_cf_options().blob_compaction_readahead_size;
     }
 
     const Version* input_version() const override {
@@ -184,14 +178,6 @@ class CompactionIterator {
       return compaction_->SupportsPerKeyPlacement();
     }
 
-    // Check if key is within penultimate level output range, to see if it's
-    // safe to output to the penultimate level for per_key_placement feature.
-    // `key` includes timestamp if user-defined timestamp is enabled.
-    bool WithinPenultimateLevelOutputRange(
-        const ParsedInternalKey& ikey) const override {
-      return compaction_->WithinPenultimateLevelOutputRange(ikey);
-    }
-
    private:
     const Compaction* compaction_;
   };
@@ -216,29 +202,29 @@ class CompactionIterator {
       const std::atomic<bool>* shutting_down = nullptr,
       const std::shared_ptr<Logger> info_log = nullptr,
       const std::string* full_history_ts_low = nullptr,
-      const SequenceNumber preserve_time_min_seqno = kMaxSequenceNumber,
-      const SequenceNumber preclude_last_level_min_seqno = kMaxSequenceNumber);
+      std::optional<SequenceNumber> preserve_seqno_min = {});
 
   // Constructor with custom CompactionProxy, used for tests.
-  CompactionIterator(
-      InternalIterator* input, const Comparator* cmp, MergeHelper* merge_helper,
-      SequenceNumber last_sequence, std::vector<SequenceNumber>* snapshots,
-      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,
-      CompactionRangeDelAggregator* range_del_agg,
-      BlobFileBuilder* blob_file_builder, bool allow_data_in_errors,
-      bool enforce_single_del_contracts,
-      const std::atomic<bool>& manual_compaction_canceled,
-      std::unique_ptr<CompactionProxy> compaction,
-      bool must_count_input_entries,
-      const CompactionFilter* compaction_filter = nullptr,
-      const std::atomic<bool>* shutting_down = nullptr,
-      const std::shared_ptr<Logger> info_log = nullptr,
-      const std::string* full_history_ts_low = nullptr,
-      const SequenceNumber preserve_time_min_seqno = kMaxSequenceNumber,
-      const SequenceNumber preclude_last_level_min_seqno = kMaxSequenceNumber);
+  CompactionIterator(InternalIterator* input, const Comparator* cmp,
+                     MergeHelper* merge_helper, SequenceNumber last_sequence,
+                     std::vector<SequenceNumber>* snapshots,
+                     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,
+                     CompactionRangeDelAggregator* range_del_agg,
+                     BlobFileBuilder* blob_file_builder,
+                     bool allow_data_in_errors,
+                     bool enforce_single_del_contracts,
+                     const std::atomic<bool>& manual_compaction_canceled,
+                     std::unique_ptr<CompactionProxy> compaction,
+                     bool must_count_input_entries,
+                     const CompactionFilter* compaction_filter = nullptr,
+                     const std::atomic<bool>* shutting_down = nullptr,
+                     const std::shared_ptr<Logger> info_log = nullptr,
+                     const std::string* full_history_ts_low = nullptr,
+                     std::optional<SequenceNumber> preserve_seqno_min = {});
 
   ~CompactionIterator();
 
@@ -269,11 +255,6 @@ class CompactionIterator {
   const CompactionIterationStats& iter_stats() const { return iter_stats_; }
   bool HasNumInputEntryScanned() const { return input_.HasNumItered(); }
   uint64_t NumInputEntryScanned() const { return input_.NumItered(); }
-  // If the current key should be placed on penultimate level, only valid if
-  // per_key_placement is supported
-  bool output_to_penultimate_level() const {
-    return output_to_penultimate_level_;
-  }
   Status InputStatus() const { return input_.status(); }
 
   bool IsDeleteRangeSentinelKey() const { return is_range_del_; }
@@ -285,10 +266,6 @@ class CompactionIterator {
   // Do final preparations before presenting the output to the callee.
   void PrepareOutput();
 
-  // Decide the current key should be output to the last level or penultimate
-  // level, only call for compaction supports per key placement
-  void DecideOutputLevel();
-
   // Passes the output value to the blob file builder (if any), and replaces it
   // with the corresponding blob reference if it has been actually written to a
   // blob file (i.e. if it passed the value size check). Returns true if the
@@ -440,16 +417,6 @@ class CompactionIterator {
   // NextFromInput()).
   ParsedInternalKey ikey_;
 
-  // When a kTypeValuePreferredSeqno entry's preferred seqno is safely swapped
-  // in in this compaction, this field saves its original sequence number for
-  // range checking whether it's safe to be placed on the penultimate level.
-  // This is to ensure when such an entry happens to be the right boundary of
-  // penultimate safe range, it won't get excluded because with the preferred
-  // seqno swapped in, it's now larger than the right boundary (itself before
-  // the swap). This is safe to do, because preferred seqno is swapped in only
-  // when no entries with the same user key exist on lower levels and this entry
-  // is already visible in the earliest snapshot.
-  std::optional<SequenceNumber> saved_seq_for_penul_check_ = kMaxSequenceNumber;
   // 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.
   bool has_current_user_key_ = false;
@@ -507,17 +474,8 @@ class CompactionIterator {
   // just been zeroed out during bottommost compaction.
   bool last_key_seq_zeroed_{false};
 
-  // True if the current key should be output to the penultimate level if
-  // possible, compaction logic makes the final decision on which level to
-  // output to.
-  bool output_to_penultimate_level_{false};
-
-  // min seqno for preserving the time information.
-  const SequenceNumber preserve_time_min_seqno_ = kMaxSequenceNumber;
-
-  // min seqno to preclude the data from the last level, if the key seqno larger
-  // than this, it will be output to penultimate level
-  const SequenceNumber preclude_last_level_min_seqno_ = kMaxSequenceNumber;
+  // Max seqno that can be zeroed out at last level (various reasons)
+  const SequenceNumber preserve_seqno_after_ = kMaxSequenceNumber;
 
   void AdvanceInputIter() { input_.Next(); }
 

db/compaction/compaction_iterator_test.cc

@@ -192,11 +192,6 @@ class FakeCompaction : public CompactionIterator::CompactionProxy {
     return supports_per_key_placement;
   }
 
-  bool WithinPenultimateLevelOutputRange(
-      const ParsedInternalKey& key) const override {
-    return (!key.user_key.starts_with("unsafe_pb"));
-  }
-
   bool key_not_exists_beyond_output_level = false;
 
   bool is_bottommost_level = false;
@@ -833,126 +828,17 @@ TEST_P(CompactionIteratorTest, ConvertToPutAtBottom) {
           true /*bottomost_level*/);
 }
 
+TEST_P(CompactionIteratorTest, ZeroSeqOfKeyAndSnapshot) {
+  AddSnapshot(0);
+  const std::vector<std::string> input_keys = {
+      test::KeyStr("a", 0, kTypeValue), test::KeyStr("b", 0, kTypeValue)};
+  const std::vector<std::string> input_values = {"a1", "b1"};
+  RunTest(input_keys, input_values, input_keys, input_values);
+}
+
 INSTANTIATE_TEST_CASE_P(CompactionIteratorTestInstance, CompactionIteratorTest,
                         testing::Values(true, false));
 
-class PerKeyPlacementCompIteratorTest : public CompactionIteratorTest {
- public:
-  bool SupportsPerKeyPlacement() const override { return true; }
-};
-
-TEST_P(PerKeyPlacementCompIteratorTest, SplitLastLevelData) {
-  std::atomic_uint64_t latest_cold_seq = 0;
-
-  SyncPoint::GetInstance()->SetCallBack(
-      "CompactionIterator::PrepareOutput.context", [&](void* arg) {
-        auto context = static_cast<PerKeyPlacementContext*>(arg);
-        context->output_to_penultimate_level =
-            context->seq_num > latest_cold_seq;
-      });
-  SyncPoint::GetInstance()->EnableProcessing();
-
-  latest_cold_seq = 5;
-
-  InitIterators(
-      {test::KeyStr("a", 7, kTypeValue), test::KeyStr("b", 6, kTypeValue),
-       test::KeyStr("c", 5, kTypeValue)},
-      {"vala", "valb", "valc"}, {}, {}, kMaxSequenceNumber, kMaxSequenceNumber,
-      nullptr, nullptr, true);
-  c_iter_->SeekToFirst();
-  ASSERT_TRUE(c_iter_->Valid());
-
-  // the first 2 keys are hot, which should has
-  // `output_to_penultimate_level()==true` and seq num not zeroed out
-  ASSERT_EQ(test::KeyStr("a", 7, kTypeValue), c_iter_->key().ToString());
-  ASSERT_TRUE(c_iter_->output_to_penultimate_level());
-  c_iter_->Next();
-  ASSERT_TRUE(c_iter_->Valid());
-  ASSERT_EQ(test::KeyStr("b", 6, kTypeValue), c_iter_->key().ToString());
-  ASSERT_TRUE(c_iter_->output_to_penultimate_level());
-  c_iter_->Next();
-  ASSERT_TRUE(c_iter_->Valid());
-  // `a` is cold data, which should be output to bottommost
-  ASSERT_EQ(test::KeyStr("c", 0, kTypeValue), c_iter_->key().ToString());
-  ASSERT_FALSE(c_iter_->output_to_penultimate_level());
-  c_iter_->Next();
-  ASSERT_OK(c_iter_->status());
-  ASSERT_FALSE(c_iter_->Valid());
-  SyncPoint::GetInstance()->DisableProcessing();
-  SyncPoint::GetInstance()->ClearAllCallBacks();
-}
-
-TEST_P(PerKeyPlacementCompIteratorTest, SnapshotData) {
-  AddSnapshot(5);
-
-  InitIterators(
-      {test::KeyStr("a", 7, kTypeValue), test::KeyStr("b", 6, kTypeDeletion),
-       test::KeyStr("b", 5, kTypeValue)},
-      {"vala", "", "valb"}, {}, {}, kMaxSequenceNumber, kMaxSequenceNumber,
-      nullptr, nullptr, true);
-  c_iter_->SeekToFirst();
-  ASSERT_TRUE(c_iter_->Valid());
-
-  // The first key and the tombstone are within snapshot, which should output
-  // to the penultimate level (and seq num cannot be zeroed out).
-  ASSERT_EQ(test::KeyStr("a", 7, kTypeValue), c_iter_->key().ToString());
-  ASSERT_TRUE(c_iter_->output_to_penultimate_level());
-  c_iter_->Next();
-  ASSERT_TRUE(c_iter_->Valid());
-  ASSERT_EQ(test::KeyStr("b", 6, kTypeDeletion), c_iter_->key().ToString());
-  ASSERT_TRUE(c_iter_->output_to_penultimate_level());
-  c_iter_->Next();
-  ASSERT_TRUE(c_iter_->Valid());
-  // `a` is not protected by the snapshot, the sequence number is zero out and
-  // should output bottommost
-  ASSERT_EQ(test::KeyStr("b", 0, kTypeValue), c_iter_->key().ToString());
-  ASSERT_FALSE(c_iter_->output_to_penultimate_level());
-  c_iter_->Next();
-  ASSERT_OK(c_iter_->status());
-  ASSERT_FALSE(c_iter_->Valid());
-}
-
-TEST_P(PerKeyPlacementCompIteratorTest, ConflictWithSnapshot) {
-  std::atomic_uint64_t latest_cold_seq = 0;
-
-  SyncPoint::GetInstance()->SetCallBack(
-      "CompactionIterator::PrepareOutput.context", [&](void* arg) {
-        auto context = static_cast<PerKeyPlacementContext*>(arg);
-        context->output_to_penultimate_level =
-            context->seq_num > latest_cold_seq;
-      });
-  SyncPoint::GetInstance()->EnableProcessing();
-
-  latest_cold_seq = 6;
-
-  AddSnapshot(5);
-
-  InitIterators({test::KeyStr("a", 7, kTypeValue),
-                 test::KeyStr("unsafe_pb", 6, kTypeValue),
-                 test::KeyStr("c", 5, kTypeValue)},
-                {"vala", "valb", "valc"}, {}, {}, kMaxSequenceNumber,
-                kMaxSequenceNumber, nullptr, nullptr, true);
-  c_iter_->SeekToFirst();
-  ASSERT_TRUE(c_iter_->Valid());
-
-  ASSERT_EQ(test::KeyStr("a", 7, kTypeValue), c_iter_->key().ToString());
-  ASSERT_TRUE(c_iter_->output_to_penultimate_level());
-  // the 2nd key is unsafe to output_to_penultimate_level, but it's within
-  // snapshot so for per_key_placement feature it has to be outputted to the
-  // penultimate level. which is a corruption. We should never see
-  // such case as the data with seq num (within snapshot) should always come
-  // from higher compaction input level, which makes it safe to
-  // output_to_penultimate_level.
-  c_iter_->Next();
-  ASSERT_TRUE(c_iter_->status().IsCorruption());
-  SyncPoint::GetInstance()->DisableProcessing();
-  SyncPoint::GetInstance()->ClearAllCallBacks();
-}
-
-INSTANTIATE_TEST_CASE_P(PerKeyPlacementCompIteratorTest,
-                        PerKeyPlacementCompIteratorTest,
-                        testing::Values(true, false));
-
 // Tests how CompactionIterator work together with SnapshotChecker.
 class CompactionIteratorWithSnapshotCheckerTest
     : public CompactionIteratorTest {
@@ -1846,6 +1732,22 @@ TEST_P(CompactionIteratorTsGcTest, SingleDeleteAllKeysOlderThanThreshold) {
   }
 }
 
+TEST_P(CompactionIteratorTsGcTest, ZeroSeqOfKeyAndSnapshot) {
+  AddSnapshot(0);
+  std::string full_history_ts_low;
+  PutFixed64(&full_history_ts_low, std::numeric_limits<uint64_t>::max());
+  const std::vector<std::string> input_keys = {
+      test::KeyStr(101, "a", 0, kTypeValue),
+      test::KeyStr(102, "b", 0, kTypeValue)};
+  const std::vector<std::string> input_values = {"a1", "b1"};
+  RunTest(input_keys, input_values, input_keys, input_values,
+          /*last_committed_seq=*/kMaxSequenceNumber,
+          /*merge_operator=*/nullptr, /*compaction_filter=*/nullptr,
+          /*bottommost_level=*/false,
+          /*earliest_write_conflict_snapshot=*/kMaxSequenceNumber,
+          /*key_not_exists_beyond_output_level=*/false, &full_history_ts_low);
+}
+
 INSTANTIATE_TEST_CASE_P(CompactionIteratorTsGcTestInstance,
                         CompactionIteratorTsGcTest,
                         testing::Values(true, false));

db/compaction/compaction_job.cc

@@ -245,7 +245,10 @@ void CompactionJob::ReportStartedCompaction(Compaction* compaction) {
   compaction_job_stats_->is_full_compaction = compaction->is_full_compaction();
 }
 
-void CompactionJob::Prepare() {
+void CompactionJob::Prepare(
+    std::optional<std::pair<std::optional<Slice>, std::optional<Slice>>>
+        known_single_subcompact) {
+  db_mutex_->AssertHeld();
   AutoThreadOperationStageUpdater stage_updater(
       ThreadStatus::STAGE_COMPACTION_PREPARE);
 
@@ -260,11 +263,12 @@ void CompactionJob::Prepare() {
   write_hint_ = storage_info->CalculateSSTWriteHint(c->output_level());
   bottommost_level_ = c->bottommost_level();
 
-  if (c->ShouldFormSubcompactions()) {
+  if (!known_single_subcompact.has_value() && c->ShouldFormSubcompactions()) {
     StopWatch sw(db_options_.clock, stats_, SUBCOMPACTION_SETUP_TIME);
     GenSubcompactionBoundaries();
   }
   if (boundaries_.size() >= 1) {
+    assert(!known_single_subcompact.has_value());
     for (size_t i = 0; i <= boundaries_.size(); i++) {
       compact_->sub_compact_states.emplace_back(
           c, (i != 0) ? std::optional<Slice>(boundaries_[i - 1]) : std::nullopt,
@@ -280,16 +284,25 @@ void CompactionJob::Prepare() {
     RecordInHistogram(stats_, NUM_SUBCOMPACTIONS_SCHEDULED,
                       compact_->sub_compact_states.size());
   } else {
-    compact_->sub_compact_states.emplace_back(c, std::nullopt, std::nullopt,
+    std::optional<Slice> start_key;
+    std::optional<Slice> end_key;
+    if (known_single_subcompact.has_value()) {
+      start_key = known_single_subcompact.value().first;
+      end_key = known_single_subcompact.value().second;
+    } else {
+      assert(!start_key.has_value() && !end_key.has_value());
+    }
+    compact_->sub_compact_states.emplace_back(c, start_key, end_key,
                                               /*sub_job_id*/ 0);
   }
 
   // collect all seqno->time information from the input files which will be used
   // to encode seqno->time to the output files.
-
+  SequenceNumber preserve_time_min_seqno = kMaxSequenceNumber;
+  SequenceNumber preclude_last_level_min_seqno = kMaxSequenceNumber;
   uint64_t preserve_time_duration =
-      std::max(c->mutable_cf_options()->preserve_internal_time_seconds,
-               c->mutable_cf_options()->preclude_last_level_data_seconds);
+      std::max(c->mutable_cf_options().preserve_internal_time_seconds,
+               c->mutable_cf_options().preclude_last_level_data_seconds);
 
   if (preserve_time_duration > 0) {
     const ReadOptions read_options(Env::IOActivity::kCompaction);
@@ -319,16 +332,16 @@ void CompactionJob::Prepare() {
                      "Failed to get current time in compaction: Status: %s",
                      s.ToString().c_str());
       // preserve all time information
-      preserve_time_min_seqno_ = 0;
-      preclude_last_level_min_seqno_ = 0;
+      preserve_time_min_seqno = 0;
+      preclude_last_level_min_seqno = 0;
       seqno_to_time_mapping_.Enforce();
     } else {
       seqno_to_time_mapping_.Enforce(_current_time);
       seqno_to_time_mapping_.GetCurrentTieringCutoffSeqnos(
           static_cast<uint64_t>(_current_time),
-          c->mutable_cf_options()->preserve_internal_time_seconds,
-          c->mutable_cf_options()->preclude_last_level_data_seconds,
-          &preserve_time_min_seqno_, &preclude_last_level_min_seqno_);
+          c->mutable_cf_options().preserve_internal_time_seconds,
+          c->mutable_cf_options().preclude_last_level_data_seconds,
+          &preserve_time_min_seqno, &preclude_last_level_min_seqno);
     }
     // For accuracy of the GetProximalSeqnoBeforeTime queries above, we only
     // limit the capacity after them.
@@ -341,6 +354,45 @@ void CompactionJob::Prepare() {
     // larger than kMaxSeqnoTimePairsPerSST.
     seqno_to_time_mapping_.SetCapacity(kMaxSeqnoToTimeEntries);
   }
+#ifndef NDEBUG
+  assert(preserve_time_min_seqno <= preclude_last_level_min_seqno);
+  TEST_SYNC_POINT_CALLBACK(
+      "CompactionJob::PrepareTimes():preclude_last_level_min_seqno",
+      static_cast<void*>(&preclude_last_level_min_seqno));
+  // Restore the invariant asserted above, in case it was broken under the
+  // callback
+  preserve_time_min_seqno =
+      std::min(preclude_last_level_min_seqno, preserve_time_min_seqno);
+#endif
+
+  // Preserve sequence numbers for preserved write times and snapshots, though
+  // the specific sequence number of the earliest snapshot can be zeroed.
+  preserve_seqno_after_ =
+      std::max(preserve_time_min_seqno, SequenceNumber{1}) - 1;
+  preserve_seqno_after_ = std::min(preserve_seqno_after_, earliest_snapshot_);
+  // If using preclude feature, also preclude snapshots from last level, just
+  // because they are heuristically more likely to be accessed than non-snapshot
+  // data.
+  if (preclude_last_level_min_seqno < kMaxSequenceNumber &&
+      earliest_snapshot_ < preclude_last_level_min_seqno) {
+    preclude_last_level_min_seqno = earliest_snapshot_;
+  }
+  // Now combine what we would like to preclude from last level with what we
+  // can safely support without dangerously moving data back up the LSM tree,
+  // to get the final seqno threshold for penultimate vs. last. In particular,
+  // when the reserved output key range for the penultimate level does not
+  // include the entire last level input key range, we need to keep entries
+  // already in the last level there. (Even allowing within-range entries to
+  // move back up could cause problems with range tombstones. Perhaps it
+  // would be better in some rare cases to keep entries in the last level
+  // one-by-one rather than based on sequence number, but that would add extra
+  // tracking and complexity to CompactionIterator that is probably not
+  // worthwhile overall. Correctness is also more clear when splitting by
+  // seqno threshold.)
+  penultimate_after_seqno_ = std::max(preclude_last_level_min_seqno,
+                                      c->GetKeepInLastLevelThroughSeqno());
+
+  options_file_number_ = versions_->options_file_number();
 }
 
 uint64_t CompactionJob::GetSubcompactionsLimit() {
@@ -469,14 +521,14 @@ void CompactionJob::GenSubcompactionBoundaries() {
   ReadOptions read_options(Env::IOActivity::kCompaction);
   read_options.rate_limiter_priority = GetRateLimiterPriority();
   auto* c = compact_->compaction;
-  if (c->mutable_cf_options()->table_factory->Name() ==
+  if (c->mutable_cf_options().table_factory->Name() ==
       TableFactory::kPlainTableName()) {
     return;
   }
 
   if (c->max_subcompactions() <= 1 &&
-      !(c->immutable_options()->compaction_pri == kRoundRobin &&
-        c->immutable_options()->compaction_style == kCompactionStyleLevel)) {
+      !(c->immutable_options().compaction_pri == kRoundRobin &&
+        c->immutable_options().compaction_style == kCompactionStyleLevel)) {
     return;
   }
   auto* cfd = c->column_family_data();
@@ -506,7 +558,7 @@ void CompactionJob::GenSubcompactionBoundaries() {
         FileMetaData* f = flevel->files[i].file_metadata;
         std::vector<TableReader::Anchor> my_anchors;
         Status s = cfd->table_cache()->ApproximateKeyAnchors(
-            read_options, icomp, *f, *c->mutable_cf_options(), my_anchors);
+            read_options, icomp, *f, c->mutable_cf_options(), my_anchors);
         if (!s.ok() || my_anchors.empty()) {
           my_anchors.emplace_back(f->largest.user_key(), f->fd.GetFileSize());
         }
@@ -545,8 +597,8 @@ void CompactionJob::GenSubcompactionBoundaries() {
   // Get the number of planned subcompactions, may update reserve threads
   // and update extra_num_subcompaction_threads_reserved_ for round-robin
   uint64_t num_planned_subcompactions;
-  if (c->immutable_options()->compaction_pri == kRoundRobin &&
-      c->immutable_options()->compaction_style == kCompactionStyleLevel) {
+  if (c->immutable_options().compaction_pri == kRoundRobin &&
+      c->immutable_options().compaction_style == kCompactionStyleLevel) {
     // For round-robin compaction prioity, we need to employ more
     // subcompactions (may exceed the max_subcompaction limit). The extra
     // subcompactions will be executed using reserved threads and taken into
@@ -585,9 +637,9 @@ void CompactionJob::GenSubcompactionBoundaries() {
   uint64_t target_range_size = std::max(
       total_size / num_planned_subcompactions,
       MaxFileSizeForLevel(
-          *(c->mutable_cf_options()), out_lvl,
-          c->immutable_options()->compaction_style, base_level,
-          c->immutable_options()->level_compaction_dynamic_level_bytes));
+          c->mutable_cf_options(), out_lvl,
+          c->immutable_options().compaction_style, base_level,
+          c->immutable_options().level_compaction_dynamic_level_bytes));
 
   if (target_range_size >= total_size) {
     return;
@@ -730,14 +782,13 @@ Status CompactionJob::Run() {
             verify_table_read_options, file_options_,
             cfd->internal_comparator(), files_output[file_idx]->meta,
             /*range_del_agg=*/nullptr,
-            *compact_->compaction->mutable_cf_options(),
+            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()),
+            MaxFileSizeForL0MetaPin(compact_->compaction->mutable_cf_options()),
             /*smallest_compaction_key=*/nullptr,
             /*largest_compaction_key=*/nullptr,
             /*allow_unprepared_value=*/false);
@@ -794,7 +845,7 @@ Status CompactionJob::Run() {
   for (const auto& state : compact_->sub_compact_states) {
     for (const auto& output : state.GetOutputs()) {
       auto fn =
-          TableFileName(state.compaction->immutable_options()->cf_paths,
+          TableFileName(state.compaction->immutable_options().cf_paths,
                         output.meta.fd.GetNumber(), output.meta.fd.GetPathId());
       compact_->compaction->SetOutputTableProperties(fn,
                                                      output.table_properties);
@@ -853,8 +904,7 @@ Status CompactionJob::Run() {
   return status;
 }
 
-Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options,
-                              bool* compaction_released) {
+Status CompactionJob::Install(bool* compaction_released) {
   assert(compact_);
 
   AutoThreadOperationStageUpdater stage_updater(
@@ -870,7 +920,7 @@ Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options,
                                             compaction_stats_);
 
   if (status.ok()) {
-    status = InstallCompactionResults(mutable_cf_options, compaction_released);
+    status = InstallCompactionResults(compaction_released);
   }
   if (!versions_->io_status().ok()) {
     io_status_ = versions_->io_status();
@@ -959,8 +1009,7 @@ Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options,
   UpdateCompactionJobStats(stats);
 
   auto stream = event_logger_->LogToBuffer(log_buffer_, 8192);
-  stream << "job" << job_id_ << "event"
-         << "compaction_finished"
+  stream << "job" << job_id_ << "event" << "compaction_finished"
          << "compaction_time_micros" << stats.micros
          << "compaction_time_cpu_micros" << stats.cpu_micros << "output_level"
          << compact_->compaction->output_level() << "num_output_files"
@@ -1050,12 +1099,10 @@ void CompactionJob::NotifyOnSubcompactionBegin(
     listener->OnSubcompactionBegin(info);
   }
   info.status.PermitUncheckedError();
-
 }
 
 void CompactionJob::NotifyOnSubcompactionCompleted(
     SubcompactionState* sub_compact) {
-
   if (db_options_.listeners.empty()) {
     return;
   }
@@ -1083,8 +1130,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
   if (db_options_.compaction_service) {
     CompactionServiceJobStatus comp_status =
         ProcessKeyValueCompactionWithCompactionService(sub_compact);
-    if (comp_status == CompactionServiceJobStatus::kSuccess ||
-        comp_status == CompactionServiceJobStatus::kFailure) {
+    if (comp_status != CompactionServiceJobStatus::kUseLocal) {
       return;
     }
     // fallback to local compaction
@@ -1098,8 +1144,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
 
   // Create compaction filter and fail the compaction if
   // IgnoreSnapshots() = false because it is not supported anymore
-  const CompactionFilter* compaction_filter =
-      cfd->ioptions()->compaction_filter;
+  const CompactionFilter* compaction_filter = cfd->ioptions().compaction_filter;
   std::unique_ptr<CompactionFilter> compaction_filter_from_factory = nullptr;
   if (compaction_filter == nullptr) {
     compaction_filter_from_factory =
@@ -1115,9 +1160,12 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
 
   NotifyOnSubcompactionBegin(sub_compact);
 
-  auto range_del_agg = std::make_unique<CompactionRangeDelAggregator>(
-      &cfd->internal_comparator(), existing_snapshots_, &full_history_ts_low_,
-      &trim_ts_);
+  // This is assigned after creation of SubcompactionState to simplify that
+  // creation across both CompactionJob and CompactionServiceCompactionJob
+  sub_compact->AssignRangeDelAggregator(
+      std::make_unique<CompactionRangeDelAggregator>(
+          &cfd->internal_comparator(), existing_snapshots_,
+          &full_history_ts_low_, &trim_ts_));
 
   // TODO: since we already use C++17, should use
   // std::optional<const Slice> instead.
@@ -1162,7 +1210,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
   // Although the v2 aggregator is what the level iterator(s) know about,
   // the AddTombstones calls will be propagated down to the v1 aggregator.
   std::unique_ptr<InternalIterator> raw_input(versions_->MakeInputIterator(
-      read_options, sub_compact->compaction, range_del_agg.get(),
+      read_options, sub_compact->compaction, sub_compact->RangeDelAgg(),
       file_options_for_read_, start, end));
   InternalIterator* input = raw_input.get();
 
@@ -1254,28 +1302,27 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
   }
 
   MergeHelper merge(
-      env_, cfd->user_comparator(), cfd->ioptions()->merge_operator.get(),
+      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);
 
-  const MutableCFOptions* mutable_cf_options =
+  const auto& mutable_cf_options =
       sub_compact->compaction->mutable_cf_options();
-  assert(mutable_cf_options);
 
   std::vector<std::string> blob_file_paths;
 
   // TODO: BlobDB to support output_to_penultimate_level compaction, which needs
   //  2 builders, so may need to move to `CompactionOutputs`
   std::unique_ptr<BlobFileBuilder> blob_file_builder(
-      (mutable_cf_options->enable_blob_files &&
+      (mutable_cf_options.enable_blob_files &&
        sub_compact->compaction->output_level() >=
-           mutable_cf_options->blob_file_starting_level)
+           mutable_cf_options.blob_file_starting_level)
           ? new BlobFileBuilder(
                 versions_, fs_.get(),
-                sub_compact->compaction->immutable_options(),
-                mutable_cf_options, &file_options_, &write_options, db_id_,
+                &sub_compact->compaction->immutable_options(),
+                &mutable_cf_options, &file_options_, &write_options, db_id_,
                 db_session_id_, job_id_, cfd->GetID(), cfd->GetName(),
                 write_hint_, io_tracer_, blob_callback_,
                 BlobFileCreationReason::kCompaction, &blob_file_paths,
@@ -1298,20 +1345,15 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
       &existing_snapshots_, earliest_snapshot_,
       earliest_write_conflict_snapshot_, job_snapshot_seq, snapshot_checker_,
       env_, ShouldReportDetailedTime(env_, stats_),
-      /*expect_valid_internal_key=*/true, range_del_agg.get(),
+      /*expect_valid_internal_key=*/true, 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
           ->DoesInputReferenceBlobFiles() /* must_count_input_entries */,
       sub_compact->compaction, compaction_filter, shutting_down_,
-      db_options_.info_log, full_history_ts_low, preserve_time_min_seqno_,
-      preclude_last_level_min_seqno_);
+      db_options_.info_log, full_history_ts_low, preserve_seqno_after_);
   c_iter->SeekToFirst();
 
-  // Assign range delete aggregator to the target output level, which makes sure
-  // it only output to single level
-  sub_compact->AssignRangeDelAggregator(std::move(range_del_agg));
-
   const auto& c_iter_stats = c_iter->iter_stats();
 
   // define the open and close functions for the compaction files, which will be
@@ -1355,12 +1397,41 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
       last_cpu_micros = cur_cpu_micros;
     }
 
+    const auto& ikey = c_iter->ikey();
+    bool use_penultimate_output = ikey.sequence > penultimate_after_seqno_;
+#ifndef NDEBUG
+    if (sub_compact->compaction->SupportsPerKeyPlacement()) {
+      // Could be overridden by unittest
+      PerKeyPlacementContext context(sub_compact->compaction->output_level(),
+                                     ikey.user_key, c_iter->value(),
+                                     ikey.sequence, use_penultimate_output);
+      TEST_SYNC_POINT_CALLBACK("CompactionIterator::PrepareOutput.context",
+                               &context);
+      if (use_penultimate_output) {
+        // Verify that entries sent to the penultimate level are within the
+        // allowed range (because the input key range of the last level could
+        // be larger than the allowed output key range of the penultimate
+        // level). This check uses user keys (ignores sequence numbers) because
+        // compaction boundaries are a "clean cut" between user keys (see
+        // CompactionPicker::ExpandInputsToCleanCut()), which is especially
+        // important when preferred sequence numbers has been swapped in for
+        // kTypeValuePreferredSeqno / TimedPut.
+        sub_compact->compaction->TEST_AssertWithinPenultimateLevelOutputRange(
+            c_iter->user_key());
+      }
+    } else {
+      assert(penultimate_after_seqno_ == kMaxSequenceNumber);
+      assert(!use_penultimate_output);
+    }
+#endif  // NDEBUG
+
     // Add current compaction_iterator key to target compaction output, if the
     // output file needs to be close or open, it will call the `open_file_func`
     // and `close_file_func`.
     // TODO: it would be better to have the compaction file open/close moved
     // into `CompactionOutputs` which has the output file information.
-    status = sub_compact->AddToOutput(*c_iter, open_file_func, close_file_func);
+    status = sub_compact->AddToOutput(*c_iter, use_penultimate_output,
+                                      open_file_func, close_file_func);
     if (!status.ok()) {
       break;
     }
@@ -1567,16 +1638,31 @@ Status CompactionJob::FinishCompactionOutputFile(
     earliest_snapshot = existing_snapshots_[0];
   }
   if (s.ok()) {
+    // Inclusive lower bound, exclusive upper bound
+    std::pair<SequenceNumber, SequenceNumber> keep_seqno_range{
+        0, kMaxSequenceNumber};
+    if (sub_compact->compaction->SupportsPerKeyPlacement()) {
+      if (outputs.IsPenultimateLevel()) {
+        keep_seqno_range.first = penultimate_after_seqno_;
+      } else {
+        keep_seqno_range.second = penultimate_after_seqno_;
+      }
+    }
     CompactionIterationStats range_del_out_stats;
-    // if the compaction supports per_key_placement, only output range dels to
-    // the penultimate level.
-    // Note: Use `bottommost_level_ = true` for both bottommost and
+    // NOTE1: Use `bottommost_level_ = true` for both bottommost and
     // output_to_penultimate_level compaction here, as it's only used to decide
-    // if range dels could be dropped.
-    if (outputs.HasRangeDel()) {
-      s = outputs.AddRangeDels(comp_start_user_key, comp_end_user_key,
-                               range_del_out_stats, bottommost_level_,
-                               cfd->internal_comparator(), earliest_snapshot,
+    // if range dels could be dropped. (Logically, we are taking a single sorted
+    // run returned from CompactionIterator and physically splitting it between
+    // two output levels.)
+    // NOTE2: with per-key placement, range tombstones will be filtered on
+    // each output level based on sequence number (traversed twice). This is
+    // CPU-inefficient for a large number of range tombstones, but that would
+    // be an unusual work load.
+    if (sub_compact->HasRangeDel()) {
+      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,
                                next_table_min_key, full_history_ts_low_);
     }
     RecordDroppedKeys(range_del_out_stats, &sub_compact->compaction_job_stats);
@@ -1586,6 +1672,8 @@ Status CompactionJob::FinishCompactionOutputFile(
   const uint64_t current_entries = outputs.NumEntries();
 
   s = outputs.Finish(s, seqno_to_time_mapping_);
+  TEST_SYNC_POINT_CALLBACK(
+      "CompactionJob::FinishCompactionOutputFile()::AfterFinish", &s);
 
   if (s.ok()) {
     // With accurate smallest and largest key, we can get a slightly more
@@ -1637,7 +1725,7 @@ Status CompactionJob::FinishCompactionOutputFile(
     // 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,
+        TableFileName(sub_compact->compaction->immutable_options().cf_paths,
                       meta->fd.GetNumber(), meta->fd.GetPathId());
 
     // TODO(AR) it is not clear if there are any larger implications if
@@ -1684,7 +1772,7 @@ Status CompactionJob::FinishCompactionOutputFile(
     }
   }
   EventHelpers::LogAndNotifyTableFileCreationFinished(
-      event_logger_, cfd->ioptions()->listeners, dbname_, cfd->GetName(), fname,
+      event_logger_, cfd->ioptions().listeners, dbname_, cfd->GetName(), fname,
       job_id_, output_fd, oldest_blob_file_number, tp,
       TableFileCreationReason::kCompaction, status_for_listener, file_checksum,
       file_checksum_func_name);
@@ -1712,8 +1800,7 @@ Status CompactionJob::FinishCompactionOutputFile(
   return s;
 }
 
-Status CompactionJob::InstallCompactionResults(
-    const MutableCFOptions& mutable_cf_options, bool* compaction_released) {
+Status CompactionJob::InstallCompactionResults(bool* compaction_released) {
   assert(compact_);
 
   db_mutex_->AssertHeld();
@@ -1787,7 +1874,7 @@ Status CompactionJob::InstallCompactionResults(
   if ((compaction->compaction_reason() ==
            CompactionReason::kLevelMaxLevelSize ||
        compaction->compaction_reason() == CompactionReason::kRoundRobinTtl) &&
-      compaction->immutable_options()->compaction_pri == kRoundRobin) {
+      compaction->immutable_options().compaction_pri == kRoundRobin) {
     int start_level = compaction->start_level();
     if (start_level > 0) {
       auto vstorage = compaction->input_version()->storage_info();
@@ -1802,11 +1889,11 @@ Status CompactionJob::InstallCompactionResults(
     *compaction_released = true;
   };
 
-  return versions_->LogAndApply(
-      compaction->column_family_data(), mutable_cf_options, read_options,
-      write_options, edit, db_mutex_, db_directory_,
-      /*new_descriptor_log=*/false,
-      /*column_family_options=*/nullptr, manifest_wcb);
+  return versions_->LogAndApply(compaction->column_family_data(), read_options,
+                                write_options, edit, db_mutex_, db_directory_,
+                                /*new_descriptor_log=*/false,
+                                /*column_family_options=*/nullptr,
+                                manifest_wcb);
 }
 
 void CompactionJob::RecordCompactionIOStats() {
@@ -1842,7 +1929,7 @@ Status CompactionJob::OpenCompactionOutputFile(SubcompactionState* sub_compact,
   // Fire events.
   ColumnFamilyData* cfd = sub_compact->compaction->column_family_data();
   EventHelpers::NotifyTableFileCreationStarted(
-      cfd->ioptions()->listeners, dbname_, cfd->GetName(), fname, job_id_,
+      cfd->ioptions().listeners, dbname_, cfd->GetName(), fname, job_id_,
       TableFileCreationReason::kCompaction);
   // Make the output file
   std::unique_ptr<FSWritableFile> writable_file;
@@ -1856,12 +1943,12 @@ Status CompactionJob::OpenCompactionOutputFile(SubcompactionState* sub_compact,
   FileOptions fo_copy = file_options_;
   Temperature temperature = sub_compact->compaction->output_temperature();
   Temperature last_level_temp =
-      sub_compact->compaction->mutable_cf_options()->last_level_temperature;
+      sub_compact->compaction->mutable_cf_options().last_level_temperature;
   // Here last_level_temperature supersedes default_write_temperature, when
   // enabled and applicable
   if (last_level_temp != Temperature::kUnknown &&
       sub_compact->compaction->is_last_level() &&
-      !sub_compact->IsCurrentPenultimateLevel()) {
+      !outputs.IsPenultimateLevel()) {
     temperature = last_level_temp;
   }
   fo_copy.temperature = temperature;
@@ -1884,7 +1971,7 @@ Status CompactionJob::OpenCompactionOutputFile(SubcompactionState* sub_compact,
         job_id_, file_number, s.ToString().c_str());
     LogFlush(db_options_.info_log);
     EventHelpers::LogAndNotifyTableFileCreationFinished(
-        event_logger_, cfd->ioptions()->listeners, dbname_, cfd->GetName(),
+        event_logger_, cfd->ioptions().listeners, dbname_, cfd->GetName(),
         fname, job_id_, FileDescriptor(), kInvalidBlobFileNumber,
         TableProperties(), TableFileCreationReason::kCompaction, s,
         kUnknownFileChecksum, kUnknownFileChecksumFuncName);
@@ -1955,7 +2042,7 @@ Status CompactionJob::OpenCompactionOutputFile(SubcompactionState* sub_compact,
   writable_file->SetPreallocationBlockSize(static_cast<size_t>(
       sub_compact->compaction->OutputFilePreallocationSize()));
   const auto& listeners =
-      sub_compact->compaction->immutable_options()->listeners;
+      sub_compact->compaction->immutable_options().listeners;
   outputs.AssignFileWriter(new WritableFileWriter(
       std::move(writable_file), fname, fo_copy, db_options_.clock, io_tracer_,
       db_options_.stats, Histograms::SST_WRITE_MICROS, listeners,
@@ -1966,7 +2053,7 @@ Status CompactionJob::OpenCompactionOutputFile(SubcompactionState* sub_compact,
   const ReadOptions read_options(Env::IOActivity::kCompaction);
   const WriteOptions write_options(Env::IOActivity::kCompaction);
   TableBuilderOptions tboptions(
-      *cfd->ioptions(), *(sub_compact->compaction->mutable_cf_options()),
+      cfd->ioptions(), sub_compact->compaction->mutable_cf_options(),
       read_options, write_options, cfd->internal_comparator(),
       cfd->internal_tbl_prop_coll_factories(),
       sub_compact->compaction->output_compression(),
@@ -1975,9 +2062,7 @@ Status CompactionJob::OpenCompactionOutputFile(SubcompactionState* sub_compact,
       bottommost_level_, TableFileCreationReason::kCompaction,
       0 /* oldest_key_time */, current_time, db_id_, db_session_id_,
       sub_compact->compaction->max_output_file_size(), file_number,
-      preclude_last_level_min_seqno_ == kMaxSequenceNumber
-          ? preclude_last_level_min_seqno_
-          : std::min(earliest_snapshot_, preclude_last_level_min_seqno_));
+      penultimate_after_seqno_ /*last_level_inclusive_max_seqno_threshold*/);
 
   outputs.NewBuilder(tboptions);
 
@@ -2034,9 +2119,8 @@ bool CompactionJob::UpdateCompactionStats(uint64_t* num_input_range_del) {
       if (file_input_entries == 0) {
         uint64_t file_number = file_meta->fd.GetNumber();
         // Try getting info from table property
-        std::string fn =
-            TableFileName(compaction->immutable_options()->cf_paths,
-                          file_number, file_meta->fd.GetPathId());
+        std::string fn = TableFileName(compaction->immutable_options().cf_paths,
+                                       file_number, file_meta->fd.GetPathId());
         const auto& tp = input_table_properties.find(fn);
         if (tp != input_table_properties.end()) {
           file_input_entries = tp->second->num_entries;
@@ -2118,7 +2202,6 @@ void CompactionJob::UpdateCompactionJobStats(
 void CompactionJob::LogCompaction() {
   Compaction* compaction = compact_->compaction;
   ColumnFamilyData* cfd = compaction->column_family_data();
-
   // Let's check if anything will get logged. Don't prepare all the info if
   // we're not logging
   if (db_options_.info_log_level <= InfoLogLevel::INFO_LEVEL) {
@@ -2133,8 +2216,7 @@ void CompactionJob::LogCompaction() {
                    cfd->GetName().c_str(), scratch);
     // build event logger report
     auto stream = event_logger_->Log();
-    stream << "job" << job_id_ << "event"
-           << "compaction_started"
+    stream << "job" << job_id_ << "event" << "compaction_started"
            << "compaction_reason"
            << GetCompactionReasonString(compaction->compaction_reason());
     for (size_t i = 0; i < compaction->num_input_levels(); ++i) {
@@ -2151,8 +2233,8 @@ void CompactionJob::LogCompaction() {
                    ? int64_t{-1}  // Use -1 for "none"
                    : static_cast<int64_t>(existing_snapshots_[0]));
     if (compaction->SupportsPerKeyPlacement()) {
-      stream << "preclude_last_level_min_seqno"
-             << preclude_last_level_min_seqno_;
+      stream << "prenultimate_after_seqno" << penultimate_after_seqno_;
+      stream << "preserve_seqno_after" << preserve_seqno_after_;
       stream << "penultimate_output_level" << compaction->GetPenultimateLevel();
       stream << "penultimate_output_range"
              << GetCompactionPenultimateOutputRangeTypeString(
@@ -2171,7 +2253,7 @@ void CompactionJob::LogCompaction() {
 }
 
 std::string CompactionJob::GetTableFileName(uint64_t file_number) {
-  return TableFileName(compact_->compaction->immutable_options()->cf_paths,
+  return TableFileName(compact_->compaction->immutable_options().cf_paths,
                        file_number, compact_->compaction->output_path_id());
 }
 

db/compaction/compaction_job.h

@@ -176,7 +176,13 @@ class CompactionJob {
 
   // REQUIRED: mutex held
   // Prepare for the compaction by setting up boundaries for each subcompaction
-  void Prepare();
+  // and organizing seqno <-> time info. `known_single_subcompact` is non-null
+  // if we already have a known single subcompaction, with optional key bounds
+  // (currently for executing a remote compaction).
+  void Prepare(
+      std::optional<std::pair<std::optional<Slice>, std::optional<Slice>>>
+          known_single_subcompact);
+
   // REQUIRED mutex not held
   // Launch threads for each subcompaction and wait for them to finish. After
   // that, verify table is usable and finally do bookkeeping to unify
@@ -187,8 +193,7 @@ class CompactionJob {
   // Add compaction input/output to the current version
   // Releases compaction file through Compaction::ReleaseCompactionFiles().
   // Sets *compaction_released to true if compaction is released.
-  Status Install(const MutableCFOptions& mutable_cf_options,
-                 bool* compaction_released);
+  Status Install(bool* compaction_released);
 
   // Return the IO status
   IOStatus io_status() const { return io_status_; }
@@ -276,8 +281,7 @@ class CompactionJob {
                                     const Slice& next_table_min_key,
                                     const Slice* comp_start_user_key,
                                     const Slice* comp_end_user_key);
-  Status InstallCompactionResults(const MutableCFOptions& mutable_cf_options,
-                                  bool* compaction_released);
+  Status InstallCompactionResults(bool* compaction_released);
   Status OpenCompactionOutputFile(SubcompactionState* sub_compact,
                                   CompactionOutputs& outputs);
 
@@ -353,15 +357,18 @@ class CompactionJob {
   // it also collects the smallest_seqno -> oldest_ancester_time from the SST.
   SeqnoToTimeMapping seqno_to_time_mapping_;
 
-  // Minimal sequence number for preserving the time information. The time info
-  // older than this sequence number won't be preserved after the compaction and
-  // if it's bottommost compaction, the seq num will be zeroed out.
-  SequenceNumber preserve_time_min_seqno_ = kMaxSequenceNumber;
+  // Max seqno that can be zeroed out in last level, including for preserving
+  // write times.
+  SequenceNumber preserve_seqno_after_ = kMaxSequenceNumber;
 
   // Minimal sequence number to preclude the data from the last level. If the
   // key has bigger (newer) sequence number than this, it will be precluded from
   // the last level (output to penultimate level).
-  SequenceNumber preclude_last_level_min_seqno_ = kMaxSequenceNumber;
+  SequenceNumber penultimate_after_seqno_ = kMaxSequenceNumber;
+
+  // Options File Number used for Remote Compaction
+  // Setting this requires DBMutex.
+  uint64_t options_file_number_ = 0;
 
   // Get table file name in where it's outputting to, which should also be in
   // `output_directory_`.
@@ -496,6 +503,10 @@ class CompactionServiceCompactionJob : private CompactionJob {
       const CompactionServiceInput& compaction_service_input,
       CompactionServiceResult* compaction_service_result);
 
+  // REQUIRED: mutex held
+  // Like CompactionJob::Prepare()
+  void Prepare();
+
   // Run the compaction in current thread and return the result
   Status Run();
 

db/compaction/compaction_job_stats_test.cc

@@ -959,7 +959,6 @@ int main(int argc, char** argv) {
   return RUN_ALL_TESTS();
 }
 
-
 #else
 
 int main(int /*argc*/, char** /*argv*/) { return 0; }

db/compaction/compaction_job_test.cc

@@ -3,7 +3,6 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).
 
-
 #include "db/compaction/compaction_job.h"
 
 #include <algorithm>
@@ -302,7 +301,7 @@ class CompactionJobTestBase : public testing::Test {
     std::unique_ptr<TableBuilder> table_builder(
         cf_options_.table_factory->NewTableBuilder(
             TableBuilderOptions(
-                *cfd_->ioptions(), mutable_cf_options_, read_options,
+                cfd_->ioptions(), mutable_cf_options_, read_options,
                 write_options, cfd_->internal_comparator(),
                 cfd_->internal_tbl_prop_coll_factories(),
                 CompressionType::kNoCompression, CompressionOptions(),
@@ -398,8 +397,8 @@ class CompactionJobTestBase : public testing::Test {
 
     mutex_.Lock();
     EXPECT_OK(versions_->LogAndApply(
-        versions_->GetColumnFamilySet()->GetDefault(), mutable_cf_options_,
-        read_options_, write_options_, &edit, &mutex_, nullptr));
+        versions_->GetColumnFamilySet()->GetDefault(), read_options_,
+        write_options_, &edit, &mutex_, nullptr));
     mutex_.Unlock();
   }
 
@@ -461,7 +460,7 @@ 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(), nullptr, FileOptions(),
                              cfd_->internal_comparator(),
                              0 /* block_protection_bytes_per_key */),
           std::move(freader), file_size, &table_reader, false);
@@ -646,12 +645,12 @@ class CompactionJobTestBase : public testing::Test {
     }
 
     Compaction compaction(
-        cfd->current()->storage_info(), *cfd->ioptions(),
-        *cfd->GetLatestMutableCFOptions(), mutable_db_options_,
+        cfd->current()->storage_info(), cfd->ioptions(),
+        cfd->GetLatestMutableCFOptions(), mutable_db_options_,
         compaction_input_files, output_level,
         mutable_cf_options_.target_file_size_base,
         mutable_cf_options_.max_compaction_bytes, 0, kNoCompression,
-        cfd->GetLatestMutableCFOptions()->compression_opts,
+        cfd->GetLatestMutableCFOptions().compression_opts,
         Temperature::kUnknown, max_subcompactions, grandparents,
         /*earliest_snapshot*/ std::nullopt, /*snapshot_checker*/ nullptr, true);
     compaction.FinalizeInputInfo(cfd->current());
@@ -678,15 +677,14 @@ class CompactionJobTestBase : public testing::Test {
         full_history_ts_low_);
     VerifyInitializationOfCompactionJobStats(compaction_job_stats_);
 
-    compaction_job.Prepare();
+    compaction_job.Prepare(std::nullopt /*subcompact to be computed*/);
     mutex_.Unlock();
     Status s = compaction_job.Run();
     ASSERT_OK(s);
     ASSERT_OK(compaction_job.io_status());
     mutex_.Lock();
     bool compaction_released = false;
-    ASSERT_OK(compaction_job.Install(*cfd->GetLatestMutableCFOptions(),
-                                     &compaction_released));
+    ASSERT_OK(compaction_job.Install(&compaction_released));
     ASSERT_OK(compaction_job.io_status());
     mutex_.Unlock();
     log_buffer.FlushBufferToLog();
@@ -1536,14 +1534,11 @@ TEST_F(CompactionJobTest, VerifyPenultimateLevelOutput) {
       /*verify_func=*/[&](Compaction& comp) {
         for (char c = 'a'; c <= 'z'; c++) {
           if (c == 'a') {
-            ParsedInternalKey pik("a", 0U, kTypeValue);
-            ASSERT_FALSE(comp.WithinPenultimateLevelOutputRange(pik));
+            comp.TEST_AssertWithinPenultimateLevelOutputRange(
+                "a", true /*expect_failure*/);
           } else {
             std::string c_str{c};
-            // WithinPenultimateLevelOutputRange checks internal key range.
-            // 'z' is the last key, so set seqno properly.
-            ParsedInternalKey pik(c_str, c == 'z' ? 12U : 0U, kTypeValue);
-            ASSERT_TRUE(comp.WithinPenultimateLevelOutputRange(pik));
+            comp.TEST_AssertWithinPenultimateLevelOutputRange(c_str);
           }
         }
       });
@@ -1900,8 +1895,8 @@ TEST_F(CompactionJobTest, CutToSkipGrandparentFile) {
   const std::vector<int> input_levels = {0, 1};
   auto lvl0_files = cfd_->current()->storage_info()->LevelFiles(0);
   auto lvl1_files = cfd_->current()->storage_info()->LevelFiles(1);
-    RunCompaction({lvl0_files, lvl1_files}, input_levels,
-                  {expected_file1, expected_file2});
+  RunCompaction({lvl0_files, lvl1_files}, input_levels,
+                {expected_file1, expected_file2});
 }
 
 TEST_F(CompactionJobTest, CutToAlignGrandparentBoundary) {
@@ -1975,8 +1970,8 @@ TEST_F(CompactionJobTest, CutToAlignGrandparentBoundary) {
   const std::vector<int> input_levels = {0, 1};
   auto lvl0_files = cfd_->current()->storage_info()->LevelFiles(0);
   auto lvl1_files = cfd_->current()->storage_info()->LevelFiles(1);
-    RunCompaction({lvl0_files, lvl1_files}, input_levels,
-                  {expected_file1, expected_file2});
+  RunCompaction({lvl0_files, lvl1_files}, input_levels,
+                {expected_file1, expected_file2});
 }
 
 TEST_F(CompactionJobTest, CutToAlignGrandparentBoundarySameKey) {
@@ -2037,8 +2032,8 @@ TEST_F(CompactionJobTest, CutToAlignGrandparentBoundarySameKey) {
   for (int i = 80; i <= 100; i++) {
     snapshots.emplace_back(i);
   }
-    RunCompaction({lvl0_files, lvl1_files}, input_levels,
-                  {expected_file1, expected_file2}, snapshots);
+  RunCompaction({lvl0_files, lvl1_files}, input_levels,
+                {expected_file1, expected_file2}, snapshots);
 }
 
 TEST_F(CompactionJobTest, CutForMaxCompactionBytesSameKey) {

db/compaction/compaction_outputs.cc

@@ -320,7 +320,7 @@ bool CompactionOutputs::ShouldStopBefore(const CompactionIterator& c_iter) {
     // More details, check PR #1963
     const size_t num_skippable_boundaries_crossed =
         being_grandparent_gap_ ? 2 : 3;
-    if (compaction_->immutable_options()->compaction_style ==
+    if (compaction_->immutable_options().compaction_style ==
             kCompactionStyleLevel &&
         num_grandparent_boundaries_crossed >=
             num_skippable_boundaries_crossed &&
@@ -341,7 +341,7 @@ bool CompactionOutputs::ShouldStopBefore(const CompactionIterator& c_iter) {
     // target file size. The test shows it can generate larger files than a
     // static threshold like 75% and has a similar write amplification
     // improvement.
-    if (compaction_->immutable_options()->compaction_style ==
+    if (compaction_->immutable_options().compaction_style ==
             kCompactionStyleLevel &&
         current_output_file_size_ >=
             ((compaction_->target_output_file_size() + 99) / 100) *
@@ -455,9 +455,11 @@ void SetMaxSeqAndTs(InternalKey& internal_key, const Slice& user_key,
 }  // namespace
 
 Status CompactionOutputs::AddRangeDels(
+    CompactionRangeDelAggregator& range_del_agg,
     const Slice* comp_start_user_key, const Slice* comp_end_user_key,
     CompactionIterationStats& range_del_out_stats, bool bottommost_level,
     const InternalKeyComparator& icmp, SequenceNumber earliest_snapshot,
+    std::pair<SequenceNumber, SequenceNumber> keep_seqno_range,
     const Slice& next_table_min_key, const std::string& full_history_ts_low) {
   // The following example does not happen since
   // CompactionOutput::ShouldStopBefore() always return false for the first
@@ -470,7 +472,7 @@ Status CompactionOutputs::AddRangeDels(
   // Then meta.smallest will be set to comp_start_user_key@seqno
   // and meta.largest will be set to comp_start_user_key@kMaxSequenceNumber
   // which violates the assumption that meta.smallest should be <= meta.largest.
-  assert(HasRangeDel());
+  assert(!range_del_agg.IsEmpty());
   FileMetaData& meta = current_output().meta;
   const Comparator* ucmp = icmp.user_comparator();
   InternalKey lower_bound_buf, upper_bound_buf;
@@ -579,13 +581,19 @@ Status CompactionOutputs::AddRangeDels(
   assert(comp_end_user_key == nullptr || upper_bound == nullptr ||
          ucmp->CompareWithoutTimestamp(ExtractUserKey(*upper_bound),
                                        *comp_end_user_key) <= 0);
-  auto it = range_del_agg_->NewIterator(lower_bound, upper_bound);
+  auto it = range_del_agg.NewIterator(lower_bound, upper_bound);
   Slice last_tombstone_start_user_key{};
   bool reached_lower_bound = false;
   const ReadOptions read_options(Env::IOActivity::kCompaction);
   for (it->SeekToFirst(); it->Valid(); it->Next()) {
     auto tombstone = it->Tombstone();
     auto kv = tombstone.Serialize();
+    // Filter out by seqno for per-key placement
+    if (tombstone.seq_ < keep_seqno_range.first ||
+        tombstone.seq_ >= keep_seqno_range.second) {
+      continue;
+    }
+
     InternalKey tombstone_end = tombstone.SerializeEndKey();
     // TODO: the underlying iterator should support clamping the bounds.
     // tombstone_end.Encode is of form user_key@kMaxSeqno
@@ -744,11 +752,10 @@ Status CompactionOutputs::AddRangeDels(
 }
 
 void CompactionOutputs::FillFilesToCutForTtl() {
-  if (compaction_->immutable_options()->compaction_style !=
+  if (compaction_->immutable_options().compaction_style !=
           kCompactionStyleLevel ||
-      compaction_->immutable_options()->compaction_pri !=
-          kMinOverlappingRatio ||
-      compaction_->mutable_cf_options()->ttl == 0 ||
+      compaction_->immutable_options().compaction_pri != kMinOverlappingRatio ||
+      compaction_->mutable_cf_options().ttl == 0 ||
       compaction_->num_input_levels() < 2 || compaction_->bottommost_level()) {
     return;
   }
@@ -756,20 +763,19 @@ void CompactionOutputs::FillFilesToCutForTtl() {
   // We define new file with the oldest ancestor time to be younger than 1/4
   // TTL, and an old one to be older than 1/2 TTL time.
   int64_t temp_current_time;
-  auto get_time_status =
-      compaction_->immutable_options()->clock->GetCurrentTime(
-          &temp_current_time);
+  auto get_time_status = compaction_->immutable_options().clock->GetCurrentTime(
+      &temp_current_time);
   if (!get_time_status.ok()) {
     return;
   }
 
   auto current_time = static_cast<uint64_t>(temp_current_time);
-  if (current_time < compaction_->mutable_cf_options()->ttl) {
+  if (current_time < compaction_->mutable_cf_options().ttl) {
     return;
   }
 
   uint64_t old_age_thres =
-      current_time - compaction_->mutable_cf_options()->ttl / 2;
+      current_time - compaction_->mutable_cf_options().ttl / 2;
   const std::vector<FileMetaData*>& olevel =
       *(compaction_->inputs(compaction_->num_input_levels() - 1));
   for (FileMetaData* file : olevel) {
@@ -779,7 +785,7 @@ void CompactionOutputs::FillFilesToCutForTtl() {
     // of small files.
     if (oldest_ancester_time < old_age_thres &&
         file->fd.GetFileSize() >
-            compaction_->mutable_cf_options()->target_file_size_base / 2) {
+            compaction_->mutable_cf_options().target_file_size_base / 2) {
       files_to_cut_for_ttl_.push_back(file);
     }
   }

db/compaction/compaction_outputs.h

@@ -45,6 +45,8 @@ class CompactionOutputs {
   explicit CompactionOutputs(const Compaction* compaction,
                              const bool is_penultimate_level);
 
+  bool IsPenultimateLevel() const { return is_penultimate_level_; }
+
   // Add generated output to the list
   void AddOutput(FileMetaData&& meta, const InternalKeyComparator& icmp,
                  bool enable_hash, bool finished = false,
@@ -182,18 +184,13 @@ class CompactionOutputs {
   // @param next_table_min_key internal key lower bound for the next compaction
   // output.
   // @param full_history_ts_low used for range tombstone garbage collection.
-  Status AddRangeDels(const Slice* comp_start_user_key,
-                      const Slice* comp_end_user_key,
-                      CompactionIterationStats& range_del_out_stats,
-                      bool bottommost_level, const InternalKeyComparator& icmp,
-                      SequenceNumber earliest_snapshot,
-                      const Slice& next_table_min_key,
-                      const std::string& full_history_ts_low);
-
-  // if the outputs have range delete, range delete is also data
-  bool HasRangeDel() const {
-    return range_del_agg_ && !range_del_agg_->IsEmpty();
-  }
+  Status AddRangeDels(
+      CompactionRangeDelAggregator& range_del_agg,
+      const Slice* comp_start_user_key, const Slice* comp_end_user_key,
+      CompactionIterationStats& range_del_out_stats, bool bottommost_level,
+      const InternalKeyComparator& icmp, SequenceNumber earliest_snapshot,
+      std::pair<SequenceNumber, SequenceNumber> keep_seqno_range,
+      const Slice& next_table_min_key, const std::string& full_history_ts_low);
 
  private:
   friend class SubcompactionState;
@@ -259,11 +256,15 @@ class CompactionOutputs {
   // Close the current output. `open_file_func` is needed for creating new file
   // for range-dels only output file.
   Status CloseOutput(const Status& curr_status,
+                     CompactionRangeDelAggregator* range_del_agg,
                      const CompactionFileOpenFunc& open_file_func,
                      const CompactionFileCloseFunc& close_file_func) {
     Status status = curr_status;
-    // handle subcompaction containing only range deletions
-    if (status.ok() && !HasBuilder() && !HasOutput() && HasRangeDel()) {
+    // Handle subcompaction containing only range deletions. They could
+    // be dropped or sent to another output level, so this is only an
+    // over-approximate check for whether opening is needed.
+    if (status.ok() && !HasBuilder() && !HasOutput() && range_del_agg &&
+        !range_del_agg->IsEmpty()) {
       status = open_file_func(*this);
     }
     if (HasBuilder()) {
@@ -288,16 +289,6 @@ class CompactionOutputs {
     return outputs_.back();
   }
 
-  // Assign the range_del_agg to the target output level. There's only one
-  // range-del-aggregator per compaction outputs, for
-  // output_to_penultimate_level compaction it is only assigned to the
-  // penultimate level.
-  void AssignRangeDelAggregator(
-      std::unique_ptr<CompactionRangeDelAggregator>&& range_del_agg) {
-    assert(range_del_agg_ == nullptr);
-    range_del_agg_ = std::move(range_del_agg);
-  }
-
   const Compaction* compaction_;
 
   // current output builder and writer
@@ -319,7 +310,6 @@ class CompactionOutputs {
   // indicate if this CompactionOutputs obj for penultimate_level, should always
   // be false if per_key_placement feature is not enabled.
   const bool is_penultimate_level_;
-  std::unique_ptr<CompactionRangeDelAggregator> range_del_agg_ = nullptr;
 
   // partitioner information
   std::string last_key_for_partitioner_;

db/compaction/compaction_picker.cc

@@ -721,14 +721,14 @@ Compaction* CompactionPicker::CompactRange(
   // two files overlap.
   if (input_level > 0) {
     const uint64_t limit = mutable_cf_options.max_compaction_bytes;
-    uint64_t input_level_total = 0;
     int hint_index = -1;
-    InternalKey* smallest = nullptr;
+    assert(!inputs.empty());
+    // Always include first file for progress.
+    uint64_t input_level_total = inputs[0]->fd.GetFileSize();
+    InternalKey* smallest = &(inputs[0]->smallest);
     InternalKey* largest = nullptr;
-    for (size_t i = 0; i + 1 < inputs.size(); ++i) {
-      if (!smallest) {
-        smallest = &inputs[i]->smallest;
-      }
+    for (size_t i = 1; i < inputs.size(); ++i) {
+      // Consider whether to include inputs[i]
       largest = &inputs[i]->largest;
 
       uint64_t input_file_size = inputs[i]->fd.GetFileSize();
@@ -744,13 +744,11 @@ Compaction* CompactionPicker::CompactRange(
 
       input_level_total += input_file_size;
 
-      if (input_level_total + output_level_total >= limit) {
+      if (input_level_total + output_level_total > limit) {
+        // To ensure compaction size is <= limit, leave out inputs from
+        // index i onwards.
         covering_the_whole_range = false;
-        // still include the current file, so the compaction could be larger
-        // than max_compaction_bytes, which is also to make sure the compaction
-        // can make progress even `max_compaction_bytes` is small (e.g. smaller
-        // than an SST file).
-        inputs.files.resize(i + 1);
+        inputs.files.resize(i);
         break;
       }
     }
@@ -819,6 +817,9 @@ Compaction* CompactionPicker::CompactRange(
     output_level = vstorage->base_level();
     assert(output_level > 0);
   }
+  for (int i = input_level + 1; i < output_level; i++) {
+    assert(vstorage->NumLevelFiles(i) == 0);
+  }
   output_level_inputs.level = output_level;
   if (input_level != output_level) {
     int parent_index = -1;

db/compaction/compaction_picker.h

@@ -73,6 +73,8 @@ class CompactionPicker {
   // compaction_end will be set to nullptr.
   // Client is responsible for compaction_end storage -- when called,
   // *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(
       const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
       const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,

db/compaction/compaction_picker_test.cc

@@ -2656,6 +2656,37 @@ TEST_F(CompactionPickerTest, HitCompactionLimitWhenAddFileFromInputLevel) {
   ASSERT_EQ(5U, compaction->input(1, 0)->fd.GetNumber());
 }
 
+TEST_F(CompactionPickerTest, CompactRangeMaxCompactionBytes) {
+  mutable_cf_options_.max_compaction_bytes = 800000U;
+  NewVersionStorage(6, kCompactionStyleLevel);
+  // We will first pick file 1 and 2 and then stop before file 3.
+  // Since picking file 3 will pull in file 4 and 5 from L2 and
+  // exceed max_compaction_bytes.
+  Add(1, 1U, "100", "110", 10000U);
+  Add(1, 2U, "200", "210", 10000U, 0, 0);
+  Add(1, 3U, "400", "410", 10000U, 0, 0);
+  Add(2, 4U, "300", "310", 400000U);
+  Add(2, 5U, "320", "330", 400000U);
+  UpdateVersionStorageInfo();
+
+  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=*/""));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(1U, compaction->num_input_levels());
+  ASSERT_EQ(2, compaction->output_level());
+  ASSERT_EQ(2U, compaction->num_input_files(0));
+  ASSERT_EQ(1U, compaction->input(0, 0)->fd.GetNumber());
+  ASSERT_EQ(2U, compaction->input(0, 1)->fd.GetNumber());
+}
+
 TEST_F(CompactionPickerTest, IsTrivialMoveOn) {
   mutable_cf_options_.max_bytes_for_level_base = 10000u;
   mutable_cf_options_.max_compaction_bytes = 10001u;

db/compaction/compaction_service_job.cc

@@ -48,10 +48,7 @@ CompactionJob::ProcessKeyValueCompactionWithCompactionService(
   compaction_input.has_end = sub_compact->end.has_value();
   compaction_input.end =
       compaction_input.has_end ? sub_compact->end->ToString() : "";
-  compaction_input.options_file_number =
-      sub_compact->compaction->input_version()
-          ->version_set()
-          ->options_file_number();
+  compaction_input.options_file_number = options_file_number_;
 
   TEST_SYNC_POINT_CALLBACK(
       "CompactionServiceJob::ProcessKeyValueCompactionWithCompactionService",
@@ -86,6 +83,14 @@ CompactionJob::ProcessKeyValueCompactionWithCompactionService(
   switch (response.status) {
     case CompactionServiceJobStatus::kSuccess:
       break;
+    case CompactionServiceJobStatus::kAborted:
+      sub_compact->status =
+          Status::Aborted("Scheduling a remote compaction job was aborted");
+      ROCKS_LOG_WARN(
+          db_options_.info_log,
+          "[%s] [JOB %d] Remote compaction was aborted at Schedule()",
+          compaction->column_family_data()->GetName().c_str(), job_id_);
+      return response.status;
     case CompactionServiceJobStatus::kFailure:
       sub_compact->status = Status::Incomplete(
           "CompactionService failed to schedule a remote compaction job.");
@@ -121,6 +126,16 @@ CompactionJob::ProcessKeyValueCompactionWithCompactionService(
     return compaction_status;
   }
 
+  if (compaction_status == CompactionServiceJobStatus::kAborted) {
+    sub_compact->status =
+        Status::Aborted("Waiting a remote compaction job was aborted");
+    ROCKS_LOG_INFO(db_options_.info_log,
+                   "[%s] [JOB %d] Remote compaction was aborted during Wait()",
+                   compaction->column_family_data()->GetName().c_str(),
+                   job_id_);
+    return compaction_status;
+  }
+
   CompactionServiceResult compaction_result;
   s = CompactionServiceResult::Read(compaction_result_binary,
                                     &compaction_result);
@@ -177,7 +192,7 @@ CompactionJob::ProcessKeyValueCompactionWithCompactionService(
   for (const auto& file : compaction_result.output_files) {
     uint64_t file_num = versions_->NewFileNumber();
     auto src_file = compaction_result.output_path + "/" + file.file_name;
-    auto tgt_file = TableFileName(compaction->immutable_options()->cf_paths,
+    auto tgt_file = TableFileName(compaction->immutable_options().cf_paths,
                                   file_num, compaction->output_path_id());
     s = fs_->RenameFile(src_file, tgt_file, IOOptions(), nullptr);
     if (!s.ok()) {
@@ -189,6 +204,10 @@ 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);
     if (!s.ok()) {
       sub_compact->status = s;
@@ -265,43 +284,38 @@ CompactionServiceCompactionJob::CompactionServiceCompactionJob(
     std::string output_path,
     const CompactionServiceInput& compaction_service_input,
     CompactionServiceResult* compaction_service_result)
-    : 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,
-          compaction->mutable_cf_options()->paranoid_file_checks,
-          compaction->mutable_cf_options()->report_bg_io_stats, dbname,
-          &(compaction_service_result->stats), Env::Priority::USER, io_tracer,
-          manual_compaction_canceled, db_id, db_session_id,
-          compaction->column_family_data()->GetFullHistoryTsLow()),
+    : 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,
+                    compaction->mutable_cf_options().paranoid_file_checks,
+                    compaction->mutable_cf_options().report_bg_io_stats, dbname,
+                    &(compaction_service_result->stats), Env::Priority::USER,
+                    io_tracer, manual_compaction_canceled, db_id, db_session_id,
+                    compaction->column_family_data()->GetFullHistoryTsLow()),
       output_path_(std::move(output_path)),
       compaction_input_(compaction_service_input),
       compaction_result_(compaction_service_result) {}
 
+void CompactionServiceCompactionJob::Prepare() {
+  std::optional<Slice> begin;
+  if (compaction_input_.has_begin) {
+    begin = compaction_input_.begin;
+  }
+  std::optional<Slice> end;
+  if (compaction_input_.has_end) {
+    end = compaction_input_.end;
+  }
+  CompactionJob::Prepare(std::make_pair(begin, end));
+}
+
 Status CompactionServiceCompactionJob::Run() {
   AutoThreadOperationStageUpdater stage_updater(
       ThreadStatus::STAGE_COMPACTION_RUN);
 
   auto* c = compact_->compaction;
-  assert(c->column_family_data() != nullptr);
-  const VersionStorageInfo* storage_info = c->input_version()->storage_info();
-  assert(storage_info);
-  assert(storage_info->NumLevelFiles(compact_->compaction->level()) > 0);
-  write_hint_ = storage_info->CalculateSSTWriteHint(c->output_level());
-
-  bottommost_level_ = c->bottommost_level();
-
-  Slice begin = compaction_input_.begin;
-  Slice end = compaction_input_.end;
-  compact_->sub_compact_states.emplace_back(
-      c,
-      compaction_input_.has_begin ? std::optional<Slice>(begin)
-                                  : std::optional<Slice>(),
-      compaction_input_.has_end ? std::optional<Slice>(end)
-                                : std::optional<Slice>(),
-      /*sub_job_id*/ 0);
 
   log_buffer_->FlushBufferToLog();
   LogCompaction();
@@ -376,16 +390,18 @@ Status CompactionServiceCompactionJob::Run() {
   // Build Output
   compaction_result_->output_level = compact_->compaction->output_level();
   compaction_result_->output_path = output_path_;
-  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);
+  if (status.ok()) {
+    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);
+    }
   }
 
   TEST_SYNC_POINT_CALLBACK("CompactionServiceCompactionJob::Run:0",

db/compaction/compaction_service_test.cc

@@ -114,6 +114,8 @@ class MyTestCompactionService : public CompactionService {
     }
   }
 
+  void CancelAwaitingJobs() override { canceled_ = true; }
+
   void OnInstallation(const std::string& /*scheduled_job_id*/,
                       CompactionServiceJobStatus status) override {
     final_updated_status_ = status;
@@ -146,6 +148,7 @@ class MyTestCompactionService : public CompactionService {
   }
 
   void SetCanceled(bool canceled) { canceled_ = canceled; }
+  bool GetCanceled() { return canceled_; }
 
   void GetResult(CompactionServiceResult* deserialized) {
     CompactionServiceResult::Read(result_, deserialized).PermitUncheckedError();
@@ -194,12 +197,15 @@ class CompactionServiceTest : public DBTestBase {
     options->statistics = primary_statistics_;
     compactor_statistics_ = CreateDBStatistics();
 
-    compaction_service_ = std::make_shared<MyTestCompactionService>(
+    auto my_cs = std::make_shared<MyTestCompactionService>(
         dbname_, *options, compactor_statistics_, remote_listeners,
         remote_table_properties_collector_factories);
+
+    compaction_service_ = my_cs;
     options->compaction_service = compaction_service_;
     DestroyAndReopen(*options);
     CreateAndReopenWithCF({"cf_1", "cf_2", "cf_3"}, *options);
+    my_cs->SetCanceled(false);
   }
 
   Statistics* GetCompactorStatistics() { return compactor_statistics_.get(); }
@@ -339,6 +345,8 @@ TEST_F(CompactionServiceTest, BasicCompactions) {
         assert(*id != kNullUniqueId64x2);
         verify_passed++;
       });
+  Close();
+  my_cs->SetCanceled(false);
   ReopenWithColumnFamilies({kDefaultColumnFamilyName, "cf_1", "cf_2", "cf_3"},
                            options);
   ASSERT_GT(verify_passed, 0);
@@ -416,6 +424,38 @@ TEST_F(CompactionServiceTest, ManualCompaction) {
   ASSERT_TRUE(result.stats.is_remote_compaction);
 }
 
+TEST_F(CompactionServiceTest, CompactionOutputFileIOError) {
+  Options options = CurrentOptions();
+  options.disable_auto_compactions = true;
+  ReopenWithCompactionService(&options);
+  GenerateTestData();
+
+  auto my_cs = GetCompactionService();
+
+  SyncPoint::GetInstance()->SetCallBack(
+      "CompactionJob::FinishCompactionOutputFile()::AfterFinish",
+      [&](void* status) {
+        // override status
+        auto s = static_cast<Status*>(status);
+        *s = Status::IOError("Injected IOError!");
+      });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  std::string start_str = Key(15);
+  std::string end_str = Key(45);
+  Slice start(start_str);
+  Slice end(end_str);
+  uint64_t comp_num = my_cs->GetCompactionNum();
+  ASSERT_NOK(db_->CompactRange(CompactRangeOptions(), &start, &end));
+  ASSERT_GE(my_cs->GetCompactionNum(), comp_num + 1);
+
+  CompactionServiceResult result;
+  my_cs->GetResult(&result);
+  ASSERT_NOK(result.status);
+  ASSERT_TRUE(result.stats.is_manual_compaction);
+  ASSERT_TRUE(result.stats.is_remote_compaction);
+}
+
 TEST_F(CompactionServiceTest, PreservedOptionsLocalCompaction) {
   Options options = CurrentOptions();
   options.level0_file_num_compaction_trigger = 2;
@@ -455,7 +495,7 @@ TEST_F(CompactionServiceTest, PreservedOptionsLocalCompaction) {
         ASSERT_TRUE(s.IsNotFound());
         // Should be old value
         ASSERT_EQ(2, compaction->mutable_cf_options()
-                         ->level0_file_num_compaction_trigger);
+                         .level0_file_num_compaction_trigger);
         ASSERT_TRUE(dbfull()->min_options_file_numbers_.empty());
       });
 
@@ -541,7 +581,7 @@ TEST_F(CompactionServiceTest, PreservedOptionsRemoteCompaction) {
       "CompactionJob::ProcessKeyValueCompaction()::Processing", [&](void* arg) {
         auto compaction = static_cast<Compaction*>(arg);
         ASSERT_EQ(2, compaction->mutable_cf_options()
-                         ->level0_file_num_compaction_trigger);
+                         .level0_file_num_compaction_trigger);
       });
 
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
@@ -964,6 +1004,41 @@ TEST_F(CompactionServiceTest, CancelCompactionOnRemoteSide) {
   VerifyTestData();
 }
 
+TEST_F(CompactionServiceTest, CancelCompactionOnPrimarySide) {
+  Options options = CurrentOptions();
+  options.disable_auto_compactions = true;
+  ReopenWithCompactionService(&options);
+  GenerateTestData();
+
+  auto my_cs = GetCompactionService();
+
+  std::string start_str = Key(15);
+  std::string end_str = Key(45);
+  Slice start(start_str);
+  Slice end(end_str);
+  uint64_t comp_num = my_cs->GetCompactionNum();
+
+  ReopenWithCompactionService(&options);
+  GenerateTestData();
+  my_cs = GetCompactionService();
+
+  // Primary DB calls CancelAllBackgroundWork() while the compaction is running
+  SyncPoint::GetInstance()->SetCallBack(
+      "CompactionJob::Run():Inprogress",
+      [&](void* /*arg*/) { CancelAllBackgroundWork(db_, false /*wait*/); });
+
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  Status s = db_->CompactRange(CompactRangeOptions(), &start, &end);
+  ASSERT_TRUE(s.IsIncomplete());
+
+  // Check canceled_ was set to true by CancelAwaitingJobs()
+  ASSERT_TRUE(my_cs->GetCanceled());
+
+  // compaction number is not increased
+  ASSERT_GE(my_cs->GetCompactionNum(), comp_num);
+}
+
 TEST_F(CompactionServiceTest, FailedToStart) {
   Options options = CurrentOptions();
   options.disable_auto_compactions = true;
@@ -1086,6 +1161,63 @@ TEST_F(CompactionServiceTest, Snapshot) {
   db_->ReleaseSnapshot(s1);
 }
 
+TEST_F(CompactionServiceTest, PrecludeLastLevel) {
+  const int kNumTrigger = 4;
+  const int kNumLevels = 7;
+  const int kNumKeys = 100;
+
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleUniversal;
+  options.last_level_temperature = Temperature::kCold;
+  options.level0_file_num_compaction_trigger = 4;
+  options.max_subcompactions = 10;
+  options.num_levels = kNumLevels;
+
+  ReopenWithCompactionService(&options);
+  // Alternate for comparison: DestroyAndReopen(options);
+
+  // This is simpler than setting up mock time to make the user option work,
+  // but is not as direct as testing with preclude option itself.
+  SyncPoint::GetInstance()->SetCallBack(
+      "Compaction::SupportsPerKeyPlacement:Enabled",
+      [&](void* arg) { *static_cast<bool*>(arg) = true; });
+  SyncPoint::GetInstance()->SetCallBack(
+      "CompactionJob::PrepareTimes():preclude_last_level_min_seqno",
+      [&](void* arg) { *static_cast<SequenceNumber*>(arg) = 100; });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  for (int i = 0; i < kNumTrigger; i++) {
+    for (int j = 0; j < kNumKeys; j++) {
+      // FIXME: need to assign outputs to levels to allow overlapping ranges:
+      // ASSERT_OK(Put(Key(j * kNumTrigger + i), "v" + std::to_string(i)));
+      // instead of this (too easy):
+      ASSERT_OK(Put(Key(i * kNumKeys + j), "v" + std::to_string(i)));
+    }
+    ASSERT_OK(Flush());
+  }
+  ASSERT_OK(dbfull()->TEST_WaitForCompact());
+
+  // Data split between penultimate (kUnknown) and last (kCold) levels
+  // FIXME: need to assign outputs to levels to get this:
+  // ASSERT_EQ("0,0,0,0,0,1,1", FilesPerLevel());
+  // ASSERT_GT(GetSstSizeHelper(Temperature::kUnknown), 0);
+  // ASSERT_GT(GetSstSizeHelper(Temperature::kCold), 0);
+  // instead of this (WRONG but currently expected):
+  ASSERT_EQ("0,0,0,0,0,0,2", FilesPerLevel());
+  // Check manifest temperatures
+  ASSERT_GT(GetSstSizeHelper(Temperature::kUnknown), 0);
+  ASSERT_EQ(GetSstSizeHelper(Temperature::kCold), 0);
+  // TODO: Check FileSystem temperatures with FileTemperatureTestFS
+
+  for (int i = 0; i < kNumTrigger; i++) {
+    for (int j = 0; j < kNumKeys; j++) {
+      // FIXME
+      // ASSERT_EQ(Get(Key(j * kNumTrigger + i)), "v" + std::to_string(i));
+      ASSERT_EQ(Get(Key(i * kNumKeys + j)), "v" + std::to_string(i));
+    }
+  }
+}
+
 TEST_F(CompactionServiceTest, ConcurrentCompaction) {
   Options options = CurrentOptions();
   options.level0_file_num_compaction_trigger = 100;
@@ -1339,6 +1471,40 @@ TEST_F(CompactionServiceTest, FallbackLocalManual) {
   VerifyTestData();
 }
 
+TEST_F(CompactionServiceTest, AbortedWhileWait) {
+  Options options = CurrentOptions();
+  options.disable_auto_compactions = true;
+  ReopenWithCompactionService(&options);
+
+  GenerateTestData();
+  VerifyTestData();
+
+  auto my_cs = GetCompactionService();
+  Statistics* compactor_statistics = GetCompactorStatistics();
+  Statistics* primary_statistics = GetPrimaryStatistics();
+
+  my_cs->ResetOverride();
+  std::string start_str = Key(15);
+  std::string end_str = Key(45);
+  Slice start(start_str);
+  Slice end(end_str);
+
+  // Override Wait() result with kAborted
+  my_cs->OverrideWaitStatus(CompactionServiceJobStatus::kAborted);
+  start_str = Key(120);
+  start = start_str;
+
+  Status s = db_->CompactRange(CompactRangeOptions(), &start, nullptr);
+  ASSERT_NOK(s);
+  ASSERT_TRUE(s.IsAborted());
+  // no remote compaction is run
+  ASSERT_EQ(my_cs->GetCompactionNum(), 0);
+  // make sure the compaction statistics is not recorded any side
+  ASSERT_EQ(primary_statistics->getTickerCount(COMPACT_WRITE_BYTES), 0);
+  ASSERT_EQ(primary_statistics->getTickerCount(REMOTE_COMPACT_WRITE_BYTES), 0);
+  ASSERT_EQ(compactor_statistics->getTickerCount(COMPACT_WRITE_BYTES), 0);
+}
+
 TEST_F(CompactionServiceTest, RemoteEventListener) {
   class RemoteEventListenerTest : public EventListener {
    public:

db/compaction/subcompaction_state.cc

@@ -15,8 +15,19 @@
 namespace ROCKSDB_NAMESPACE {
 void SubcompactionState::AggregateCompactionOutputStats(
     InternalStats::CompactionStatsFull& compaction_stats) const {
+  // Outputs should be closed. By extension, any files created just for
+  // range deletes have already been written also.
+  assert(compaction_outputs_.HasBuilder() == false);
+  assert(penultimate_level_outputs_.HasBuilder() == false);
+
+  // FIXME: These stats currently include abandonned output files
+  // assert(compaction_outputs_.stats_.num_output_files ==
+  //        compaction_outputs_.outputs_.size());
+  // assert(penultimate_level_outputs_.stats_.num_output_files ==
+  //        penultimate_level_outputs_.outputs_.size());
+
   compaction_stats.stats.Add(compaction_outputs_.stats_);
-  if (HasPenultimateLevelOutputs()) {
+  if (penultimate_level_outputs_.HasOutput()) {
     compaction_stats.has_penultimate_level_output = true;
     compaction_stats.penultimate_level_stats.Add(
         penultimate_level_outputs_.stats_);
@@ -43,7 +54,7 @@ void SubcompactionState::Cleanup(Cache* cache) {
       // be picked up by scanning the DB directory.
       TableCache::ReleaseObsolete(
           cache, out.meta.fd.GetNumber(), nullptr /*handle*/,
-          compaction->mutable_cf_options()->uncache_aggressiveness);
+          compaction->mutable_cf_options().uncache_aggressiveness);
     }
   }
   // TODO: sub_compact.io_status is not checked like status. Not sure if thats
@@ -52,7 +63,7 @@ void SubcompactionState::Cleanup(Cache* cache) {
 }
 
 Slice SubcompactionState::SmallestUserKey() const {
-  if (has_penultimate_level_outputs_) {
+  if (penultimate_level_outputs_.HasOutput()) {
     Slice a = compaction_outputs_.SmallestUserKey();
     Slice b = penultimate_level_outputs_.SmallestUserKey();
     if (a.empty()) {
@@ -74,7 +85,7 @@ Slice SubcompactionState::SmallestUserKey() const {
 }
 
 Slice SubcompactionState::LargestUserKey() const {
-  if (has_penultimate_level_outputs_) {
+  if (penultimate_level_outputs_.HasOutput()) {
     Slice a = compaction_outputs_.LargestUserKey();
     Slice b = penultimate_level_outputs_.LargestUserKey();
     if (a.empty()) {
@@ -96,18 +107,13 @@ Slice SubcompactionState::LargestUserKey() const {
 }
 
 Status SubcompactionState::AddToOutput(
-    const CompactionIterator& iter,
+    const CompactionIterator& iter, bool use_penultimate_output,
     const CompactionFileOpenFunc& open_file_func,
     const CompactionFileCloseFunc& close_file_func) {
-  // update target output first
-  is_current_penultimate_level_ = iter.output_to_penultimate_level();
-  current_outputs_ = is_current_penultimate_level_ ? &penultimate_level_outputs_
-                                                   : &compaction_outputs_;
-  if (is_current_penultimate_level_) {
-    has_penultimate_level_outputs_ = true;
-  }
-
-  return Current().AddToOutput(iter, open_file_func, close_file_func);
+  // update target output
+  current_outputs_ = use_penultimate_output ? &penultimate_level_outputs_
+                                            : &compaction_outputs_;
+  return current_outputs_->AddToOutput(iter, open_file_func, close_file_func);
 }
 
 }  // namespace ROCKSDB_NAMESPACE

db/compaction/subcompaction_state.h

@@ -81,22 +81,13 @@ class SubcompactionState {
   // it returns both the last level outputs and penultimate level outputs.
   OutputIterator GetOutputs() const;
 
-  // Assign range dels aggregator, for each range_del, it can only be assigned
-  // to one output level, for per_key_placement, it's going to be the
-  // penultimate level.
-  // TODO: This does not work for per_key_placement + user-defined timestamp +
-  //  DeleteRange() combo. If user-defined timestamp is enabled,
-  //  it is possible for a range tombstone to belong to bottommost level (
-  //  seqno < earliest snapshot) without being dropped (garbage collection
-  //  for user-defined timestamp).
+  // Assign range dels aggregator. The various tombstones will potentially
+  // be filtered to different outputs.
   void AssignRangeDelAggregator(
       std::unique_ptr<CompactionRangeDelAggregator>&& range_del_agg) {
-    if (compaction->SupportsPerKeyPlacement()) {
-      penultimate_level_outputs_.AssignRangeDelAggregator(
-          std::move(range_del_agg));
-    } else {
-      compaction_outputs_.AssignRangeDelAggregator(std::move(range_del_agg));
-    }
+    assert(range_del_agg_ == nullptr);
+    assert(range_del_agg);
+    range_del_agg_ = std::move(range_del_agg);
   }
 
   void RemoveLastEmptyOutput() {
@@ -151,20 +142,11 @@ class SubcompactionState {
         sub_job_id(state.sub_job_id),
         compaction_outputs_(std::move(state.compaction_outputs_)),
         penultimate_level_outputs_(std::move(state.penultimate_level_outputs_)),
-        is_current_penultimate_level_(state.is_current_penultimate_level_),
-        has_penultimate_level_outputs_(state.has_penultimate_level_outputs_) {
-    current_outputs_ = is_current_penultimate_level_
-                           ? &penultimate_level_outputs_
-                           : &compaction_outputs_;
-  }
-
-  bool HasPenultimateLevelOutputs() const {
-    return has_penultimate_level_outputs_ ||
-           penultimate_level_outputs_.HasRangeDel();
-  }
-
-  bool IsCurrentPenultimateLevel() const {
-    return is_current_penultimate_level_;
+        range_del_agg_(std::move(state.range_del_agg_)) {
+    current_outputs_ =
+        state.current_outputs_ == &state.penultimate_level_outputs_
+            ? &penultimate_level_outputs_
+            : &compaction_outputs_;
   }
 
   // Add all the new files from this compaction to version_edit
@@ -187,8 +169,18 @@ class SubcompactionState {
     return *current_outputs_;
   }
 
+  CompactionRangeDelAggregator* RangeDelAgg() const {
+    return range_del_agg_.get();
+  }
+
+  // if the outputs have range delete, range delete is also data
+  bool HasRangeDel() const {
+    return range_del_agg_ && !range_del_agg_->IsEmpty();
+  }
+
   // Add compaction_iterator key/value to the `Current` output group.
   Status AddToOutput(const CompactionIterator& iter,
+                     bool use_penultimate_output,
                      const CompactionFileOpenFunc& open_file_func,
                      const CompactionFileCloseFunc& close_file_func);
 
@@ -197,14 +189,20 @@ class SubcompactionState {
   Status CloseCompactionFiles(const Status& curr_status,
                               const CompactionFileOpenFunc& open_file_func,
                               const CompactionFileCloseFunc& close_file_func) {
+    auto per_key = compaction->SupportsPerKeyPlacement();
     // Call FinishCompactionOutputFile() even if status is not ok: it needs to
     // close the output file.
     // CloseOutput() may open new compaction output files.
-    is_current_penultimate_level_ = true;
-    Status s = penultimate_level_outputs_.CloseOutput(
-        curr_status, open_file_func, close_file_func);
-    is_current_penultimate_level_ = false;
-    s = compaction_outputs_.CloseOutput(s, open_file_func, close_file_func);
+    Status s = curr_status;
+    if (per_key) {
+      s = penultimate_level_outputs_.CloseOutput(
+          s, range_del_agg_.get(), open_file_func, close_file_func);
+    } else {
+      assert(penultimate_level_outputs_.HasBuilder() == false);
+      assert(penultimate_level_outputs_.HasOutput() == false);
+    }
+    s = compaction_outputs_.CloseOutput(s, range_del_agg_.get(), open_file_func,
+                                        close_file_func);
     return s;
   }
 
@@ -213,8 +211,7 @@ class SubcompactionState {
   CompactionOutputs compaction_outputs_;
   CompactionOutputs penultimate_level_outputs_;
   CompactionOutputs* current_outputs_ = &compaction_outputs_;
-  bool is_current_penultimate_level_ = false;
-  bool has_penultimate_level_outputs_ = false;
+  std::unique_ptr<CompactionRangeDelAggregator> range_del_agg_;
 };
 
 }  // namespace ROCKSDB_NAMESPACE

db/compaction/tiered_compaction_test.cc

@@ -195,10 +195,9 @@ TEST_F(TieredCompactionTest, SequenceBasedTieredStorageUniversal) {
   std::vector<SequenceNumber> seq_history;
 
   SyncPoint::GetInstance()->SetCallBack(
-      "CompactionIterator::PrepareOutput.context", [&](void* arg) {
-        auto context = static_cast<PerKeyPlacementContext*>(arg);
-        context->output_to_penultimate_level =
-            context->seq_num > latest_cold_seq;
+      "CompactionJob::PrepareTimes():preclude_last_level_min_seqno",
+      [&](void* arg) {
+        *static_cast<SequenceNumber*>(arg) = latest_cold_seq.load();
       });
   SyncPoint::GetInstance()->EnableProcessing();
 
@@ -339,7 +338,10 @@ TEST_F(TieredCompactionTest, SequenceBasedTieredStorageUniversal) {
   ASSERT_GT(GetSstSizeHelper(Temperature::kCold), 0);
 }
 
-TEST_F(TieredCompactionTest, RangeBasedTieredStorageUniversal) {
+// This test was essentially for a hacked-up version on future functionality.
+// It can be resurrected if/when a form of range-based tiering is properly
+// implemented.
+TEST_F(TieredCompactionTest, DISABLED_RangeBasedTieredStorageUniversal) {
   const int kNumTrigger = 4;
   const int kNumLevels = 7;
   const int kNumKeys = 100;
@@ -440,12 +442,12 @@ TEST_F(TieredCompactionTest, RangeBasedTieredStorageUniversal) {
   {
     MutexLock l(&mutex);
     hot_start = Key(1);
-    hot_end = Key(1000);
+    hot_end = Key(300);
   }
 
   // generate files just enough to trigger compaction
   for (int i = 0; i < kNumTrigger - 1; i++) {
-    for (int j = 0; j < 1000; j++) {
+    for (int j = 0; j < 300; j++) {
       ASSERT_OK(Put(Key(j), "value" + std::to_string(j)));
     }
     ASSERT_OK(Flush());
@@ -494,6 +496,7 @@ TEST_F(TieredCompactionTest, RangeBasedTieredStorageUniversal) {
   // verify data
   std::string value;
   for (int i = 0; i < kNumKeys; i++) {
+    SCOPED_TRACE(Key(i));
     if (i < 70) {
       ASSERT_TRUE(db_->Get(ReadOptions(), Key(i), &value).IsNotFound());
     } else {
@@ -527,13 +530,13 @@ TEST_F(TieredCompactionTest, LevelColdRangeDelete) {
   options.max_subcompactions = 10;
   DestroyAndReopen(options);
 
-  std::atomic_uint64_t latest_cold_seq = 0;
+  // Initially let everything into cold
+  std::atomic_uint64_t latest_cold_seq = kMaxSequenceNumber;
 
   SyncPoint::GetInstance()->SetCallBack(
-      "CompactionIterator::PrepareOutput.context", [&](void* arg) {
-        auto context = static_cast<PerKeyPlacementContext*>(arg);
-        context->output_to_penultimate_level =
-            context->seq_num > latest_cold_seq;
+      "CompactionJob::PrepareTimes():preclude_last_level_min_seqno",
+      [&](void* arg) {
+        *static_cast<SequenceNumber*>(arg) = latest_cold_seq.load();
       });
   SyncPoint::GetInstance()->EnableProcessing();
 
@@ -634,13 +637,13 @@ TEST_F(TieredCompactionTest, LevelOutofBoundaryRangeDelete) {
   options.max_subcompactions = 10;
   DestroyAndReopen(options);
 
-  std::atomic_uint64_t latest_cold_seq = 0;
+  // Initially let everything into cold
+  std::atomic_uint64_t latest_cold_seq = kMaxSequenceNumber;
 
   SyncPoint::GetInstance()->SetCallBack(
-      "CompactionIterator::PrepareOutput.context", [&](void* arg) {
-        auto context = static_cast<PerKeyPlacementContext*>(arg);
-        context->output_to_penultimate_level =
-            context->seq_num > latest_cold_seq;
+      "CompactionJob::PrepareTimes():preclude_last_level_min_seqno",
+      [&](void* arg) {
+        *static_cast<SequenceNumber*>(arg) = latest_cold_seq.load();
       });
   SyncPoint::GetInstance()->EnableProcessing();
 
@@ -654,6 +657,8 @@ TEST_F(TieredCompactionTest, LevelOutofBoundaryRangeDelete) {
   ASSERT_GT(GetSstSizeHelper(Temperature::kCold), 0);
   ASSERT_EQ("0,0,10", FilesPerLevel());
 
+  // Stop admitting to cold tier
+  latest_cold_seq = dbfull()->GetLatestSequenceNumber();
   auto snap = db_->GetSnapshot();
 
   // only range delete
@@ -767,10 +772,9 @@ TEST_F(TieredCompactionTest, UniversalRangeDelete) {
   std::atomic_uint64_t latest_cold_seq = 0;
 
   SyncPoint::GetInstance()->SetCallBack(
-      "CompactionIterator::PrepareOutput.context", [&](void* arg) {
-        auto context = static_cast<PerKeyPlacementContext*>(arg);
-        context->output_to_penultimate_level =
-            context->seq_num > latest_cold_seq;
+      "CompactionJob::PrepareTimes():preclude_last_level_min_seqno",
+      [&](void* arg) {
+        *static_cast<SequenceNumber*>(arg) = latest_cold_seq.load();
       });
   SyncPoint::GetInstance()->EnableProcessing();
 
@@ -894,14 +898,13 @@ TEST_F(TieredCompactionTest, SequenceBasedTieredStorageLevel) {
   options.max_subcompactions = 10;
   DestroyAndReopen(options);
 
-  std::atomic_uint64_t latest_cold_seq = 0;
+  std::atomic_uint64_t latest_cold_seq = kMaxSequenceNumber;
   std::vector<SequenceNumber> seq_history;
 
   SyncPoint::GetInstance()->SetCallBack(
-      "CompactionIterator::PrepareOutput.context", [&](void* arg) {
-        auto context = static_cast<PerKeyPlacementContext*>(arg);
-        context->output_to_penultimate_level =
-            context->seq_num > latest_cold_seq;
+      "CompactionJob::PrepareTimes():preclude_last_level_min_seqno",
+      [&](void* arg) {
+        *static_cast<SequenceNumber*>(arg) = latest_cold_seq.load();
       });
   SyncPoint::GetInstance()->EnableProcessing();
 
@@ -952,6 +955,7 @@ TEST_F(TieredCompactionTest, SequenceBasedTieredStorageLevel) {
   VerifyCompactionStats(expect_stats, expect_pl_stats);
 
   // Add new data, which is all hot and overriding all existing data
+  latest_cold_seq = dbfull()->GetLatestSequenceNumber();
   for (int i = 0; i < kNumTrigger; i++) {
     for (int j = 0; j < kNumKeys; j++) {
       ASSERT_OK(Put(Key(i * 10 + j), "value" + std::to_string(i)));
@@ -1104,7 +1108,13 @@ TEST_F(TieredCompactionTest, SequenceBasedTieredStorageLevel) {
   ASSERT_GT(GetSstSizeHelper(Temperature::kCold), 0);
 }
 
-TEST_F(TieredCompactionTest, RangeBasedTieredStorageLevel) {
+// This test was essentially for a hacked-up version on future functionality.
+// It can be resurrected if/when a form of range-based tiering is properly
+// implemented.
+// FIXME: aside from that, this test reproduces a near-endless compaction
+// cycle that needs to be reproduced independently and fixed before
+// leveled compaction can be used with the preclude feature in production.
+TEST_F(TieredCompactionTest, DISABLED_RangeBasedTieredStorageLevel) {
   const int kNumTrigger = 4;
   const int kNumLevels = 7;
   const int kNumKeys = 100;
@@ -1213,6 +1223,9 @@ TEST_F(TieredCompactionTest, RangeBasedTieredStorageLevel) {
 
   // Tests that we only compact keys up to penultimate level
   // that are within penultimate level input's internal key range.
+  // UPDATE: this functionality has changed. With penultimate-enabled
+  // compaction, the expanded potential output range in the penultimate
+  // level is reserved so should be safe to use.
   {
     MutexLock l(&mutex);
     hot_start = Key(0);
@@ -1221,7 +1234,6 @@ TEST_F(TieredCompactionTest, RangeBasedTieredStorageLevel) {
   const Snapshot* temp_snap = db_->GetSnapshot();
   // Key(0) and Key(1) here are inserted with higher sequence number
   // than Key(0) and Key(1) inserted above.
-  // Only Key(0) in last level will be compacted up, not Key(1).
   ASSERT_OK(Put(Key(0), "value" + std::to_string(0)));
   ASSERT_OK(Put(Key(1), "value" + std::to_string(100)));
   ASSERT_OK(db_->CompactRange(cro, nullptr, nullptr));
@@ -1231,14 +1243,17 @@ TEST_F(TieredCompactionTest, RangeBasedTieredStorageLevel) {
     db_->GetLiveFilesMetaData(&metas);
     for (const auto& f : metas) {
       if (f.temperature == Temperature::kUnknown) {
-        // Expect Key(0), Key(0), Key(1)
-        ASSERT_EQ(f.num_entries, 3);
+        // UPDATED (was 3 entries, Key0..Key1)
+        ASSERT_EQ(f.num_entries, 52);
         ASSERT_EQ(f.smallestkey, Key(0));
-        ASSERT_EQ(f.largestkey, Key(1));
+        ASSERT_EQ(f.largestkey, Key(49));
       } else {
         ASSERT_EQ(f.temperature, Temperature::kCold);
-        // Key(2)-Key(49) and Key(100).
-        ASSERT_EQ(f.num_entries, 50);
+        // UPDATED (was 50 entries, Key0..Key49)
+        // Key(100) is outside the hot range
+        ASSERT_EQ(f.num_entries, 1);
+        ASSERT_EQ(f.smallestkey, Key(100));
+        ASSERT_EQ(f.largestkey, Key(100));
       }
     }
   }
@@ -1279,8 +1294,6 @@ class PrecludeLastLevelTestBase : public DBTestBase {
       const std::unordered_map<std::string, std::string>& db_config_change) {
     if (dynamic) {
       if (config_change.size() > 0) {
-        // FIXME: temporary while preserve/preclude options are not user mutable
-        SaveAndRestore<bool> m(&TEST_allowSetOptionsImmutableInMutable, true);
         ASSERT_OK(db_->SetOptions(config_change));
       }
       if (db_config_change.size() > 0) {
@@ -1616,7 +1629,8 @@ TEST_P(PrecludeLastLevelTest, CheckInternalKeyRange) {
 
   ASSERT_EQ("0,0,0,0,0,2,2", FilesPerLevel());
 
-  auto VerifyLogicalState = [&]() {
+  auto VerifyLogicalState = [&](int line) {
+    SCOPED_TRACE("Called from line " + std::to_string(line));
     // First with snapshot
     ASSERT_EQ("val2", Get(Key(2), snapshot));
     ASSERT_EQ("val3", Get(Key(3), snapshot));
@@ -1630,7 +1644,7 @@ TEST_P(PrecludeLastLevelTest, CheckInternalKeyRange) {
     ASSERT_EQ("val100", Get(Key(100)));
   };
 
-  VerifyLogicalState();
+  VerifyLogicalState(__LINE__);
 
   // Try to compact keys up
   CompactRangeOptions cro;
@@ -1641,24 +1655,40 @@ TEST_P(PrecludeLastLevelTest, CheckInternalKeyRange) {
   // type:1 vs. file #15 smallest key: '6B6579303030303033' seq:104, type:1
   ASSERT_OK(CompactRange(cro, Key(1), Key(2)));
 
-  VerifyLogicalState();
+  VerifyLogicalState(__LINE__);
 
   db_->ReleaseSnapshot(snapshot);
   Close();
 }
 
-TEST_P(PrecludeLastLevelTest, RangeTombstoneSnapshotMigrateFromLast) {
+INSTANTIATE_TEST_CASE_P(PrecludeLastLevelTest, PrecludeLastLevelTest,
+                        ::testing::Bool());
+
+class PrecludeWithCompactStyleTest : public PrecludeLastLevelTestBase,
+                                     public testing::WithParamInterface<bool> {
+ public:
+  void ApplyConfigChange(
+      Options* options,
+      const std::unordered_map<std::string, std::string>& config_change,
+      const std::unordered_map<std::string, std::string>& db_config_change =
+          {}) {
+    // Depends on dynamic config change while holding a snapshot
+    ApplyConfigChangeImpl(true /*dynamic*/, options, config_change,
+                          db_config_change);
+  }
+};
+
+TEST_P(PrecludeWithCompactStyleTest, RangeTombstoneSnapshotMigrateFromLast) {
   // Reproducer for issue originally described in
   // https://github.com/facebook/rocksdb/pull/9964/files#r1024449523
-  if (!UseDynamicConfig()) {
-    // Depends on config change while holding a snapshot
-    return;
-  }
+  const bool universal = GetParam();
   const int kNumLevels = 7;
   auto options = CurrentOptions();
   options.env = mock_env_.get();
   options.last_level_temperature = Temperature::kCold;
-  options.level_compaction_dynamic_level_bytes = true;
+  options.compaction_style =
+      universal ? kCompactionStyleUniversal : kCompactionStyleLevel;
+  options.compaction_options_universal.allow_trivial_move = true;
   options.num_levels = kNumLevels;
   options.statistics = CreateDBStatistics();
   options.max_subcompactions = 10;
@@ -1678,7 +1708,13 @@ TEST_P(PrecludeLastLevelTest, RangeTombstoneSnapshotMigrateFromLast) {
   ASSERT_OK(db_->DeleteRange({}, db_->DefaultColumnFamily(), Key(1), Key(5)));
   ASSERT_OK(Put(Key(1), "val1"));
   ASSERT_OK(Flush());
-  MoveFilesToLevel(6);
+
+  // Send to last level
+  if (universal) {
+    ASSERT_OK(CompactRange({}, {}, {}));
+  } else {
+    MoveFilesToLevel(6);
+  }
   ASSERT_EQ("0,0,0,0,0,0,1", FilesPerLevel());
 
   ApplyConfigChange(&options, {{"preclude_last_level_data_seconds", "10000"}});
@@ -1697,10 +1733,18 @@ TEST_P(PrecludeLastLevelTest, RangeTombstoneSnapshotMigrateFromLast) {
       [&] { mock_clock_->MockSleepForSeconds(static_cast<int>(1000)); });
 
   ASSERT_OK(Flush());
-  MoveFilesToLevel(5);
+
+  // Send three files to next-to-last level (if explicitly needed)
+  if (universal) {
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+  } else {
+    MoveFilesToLevel(5);
+  }
+
   ASSERT_EQ("0,0,0,0,0,3,1", FilesPerLevel());
 
-  auto VerifyLogicalState = [&]() {
+  auto VerifyLogicalState = [&](int line) {
+    SCOPED_TRACE("Called from line " + std::to_string(line));
     // First with snapshot
     if (snapshot) {
       ASSERT_EQ("NOT_FOUND", Get(Key(0), snapshot));
@@ -1720,24 +1764,31 @@ TEST_P(PrecludeLastLevelTest, RangeTombstoneSnapshotMigrateFromLast) {
     ASSERT_EQ("val7", Get(Key(7)));
   };
 
-  VerifyLogicalState();
+  VerifyLogicalState(__LINE__);
 
   // Try a limited range compaction
-  // FIXME: this currently hits the "Unsafe to store Seq later than snapshot"
-  // error. Needs to work safely for preclude option to be user mutable.
-  // ASSERT_OK(CompactRange({}, Key(3), Key(4)));
+  // (These would previously hit "Unsafe to store Seq later than snapshot")
+  if (universal) {
+    uint64_t middle_l5 = GetLevelFileMetadatas(5)[1]->fd.GetNumber();
+    ASSERT_OK(db_->CompactFiles({}, {MakeTableFileName(middle_l5)}, 6));
+  } else {
+    ASSERT_OK(CompactRange({}, Key(3), Key(4)));
+  }
   EXPECT_EQ("0,0,0,0,0,3,1", FilesPerLevel());
-  VerifyLogicalState();
+  VerifyLogicalState(__LINE__);
 
   // Compact everything, but some data still goes to both penultimate and last
   // levels. A full-range compaction should be safe to "migrate" data from the
   // last level to penultimate (because of preclude setting change).
   ASSERT_OK(CompactRange({}, {}, {}));
   EXPECT_EQ("0,0,0,0,0,1,1", FilesPerLevel());
-  VerifyLogicalState();
+  VerifyLogicalState(__LINE__);
   // Key1 should have been migrated out of the last level
-  auto& meta = *GetLevelFileMetadatas(6)[0];
-  ASSERT_LT(Key(1), meta.smallest.user_key().ToString());
+  // FIXME: doesn't yet work with leveled compaction
+  if (universal) {
+    auto& meta = *GetLevelFileMetadatas(6)[0];
+    ASSERT_LT(Key(1), meta.smallest.user_key().ToString());
+  }
 
   // Make data eligible for last level
   db_->ReleaseSnapshot(snapshot);
@@ -1746,13 +1797,13 @@ TEST_P(PrecludeLastLevelTest, RangeTombstoneSnapshotMigrateFromLast) {
 
   ASSERT_OK(CompactRange({}, {}, {}));
   EXPECT_EQ("0,0,0,0,0,0,1", FilesPerLevel());
-  VerifyLogicalState();
+  VerifyLogicalState(__LINE__);
 
   Close();
 }
 
-INSTANTIATE_TEST_CASE_P(PrecludeLastLevelTest, PrecludeLastLevelTest,
-                        ::testing::Bool());
+INSTANTIATE_TEST_CASE_P(PrecludeWithCompactStyleTest,
+                        PrecludeWithCompactStyleTest, ::testing::Bool());
 
 class TimedPutPrecludeLastLevelTest
     : public PrecludeLastLevelTestBase,

db/convenience.cc

@@ -4,7 +4,6 @@
 //  (found in the LICENSE.Apache file in the root directory).
 //
 
-
 #include "rocksdb/convenience.h"
 
 #include "db/convenience_impl.h"

db/corruption_test.cc

@@ -7,8 +7,6 @@
 // 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 "rocksdb/options.h"
-
 #include <fcntl.h>
 #include <sys/stat.h>
 #include <sys/types.h>
@@ -25,6 +23,7 @@
 #include "rocksdb/convenience.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
+#include "rocksdb/options.h"
 #include "rocksdb/table.h"
 #include "rocksdb/utilities/transaction_db.h"
 #include "rocksdb/write_batch.h"

db/cuckoo_table_db_test.cc

@@ -3,7 +3,6 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).
 
-
 #include "db/db_impl/db_impl.h"
 #include "db/db_test_util.h"
 #include "rocksdb/db.h"
@@ -349,4 +348,3 @@ int main(int argc, char** argv) {
     return 0;
   }
 }
-

db/db_basic_test.cc

@@ -878,7 +878,6 @@ TEST_F(DBBasicTest, Snapshot) {
   } while (ChangeOptions());
 }
 
-
 class DBBasicMultiConfigs : public DBBasicTest,
                             public ::testing::WithParamInterface<int> {
  public:
@@ -2604,8 +2603,7 @@ class DBMultiGetAsyncIOTest : public DBBasicTest,
     // Warm up the block cache so we don't need to use the IO uring
     Iterator* iter = dbfull()->NewIterator(ReadOptions());
     for (iter->SeekToFirst(); iter->Valid() && iter->status().ok();
-         iter->Next())
-      ;
+         iter->Next());
     EXPECT_OK(iter->status());
     delete iter;
 #endif  // ROCKSDB_IOURING_PRESENT
@@ -4996,6 +4994,97 @@ TEST_F(DBBasicTest, VerifyFileChecksumsReadahead) {
             (sst_size + alignment - 1) / (alignment));
 }
 
+TEST_F(DBBasicTest, DisallowMemtableWrite) {
+  // This test is mostly about what you can't do with memtable writes
+  // disallowed. For what you can do, see
+  // ExternalSSTFileBasicTest.FailIfNotBottommostLevelAndDisallowMemtable
+  Options options_allow = GetDefaultOptions();
+  options_allow.create_if_missing = true;
+  Options options_disallow = options_allow;
+  options_disallow.disallow_memtable_writes = true;
+
+  DestroyAndReopen(options_allow);
+  // CFs allowing and disallowing memtable write
+  CreateColumnFamilies({"cf1", "cf2"}, options_allow);
+  CreateColumnFamilies({"cf3"}, options_disallow);
+  // XXX: needed to get consistent handles_ mappings
+  ReopenWithColumnFamilies(
+      {"default", "cf1", "cf2", "cf3"},
+      {options_allow, options_allow, options_allow, options_disallow});
+
+  EXPECT_EQ(Put(0, "a0", "1").code(), Status::Code::kOk);
+  EXPECT_EQ(Put(1, "a1", "1").code(), Status::Code::kOk);
+  EXPECT_EQ(Put(2, "a2", "1").code(), Status::Code::kOk);
+  EXPECT_EQ(Put(3, "a3", "1").code(), Status::Code::kInvalidArgument);
+
+  EXPECT_EQ(Get(0, "a0"), "1");
+  EXPECT_EQ(Get(1, "a1"), "1");
+  EXPECT_EQ(Get(2, "a2"), "1");
+  EXPECT_EQ(Get(3, "a3"), "NOT_FOUND");
+
+  EXPECT_EQ(Delete(0, "z0").code(), Status::Code::kOk);
+  EXPECT_EQ(Delete(1, "z1").code(), Status::Code::kOk);
+  EXPECT_EQ(Delete(2, "z2").code(), Status::Code::kOk);
+  EXPECT_EQ(Delete(3, "z3").code(), Status::Code::kInvalidArgument);
+
+  WriteBatch wb;
+  EXPECT_EQ(wb.Put(handles_[0], "b0", "2").code(), Status::Code::kOk);
+  EXPECT_EQ(wb.Put(handles_[1], "b1", "2").code(), Status::Code::kOk);
+  EXPECT_EQ(wb.Put(handles_[2], "b2", "2").code(), Status::Code::kOk);
+  EXPECT_EQ(wb.Put(handles_[3], "b3", "2").code(),
+            Status::Code::kInvalidArgument);
+  ASSERT_OK(db_->Write({}, &wb));
+  wb.Clear();
+
+  EXPECT_EQ(Get(0, "b0"), "2");
+  EXPECT_EQ(Get(1, "b1"), "2");
+  EXPECT_EQ(Get(2, "b2"), "2");
+  EXPECT_EQ(Get(3, "b3"), "NOT_FOUND");
+
+  // 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(
+                {"default", "cf1", "cf2", "cf3"},
+                {options_allow, options_allow, options_disallow, options_allow})
+                .code(),
+            Status::Code::kInvalidArgument);
+
+  // Successfully opening with allow creates L0 files from the WAL
+  ReopenWithColumnFamilies({"default", "cf1", "cf2", "cf3"}, options_allow);
+
+  EXPECT_EQ(Get(0, "a0"), "1");
+  EXPECT_EQ(Get(1, "a1"), "1");
+  EXPECT_EQ(Get(2, "a2"), "1");
+  EXPECT_EQ(Get(3, "a3"), "NOT_FOUND");
+
+  // Now able to disallow on CF2 because no relevant WAL entries
+  ReopenWithColumnFamilies(
+      {"default", "cf1", "cf2", "cf3"},
+      {options_allow, options_allow, options_disallow, options_allow});
+
+  EXPECT_EQ(Get(0, "a0"), "1");
+  EXPECT_EQ(Get(1, "a1"), "1");
+  EXPECT_EQ(Get(2, "a2"), "1");
+  EXPECT_EQ(Get(3, "a3"), "NOT_FOUND");
+
+  // Now able to write to CF 3 but not CF 2
+  EXPECT_EQ(Put(0, "c0", "3").code(), Status::Code::kOk);
+  EXPECT_EQ(Put(1, "c1", "3").code(), Status::Code::kOk);
+  EXPECT_EQ(Put(2, "c2", "3").code(), Status::Code::kInvalidArgument);
+  EXPECT_EQ(Put(3, "c3", "3").code(), Status::Code::kOk);
+
+  EXPECT_EQ(Get(0, "c0"), "3");
+  EXPECT_EQ(Get(1, "c1"), "3");
+  EXPECT_EQ(Get(2, "c2"), "NOT_FOUND");
+  EXPECT_EQ(Get(3, "c3"), "3");
+
+  // disallow_memtable_writes not supported on default column family.
+  // (Would be complicated to make a WriteBatch aware of the setting in order
+  // to reject the write before entering the write path.)
+  Destroy(options_allow);
+  EXPECT_EQ(TryReopen(options_disallow).code(), Status::Code::kInvalidArgument);
+}
+
 // TODO: re-enable after we provide finer-grained control for WAL tracking to
 // meet the needs of different use cases, durability levels and recovery modes.
 TEST_F(DBBasicTest, DISABLED_ManualWalSync) {

db/db_block_cache_test.cc

@@ -305,7 +305,6 @@ class ReadOnlyCacheWrapper : public CacheWrapper {
 }  // anonymous namespace
 #endif  // SNAPPY
 
-
 // Make sure that when options.block_cache is set, after a new table is
 // created its index/filter blocks are added to block cache.
 TEST_F(DBBlockCacheTest, IndexAndFilterBlocksOfNewTableAddedToCache) {
@@ -836,8 +835,6 @@ TEST_F(DBBlockCacheTest, CacheCompressionDict) {
   }
   if (ZSTD_Supported()) {
     compression_types.push_back(kZSTD);
-  } else if (ZSTDNotFinal_Supported()) {
-    compression_types.push_back(kZSTDNotFinalCompression);
   }
   Random rnd(301);
   for (auto compression_type : compression_types) {
@@ -1758,8 +1755,8 @@ class CacheKeyTest : public testing::Test {
     tp_.db_id = std::to_string(db_id_);
     tp_.orig_file_number = file_number;
     bool is_stable;
-    std::string cur_session_id;       // ignored
-    uint64_t cur_file_number = 42;    // ignored
+    std::string cur_session_id;     // ignored
+    uint64_t cur_file_number = 42;  // ignored
     OffsetableCacheKey rv;
     BlockBasedTable::SetupBaseCacheKey(&tp_, cur_session_id, cur_file_number,
                                        &rv, &is_stable);

db/db_bloom_filter_test.cc

@@ -2244,6 +2244,54 @@ TEST_F(DBBloomFilterTest, MemtableWholeKeyBloomFilterMultiGet) {
 
   db_->ReleaseSnapshot(snapshot);
 }
+
+TEST_F(DBBloomFilterTest, TestMemtableBloomAndWBM) {
+  Options options = CurrentOptions();
+  options.arena_block_size = 4096;
+  options.write_buffer_size = 4000000;
+  std::shared_ptr<Cache> cache = NewLRUCache(LRUCacheOptions(
+      options.write_buffer_size * 3 /* capacity */, 1 /* num_shard_bits */,
+      false /* strict_capacity_limit */, 0.0 /* high_pri_pool_ratio */,
+      nullptr /* memory_allocator */, kDefaultToAdaptiveMutex,
+      kDontChargeCacheMetadata));
+
+  options.write_buffer_manager.reset(
+      new WriteBufferManager(options.write_buffer_size, cache));
+  Reopen(options);
+  ASSERT_OK(Put("foo", "bar"));
+
+  const auto kDummyEntrySize =
+      CacheReservationManagerImpl<CacheEntryRole::kMisc>::GetDummyEntrySize();
+
+  // Just the start of a memtable, no Bloom
+  ASSERT_GE(cache->GetUsage(), options.arena_block_size);
+  ASSERT_LE(cache->GetUsage(), kDummyEntrySize);
+
+  // Now enable memtable bloom filter
+  const double kRatio = 0.25;
+  options.memtable_prefix_bloom_size_ratio = kRatio;
+  options.memtable_whole_key_filtering = true;
+  Reopen(options);
+  ASSERT_OK(Put("foo2", "bar2"));
+
+  // Expecting a memtable Bloom of ratio times write_buffer_size, memory tracked
+  auto bloom_size = static_cast<size_t>(kRatio * options.write_buffer_size);
+  ASSERT_GE(cache->GetUsage(), bloom_size);
+  ASSERT_LE(cache->GetUsage(), bloom_size + 2U * kDummyEntrySize);
+
+  // Now get another memtable and test
+  // Pin this memtable with an iterator
+  std::unique_ptr<Iterator> iter{db_->NewIterator({})};
+  iter->Seek("foo2");
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("foo2", "bar2"));
+
+  // Expecting twice as much
+  ASSERT_GE(cache->GetUsage(), 2U * bloom_size);
+  ASSERT_LE(cache->GetUsage(), 2U * bloom_size + 2U * kDummyEntrySize);
+}
+
 namespace {
 std::pair<uint64_t, uint64_t> GetBloomStat(const Options& options, bool sst) {
   if (sst) {

db/db_compaction_test.cc

@@ -5856,6 +5856,26 @@ TEST_F(DBCompactionTest, CompactionStatsTest) {
   options.listeners.emplace_back(collector);
   DestroyAndReopen(options);
 
+  // Verify that the internal statistics for num_running_compactions and
+  // num_running_compaction_sorted_runs start and end at valid states
+  uint64_t num_running_compactions = 0;
+  ASSERT_TRUE(db_->GetIntProperty(DB::Properties::kNumRunningCompactions,
+                                  &num_running_compactions));
+  ASSERT_EQ(num_running_compactions, 0);
+  uint64_t num_running_compaction_sorted_runs = 0;
+  ASSERT_TRUE(
+      db_->GetIntProperty(DB::Properties::kNumRunningCompactionSortedRuns,
+                          &num_running_compaction_sorted_runs));
+  ASSERT_EQ(num_running_compaction_sorted_runs, 0);
+  // Check that the stat actually gets changed some time between the start and
+  // end of compaction
+  std::atomic<bool> sorted_runs_count_incremented = false;
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+      "CompactionMergingIterator::UpdateInternalStats",
+      [&](void*) { sorted_runs_count_incremented = true; });
+
+  ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
   for (int i = 0; i < 32; i++) {
     for (int j = 0; j < 5000; j++) {
       ASSERT_OK(Put(std::to_string(j), std::string(1, 'A')));
@@ -5869,6 +5889,19 @@ TEST_F(DBCompactionTest, CompactionStatsTest) {
   ColumnFamilyData* cfd = cfh->cfd();
 
   VerifyCompactionStats(*cfd, *collector);
+  // There should be no more running compactions, and thus no more sorted runs
+  // to process
+  ASSERT_TRUE(db_->GetIntProperty(DB::Properties::kNumRunningCompactions,
+                                  &num_running_compactions));
+  ASSERT_EQ(num_running_compactions, 0);
+  ASSERT_TRUE(
+      db_->GetIntProperty(DB::Properties::kNumRunningCompactionSortedRuns,
+                          &num_running_compaction_sorted_runs));
+  ASSERT_EQ(num_running_compaction_sorted_runs, 0);
+  ASSERT_TRUE(sorted_runs_count_incremented);
+
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
 }
 
 TEST_F(DBCompactionTest, SubcompactionEvent) {
@@ -6407,20 +6440,40 @@ TEST_F(DBCompactionTest, PersistRoundRobinCompactCursor) {
   }
 }
 
-TEST_P(RoundRobinSubcompactionsAgainstPressureToken, PressureTokenTest) {
+// FIXME: the test is flaky and failing the assertion
+// ASSERT_EQ(num_planned_subcompactions, kNumSubcompactions);
+// It's likely a test set up issue, fix if we are to use RoubdRobin compaction.
+TEST_P(RoundRobinSubcompactionsAgainstPressureToken,
+       DISABLED_PressureTokenTest) {
   const int kKeysPerBuffer = 100;
   const int kNumSubcompactions = 2;
   const int kFilesPerLevel = 50;
+  SyncPoint::GetInstance()->LoadDependency({
+      // SetBackgroundThreads() only starts the bg threads. Background
+      // threads may not be immediately available after SetBackgroundThreads
+      // returns. There are some initialization before they start waiting for
+      // new jobs, see ThreadPoolImpl::Impl::BGThread(). Here is a hacky way
+      // to wait until there are at least two background threads (thread id
+      // starts from 0) start waiting to accept jobs.
+      {"ThreadPoolImpl::BGThread::Start:th1", "WaitForThreadAvailable"},
+  });
+  SyncPoint::GetInstance()->EnableProcessing();
+  env_->SetBackgroundThreads(kNumSubcompactions, Env::LOW);
+  TEST_SYNC_POINT("WaitForThreadAvailable");
+  SyncPoint::GetInstance()->DisableProcessing();
   Options options = CurrentOptions();
   options.num_levels = 3;
   options.max_bytes_for_level_multiplier = 2;
   options.level0_file_num_compaction_trigger = 4;
   options.target_file_size_base = kKeysPerBuffer * 1024;
   options.compaction_pri = CompactionPri::kRoundRobin;
-  // Target size is chosen so that filling the level with `kFilesPerLevel` files
-  // will make it oversized by `kNumSubcompactions` files.
+  // Pick a small target level size so that round-robin compaction will pick
+  // more files to compact and trigger subcompactions. Actual number of
+  // subcompactions will be limited by number of bg threads available.
+  // Cannot be too small to cause compaction pressure. See
+  // GetPendingCompactionBytesForCompactionSpeedup().
   options.max_bytes_for_level_base =
-      (kFilesPerLevel - kNumSubcompactions) * kKeysPerBuffer * 1024;
+      (kFilesPerLevel - 10) * kKeysPerBuffer * 1024;
   options.disable_auto_compactions = true;
   // Setup `kNumSubcompactions` threads but limited subcompactions so
   // that RoundRobin requires extra compactions from reserved threads
@@ -6428,7 +6481,6 @@ TEST_P(RoundRobinSubcompactionsAgainstPressureToken, PressureTokenTest) {
   options.max_background_compactions = kNumSubcompactions;
   options.max_compaction_bytes = 100000000;
   DestroyAndReopen(options);
-  env_->SetBackgroundThreads(kNumSubcompactions, Env::LOW);
 
   Random rnd(301);
   for (int lvl = 2; lvl > 0; lvl--) {
@@ -6446,21 +6498,34 @@ TEST_P(RoundRobinSubcompactionsAgainstPressureToken, PressureTokenTest) {
     ASSERT_EQ(kFilesPerLevel, NumTableFilesAtLevel(lvl, 0));
   }
 
+  bool compaction_num_input_file_verified = false;
+  SyncPoint::GetInstance()->SetCallBack(
+      "LevelCompactionPicker::PickCompaction:Return", [&](void* arg) {
+        if (!compaction_num_input_file_verified) {
+          Compaction* compaction = static_cast<Compaction*>(arg);
+          // In Round-Robin, # of subcompactions needed is the # of input files
+          ASSERT_GT(compaction->num_input_files(0), 1);
+          compaction_num_input_file_verified = true;
+        }
+      });
+
   // This is a variable for making sure the following callback is called
   // and the assertions in it are indeed excuted.
   bool num_planned_subcompactions_verified = false;
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
       "CompactionJob::GenSubcompactionBoundaries:0", [&](void* arg) {
-        uint64_t num_planned_subcompactions = *(static_cast<uint64_t*>(arg));
-        if (grab_pressure_token_) {
-          // `kNumSubcompactions` files are selected for round-robin under auto
-          // compaction. The number of planned subcompaction is restricted by
-          // the limited number of max_background_compactions
-          ASSERT_EQ(num_planned_subcompactions, kNumSubcompactions);
-        } else {
-          ASSERT_EQ(num_planned_subcompactions, 1);
+        if (!num_planned_subcompactions_verified) {
+          uint64_t num_planned_subcompactions = *(static_cast<uint64_t*>(arg));
+          if (grab_pressure_token_) {
+            // `kNumSubcompactions` files are selected for round-robin under
+            // auto compaction. The number of planned subcompaction is
+            // restricted by the limited number of max_background_compactions
+            ASSERT_EQ(num_planned_subcompactions, kNumSubcompactions);
+          } else {
+            ASSERT_EQ(num_planned_subcompactions, 1);
+          }
+          num_planned_subcompactions_verified = true;
         }
-        num_planned_subcompactions_verified = true;
       });
 
   // The following 3 dependencies have to be added to ensure the auto
@@ -6911,7 +6976,8 @@ TEST_F(DBCompactionTest, ManualCompactionMax) {
   DestroyAndReopen(opts);
   generate_sst_func();
   uint64_t total_size = (l1_avg_size * 10) + (l2_avg_size * 100);
-  opts.max_compaction_bytes = total_size / num_split;
+  // Slightly inflate max_compaction_bytes since it's usually a strict bound.
+  opts.max_compaction_bytes = total_size / num_split + l2_avg_size * 10;
   opts.target_file_size_base = total_size / num_split;
   Reopen(opts);
   num_compactions.store(0);
@@ -6933,8 +6999,10 @@ TEST_F(DBCompactionTest, ManualCompactionMax) {
   DestroyAndReopen(opts);
   generate_sst_func();
   total_size = (l1_avg_size * 10) + (l2_avg_size * 100);
+  // Slightly inflate max_compaction_bytes since it's usually a strict bound.
   Status s = db_->SetOptions(
-      {{"max_compaction_bytes", std::to_string(total_size / num_split)},
+      {{"max_compaction_bytes",
+        std::to_string(total_size / num_split + 10 * l2_avg_size)},
        {"target_file_size_base", std::to_string(total_size / num_split)}});
   ASSERT_OK(s);
 

db/db_dynamic_level_test.cc

@@ -491,7 +491,6 @@ TEST_F(DBTestDynamicLevel, DISABLED_MigrateToDynamicLevelMaxBytesBase) {
 }
 }  // namespace ROCKSDB_NAMESPACE
 
-
 int main(int argc, char** argv) {
   ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
   ::testing::InitGoogleTest(&argc, argv);

db/db_encryption_test.cc

@@ -3,12 +3,13 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).
 //
+#include <iostream>
+#include <string>
+
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
 #include "rocksdb/perf_context.h"
 #include "test_util/sync_point.h"
-#include <iostream>
-#include <string>
 
 namespace ROCKSDB_NAMESPACE {
 
@@ -25,7 +26,6 @@ class DBEncryptionTest : public DBTestBase {
   }
 };
 
-
 TEST_F(DBEncryptionTest, CheckEncrypted) {
   ASSERT_OK(Put("foo567", "v1.fetdq"));
   ASSERT_OK(Put("bar123", "v2.dfgkjdfghsd"));
@@ -116,7 +116,6 @@ TEST_F(DBEncryptionTest, ReadEmptyFile) {
   ASSERT_TRUE(data.empty());
 }
 
-
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/db_filesnapshot.cc

@@ -4,7 +4,6 @@
 //  (found in the LICENSE.Apache file in the root directory).
 //
 
-
 #include <algorithm>
 #include <cstdint>
 #include <memory>
@@ -266,7 +265,7 @@ Status DBImpl::GetLiveFilesStorageInfo(
       continue;
     }
     VersionStorageInfo& vsi = *cfd->current()->storage_info();
-    auto& cf_paths = cfd->ioptions()->cf_paths;
+    auto& cf_paths = cfd->ioptions().cf_paths;
 
     auto GetDir = [&](size_t path_id) {
       // Matching TableFileName() behavior

db/db_impl/compacted_db_impl.cc

@@ -248,7 +248,7 @@ Status CompactedDBImpl::Init(const Options& options) {
 }
 
 Status CompactedDBImpl::Open(const Options& options, const std::string& dbname,
-                             DB** dbptr) {
+                             std::unique_ptr<DB>* dbptr) {
   *dbptr = nullptr;
 
   if (options.max_open_files != -1) {
@@ -267,7 +267,7 @@ Status CompactedDBImpl::Open(const Options& options, const std::string& dbname,
     ROCKS_LOG_INFO(db->immutable_db_options_.info_log,
                    "Opened the db as fully compacted mode");
     LogFlush(db->immutable_db_options_.info_log);
-    *dbptr = db.release();
+    *dbptr = std::move(db);
   }
   return s;
 }

db/db_impl/compacted_db_impl.h

@@ -22,7 +22,7 @@ class CompactedDBImpl : public DBImpl {
   ~CompactedDBImpl() override;
 
   static Status Open(const Options& options, const std::string& dbname,
-                     DB** dbptr);
+                     std::unique_ptr<DB>* dbptr);
 
   // Implementations of the DB interface
   using DB::Get;

db/db_impl/db_impl.cc

@@ -386,9 +386,8 @@ Status DBImpl::ResumeImpl(DBRecoverContext context) {
           static_cast_with_check<ColumnFamilyHandleImpl>(default_cf_handle_);
       assert(cfh);
       ColumnFamilyData* cfd = cfh->cfd();
-      const MutableCFOptions& cf_opts = *cfd->GetLatestMutableCFOptions();
-      s = versions_->LogAndApply(cfd, cf_opts, read_options, write_options,
-                                 &edit, &mutex_, directories_.GetDbDir());
+      s = versions_->LogAndApply(cfd, read_options, write_options, &edit,
+                                 &mutex_, directories_.GetDbDir());
       if (!s.ok()) {
         io_s = versions_->io_status();
         if (!io_s.ok()) {
@@ -418,10 +417,19 @@ Status DBImpl::ResumeImpl(DBRecoverContext context) {
     }
   }
 
+  JobContext job_context(0);
+  FindObsoleteFiles(&job_context, true);
+  mutex_.Unlock();
+  // If DB shutdown initiated here, it will wait for this ongoing recovery.
+  job_context.manifest_file_number = 1;
+  if (job_context.HaveSomethingToDelete()) {
+    PurgeObsoleteFiles(job_context);
+  }
+  job_context.Clean();
+  mutex_.Lock();
+
   if (s.ok()) {
-    // This will notify and unblock threads waiting for error recovery to
-    // finish. Those previouly waiting threads can now proceed, which may
-    // include closing the db.
+    // Will notify and unblock threads waiting for error recovery to finish.
     s = error_handler_.ClearBGError();
   } else {
     // NOTE: this is needed to pass ASSERT_STATUS_CHECKED
@@ -430,15 +438,6 @@ Status DBImpl::ResumeImpl(DBRecoverContext context) {
     error_handler_.GetRecoveryError().PermitUncheckedError();
   }
 
-  JobContext job_context(0);
-  FindObsoleteFiles(&job_context, true);
-  mutex_.Unlock();
-  job_context.manifest_file_number = 1;
-  if (job_context.HaveSomethingToDelete()) {
-    PurgeObsoleteFiles(job_context);
-  }
-  job_context.Clean();
-
   if (s.ok()) {
     ROCKS_LOG_INFO(immutable_db_options_.info_log, "Successfully resumed DB");
   } else {
@@ -446,7 +445,6 @@ Status DBImpl::ResumeImpl(DBRecoverContext context) {
                    s.ToString().c_str());
   }
 
-  mutex_.Lock();
   // Check for shutdown again before scheduling further compactions,
   // since we released and re-acquired the lock above
   if (shutdown_initiated_) {
@@ -509,6 +507,11 @@ void DBImpl::CancelAllBackgroundWork(bool wait) {
     s.PermitUncheckedError();  //**TODO: What to do on error?
   }
 
+  // Cancel awaiting remote compactions
+  if (immutable_db_options_.compaction_service) {
+    immutable_db_options_.compaction_service->CancelAwaitingJobs();
+  }
+
   shutting_down_.store(true, std::memory_order_release);
   bg_cv_.SignalAll();
   if (!wait) {
@@ -540,8 +543,8 @@ Status DBImpl::CloseHelper() {
   // continuing with the shutdown
   mutex_.Lock();
   shutdown_initiated_ = true;
-  error_handler_.CancelErrorRecovery();
-  while (error_handler_.IsRecoveryInProgress()) {
+  error_handler_.CancelErrorRecoveryForShutDown();
+  while (!error_handler_.ReadyForShutdown()) {
     bg_cv_.Wait();
   }
   mutex_.Unlock();
@@ -852,7 +855,7 @@ Status DBImpl::RegisterRecordSeqnoTimeWorker(const ReadOptions& read_options,
     InstrumentedMutexLock l(&mutex_);
 
     for (auto cfd : *versions_->GetColumnFamilySet()) {
-      auto& mopts = *cfd->GetLatestMutableCFOptions();
+      auto& mopts = cfd->GetLatestMutableCFOptions();
       // preserve time is the max of 2 options.
       uint64_t preserve_seconds =
           std::max(mopts.preserve_internal_time_seconds,
@@ -1167,7 +1170,7 @@ void DBImpl::DumpStats() {
       }
 
       auto* table_factory =
-          cfd->GetCurrentMutableCFOptions()->table_factory.get();
+          cfd->GetCurrentMutableCFOptions().table_factory.get();
       assert(table_factory != nullptr);
       // FIXME: need to a shared_ptr if/when block_cache is going to be mutable
       Cache* cache =
@@ -1286,28 +1289,64 @@ Status DBImpl::SetOptions(
   }
 
   InstrumentedMutexLock ol(&options_mutex_);
-  MutableCFOptions new_options;
+  MutableCFOptions new_options_copy;  // For logging outside of DB mutex
   Status s;
   Status persist_options_status;
   SuperVersionContext sv_context(/* create_superversion */ true);
   {
     auto db_options = GetDBOptions();
     InstrumentedMutexLock l(&mutex_);
-    s = cfd->SetOptions(db_options, options_map);
+    // Manifest writers + Version appenders like flush and compaction use
+    // LogAndApply, which releases DB mutex to wait for other manifest writers
+    // and for the manifest write. We need to append a Version for the options
+    // to take full effect (e.g. compaction scores), but we don't want to
+    // interleave with other callers of LogAndApply, which could at least
+    // temporarily roll back option changes. Thus, we use a special call to
+    // LogAndApply that allows us to
+    //
+    // (a) Apply the options update when we know we are the exclusive version
+    // appender + (fake) manifest writer, and
+    //
+    // (b) Append a new Version without manifest write nor DB mutex release
+    //
+    // Thus aren't releasing the DB mutex again until the end of this block,
+    // after installing the new SuperVersion.
+    auto pre_cb = [&]() -> Status {
+      Status cb_s = cfd->SetOptions(db_options, options_map);
+      if (cb_s.ok()) {
+        new_options_copy = cfd->GetLatestMutableCFOptions();
+      }
+      return cb_s;
+    };
+    VersionEdit dummy_edit;
+    dummy_edit.MarkNoManifestWriteDummy();
+    TEST_SYNC_POINT_CALLBACK("DBImpl::SetOptions:dummy_edit", &dummy_edit);
+    s = versions_->LogAndApply(
+        cfd, read_options, write_options, &dummy_edit, &mutex_,
+        directories_.GetDbDir(), false /*new_descriptor_log=*/,
+        nullptr /*new_opts*/, {} /*manifest_wcb*/, pre_cb);
+    if (!versions_->io_status().ok()) {
+      assert(!s.ok());
+      error_handler_.SetBGError(versions_->io_status(),
+                                BackgroundErrorReason::kManifestWrite);
+    }
+
     if (s.ok()) {
-      new_options = *cfd->GetLatestMutableCFOptions();
-      // Append new version to recompute compaction score.
-      VersionEdit dummy_edit;
-      s = versions_->LogAndApply(cfd, new_options, read_options, write_options,
-                                 &dummy_edit, &mutex_, directories_.GetDbDir());
       // Trigger possible flush/compactions. This has to be before we persist
       // options to file, otherwise there will be a deadlock with writer
       // thread.
-      InstallSuperVersionAndScheduleWork(cfd, &sv_context, new_options);
-
+      InstallSuperVersionAndScheduleWork(cfd, &sv_context);
       persist_options_status =
           WriteOptionsFile(write_options, true /*db_mutex_already_held*/);
       bg_cv_.SignalAll();
+
+#if __cplusplus >= 202002L
+      assert(new_options_copy == cfd->GetLatestMutableCFOptions());
+      assert(cfd->GetLatestMutableCFOptions() ==
+             cfd->GetCurrentMutableCFOptions());
+      assert(cfd->GetCurrentMutableCFOptions() ==
+             cfd->current()->GetMutableCFOptions());
+#endif
     }
   }
   sv_context.Clean();
@@ -1328,7 +1367,7 @@ Status DBImpl::SetOptions(
   if (s.ok()) {
     ROCKS_LOG_INFO(immutable_db_options_.info_log,
                    "[%s] SetOptions() succeeded", cfd->GetName().c_str());
-    new_options.Dump(immutable_db_options_.info_log.get());
+    new_options_copy.Dump(immutable_db_options_.info_log.get());
     if (!persist_options_status.ok()) {
       // NOTE: WriteOptionsFile already logs on failure
       s = persist_options_status;
@@ -2098,7 +2137,8 @@ InternalIterator* DBImpl::NewInternalIterator(
   // Collect iterator for mutable memtable
   auto mem_iter = super_version->mem->NewIterator(
       read_options, super_version->GetSeqnoToTimeMapping(), arena,
-      super_version->mutable_cf_options.prefix_extractor.get());
+      super_version->mutable_cf_options.prefix_extractor.get(),
+      /*for_flush=*/false);
   Status s;
   if (!read_options.ignore_range_deletions) {
     std::unique_ptr<TruncatedRangeDelIterator> mem_tombstone_iter;
@@ -2109,7 +2149,7 @@ InternalIterator* DBImpl::NewInternalIterator(
     } else {
       mem_tombstone_iter = std::make_unique<TruncatedRangeDelIterator>(
           std::unique_ptr<FragmentedRangeTombstoneIterator>(range_del_iter),
-          &cfd->ioptions()->internal_comparator, nullptr /* smallest */,
+          &cfd->ioptions().internal_comparator, nullptr /* smallest */,
           nullptr /* largest */);
     }
     merge_iter_builder.AddPointAndTombstoneIterator(
@@ -2558,6 +2598,8 @@ Status DBImpl::GetImpl(const ReadOptions& read_options, const Slice& key,
         // Return all merge operands for get_impl_options.key
         *get_impl_options.number_of_operands =
             static_cast<int>(merge_context.GetNumOperands());
+        // OK status is returned, some merge operand is found.
+        assert(*get_impl_options.number_of_operands > 0);
         if (*get_impl_options.number_of_operands >
             get_impl_options.get_merge_operands_options
                 ->expected_max_number_of_operands) {
@@ -3247,8 +3289,13 @@ Status DBImpl::MultiGetImpl(
       s = Status::Aborted();
       break;
     }
+
     // This could be a long-running operation
-    ROCKSDB_THREAD_YIELD_HOOK();
+    bool aborted = ROCKSDB_THREAD_YIELD_CHECK_ABORT();
+    if (aborted) {
+      s = Status::Aborted("Query abort.");
+      break;
+    }
   }
 
   // Post processing (decrement reference counts and record statistics)
@@ -3637,9 +3684,9 @@ Status DBImpl::CreateColumnFamilyImpl(const ReadOptions& read_options,
       write_thread_.EnterUnbatched(&w, &mutex_);
       // LogAndApply will both write the creation in MANIFEST and create
       // ColumnFamilyData object
-      s = versions_->LogAndApply(nullptr, MutableCFOptions(cf_options),
-                                 read_options, write_options, &edit, &mutex_,
-                                 directories_.GetDbDir(), false, &cf_options);
+      s = versions_->LogAndApply(nullptr, read_options, write_options, &edit,
+                                 &mutex_, directories_.GetDbDir(), false,
+                                 &cf_options);
       write_thread_.ExitUnbatched(&w);
     }
     if (s.ok()) {
@@ -3653,8 +3700,7 @@ Status DBImpl::CreateColumnFamilyImpl(const ReadOptions& read_options,
       auto* cfd =
           versions_->GetColumnFamilySet()->GetColumnFamily(column_family_name);
       assert(cfd != nullptr);
-      InstallSuperVersionAndScheduleWork(cfd, &sv_context,
-                                         *cfd->GetLatestMutableCFOptions());
+      InstallSuperVersionAndScheduleWork(cfd, &sv_context);
 
       if (!cfd->mem()->IsSnapshotSupported()) {
         is_snapshot_supported_ = false;
@@ -3748,13 +3794,12 @@ Status DBImpl::DropColumnFamilyImpl(ColumnFamilyHandle* column_family) {
       // we drop column family from a single write thread
       WriteThread::Writer w;
       write_thread_.EnterUnbatched(&w, &mutex_);
-      s = versions_->LogAndApply(cfd, *cfd->GetLatestMutableCFOptions(),
-                                 read_options, write_options, &edit, &mutex_,
-                                 directories_.GetDbDir());
+      s = versions_->LogAndApply(cfd, read_options, write_options, &edit,
+                                 &mutex_, directories_.GetDbDir());
       write_thread_.ExitUnbatched(&w);
     }
     if (s.ok()) {
-      auto& moptions = *cfd->GetLatestMutableCFOptions();
+      auto& moptions = cfd->GetLatestMutableCFOptions();
       max_total_in_memory_state_ -=
           moptions.write_buffer_size * moptions.max_write_buffer_number;
       used_preserve_preclude = moptions.preserve_internal_time_seconds > 0 ||
@@ -3802,7 +3847,6 @@ bool DBImpl::KeyMayExist(const ReadOptions& read_options,
                          ColumnFamilyHandle* column_family, const Slice& key,
                          std::string* value, std::string* timestamp,
                          bool* value_found) {
-  assert(value != nullptr);
   assert(read_options.io_activity == Env::IOActivity::kUnknown);
 
   if (value_found != nullptr) {
@@ -3819,7 +3863,9 @@ bool DBImpl::KeyMayExist(const ReadOptions& read_options,
   get_impl_options.value_found = value_found;
   get_impl_options.timestamp = timestamp;
   auto s = GetImpl(roptions, key, get_impl_options);
-  value->assign(pinnable_val.data(), pinnable_val.size());
+  if (value_found && *value_found && value) {
+    value->assign(pinnable_val.data(), pinnable_val.size());
+  }
 
   // If block_cache is enabled and the index block of the table didn't
   // not present in block_cache, the return value will be Status::Incomplete.
@@ -3884,7 +3930,7 @@ 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,
+        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);
@@ -3966,7 +4012,7 @@ ArenaWrappedDBIter* DBImpl::NewIteratorImpl(
   // 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,
+      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);
@@ -4101,7 +4147,7 @@ Status DBImpl::NewIterators(
                                       cf_sv_pair.super_version,
                                       /* allow_unprepared_value */ true);
       iterators->push_back(
-          NewDBIterator(env_, read_options, *cf_sv_pair.cfd->ioptions(),
+          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,
@@ -4339,7 +4385,7 @@ 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->ioptions().allow_ingest_behind) {
           cfd->current()->storage_info()->UpdateOldestSnapshot(
               oldest_snapshot, /*allow_ingest_behind=*/false);
           if (!cfd->current()
@@ -4360,7 +4406,7 @@ void DBImpl::ReleaseSnapshot(const Snapshot* s) {
       SequenceNumber new_bottommost_files_mark_threshold = kMaxSequenceNumber;
       for (auto* cfd : *versions_->GetColumnFamilySet()) {
         if (CfdListContains(cf_scheduled, cfd) ||
-            cfd->ioptions()->allow_ingest_behind) {
+            cfd->ioptions().allow_ingest_behind) {
           continue;
         }
         new_bottommost_files_mark_threshold = std::min(
@@ -4871,110 +4917,6 @@ Status DBImpl::GetUpdatesSince(
   return wal_manager_.GetUpdatesSince(seq, iter, read_options, versions_.get());
 }
 
-Status DBImpl::DeleteFile(std::string name) {
-  // TODO: plumb Env::IOActivity, Env::IOPriority
-  const ReadOptions read_options;
-  const WriteOptions write_options;
-
-  uint64_t number;
-  FileType type;
-  WalFileType log_type;
-  if (!ParseFileName(name, &number, &type, &log_type) ||
-      (type != kTableFile && type != kWalFile)) {
-    ROCKS_LOG_ERROR(immutable_db_options_.info_log, "DeleteFile %s failed.\n",
-                    name.c_str());
-    return Status::InvalidArgument("Invalid file name");
-  }
-
-  if (type == kWalFile) {
-    // Only allow deleting archived log files
-    if (log_type != kArchivedLogFile) {
-      ROCKS_LOG_ERROR(immutable_db_options_.info_log,
-                      "DeleteFile %s failed - not archived log.\n",
-                      name.c_str());
-      return Status::NotSupported("Delete only supported for archived logs");
-    }
-    Status status = wal_manager_.DeleteFile(name, number);
-    if (!status.ok()) {
-      ROCKS_LOG_ERROR(immutable_db_options_.info_log,
-                      "DeleteFile %s failed -- %s.\n", name.c_str(),
-                      status.ToString().c_str());
-    }
-    return status;
-  }
-
-  Status status;
-  int level;
-  FileMetaData* metadata;
-  ColumnFamilyData* cfd;
-  VersionEdit edit;
-  JobContext job_context(next_job_id_.fetch_add(1), true);
-  {
-    InstrumentedMutexLock l(&mutex_);
-    status = versions_->GetMetadataForFile(number, &level, &metadata, &cfd);
-    if (!status.ok()) {
-      ROCKS_LOG_WARN(immutable_db_options_.info_log,
-                     "DeleteFile %s failed. File not found\n", name.c_str());
-      job_context.Clean();
-      return Status::InvalidArgument("File not found");
-    }
-    assert(level < cfd->NumberLevels());
-
-    // If the file is being compacted no need to delete.
-    if (metadata->being_compacted) {
-      ROCKS_LOG_INFO(immutable_db_options_.info_log,
-                     "DeleteFile %s Skipped. File about to be compacted\n",
-                     name.c_str());
-      job_context.Clean();
-      return Status::OK();
-    }
-
-    // Only the files in the last level can be deleted externally.
-    // This is to make sure that any deletion tombstones are not
-    // lost. Check that the level passed is the last level.
-    auto* vstoreage = cfd->current()->storage_info();
-    for (int i = level + 1; i < cfd->NumberLevels(); i++) {
-      if (vstoreage->NumLevelFiles(i) != 0) {
-        ROCKS_LOG_WARN(immutable_db_options_.info_log,
-                       "DeleteFile %s FAILED. File not in last level\n",
-                       name.c_str());
-        job_context.Clean();
-        return Status::InvalidArgument("File not in last level");
-      }
-    }
-    // if level == 0, it has to be the oldest file
-    if (level == 0 &&
-        vstoreage->LevelFiles(0).back()->fd.GetNumber() != number) {
-      ROCKS_LOG_WARN(immutable_db_options_.info_log,
-                     "DeleteFile %s failed ---"
-                     " target file in level 0 must be the oldest.",
-                     name.c_str());
-      job_context.Clean();
-      return Status::InvalidArgument("File in level 0, but not oldest");
-    }
-    edit.SetColumnFamily(cfd->GetID());
-    edit.DeleteFile(level, number);
-    status = versions_->LogAndApply(cfd, *cfd->GetLatestMutableCFOptions(),
-                                    read_options, write_options, &edit, &mutex_,
-                                    directories_.GetDbDir());
-    if (status.ok()) {
-      InstallSuperVersionAndScheduleWork(
-          cfd, job_context.superversion_contexts.data(),
-          *cfd->GetLatestMutableCFOptions());
-    }
-    FindObsoleteFiles(&job_context, false);
-  }  // lock released here
-
-  LogFlush(immutable_db_options_.info_log);
-  // remove files outside the db-lock
-  if (job_context.HaveSomethingToDelete()) {
-    // Call PurgeObsoleteFiles() without holding mutex.
-    PurgeObsoleteFiles(job_context);
-  }
-  job_context.Clean();
-  return status;
-}
-
 Status DBImpl::DeleteFilesInRanges(ColumnFamilyHandle* column_family,
                                    const RangePtr* ranges, size_t n,
                                    bool include_end) {
@@ -5059,21 +5001,19 @@ Status DBImpl::DeleteFilesInRanges(ColumnFamilyHandle* column_family,
       }
     }
     if (!deleted_files.empty()) {
-      vstorage->ComputeCompactionScore(*cfd->ioptions(),
-                                       *cfd->GetLatestMutableCFOptions());
+      vstorage->ComputeCompactionScore(cfd->ioptions(),
+                                       cfd->GetLatestMutableCFOptions());
     }
     if (edit.GetDeletedFiles().empty()) {
       job_context.Clean();
       return status;
     }
     input_version->Ref();
-    status = versions_->LogAndApply(cfd, *cfd->GetLatestMutableCFOptions(),
-                                    read_options, write_options, &edit, &mutex_,
-                                    directories_.GetDbDir());
+    status = versions_->LogAndApply(cfd, read_options, write_options, &edit,
+                                    &mutex_, directories_.GetDbDir());
     if (status.ok()) {
       InstallSuperVersionAndScheduleWork(
-          cfd, job_context.superversion_contexts.data(),
-          *cfd->GetLatestMutableCFOptions());
+          cfd, job_context.superversion_contexts.data());
     }
     for (auto* deleted_file : deleted_files) {
       deleted_file->being_compacted = false;
@@ -5672,7 +5612,7 @@ Status DBImpl::RenameTempFileToOptionsFile(const std::string& file_name,
 void DBImpl::NewThreadStatusCfInfo(ColumnFamilyData* cfd) const {
   if (immutable_db_options_.enable_thread_tracking) {
     ThreadStatusUtil::NewColumnFamilyInfo(this, cfd, cfd->GetName(),
-                                          cfd->ioptions()->env);
+                                          cfd->ioptions().env);
   }
 }
 
@@ -5890,6 +5830,7 @@ Status DBImpl::IngestExternalFile(
 
 Status DBImpl::IngestExternalFiles(
     const std::vector<IngestExternalFileArg>& args) {
+  PERF_TIMER_GUARD(file_ingestion_nanos);
   // TODO: plumb Env::IOActivity, Env::IOPriority
   const WriteOptions write_options;
 
@@ -5912,8 +5853,8 @@ Status DBImpl::IngestExternalFiles(
   const size_t num_cfs = args.size();
   for (size_t i = 0; i != num_cfs; ++i) {
     if (args[i].external_files.empty()) {
-      char err_msg[128] = {0};
-      snprintf(err_msg, 128, "external_files[%zu] is empty", i);
+      std::string err_msg =
+          "external_files[" + std::to_string(i) + "] is empty";
       return Status::InvalidArgument(err_msg);
     }
     if (i && args[i].options.fill_cache != args[i - 1].options.fill_cache) {
@@ -6034,6 +5975,7 @@ Status DBImpl::IngestExternalFiles(
     if (two_write_queues_) {
       nonmem_write_thread_.EnterUnbatched(&nonmem_w, &mutex_);
     }
+    PERF_TIMER_GUARD(file_ingestion_blocking_live_writes_nanos);
 
     // When unordered_write is enabled, the keys are writing to memtable in an
     // unordered way. If the ingestion job checks memtable key range before the
@@ -6044,6 +5986,7 @@ Status DBImpl::IngestExternalFiles(
 
     num_running_ingest_file_ += static_cast<int>(num_cfs);
     TEST_SYNC_POINT("DBImpl::IngestExternalFile:AfterIncIngestFileCounter");
+    TEST_SYNC_POINT("DBImpl::IngestExternalFile:AfterIncIngestFileCounter:2");
 
     bool at_least_one_cf_need_flush = false;
     std::vector<bool> need_flush(num_cfs, false);
@@ -6114,14 +6057,12 @@ Status DBImpl::IngestExternalFiles(
       ReadOptions read_options;
       read_options.fill_cache = args[0].options.fill_cache;
       autovector<ColumnFamilyData*> cfds_to_commit;
-      autovector<const MutableCFOptions*> mutable_cf_options_list;
       autovector<autovector<VersionEdit*>> edit_lists;
       uint32_t num_entries = 0;
       for (size_t i = 0; i != num_cfs; ++i) {
         auto* cfd = ingestion_jobs[i].GetColumnFamilyData();
         assert(!cfd->IsDropped());
         cfds_to_commit.push_back(cfd);
-        mutable_cf_options_list.push_back(cfd->GetLatestMutableCFOptions());
         autovector<VersionEdit*> edit_list;
         edit_list.push_back(ingestion_jobs[i].edit());
         edit_lists.push_back(edit_list);
@@ -6136,10 +6077,10 @@ Status DBImpl::IngestExternalFiles(
         }
         assert(0 == num_entries);
       }
-      status = versions_->LogAndApply(
-          cfds_to_commit, mutable_cf_options_list, read_options, write_options,
+      status =
+          versions_->LogAndApply(cfds_to_commit, read_options, write_options,
 
-          edit_lists, &mutex_, directories_.GetDbDir());
+                                 edit_lists, &mutex_, directories_.GetDbDir());
       // It is safe to update VersionSet last seqno here after LogAndApply since
       // LogAndApply persists last sequence number from VersionEdits,
       // which are from file's largest seqno and not from VersionSet.
@@ -6171,8 +6112,7 @@ Status DBImpl::IngestExternalFiles(
       for (size_t i = 0; i != num_cfs; ++i) {
         auto* cfd = ingestion_jobs[i].GetColumnFamilyData();
         assert(!cfd->IsDropped());
-        InstallSuperVersionAndScheduleWork(cfd, &sv_ctxs[i],
-                                           *cfd->GetLatestMutableCFOptions());
+        InstallSuperVersionAndScheduleWork(cfd, &sv_ctxs[i]);
 #ifndef NDEBUG
         if (0 == i && num_cfs > 1) {
           TEST_SYNC_POINT("DBImpl::IngestExternalFiles:InstallSVForFirstCF:0");
@@ -6196,6 +6136,7 @@ Status DBImpl::IngestExternalFiles(
       nonmem_write_thread_.ExitUnbatched(&nonmem_w);
     }
     write_thread_.ExitUnbatched(&w);
+    PERF_TIMER_STOP(file_ingestion_blocking_live_writes_nanos);
 
     if (status.ok()) {
       for (auto& job : ingestion_jobs) {
@@ -6290,12 +6231,11 @@ Status DBImpl::CreateColumnFamilyWithImport(
       // and this will overwrite the external file. To protect the external
       // file, we have to make sure the file number will never being reused.
       next_file_number = versions_->FetchAddFileNumber(total_file_num);
-      auto cf_options = cfd->GetLatestMutableCFOptions();
       status =
-          versions_->LogAndApply(cfd, *cf_options, read_options, write_options,
-                                 &dummy_edit, &mutex_, directories_.GetDbDir());
+          versions_->LogAndApply(cfd, read_options, write_options, &dummy_edit,
+                                 &mutex_, directories_.GetDbDir());
       if (status.ok()) {
-        InstallSuperVersionAndScheduleWork(cfd, &dummy_sv_ctx, *cf_options);
+        InstallSuperVersionAndScheduleWork(cfd, &dummy_sv_ctx);
       }
     }
   }
@@ -6328,12 +6268,11 @@ Status DBImpl::CreateColumnFamilyWithImport(
 
       // Install job edit [Mutex will be unlocked here]
       if (status.ok()) {
-        auto cf_options = cfd->GetLatestMutableCFOptions();
-        status = versions_->LogAndApply(cfd, *cf_options, read_options,
-                                        write_options, import_job.edit(),
-                                        &mutex_, directories_.GetDbDir());
+        status = versions_->LogAndApply(cfd, read_options, write_options,
+                                        import_job.edit(), &mutex_,
+                                        directories_.GetDbDir());
         if (status.ok()) {
-          InstallSuperVersionAndScheduleWork(cfd, &sv_context, *cf_options);
+          InstallSuperVersionAndScheduleWork(cfd, &sv_context);
         }
       }
 
@@ -6534,7 +6473,7 @@ Status DBImpl::VerifyChecksumInternal(const ReadOptions& read_options,
         const auto& fd = fd_with_krange.fd;
         const FileMetaData* fmeta = fd_with_krange.file_metadata;
         assert(fmeta);
-        std::string fname = TableFileName(cfd->ioptions()->cf_paths,
+        std::string fname = TableFileName(cfd->ioptions().cf_paths,
                                           fd.GetNumber(), fd.GetPathId());
         if (use_file_checksum) {
           s = VerifyFullFileChecksum(fmeta->file_checksum,
@@ -6558,7 +6497,7 @@ Status DBImpl::VerifyChecksumInternal(const ReadOptions& read_options,
         const uint64_t blob_file_number = meta->GetBlobFileNumber();
 
         const std::string blob_file_name = BlobFileName(
-            cfd->ioptions()->cf_paths.front().path, blob_file_number);
+            cfd->ioptions().cf_paths.front().path, blob_file_number);
         s = VerifyFullFileChecksum(meta->GetChecksumValue(),
                                    meta->GetChecksumMethod(), blob_file_name,
                                    read_options);
@@ -6751,16 +6690,15 @@ Status DBImpl::ReserveFileNumbersBeforeIngestion(
   pending_output_elem.reset(new std::list<uint64_t>::iterator(
       CaptureCurrentFileNumberInPendingOutputs()));
   *next_file_number = versions_->FetchAddFileNumber(static_cast<uint64_t>(num));
-  auto cf_options = cfd->GetLatestMutableCFOptions();
   VersionEdit dummy_edit;
   // If crash happen after a hard link established, Recover function may
   // reuse the file number that has already assigned to the internal file,
   // and this will overwrite the external file. To protect the external
   // file, we have to make sure the file number will never being reused.
-  s = versions_->LogAndApply(cfd, *cf_options, read_options, write_options,
-                             &dummy_edit, &mutex_, directories_.GetDbDir());
+  s = versions_->LogAndApply(cfd, read_options, write_options, &dummy_edit,
+                             &mutex_, directories_.GetDbDir());
   if (s.ok()) {
-    InstallSuperVersionAndScheduleWork(cfd, &dummy_sv_ctx, *cf_options);
+    InstallSuperVersionAndScheduleWork(cfd, &dummy_sv_ctx);
   }
   dummy_sv_ctx.Clean();
   return s;
@@ -6799,17 +6737,24 @@ void DBImpl::RecordSeqnoToTimeMapping(uint64_t populate_historical_seconds) {
   // for SeqnoToTimeMapping. We don't know how long it has been since the last
   // sequence number was written, so we at least have a one-sided bound by
   // sampling in this order.
-  SequenceNumber seqno = GetLatestSequenceNumber();
-  int64_t unix_time_signed = 0;
-  immutable_db_options_.clock->GetCurrentTime(&unix_time_signed)
-      .PermitUncheckedError();  // Ignore error
-  uint64_t unix_time = static_cast<uint64_t>(unix_time_signed);
-
+  // ALSO, to avoid out-of-order mappings, we need to get the seqno and times
+  // while holding the DB mutex. (This is really to make testing happy because
+  // it's fine to throw out extra close-but-not-quite-consistent mappings in
+  // production.)
   std::vector<SuperVersionContext> sv_contexts;
-  if (populate_historical_seconds > 0) {
-    bool success = true;
-    {
-      InstrumentedMutexLock l(&mutex_);
+  bool success = true;
+  SequenceNumber seqno;
+  uint64_t unix_time;
+  {
+    InstrumentedMutexLock l(&mutex_);
+
+    seqno = GetLatestSequenceNumber();
+    int64_t unix_time_signed = 0;
+    immutable_db_options_.clock->GetCurrentTime(&unix_time_signed)
+        .PermitUncheckedError();  // Ignore error
+    unix_time = static_cast<uint64_t>(unix_time_signed);
+
+    if (populate_historical_seconds > 0) {
       if (seqno > 1 && unix_time > populate_historical_seconds) {
         // seqno=0 is reserved
         SequenceNumber from_seqno = 1;
@@ -6823,7 +6768,20 @@ void DBImpl::RecordSeqnoToTimeMapping(uint64_t populate_historical_seconds) {
         assert(unix_time > populate_historical_seconds);
         success = false;
       }
+    } else {
+      // FIXME: assert(seqno > 0);
+      // Always successful assuming seqno never go backwards
+      seqno_to_time_mapping_.Append(seqno, unix_time);
+      InstallSeqnoToTimeMappingInSV(&sv_contexts);
     }
+  }
+
+  // clean up & report outside db mutex
+  for (SuperVersionContext& sv_context : sv_contexts) {
+    sv_context.Clean();
+  }
+
+  if (populate_historical_seconds > 0) {
     if (success) {
       ROCKS_LOG_INFO(
           immutable_db_options_.info_log,
@@ -6838,16 +6796,7 @@ void DBImpl::RecordSeqnoToTimeMapping(uint64_t populate_historical_seconds) {
           seqno, unix_time - populate_historical_seconds, unix_time);
     }
   } else {
-    InstrumentedMutexLock l(&mutex_);
-    // FIXME: assert(seqno > 0);
-    // Always successful assuming seqno never go backwards
-    seqno_to_time_mapping_.Append(seqno, unix_time);
-    InstallSeqnoToTimeMappingInSV(&sv_contexts);
-  }
-
-  // clean up outside db mutex
-  for (SuperVersionContext& sv_context : sv_contexts) {
-    sv_context.Clean();
+    assert(success);
   }
 }
 
@@ -6920,7 +6869,7 @@ void DBImpl::InstallSeqnoToTimeMappingInSV(
     sv_contexts->emplace_back(/*create_superversion=*/true);
     sv_contexts->back().new_seqno_to_time_mapping = new_seqno_to_time_mapping;
     cfd->InstallSuperVersion(&sv_contexts->back(),
-                             *(cfd->GetLatestMutableCFOptions()));
+                             cfd->GetLatestMutableCFOptions());
   }
   bg_cv_.SignalAll();
 }

db/db_impl/db_impl.h

@@ -48,6 +48,7 @@
 #include "db/write_controller.h"
 #include "db/write_thread.h"
 #include "logging/event_logger.h"
+#include "memtable/wbwi_memtable.h"
 #include "monitoring/instrumented_mutex.h"
 #include "options/db_options.h"
 #include "port/port.h"
@@ -60,6 +61,7 @@
 #include "rocksdb/transaction_log.h"
 #include "rocksdb/user_write_callback.h"
 #include "rocksdb/utilities/replayer.h"
+#include "rocksdb/utilities/write_batch_with_index.h"
 #include "rocksdb/write_buffer_manager.h"
 #include "table/merging_iterator.h"
 #include "util/autovector.h"
@@ -160,6 +162,23 @@ class Directories {
   std::unique_ptr<FSDirectory> wal_dir_;
 };
 
+struct DBOpenLogRecordReadReporter : public log::Reader::Reporter {
+  Env* env;
+  Logger* info_log;
+  const char* fname;
+  Status* status;  // nullptr if immutable_db_options_.paranoid_checks==false
+  bool* old_log_record;
+  void Corruption(size_t bytes, const Status& s,
+                  uint64_t log_number = kMaxSequenceNumber) override;
+
+  void OldLogRecord(size_t bytes) override;
+
+  uint64_t GetCorruptedLogNumber() const { return corrupted_log_number_; }
+
+ private:
+  uint64_t corrupted_log_number_ = kMaxSequenceNumber;
+};
+
 // While DB is the public interface of RocksDB, and DBImpl is the actual
 // class implementing it. It's the entrance of the core RocksdB engine.
 // All other DB implementations, e.g. TransactionDB, BlobDB, etc, wrap a
@@ -522,7 +541,6 @@ class DBImpl : public DB {
       SequenceNumber seq_number, std::unique_ptr<TransactionLogIterator>* iter,
       const TransactionLogIterator::ReadOptions& read_options =
           TransactionLogIterator::ReadOptions()) override;
-  Status DeleteFile(std::string name) override;
   Status DeleteFilesInRanges(ColumnFamilyHandle* column_family,
                              const RangePtr* ranges, size_t n,
                              bool include_end = true);
@@ -937,13 +955,6 @@ class DBImpl : public DB {
     return num_running_flushes_;
   }
 
-  // Returns the number of currently running compactions.
-  // REQUIREMENT: mutex_ must be held when calling this function.
-  int num_running_compactions() {
-    mutex_.AssertHeld();
-    return num_running_compactions_;
-  }
-
   const WriteController& write_controller() { return write_controller_; }
 
   // hollow transactions shell used for recovery.
@@ -1079,9 +1090,10 @@ class DBImpl : public DB {
   // This is to be used only by internal rocksdb classes.
   static Status Open(const DBOptions& db_options, const std::string& name,
                      const std::vector<ColumnFamilyDescriptor>& column_families,
-                     std::vector<ColumnFamilyHandle*>* handles, DB** dbptr,
-                     const bool seq_per_batch, const bool batch_per_txn,
-                     const bool is_retry, bool* can_retry);
+                     std::vector<ColumnFamilyHandle*>* handles,
+                     std::unique_ptr<DB>* dbptr, const bool seq_per_batch,
+                     const bool batch_per_txn, const bool is_retry,
+                     bool* can_retry);
 
   static IOStatus CreateAndNewDirectory(
       FileSystem* fs, const std::string& dirname,
@@ -1438,7 +1450,6 @@ class DBImpl : public DB {
         uint32_t size = static_cast<uint32_t>(map_.size());
         map_.emplace(cfd->GetID(), size);
         cfds_.emplace_back(cfd);
-        mutable_cf_opts_.emplace_back(cfd->GetLatestMutableCFOptions());
         edit_lists_.emplace_back(autovector<VersionEdit*>());
       }
       uint32_t i = map_[cfd->GetID()];
@@ -1447,7 +1458,6 @@ class DBImpl : public DB {
 
     std::unordered_map<uint32_t, uint32_t> map_;  // cf_id to index;
     autovector<ColumnFamilyData*> cfds_;
-    autovector<const MutableCFOptions*> mutable_cf_opts_;
     autovector<autovector<VersionEdit*>> edit_lists_;
     // All existing data files (SST files and Blob files) found during DB::Open.
     std::vector<std::string> existing_data_files_;
@@ -1507,6 +1517,23 @@ class DBImpl : public DB {
 
   void EraseThreadStatusDbInfo() const;
 
+  // For CFs that has updates in `wbwi`, their memtable will be switched,
+  // and `wbwi` will be added as the latest immutable memtable.
+  //
+  // REQUIRES: this thread is currently at the front of the main writer queue.
+  // @param prep_log refers to the WAL that contains prepare record
+  // for the transaction based on wbwi.
+  // @param assigned_seqno Sequence numbers for the ingested memtable.
+  // @param last_seqno the value of versions_->LastSequence() after the write
+  // 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);
+
   // If disable_memtable is set the application logic must guarantee that the
   // batch will still be skipped from memtable during the recovery. An excption
   // to this is seq_per_batch_ mode, in which since each batch already takes one
@@ -1522,6 +1549,16 @@ class DBImpl : public DB {
   // batch_cnt is expected to be non-zero in seq_per_batch mode and
   // indicates the number of sub-patches. A sub-patch is a subset of the write
   // batch that does not have duplicate keys.
+  // `callback` is called before WAL write.
+  // See more in comment above WriteCallback::Callback().
+  // pre_release_callback is called after WAL write and before memtable write.
+  // See more in comment above PreReleaseCallback::Callback().
+  // post_memtable_callback is called after memtable write but before publishing
+  // the sequence number to readers.
+  //
+  // The main write queue. This is the only write queue that updates
+  // LastSequence. When using one write queue, the same sequence also indicates
+  // the last published sequence.
   Status WriteImpl(const WriteOptions& options, WriteBatch* updates,
                    WriteCallback* callback = nullptr,
                    UserWriteCallback* user_write_cb = nullptr,
@@ -1529,7 +1566,9 @@ class DBImpl : public DB {
                    bool disable_memtable = false, uint64_t* seq_used = nullptr,
                    size_t batch_cnt = 0,
                    PreReleaseCallback* pre_release_callback = nullptr,
-                   PostMemTableCallback* post_memtable_callback = nullptr);
+                   PostMemTableCallback* post_memtable_callback = nullptr,
+                   std::shared_ptr<WriteBatchWithIndex> wbwi = nullptr,
+                   uint64_t min_prep_log = 0);
 
   Status PipelinedWriteImpl(const WriteOptions& options, WriteBatch* updates,
                             WriteCallback* callback = nullptr,
@@ -1879,8 +1918,8 @@ class DBImpl : public DB {
     const InternalKey* begin = nullptr;  // nullptr means beginning of key range
     const InternalKey* end = nullptr;    // nullptr means end of key range
     InternalKey* manual_end = nullptr;   // how far we are compacting
-    InternalKey tmp_storage;      // Used to keep track of compaction progress
-    InternalKey tmp_storage1;     // Used to keep track of compaction progress
+    InternalKey tmp_storage;   // Used to keep track of compaction progress
+    InternalKey tmp_storage1;  // Used to keep track of compaction progress
 
     // When the user provides a canceled pointer in CompactRangeOptions, the
     // above varaibe is the reference of the user-provided
@@ -2012,6 +2051,94 @@ class DBImpl : public DB {
                          bool is_retry, bool* corrupted_log_found,
                          RecoveryContext* recovery_ctx);
 
+  void SetupLogFilesRecovery(
+      const std::vector<uint64_t>& wal_numbers,
+      std::unordered_map<int, VersionEdit>* version_edits, int* job_id,
+      uint64_t* min_wal_number);
+
+  Status ProcessLogFiles(const std::vector<uint64_t>& wal_numbers,
+                         bool read_only, bool is_retry, uint64_t min_wal_number,
+                         int job_id, SequenceNumber* next_sequence,
+                         std::unordered_map<int, VersionEdit>* version_edits,
+                         bool* corrupted_wal_found,
+                         RecoveryContext* recovery_ctx);
+
+  Status ProcessLogFile(
+      uint64_t wal_number, uint64_t min_wal_number, bool is_retry,
+      bool read_only, int job_id, SequenceNumber* next_sequence,
+      bool* stop_replay_for_corruption, bool* stop_replay_by_wal_filter,
+      uint64_t* corrupted_wal_number, bool* corrupted_wal_found,
+      std::unordered_map<int, VersionEdit>* version_edits, bool* flushed,
+      PredecessorWALInfo& predecessor_wal_info);
+
+  void SetupLogFileProcessing(uint64_t wal_number);
+
+  Status InitializeLogReader(uint64_t wal_number, bool is_retry,
+                             std::string& fname,
+
+                             bool stop_replay_for_corruption,
+                             uint64_t min_wal_number,
+                             const PredecessorWALInfo& predecessor_wal_info,
+                             bool* const old_log_record,
+                             Status* const reporter_status,
+                             DBOpenLogRecordReadReporter* reporter,
+                             std::unique_ptr<log::Reader>& reader);
+  Status ProcessLogRecord(
+      Slice record, const std::unique_ptr<log::Reader>& reader,
+      const UnorderedMap<uint32_t, size_t>& running_ts_sz, uint64_t wal_number,
+      const std::string& fname, bool read_only, int job_id,
+      const std::function<void()>& logFileDropped,
+      DBOpenLogRecordReadReporter* reporter, uint64_t* record_checksum,
+      SequenceNumber* last_seqno_observed, SequenceNumber* next_sequence,
+      bool* stop_replay_for_corruption, Status* status,
+      bool* stop_replay_by_wal_filter,
+      std::unordered_map<int, VersionEdit>* version_edits, bool* flushed);
+
+  Status InitializeWriteBatchForLogRecord(
+      Slice record, const std::unique_ptr<log::Reader>& reader,
+      const UnorderedMap<uint32_t, size_t>& running_ts_sz, WriteBatch* batch,
+      std::unique_ptr<WriteBatch>& new_batch, WriteBatch*& batch_to_use,
+      uint64_t* record_checksum);
+
+  void MaybeReviseStopReplayForCorruption(
+      SequenceNumber sequence, SequenceNumber const* const next_sequence,
+      bool* stop_replay_for_corruption);
+
+  Status InsertLogRecordToMemtable(WriteBatch* batch_to_use,
+                                   uint64_t wal_number,
+                                   SequenceNumber* next_sequence,
+                                   bool* has_valid_writes);
+
+  Status MaybeWriteLevel0TableForRecovery(
+      bool has_valid_writes, bool read_only, uint64_t wal_number, int job_id,
+      SequenceNumber const* const next_sequence,
+      std::unordered_map<int, VersionEdit>* version_edits, bool* flushed);
+
+  Status HandleNonOkStatusOrOldLogRecord(
+      uint64_t wal_number, SequenceNumber const* const next_sequence,
+      Status status, const DBOpenLogRecordReadReporter& reporter,
+      bool* old_log_record, bool* stop_replay_for_corruption,
+      uint64_t* corrupted_wal_number, bool* corrupted_wal_found);
+
+  Status UpdatePredecessorWALInfo(uint64_t wal_number,
+                                  const SequenceNumber last_seqno_observed,
+                                  const std::string& fname,
+                                  PredecessorWALInfo& predecessor_wal_info);
+
+  void FinishLogFileProcessing(const Status& status,
+                               const SequenceNumber* next_sequence);
+
+  // Return `Status::Corruption()` when `stop_replay_for_corruption == true` and
+  // exits inconsistency between SST and WAL data
+  Status MaybeHandleStopReplayForCorruptionForInconsistency(
+      bool stop_replay_for_corruption, uint64_t corrupted_wal_number);
+
+  Status MaybeFlushFinalMemtableOrRestoreActiveLogFiles(
+      const std::vector<uint64_t>& wal_numbers, bool read_only, int job_id,
+      bool flushed, std::unordered_map<int, VersionEdit>* version_edits,
+      RecoveryContext* recovery_ctx);
+
+  void FinishLogFilesRecovery(int job_id, const Status& status);
   // The following two methods are used to flush a memtable to
   // storage. The first one is used at database RecoveryTime (when the
   // database is opened) and is heavyweight because it holds the mutex
@@ -2053,7 +2180,19 @@ class DBImpl : public DB {
 
   // Switches the current live memtable to immutable/read-only memtable.
   // A new WAL is created if the current WAL is not empty.
-  Status SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context);
+  // If `new_imm` is not nullptr, it will be added as the newest immutable
+  // memtable, if and only if OK status is returned.
+  // `last_seqno` needs to be provided if `new_imm` is not nullptr. It is
+  // the value of versions_->LastSequence() after the write that ingests new_imm
+  // is done.
+  //
+  // REQUIRES: mutex_ is held
+  // REQUIRES: this thread is currently at the front of the writer queue
+  // REQUIRES: this thread is currently at the front of the 2nd writer queue if
+  // two_write_queues_ is true (This is to simplify the reasoning.)
+  Status SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context,
+                        ReadOnlyMemTable* new_imm = nullptr,
+                        SequenceNumber last_seqno = 0);
 
   // Select and output column families qualified for atomic flush in
   // `selected_cfds`. If `provided_candidate_cfds` is non-empty, it will be used
@@ -2179,7 +2318,8 @@ class DBImpl : public DB {
                       const WriteOptions& write_options,
                       log::Writer* log_writer, uint64_t* log_used,
                       uint64_t* log_size,
-                      LogFileNumberSize& log_file_number_size);
+                      LogFileNumberSize& log_file_number_size,
+                      SequenceNumber sequence);
 
   IOStatus WriteToWAL(const WriteThread::WriteGroup& write_group,
                       log::Writer* log_writer, uint64_t* log_used,
@@ -2383,9 +2523,8 @@ class DBImpl : public DB {
   // All ColumnFamily state changes go through this function. Here we analyze
   // the new state and we schedule background work if we detect that the new
   // state needs flush or compaction.
-  void InstallSuperVersionAndScheduleWork(
-      ColumnFamilyData* cfd, SuperVersionContext* sv_context,
-      const MutableCFOptions& mutable_cf_options);
+  void InstallSuperVersionAndScheduleWork(ColumnFamilyData* cfd,
+                                          SuperVersionContext* sv_context);
 
   bool GetIntPropertyInternal(ColumnFamilyData* cfd,
                               const DBPropertyInfo& property_info,
@@ -2424,6 +2563,7 @@ class DBImpl : public DB {
 
   IOStatus CreateWAL(const WriteOptions& write_options, uint64_t log_file_num,
                      uint64_t recycle_log_number, size_t preallocate_block_size,
+                     const PredecessorWALInfo& predecessor_wal_info,
                      log::Writer** new_log);
 
   // Validate self-consistency of DB options

db/db_impl/db_impl_compaction_flush.cc

@@ -66,7 +66,7 @@ bool DBImpl::RequestCompactionToken(ColumnFamilyData* cfd, bool force,
                                     LogBuffer* log_buffer) {
   assert(*token == nullptr);
   auto limiter = static_cast<ConcurrentTaskLimiterImpl*>(
-      cfd->ioptions()->compaction_thread_limiter.get());
+      cfd->ioptions().compaction_thread_limiter.get());
   if (limiter == nullptr) {
     return true;
   }
@@ -99,7 +99,7 @@ bool DBImpl::ShouldRescheduleFlushRequestToRetainUDT(
   // Check if holding on the flush will cause entering write stall mode.
   // Write stall entered because of the accumulation of write buffers can be
   // alleviated if we continue with the flush instead of postponing it.
-  const auto& mutable_cf_options = *cfd->GetLatestMutableCFOptions();
+  const auto& mutable_cf_options = cfd->GetLatestMutableCFOptions();
 
   // Use the same criteria as WaitUntilFlushWouldNotStallWrites does w.r.t
   // defining what a write stall is about to happen means. If this uses a
@@ -116,7 +116,7 @@ bool DBImpl::ShouldRescheduleFlushRequestToRetainUDT(
           cfd->GetUnflushedMemTableCountForWriteStallCheck(),
           /*num_l0_files=*/0,
           /*num_compaction_needed_bytes=*/0, mutable_cf_options,
-          *cfd->ioptions())
+          cfd->ioptions())
           .first;
   if (write_stall != WriteStallCondition::kNormal) {
     return false;
@@ -213,7 +213,7 @@ Status DBImpl::FlushMemTableToOutputFile(
       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_,
+      GetCompressionFlush(cfd->ioptions(), mutable_cf_options), stats_,
       &event_logger_, mutable_cf_options.report_bg_io_stats,
       true /* sync_output_directory */, true /* write_manifest */, thread_pri,
       io_tracer_, cfd->GetSuperVersion()->ShareSeqnoToTimeMapping(), db_id_,
@@ -299,8 +299,7 @@ Status DBImpl::FlushMemTableToOutputFile(
   }
 
   if (s.ok()) {
-    InstallSuperVersionAndScheduleWork(cfd, superversion_context,
-                                       mutable_cf_options);
+    InstallSuperVersionAndScheduleWork(cfd, superversion_context);
     if (made_progress) {
       *made_progress = true;
     }
@@ -369,7 +368,7 @@ Status DBImpl::FlushMemTableToOutputFile(
     if (sfm) {
       // Notify sst_file_manager that a new file was added
       std::string file_path = MakeTableFileName(
-          cfd->ioptions()->cf_paths[0].path, file_meta.fd.GetNumber());
+          cfd->ioptions().cf_paths[0].path, file_meta.fd.GetNumber());
       // TODO (PR7798).  We should only add the file to the FileManager if it
       // exists. Otherwise, some tests may fail.  Ignore the error in the
       // interim.
@@ -404,7 +403,8 @@ Status DBImpl::FlushMemTablesToOutputFiles(
   const auto& bg_flush_arg = bg_flush_args[0];
   ColumnFamilyData* cfd = bg_flush_arg.cfd_;
   // intentional infrequent copy for each flush
-  MutableCFOptions mutable_cf_options_copy = *cfd->GetLatestMutableCFOptions();
+  const MutableCFOptions mutable_cf_options_copy =
+      cfd->GetLatestMutableCFOptions();
   SuperVersionContext* superversion_context =
       bg_flush_arg.superversion_context_;
   FlushReason flush_reason = bg_flush_arg.flush_reason_;
@@ -461,7 +461,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
   for (int i = 0; i < num_cfs; ++i) {
     auto cfd = cfds[i];
     FSDirectory* data_dir = GetDataDir(cfd, 0U);
-    const std::string& curr_path = cfd->ioptions()->cf_paths[0].path;
+    const std::string& curr_path = cfd->ioptions().cf_paths[0].path;
 
     // Add to distinct output directories if eligible. Use linear search. Since
     // the number of elements in the vector is not large, performance should be
@@ -478,7 +478,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
       distinct_output_dirs.emplace_back(data_dir);
     }
 
-    all_mutable_cf_options.emplace_back(*cfd->GetLatestMutableCFOptions());
+    all_mutable_cf_options.emplace_back(cfd->GetLatestMutableCFOptions());
     const MutableCFOptions& mutable_cf_options = all_mutable_cf_options.back();
     uint64_t max_memtable_id = bg_flush_args[i].max_memtable_id_;
     FlushReason flush_reason = bg_flush_args[i].flush_reason_;
@@ -488,7 +488,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
         &shutting_down_, snapshot_seqs, earliest_write_conflict_snapshot,
         snapshot_checker, job_context, flush_reason, log_buffer,
         directories_.GetDbDir(), data_dir,
-        GetCompressionFlush(*cfd->ioptions(), mutable_cf_options), stats_,
+        GetCompressionFlush(cfd->ioptions(), mutable_cf_options), stats_,
         &event_logger_, mutable_cf_options.report_bg_io_stats,
         false /* sync_output_directory */, false /* write_manifest */,
         thread_pri, io_tracer_,
@@ -754,7 +754,6 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
   if (s.ok()) {
     autovector<ColumnFamilyData*> tmp_cfds;
     autovector<const autovector<ReadOnlyMemTable*>*> mems_list;
-    autovector<const MutableCFOptions*> mutable_cf_options_list;
     autovector<FileMetaData*> tmp_file_meta;
     autovector<std::list<std::unique_ptr<FlushJobInfo>>*>
         committed_flush_jobs_info;
@@ -763,7 +762,6 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
       if (!cfds[i]->IsDropped() && !mems.empty()) {
         tmp_cfds.emplace_back(cfds[i]);
         mems_list.emplace_back(&mems);
-        mutable_cf_options_list.emplace_back(&all_mutable_cf_options[i]);
         tmp_file_meta.emplace_back(&file_meta[i]);
         committed_flush_jobs_info.emplace_back(
             jobs[i]->GetCommittedFlushJobsInfo());
@@ -771,8 +769,8 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
     }
 
     s = InstallMemtableAtomicFlushResults(
-        nullptr /* imm_lists */, tmp_cfds, mutable_cf_options_list, mems_list,
-        versions_.get(), &logs_with_prep_tracker_, &mutex_, tmp_file_meta,
+        nullptr /* imm_lists */, tmp_cfds, mems_list, versions_.get(),
+        &logs_with_prep_tracker_, &mutex_, tmp_file_meta,
         committed_flush_jobs_info, &job_context->memtables_to_free,
         directories_.GetDbDir(), log_buffer);
   }
@@ -786,9 +784,8 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
       if (cfds[i]->IsDropped()) {
         continue;
       }
-      InstallSuperVersionAndScheduleWork(cfds[i],
-                                         &job_context->superversion_contexts[i],
-                                         all_mutable_cf_options[i]);
+      InstallSuperVersionAndScheduleWork(
+          cfds[i], &job_context->superversion_contexts[i]);
 
       const std::string& column_family_name = cfds[i]->GetName();
 
@@ -834,7 +831,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
                              jobs[i]->GetCommittedFlushJobsInfo());
       if (sfm) {
         std::string file_path = MakeTableFileName(
-            cfds[i]->ioptions()->cf_paths[0].path, file_meta[i].fd.GetNumber());
+            cfds[i]->ioptions().cf_paths[0].path, file_meta[i].fd.GetNumber());
         // TODO (PR7798).  We should only add the file to the FileManager if it
         // exists. Otherwise, some tests may fail.  Ignore the error in the
         // interim.
@@ -908,7 +905,7 @@ void DBImpl::NotifyOnFlushBegin(ColumnFamilyData* cfd, FileMetaData* file_meta,
     //                 go to L0 in the future.
     const uint64_t file_number = file_meta->fd.GetNumber();
     info.file_path =
-        MakeTableFileName(cfd->ioptions()->cf_paths[0].path, file_number);
+        MakeTableFileName(cfd->ioptions().cf_paths[0].path, file_number);
     info.file_number = file_number;
     info.thread_id = env_->GetThreadID();
     info.job_id = job_id;
@@ -1040,9 +1037,8 @@ Status DBImpl::IncreaseFullHistoryTsLowImpl(ColumnFamilyData* cfd,
     return Status::InvalidArgument(oss.str());
   }
 
-  Status s = versions_->LogAndApply(cfd, *cfd->GetLatestMutableCFOptions(),
-                                    read_options, write_options, &edit, &mutex_,
-                                    directories_.GetDbDir());
+  Status s = versions_->LogAndApply(cfd, read_options, write_options, &edit,
+                                    &mutex_, directories_.GetDbDir());
   if (!s.ok()) {
     return s;
   }
@@ -1066,11 +1062,11 @@ Status DBImpl::CompactRangeInternal(const CompactRangeOptions& options,
   auto cfh = static_cast_with_check<ColumnFamilyHandleImpl>(column_family);
   auto cfd = cfh->cfd();
 
-  if (options.target_path_id >= cfd->ioptions()->cf_paths.size()) {
+  if (options.target_path_id >= cfd->ioptions().cf_paths.size()) {
     return Status::InvalidArgument("Invalid target path ID");
   }
   if (options.change_level &&
-      cfd->ioptions()->compaction_style == kCompactionStyleFIFO) {
+      cfd->ioptions().compaction_style == kCompactionStyleFIFO) {
     return Status::NotSupported(
         "FIFO compaction does not support change_level.");
   }
@@ -1122,7 +1118,7 @@ Status DBImpl::CompactRangeInternal(const CompactRangeOptions& options,
   constexpr int kInvalidLevel = -1;
   int final_output_level = kInvalidLevel;
   bool exclusive = options.exclusive_manual_compaction;
-  if (cfd->ioptions()->compaction_style == kCompactionStyleUniversal &&
+  if (cfd->ioptions().compaction_style == kCompactionStyleUniversal &&
       cfd->NumberLevels() > 1) {
     // Always compact all files together.
     final_output_level = cfd->NumberLevels() - 1;
@@ -1134,7 +1130,7 @@ Status DBImpl::CompactRangeInternal(const CompactRangeOptions& options,
                             final_output_level, options, begin, end, exclusive,
                             false /* disable_trivial_move */,
                             std::numeric_limits<uint64_t>::max(), trim_ts);
-  } else if (cfd->ioptions()->compaction_style == kCompactionStyleFIFO) {
+  } else if (cfd->ioptions().compaction_style == kCompactionStyleFIFO) {
     // FIFOCompactionPicker::CompactRange() will ignore the input key range
     // [begin, end] and just try to pick compaction based on the configured
     // option `compaction_options_fifo`. So we skip checking if [begin, end]
@@ -1152,7 +1148,7 @@ Status DBImpl::CompactRangeInternal(const CompactRangeOptions& options,
 
       // Might need to query the partitioner
       SstPartitionerFactory* partitioner_factory =
-          current_version->cfd()->ioptions()->sst_partitioner_factory.get();
+          current_version->cfd()->ioptions().sst_partitioner_factory.get();
       std::unique_ptr<SstPartitioner> partitioner;
       if (partitioner_factory && begin != nullptr && end != nullptr) {
         SstPartitioner::Context context;
@@ -1228,7 +1224,7 @@ Status DBImpl::CompactRangeInternal(const CompactRangeOptions& options,
       CleanupSuperVersion(super_version);
     }
     if (s.ok() && first_overlapped_level != kInvalidLevel) {
-      if (cfd->ioptions()->compaction_style == kCompactionStyleUniversal) {
+      if (cfd->ioptions().compaction_style == kCompactionStyleUniversal) {
         assert(first_overlapped_level == 0);
         s = RunManualCompaction(
             cfd, first_overlapped_level, first_overlapped_level, options, begin,
@@ -1237,7 +1233,7 @@ Status DBImpl::CompactRangeInternal(const CompactRangeOptions& options,
             trim_ts);
         final_output_level = first_overlapped_level;
       } else {
-        assert(cfd->ioptions()->compaction_style == kCompactionStyleLevel);
+        assert(cfd->ioptions().compaction_style == kCompactionStyleLevel);
         uint64_t next_file_number = versions_->current_next_file_number();
         // Start compaction from `first_overlapped_level`, one level down at a
         // time, until output level >= max_overlapped_level.
@@ -1250,9 +1246,9 @@ Status DBImpl::CompactRangeInternal(const CompactRangeOptions& options,
         for (;;) {
           // Always allow L0 -> L1 compaction
           if (level > 0) {
-            if (cfd->ioptions()->level_compaction_dynamic_level_bytes) {
-              assert(final_output_level < cfd->ioptions()->num_levels);
-              if (final_output_level + 1 == cfd->ioptions()->num_levels) {
+            if (cfd->ioptions().level_compaction_dynamic_level_bytes) {
+              assert(final_output_level < cfd->ioptions().num_levels);
+              if (final_output_level + 1 == cfd->ioptions().num_levels) {
                 break;
               }
             } else {
@@ -1271,7 +1267,7 @@ Status DBImpl::CompactRangeInternal(const CompactRangeOptions& options,
             }
           }
           output_level = level + 1;
-          if (cfd->ioptions()->level_compaction_dynamic_level_bytes &&
+          if (cfd->ioptions().level_compaction_dynamic_level_bytes &&
               level == 0) {
             output_level = ColumnFamilyData::kCompactToBaseLevel;
           }
@@ -1303,8 +1299,8 @@ Status DBImpl::CompactRangeInternal(const CompactRangeOptions& options,
           // bottommost level intra-level compaction
           if ((options.bottommost_level_compaction ==
                    BottommostLevelCompaction::kIfHaveCompactionFilter &&
-               (cfd->ioptions()->compaction_filter != nullptr ||
-                cfd->ioptions()->compaction_filter_factory != nullptr)) ||
+               (cfd->ioptions().compaction_filter != nullptr ||
+                cfd->ioptions().compaction_filter_factory != nullptr)) ||
               options.bottommost_level_compaction ==
                   BottommostLevelCompaction::kForceOptimized ||
               options.bottommost_level_compaction ==
@@ -1458,7 +1454,7 @@ Status DBImpl::CompactFilesImpl(
   }
 
   if (output_path_id < 0) {
-    if (cfd->ioptions()->cf_paths.size() == 1U) {
+    if (cfd->ioptions().cf_paths.size() == 1U) {
       output_path_id = 0;
     } else {
       return Status::NotSupported(
@@ -1467,12 +1463,12 @@ Status DBImpl::CompactFilesImpl(
     }
   }
 
-  if (cfd->ioptions()->allow_ingest_behind &&
-      output_level >= cfd->ioptions()->num_levels - 1) {
+  if (cfd->ioptions().allow_ingest_behind &&
+      output_level >= cfd->ioptions().num_levels - 1) {
     return Status::InvalidArgument(
         "Exceed the maximum output level defined by "
         "the current compaction algorithm with ingest_behind --- " +
-        std::to_string(cfd->ioptions()->num_levels - 1));
+        std::to_string(cfd->ioptions().num_levels - 1));
   }
 
   std::vector<CompactionInputFiles> input_files;
@@ -1509,7 +1505,7 @@ Status DBImpl::CompactFilesImpl(
   assert(cfd->compaction_picker());
   c.reset(cfd->compaction_picker()->CompactFiles(
       compact_options, input_files, output_level, version->storage_info(),
-      *cfd->GetLatestMutableCFOptions(), mutable_db_options_, output_path_id));
+      cfd->GetLatestMutableCFOptions(), mutable_db_options_, output_path_id));
   // we already sanitized the set of input files and checked for conflicts
   // without releasing the lock, so we're guaranteed a compaction can be formed.
   assert(c != nullptr);
@@ -1539,8 +1535,8 @@ Status DBImpl::CompactFilesImpl(
       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_,
+      c->mutable_cf_options().paranoid_file_checks,
+      c->mutable_cf_options().report_bg_io_stats, dbname_,
       &compaction_job_stats, Env::Priority::USER, io_tracer_,
       kManualCompactionCanceledFalse_, db_id_, db_session_id_,
       c->column_family_data()->GetFullHistoryTsLow(), c->trim_ts(),
@@ -1551,10 +1547,10 @@ Status DBImpl::CompactFilesImpl(
   // takes running compactions into account (by skipping files that are already
   // being compacted). Since we just changed compaction score, we recalculate it
   // here.
-  version->storage_info()->ComputeCompactionScore(*cfd->ioptions(),
-                                                  *c->mutable_cf_options());
+  version->storage_info()->ComputeCompactionScore(cfd->ioptions(),
+                                                  c->mutable_cf_options());
 
-  compaction_job.Prepare();
+  compaction_job.Prepare(std::nullopt /*subcompact to be computed*/);
 
   std::unique_ptr<std::list<uint64_t>::iterator> min_options_file_number_elem;
   if (immutable_db_options().compaction_service != nullptr) {
@@ -1576,16 +1572,14 @@ Status DBImpl::CompactFilesImpl(
   }
 
   bool compaction_released = false;
-  Status status =
-      compaction_job.Install(*c->mutable_cf_options(), &compaction_released);
+  Status status = compaction_job.Install(&compaction_released);
   if (!compaction_released) {
     c->ReleaseCompactionFiles(s);
   }
   if (status.ok()) {
     assert(compaction_job.io_status().ok());
     InstallSuperVersionAndScheduleWork(
-        c->column_family_data(), job_context->superversion_contexts.data(),
-        *c->mutable_cf_options());
+        c->column_family_data(), job_context->superversion_contexts.data());
   }
   // status above captures any error during compaction_job.Install, so its ok
   // not check compaction_job.io_status() explicitly if we're not calling
@@ -1633,13 +1627,13 @@ Status DBImpl::CompactFilesImpl(
   if (output_file_names != nullptr) {
     for (const auto& newf : c->edit()->GetNewFiles()) {
       output_file_names->push_back(TableFileName(
-          c->immutable_options()->cf_paths, newf.second.fd.GetNumber(),
+          c->immutable_options().cf_paths, newf.second.fd.GetNumber(),
           newf.second.fd.GetPathId()));
     }
 
     for (const auto& blob_file : c->edit()->GetBlobFileAdditions()) {
       output_file_names->push_back(
-          BlobFileName(c->immutable_options()->cf_paths.front().path,
+          BlobFileName(c->immutable_options().cf_paths.front().path,
                        blob_file.GetBlobFileNumber()));
     }
   }
@@ -1774,7 +1768,7 @@ Status DBImpl::ReFitLevel(ColumnFamilyData* cfd, int level, int target_level) {
   }
   refitting_level_ = true;
 
-  const MutableCFOptions mutable_cf_options = *cfd->GetLatestMutableCFOptions();
+  const auto& mutable_cf_options = cfd->GetLatestMutableCFOptions();
   // move to a smaller level
   int to_level = target_level;
   if (target_level < 0) {
@@ -1845,12 +1839,12 @@ Status DBImpl::ReFitLevel(ColumnFamilyData* cfd, int level, int target_level) {
                     cfd->current()->DebugString().data());
 
     std::unique_ptr<Compaction> c(new Compaction(
-        vstorage, *cfd->ioptions(), mutable_cf_options, mutable_db_options_,
+        vstorage, cfd->ioptions(), mutable_cf_options, mutable_db_options_,
         {input}, to_level,
         MaxFileSizeForLevel(
             mutable_cf_options, to_level,
             cfd->ioptions()
-                ->compaction_style) /* output file size limit, not applicable */
+                .compaction_style) /* output file size limit, not applicable */
         ,
         LLONG_MAX /* max compaction bytes, not applicable */,
         0 /* output path ID, not applicable */, mutable_cf_options.compression,
@@ -1883,14 +1877,14 @@ Status DBImpl::ReFitLevel(ColumnFamilyData* cfd, int level, int target_level) {
                     "[%s] Apply version edit:\n%s", cfd->GetName().c_str(),
                     edit.DebugString().data());
 
-    Status status = versions_->LogAndApply(cfd, mutable_cf_options,
-                                           read_options, write_options, &edit,
-                                           &mutex_, directories_.GetDbDir());
+    Status status =
+        versions_->LogAndApply(cfd, read_options, write_options, &edit, &mutex_,
+                               directories_.GetDbDir());
     c->MarkFilesBeingCompacted(false);
     cfd->compaction_picker()->UnregisterCompaction(c.get());
     c.reset();
 
-    InstallSuperVersionAndScheduleWork(cfd, &sv_context, mutable_cf_options);
+    InstallSuperVersionAndScheduleWork(cfd, &sv_context);
 
     ROCKS_LOG_DEBUG(immutable_db_options_.info_log, "[%s] LogAndApply: %s\n",
                     cfd->GetName().c_str(), status.ToString().data());
@@ -2043,16 +2037,16 @@ Status DBImpl::RunManualCompaction(
   // For universal compaction, we enforce every manual compaction to compact
   // all files.
   if (begin == nullptr ||
-      cfd->ioptions()->compaction_style == kCompactionStyleUniversal ||
-      cfd->ioptions()->compaction_style == kCompactionStyleFIFO) {
+      cfd->ioptions().compaction_style == kCompactionStyleUniversal ||
+      cfd->ioptions().compaction_style == kCompactionStyleFIFO) {
     manual.begin = nullptr;
   } else {
     begin_storage.SetMinPossibleForUserKey(*begin);
     manual.begin = &begin_storage;
   }
   if (end == nullptr ||
-      cfd->ioptions()->compaction_style == kCompactionStyleUniversal ||
-      cfd->ioptions()->compaction_style == kCompactionStyleFIFO) {
+      cfd->ioptions().compaction_style == kCompactionStyleUniversal ||
+      cfd->ioptions().compaction_style == kCompactionStyleFIFO) {
     manual.end = nullptr;
   } else {
     end_storage.SetMaxPossibleForUserKey(*end);
@@ -2133,7 +2127,7 @@ Status DBImpl::RunManualCompaction(
         scheduled ||
         (((manual.manual_end = &manual.tmp_storage1) != nullptr) &&
          ((compaction = manual.cfd->CompactRange(
-               *manual.cfd->GetLatestMutableCFOptions(), mutable_db_options_,
+               manual.cfd->GetLatestMutableCFOptions(), mutable_db_options_,
                manual.input_level, manual.output_level, compact_range_options,
                manual.begin, manual.end, &manual.manual_end, &manual_conflict,
                max_file_num_to_ignore, trim_ts)) == nullptr &&
@@ -2676,7 +2670,7 @@ Status DBImpl::WaitUntilFlushWouldNotStallWrites(ColumnFamilyData* cfd,
         return Status::OK();
       }
 
-      const auto& mutable_cf_options = *cfd->GetLatestMutableCFOptions();
+      const auto& mutable_cf_options = cfd->GetLatestMutableCFOptions();
       const auto* vstorage = cfd->current()->storage_info();
 
       // Skip stalling check if we're below auto-flush and auto-compaction
@@ -2684,7 +2678,7 @@ Status DBImpl::WaitUntilFlushWouldNotStallWrites(ColumnFamilyData* cfd,
       // triggers are so low that stalling is needed for any background work. In
       // that case we shouldn't wait since background work won't be scheduled.
       if (cfd->imm()->NumNotFlushed() <
-              cfd->ioptions()->min_write_buffer_number_to_merge &&
+              cfd->ioptions().min_write_buffer_number_to_merge &&
           vstorage->l0_delay_trigger_count() <
               mutable_cf_options.level0_file_num_compaction_trigger) {
         break;
@@ -2700,7 +2694,7 @@ Status DBImpl::WaitUntilFlushWouldNotStallWrites(ColumnFamilyData* cfd,
               cfd->GetUnflushedMemTableCountForWriteStallCheck(),
               vstorage->l0_delay_trigger_count() + 1,
               vstorage->estimated_compaction_needed_bytes(), mutable_cf_options,
-              *cfd->ioptions())
+              cfd->ioptions())
               .first;
     } while (write_stall_condition != WriteStallCondition::kNormal);
   }
@@ -3528,6 +3522,7 @@ void DBImpl::BackgroundCallCompaction(PrepickedCompaction* prepicked_compaction,
   }
 }
 
+// Precondition: mutex_ must be held when calling this function.
 Status DBImpl::BackgroundCompaction(bool* made_progress,
                                     JobContext* job_context,
                                     LogBuffer* log_buffer,
@@ -3679,8 +3674,8 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
     // Compaction makes a copy of the latest MutableCFOptions. It should be used
     // throughout the compaction procedure to make sure consistency. It will
     // eventually be installed into SuperVersion
-    auto* mutable_cf_options = cfd->GetLatestMutableCFOptions();
-    if (!mutable_cf_options->disable_auto_compactions && !cfd->IsDropped()) {
+    const auto& mutable_cf_options = cfd->GetLatestMutableCFOptions();
+    if (!mutable_cf_options.disable_auto_compactions && !cfd->IsDropped()) {
       // NOTE: try to avoid unnecessary copy of MutableCFOptions if
       // compaction is not necessary. Need to make sure mutex is held
       // until we make a copy in the following code
@@ -3689,7 +3684,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
       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 &&
+      if (cfd->ioptions().compaction_style == kCompactionStyleUniversal &&
           cfd->user_comparator()->timestamp_size() == 0) {
         SequenceNumber earliest_write_conflict_snapshot;
         GetSnapshotContext(job_context, &snapshot_seqs,
@@ -3697,7 +3692,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
                            &snapshot_checker);
         assert(is_snapshot_supported_ || snapshots_.empty());
       }
-      c.reset(cfd->PickCompaction(*mutable_cf_options, mutable_db_options_,
+      c.reset(cfd->PickCompaction(mutable_cf_options, mutable_db_options_,
                                   snapshot_seqs, snapshot_checker, log_buffer));
       TEST_SYNC_POINT("DBImpl::BackgroundCompaction():AfterPickCompaction");
 
@@ -3711,8 +3706,8 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
           c->column_family_data()
               ->current()
               ->storage_info()
-              ->ComputeCompactionScore(*(c->immutable_options()),
-                                       *(c->mutable_cf_options()));
+              ->ComputeCompactionScore(c->immutable_options(),
+                                       c->mutable_cf_options());
           AddToCompactionQueue(cfd);
 
           c.reset();
@@ -3760,7 +3755,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
     TEST_SYNC_POINT_CALLBACK("DBImpl::BackgroundCompaction:BeforeCompaction",
                              c->column_family_data());
     assert(c->num_input_files(1) == 0);
-    assert(c->column_family_data()->ioptions()->compaction_style ==
+    assert(c->column_family_data()->ioptions().compaction_style ==
            kCompactionStyleFIFO);
 
     compaction_job_stats.num_input_files = c->num_input_files(0);
@@ -3772,8 +3767,8 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
       c->edit()->DeleteFile(c->level(), f->fd.GetNumber());
     }
     status = versions_->LogAndApply(
-        c->column_family_data(), *c->mutable_cf_options(), read_options,
-        write_options, c->edit(), &mutex_, directories_.GetDbDir(),
+        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);
@@ -3781,8 +3776,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
         });
     io_s = versions_->io_status();
     InstallSuperVersionAndScheduleWork(
-        c->column_family_data(), job_context->superversion_contexts.data(),
-        *c->mutable_cf_options());
+        c->column_family_data(), job_context->superversion_contexts.data());
     ROCKS_LOG_BUFFER(log_buffer, "[%s] Deleted %d files\n",
                      c->column_family_data()->GetName().c_str(),
                      c->num_input_files(0));
@@ -3838,7 +3832,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
       }
     }
     if (c->compaction_reason() == CompactionReason::kLevelMaxLevelSize &&
-        c->immutable_options()->compaction_pri == kRoundRobin) {
+        c->immutable_options().compaction_pri == kRoundRobin) {
       int start_level = c->start_level();
       if (start_level > 0) {
         auto vstorage = c->input_version()->storage_info();
@@ -3848,26 +3842,23 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
       }
     }
     status = versions_->LogAndApply(
-        c->column_family_data(), *c->mutable_cf_options(), read_options,
-        write_options, c->edit(), &mutex_, directories_.GetDbDir(),
+        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;
         });
     io_s = versions_->io_status();
-    // Use latest MutableCFOptions
     InstallSuperVersionAndScheduleWork(
-        c->column_family_data(), job_context->superversion_contexts.data(),
-        *c->mutable_cf_options());
+        c->column_family_data(), job_context->superversion_contexts.data());
 
     VersionStorageInfo::LevelSummaryStorage tmp;
     c->column_family_data()->internal_stats()->IncBytesMoved(c->output_level(),
                                                              moved_bytes);
     {
       event_logger_.LogToBuffer(log_buffer)
-          << "job" << job_context->job_id << "event"
-          << "trivial_move"
+          << "job" << job_context->job_id << "event" << "trivial_move"
           << "destination_level" << c->output_level() << "files" << moved_files
           << "total_files_size" << moved_bytes;
     }
@@ -3929,15 +3920,15 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
         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_,
+        c->mutable_cf_options().paranoid_file_checks,
+        c->mutable_cf_options().report_bg_io_stats, dbname_,
         &compaction_job_stats, thread_pri, io_tracer_,
         is_manual ? manual_compaction->canceled
                   : kManualCompactionCanceledFalse_,
         db_id_, db_session_id_, c->column_family_data()->GetFullHistoryTsLow(),
         c->trim_ts(), &blob_callback_, &bg_compaction_scheduled_,
         &bg_bottom_compaction_scheduled_);
-    compaction_job.Prepare();
+    compaction_job.Prepare(std::nullopt /*subcompact to be computed*/);
 
     std::unique_ptr<std::list<uint64_t>::iterator> min_options_file_number_elem;
     if (immutable_db_options().compaction_service != nullptr) {
@@ -3959,13 +3950,11 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
       ReleaseOptionsFileNumber(min_options_file_number_elem);
     }
 
-    status =
-        compaction_job.Install(*c->mutable_cf_options(), &compaction_released);
+    status = compaction_job.Install(&compaction_released);
     io_s = compaction_job.io_status();
     if (status.ok()) {
       InstallSuperVersionAndScheduleWork(
-          c->column_family_data(), job_context->superversion_contexts.data(),
-          *c->mutable_cf_options());
+          c->column_family_data(), job_context->superversion_contexts.data());
     }
     *made_progress = true;
     TEST_SYNC_POINT_CALLBACK("DBImpl::BackgroundCompaction:AfterCompaction",
@@ -4043,8 +4032,8 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
       c->column_family_data()
           ->current()
           ->storage_info()
-          ->ComputeCompactionScore(*(c->immutable_options()),
-                                   *(c->mutable_cf_options()));
+          ->ComputeCompactionScore(c->immutable_options(),
+                                   c->mutable_cf_options());
       if (!cfd->queued_for_compaction()) {
         AddToCompactionQueue(cfd);
       }
@@ -4079,10 +4068,9 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
       // We only compacted part of the requested range.  Update *m
       // to the range that is left to be compacted.
       // Universal and FIFO compactions should always compact the whole range
-      assert(m->cfd->ioptions()->compaction_style !=
-                 kCompactionStyleUniversal ||
-             m->cfd->ioptions()->num_levels > 1);
-      assert(m->cfd->ioptions()->compaction_style != kCompactionStyleFIFO);
+      assert(m->cfd->ioptions().compaction_style != kCompactionStyleUniversal ||
+             m->cfd->ioptions().num_levels > 1);
+      assert(m->cfd->ioptions().compaction_style != kCompactionStyleFIFO);
       m->tmp_storage = *m->manual_end;
       m->begin = &m->tmp_storage;
       m->incomplete = true;
@@ -4229,7 +4217,7 @@ void DBImpl::BuildCompactionJobInfo(
     for (const auto fmd : *c->inputs(i)) {
       const FileDescriptor& desc = fmd->fd;
       const uint64_t file_number = desc.GetNumber();
-      auto fn = TableFileName(c->immutable_options()->cf_paths, file_number,
+      auto fn = TableFileName(c->immutable_options().cf_paths, file_number,
                               desc.GetPathId());
       compaction_job_info->input_files.push_back(fn);
       compaction_job_info->input_file_infos.push_back(CompactionFileInfo{
@@ -4242,17 +4230,17 @@ void DBImpl::BuildCompactionJobInfo(
     const FileDescriptor& desc = meta.fd;
     const uint64_t file_number = desc.GetNumber();
     compaction_job_info->output_files.push_back(TableFileName(
-        c->immutable_options()->cf_paths, file_number, desc.GetPathId()));
+        c->immutable_options().cf_paths, file_number, desc.GetPathId()));
     compaction_job_info->output_file_infos.push_back(CompactionFileInfo{
         newf.first, file_number, meta.oldest_blob_file_number});
   }
   compaction_job_info->blob_compression_type =
-      c->mutable_cf_options()->blob_compression_type;
+      c->mutable_cf_options().blob_compression_type;
 
   // Update BlobFilesInfo.
   for (const auto& blob_file : c->edit()->GetBlobFileAdditions()) {
     BlobFileAdditionInfo blob_file_addition_info(
-        BlobFileName(c->immutable_options()->cf_paths.front().path,
+        BlobFileName(c->immutable_options().cf_paths.front().path,
                      blob_file.GetBlobFileNumber()) /*blob_file_path*/,
         blob_file.GetBlobFileNumber(), blob_file.GetTotalBlobCount(),
         blob_file.GetTotalBlobBytes());
@@ -4263,7 +4251,7 @@ void DBImpl::BuildCompactionJobInfo(
   // Update BlobFilesGarbageInfo.
   for (const auto& blob_file : c->edit()->GetBlobFileGarbages()) {
     BlobFileGarbageInfo blob_file_garbage_info(
-        BlobFileName(c->immutable_options()->cf_paths.front().path,
+        BlobFileName(c->immutable_options().cf_paths.front().path,
                      blob_file.GetBlobFileNumber()) /*blob_file_path*/,
         blob_file.GetBlobFileNumber(), blob_file.GetGarbageBlobCount(),
         blob_file.GetGarbageBlobBytes());
@@ -4285,9 +4273,9 @@ void DBImpl::BuildCompactionJobInfo(
 // for superversion_to_free
 
 void DBImpl::InstallSuperVersionAndScheduleWork(
-    ColumnFamilyData* cfd, SuperVersionContext* sv_context,
-    const MutableCFOptions& mutable_cf_options) {
+    ColumnFamilyData* cfd, SuperVersionContext* sv_context) {
   mutex_.AssertHeld();
+  const auto& mutable_cf_options = cfd->GetLatestMutableCFOptions();
 
   // Update max_total_in_memory_state_
   size_t old_memtable_size = 0;
@@ -4310,7 +4298,7 @@ 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->ioptions().allow_ingest_behind) {
       bottommost_files_mark_threshold_ = std::min(
           bottommost_files_mark_threshold_,
           my_cfd->current()->storage_info()->bottommost_files_mark_threshold());

db/db_impl/db_impl_debug.cc

@@ -6,8 +6,8 @@
 // Copyright (c) 2011 The LevelDB Authors. All rights reserved.
 // 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.
-
 #ifndef NDEBUG
+#include <iostream>
 
 #include "db/blob/blob_file_cache.h"
 #include "db/column_family.h"
@@ -103,8 +103,8 @@ Status DBImpl::TEST_CompactRange(int level, const Slice* begin,
     cfd = cfh->cfd();
   }
   int output_level =
-      (cfd->ioptions()->compaction_style == kCompactionStyleUniversal ||
-       cfd->ioptions()->compaction_style == kCompactionStyleFIFO)
+      (cfd->ioptions().compaction_style == kCompactionStyleUniversal ||
+       cfd->ioptions().compaction_style == kCompactionStyleFIFO)
           ? level
           : level + 1;
   return RunManualCompaction(
@@ -239,7 +239,7 @@ void DBImpl::TEST_GetAllBlockCaches(
   for (auto cfd : *versions_->GetColumnFamilySet()) {
     if (const auto bbto =
             cfd->GetCurrentMutableCFOptions()
-                ->table_factory->GetOptions<BlockBasedTableOptions>()) {
+                .table_factory->GetOptions<BlockBasedTableOptions>()) {
       cache_set->insert(bbto->block_cache.get());
     }
   }
@@ -267,7 +267,7 @@ Status DBImpl::TEST_GetLatestMutableCFOptions(
   InstrumentedMutexLock l(&mutex_);
 
   auto cfh = static_cast_with_check<ColumnFamilyHandleImpl>(column_family);
-  *mutable_cf_options = *cfh->cfd()->GetLatestMutableCFOptions();
+  *mutable_cf_options = cfh->cfd()->GetLatestMutableCFOptions();
   return Status::OK();
 }
 
@@ -379,7 +379,11 @@ void DBImpl::TEST_VerifyNoObsoleteFilesCached(
     uint64_t file_number;
     GetUnaligned(reinterpret_cast<const uint64_t*>(key.data()), &file_number);
     // Assert file is in live/quarantined set
-    assert(live_and_quar_files.find(file_number) != live_and_quar_files.end());
+    if (live_and_quar_files.find(file_number) == live_and_quar_files.end()) {
+      std::cerr << "File " << file_number << " is not live nor quarantined"
+                << std::endl;
+      assert(false);
+    }
   };
   table_cache_->ApplyToAllEntries(fn, {});
 }

db/db_impl/db_impl_experimental.cc

@@ -45,8 +45,8 @@ Status DBImpl::SuggestCompactRange(ColumnFamilyHandle* column_family,
     }
     // Since we have some more files to compact, we should also recompute
     // compaction score
-    vstorage->ComputeCompactionScore(*cfd->ioptions(),
-                                     *cfd->GetLatestMutableCFOptions());
+    vstorage->ComputeCompactionScore(cfd->ioptions(),
+                                     cfd->GetLatestMutableCFOptions());
     EnqueuePendingCompaction(cfd);
     MaybeScheduleFlushOrCompaction();
   }
@@ -146,13 +146,11 @@ Status DBImpl::PromoteL0(ColumnFamilyHandle* column_family, int target_level) {
                    f->user_defined_timestamps_persisted);
     }
 
-    status = versions_->LogAndApply(cfd, *cfd->GetLatestMutableCFOptions(),
-                                    read_options, write_options, &edit, &mutex_,
-                                    directories_.GetDbDir());
+    status = versions_->LogAndApply(cfd, read_options, write_options, &edit,
+                                    &mutex_, directories_.GetDbDir());
     if (status.ok()) {
       InstallSuperVersionAndScheduleWork(
-          cfd, job_context.superversion_contexts.data(),
-          *cfd->GetLatestMutableCFOptions());
+          cfd, job_context.superversion_contexts.data());
     }
   }  // lock released here
   LogFlush(immutable_db_options_.info_log);

db/db_impl/db_impl_files.cc

@@ -159,6 +159,7 @@ void DBImpl::FindObsoleteFiles(JobContext* job_context, bool force,
 
   // Get obsolete files.  This function will also update the list of
   // pending files in VersionSet().
+  assert(versions_);
   versions_->GetObsoleteFiles(
       &job_context->sst_delete_files, &job_context->blob_delete_files,
       &job_context->manifest_delete_files, job_context->min_pending_output);
@@ -1039,7 +1040,7 @@ std::set<std::string> DBImpl::CollectAllDBPaths() {
     all_db_paths.insert(NormalizePath(db_path.path));
   }
   for (const auto* cfd : *versions_->GetColumnFamilySet()) {
-    for (const auto& cf_path : cfd->ioptions()->cf_paths) {
+    for (const auto& cf_path : cfd->ioptions().cf_paths) {
       all_db_paths.insert(NormalizePath(cf_path.path));
     }
   }

db/db_impl/db_impl_follower.cc

@@ -140,10 +140,8 @@ Status DBImplFollower::TryCatchUpWithLeader() {
           // to super versions in a lock free manner, it checks the earliest
           // sequence number to detect if there was a change in version in
           // the meantime.
-          const MutableCFOptions mutable_cf_options =
-              *cfd->GetLatestMutableCFOptions();
           MemTable* new_mem = cfd->ConstructNewMemtable(
-              mutable_cf_options, versions_->LastSequence());
+              cfd->GetLatestMutableCFOptions(), versions_->LastSequence());
           cfd->mem()->SetNextLogNumber(cfd->GetLogNumber());
           cfd->mem()->ConstructFragmentedRangeTombstones();
           cfd->imm()->Add(cfd->mem(), &job_context.memtables_to_free);

db/db_impl/db_impl_readonly.cc

@@ -8,7 +8,7 @@
 #include "db/arena_wrapped_db_iter.h"
 #include "db/db_impl/compacted_db_impl.h"
 #include "db/db_impl/db_impl.h"
-#include "db/db_iter.h"
+#include "db/manifest_ops.h"
 #include "db/merge_context.h"
 #include "logging/logging.h"
 #include "monitoring/perf_context_imp.h"
@@ -16,7 +16,6 @@
 
 namespace ROCKSDB_NAMESPACE {
 
-
 DBImplReadOnly::DBImplReadOnly(const DBOptions& db_options,
                                const std::string& dbname)
     : DBImpl(db_options, dbname, /*seq_per_batch*/ false,
@@ -33,7 +32,8 @@ Status DBImplReadOnly::GetImpl(const ReadOptions& read_options,
                                const Slice& key,
                                GetImplOptions& get_impl_options) {
   assert(get_impl_options.value != nullptr ||
-         get_impl_options.columns != nullptr);
+         get_impl_options.columns != nullptr ||
+         get_impl_options.merge_operands != nullptr);
   assert(get_impl_options.column_family);
 
   Status s;
@@ -87,7 +87,11 @@ Status DBImplReadOnly::GetImpl(const ReadOptions& read_options,
       return s;
     }
   }
+  // Prepare to store a list of merge operations if merge occurs.
   MergeContext merge_context;
+  // TODO - Large Result Optimization for Read Only DB
+  // (https://github.com/facebook/rocksdb/pull/10458)
+
   SequenceNumber max_covering_tombstone_seq = 0;
   LookupKey lkey(key, snapshot, read_options.timestamp);
   PERF_TIMER_STOP(get_snapshot_time);
@@ -98,7 +102,8 @@ Status DBImplReadOnly::GetImpl(const ReadOptions& read_options,
           get_impl_options.value ? get_impl_options.value->GetSelf() : nullptr,
           get_impl_options.columns, ts, &s, &merge_context,
           &max_covering_tombstone_seq, read_options,
-          false /* immutable_memtable */, &read_cb)) {
+          false /* immutable_memtable */, &read_cb,
+          /*is_blob_index=*/nullptr, /*do_merge=*/get_impl_options.get_value)) {
     if (get_impl_options.value) {
       get_impl_options.value->PinSelf();
     }
@@ -112,7 +117,7 @@ Status DBImplReadOnly::GetImpl(const ReadOptions& read_options,
         /*value_found*/ nullptr,
         /*key_exists*/ nullptr, /*seq*/ nullptr, &read_cb,
         /*is_blob*/ nullptr,
-        /*do_merge*/ true);
+        /*do_merge=*/get_impl_options.get_value);
     RecordTick(stats_, MEMTABLE_MISS);
   }
   {
@@ -122,6 +127,14 @@ Status DBImplReadOnly::GetImpl(const ReadOptions& read_options,
       size = get_impl_options.value->size();
     } else if (get_impl_options.columns) {
       size = get_impl_options.columns->serialized_size();
+    } else if (get_impl_options.merge_operands) {
+      *get_impl_options.number_of_operands =
+          static_cast<int>(merge_context.GetNumOperands());
+      for (const Slice& sl : merge_context.GetOperands()) {
+        size += sl.size();
+        get_impl_options.merge_operands->PinSelf(sl);
+        get_impl_options.merge_operands++;
+      }
     }
     RecordTick(stats_, BYTES_READ, size);
     RecordInHistogram(stats_, BYTES_PER_READ, size);
@@ -173,7 +186,7 @@ Iterator* DBImplReadOnly::NewIterator(const ReadOptions& _read_options,
           : latest_snapshot;
   ReadCallback* read_callback = nullptr;  // No read callback provided.
   auto db_iter = NewArenaWrappedDbIterator(
-      env_, read_options, *cfd->ioptions(), super_version->mutable_cf_options,
+      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);
@@ -240,7 +253,7 @@ Status DBImplReadOnly::NewIterators(
   assert(cfd_to_sv.size() == column_families.size());
   for (auto [cfd, sv] : cfd_to_sv) {
     auto* db_iter = NewArenaWrappedDbIterator(
-        env_, read_options, *cfd->ioptions(), sv->mutable_cf_options,
+        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);
@@ -265,9 +278,8 @@ Status OpenForReadOnlyCheckExistence(const DBOptions& db_options,
     const std::shared_ptr<FileSystem>& fs = db_options.env->GetFileSystem();
     std::string manifest_path;
     uint64_t manifest_file_number;
-    s = VersionSet::GetCurrentManifestPath(dbname, fs.get(), /*is_retry=*/false,
-                                           &manifest_path,
-                                           &manifest_file_number);
+    s = GetCurrentManifestPath(dbname, fs.get(), /*is_retry=*/false,
+                               &manifest_path, &manifest_file_number);
   } else {
     // Historic behavior that doesn't necessarily make sense
     s = db_options.env->CreateDirIfMissing(dbname);
@@ -277,7 +289,8 @@ Status OpenForReadOnlyCheckExistence(const DBOptions& db_options,
 }  // namespace
 
 Status DB::OpenForReadOnly(const Options& options, const std::string& dbname,
-                           DB** dbptr, bool /*error_if_wal_file_exists*/) {
+                           std::unique_ptr<DB>* dbptr,
+                           bool /*error_if_wal_file_exists*/) {
   Status s = OpenForReadOnlyCheckExistence(options, dbname);
   if (!s.ok()) {
     return s;
@@ -311,7 +324,7 @@ Status DB::OpenForReadOnly(const Options& options, const std::string& dbname,
 Status DB::OpenForReadOnly(
     const DBOptions& db_options, const std::string& dbname,
     const std::vector<ColumnFamilyDescriptor>& column_families,
-    std::vector<ColumnFamilyHandle*>* handles, DB** dbptr,
+    std::vector<ColumnFamilyHandle*>* handles, std::unique_ptr<DB>* dbptr,
     bool error_if_wal_file_exists) {
   // If dbname does not exist in the file system, should not do anything
   Status s = OpenForReadOnlyCheckExistence(db_options, dbname);
@@ -327,7 +340,7 @@ Status DB::OpenForReadOnly(
 Status DBImplReadOnly::OpenForReadOnlyWithoutCheck(
     const DBOptions& db_options, const std::string& dbname,
     const std::vector<ColumnFamilyDescriptor>& column_families,
-    std::vector<ColumnFamilyHandle*>* handles, DB** dbptr,
+    std::vector<ColumnFamilyHandle*>* handles, std::unique_ptr<DB>* dbptr,
     bool error_if_wal_file_exists) {
   *dbptr = nullptr;
   handles->clear();
@@ -358,7 +371,7 @@ Status DBImplReadOnly::OpenForReadOnlyWithoutCheck(
   impl->mutex_.Unlock();
   sv_context.Clean();
   if (s.ok()) {
-    *dbptr = impl;
+    dbptr->reset(impl);
     for (auto* h : *handles) {
       impl->NewThreadStatusCfInfo(
           static_cast_with_check<ColumnFamilyHandleImpl>(h)->cfd());
@@ -373,5 +386,4 @@ Status DBImplReadOnly::OpenForReadOnlyWithoutCheck(
   return s;
 }
 
-
 }  // namespace ROCKSDB_NAMESPACE

db/db_impl/db_impl_readonly.h

@@ -5,7 +5,6 @@
 
 #pragma once
 
-
 #include <string>
 #include <vector>
 
@@ -172,7 +171,7 @@ class DBImplReadOnly : public DBImpl {
   static Status OpenForReadOnlyWithoutCheck(
       const DBOptions& db_options, const std::string& dbname,
       const std::vector<ColumnFamilyDescriptor>& column_families,
-      std::vector<ColumnFamilyHandle*>* handles, DB** dbptr,
+      std::vector<ColumnFamilyHandle*>* handles, std::unique_ptr<DB>* dbptr,
       bool error_if_wal_file_exists = false);
   friend class DB;
 };

db/db_impl/db_impl_secondary.cc

@@ -55,9 +55,9 @@ Status DBImplSecondary::Recover(
   // Initial max_total_in_memory_state_ before recovery logs.
   max_total_in_memory_state_ = 0;
   for (auto cfd : *versions_->GetColumnFamilySet()) {
-    auto* mutable_cf_options = cfd->GetLatestMutableCFOptions();
-    max_total_in_memory_state_ += mutable_cf_options->write_buffer_size *
-                                  mutable_cf_options->max_write_buffer_number;
+    const auto& mutable_cf_options = cfd->GetLatestMutableCFOptions();
+    max_total_in_memory_state_ += mutable_cf_options.write_buffer_size *
+                                  mutable_cf_options.max_write_buffer_number;
   }
   if (s.ok()) {
     default_cf_handle_ = new ColumnFamilyHandleImpl(
@@ -270,10 +270,8 @@ Status DBImplSecondary::RecoverLogFiles(
           if (!cfd->mem()->IsEmpty() &&
               (curr_log_num == std::numeric_limits<uint64_t>::max() ||
                curr_log_num != log_number)) {
-            const MutableCFOptions mutable_cf_options =
-                *cfd->GetLatestMutableCFOptions();
-            MemTable* new_mem =
-                cfd->ConstructNewMemtable(mutable_cf_options, seq_of_batch);
+            MemTable* new_mem = cfd->ConstructNewMemtable(
+                cfd->GetLatestMutableCFOptions(), seq_of_batch);
             cfd->mem()->SetNextLogNumber(log_number);
             cfd->mem()->ConstructFragmentedRangeTombstones();
             cfd->imm()->Add(cfd->mem(), &job_context->memtables_to_free);
@@ -344,7 +342,8 @@ Status DBImplSecondary::GetImpl(const ReadOptions& read_options,
                                 const Slice& key,
                                 GetImplOptions& get_impl_options) {
   assert(get_impl_options.value != nullptr ||
-         get_impl_options.columns != nullptr);
+         get_impl_options.columns != nullptr ||
+         get_impl_options.merge_operands != nullptr);
   assert(get_impl_options.column_family);
 
   Status s;
@@ -399,37 +398,64 @@ Status DBImplSecondary::GetImpl(const ReadOptions& read_options,
     }
   }
   MergeContext merge_context;
+  // TODO - Large Result Optimization for Secondary DB
+  // (https://github.com/facebook/rocksdb/pull/10458)
+
   SequenceNumber max_covering_tombstone_seq = 0;
   LookupKey lkey(key, snapshot, read_options.timestamp);
   PERF_TIMER_STOP(get_snapshot_time);
   bool done = false;
 
   // Look up starts here
-  if (super_version->mem->Get(
-          lkey,
-          get_impl_options.value ? get_impl_options.value->GetSelf() : nullptr,
-          get_impl_options.columns, ts, &s, &merge_context,
-          &max_covering_tombstone_seq, read_options,
-          false /* immutable_memtable */, &read_cb)) {
-    done = true;
-    if (get_impl_options.value) {
-      get_impl_options.value->PinSelf();
+  if (get_impl_options.get_value) {
+    if (super_version->mem->Get(
+            lkey,
+            get_impl_options.value ? get_impl_options.value->GetSelf()
+                                   : nullptr,
+            get_impl_options.columns, ts, &s, &merge_context,
+            &max_covering_tombstone_seq, read_options,
+            false /* immutable_memtable */, &read_cb,
+            /*is_blob_index=*/nullptr, /*do_merge=*/true)) {
+      done = true;
+      if (get_impl_options.value) {
+        get_impl_options.value->PinSelf();
+      }
+      RecordTick(stats_, MEMTABLE_HIT);
+    } else if ((s.ok() || s.IsMergeInProgress()) &&
+               super_version->imm->Get(
+                   lkey,
+                   get_impl_options.value ? get_impl_options.value->GetSelf()
+                                          : nullptr,
+                   get_impl_options.columns, ts, &s, &merge_context,
+                   &max_covering_tombstone_seq, read_options, &read_cb)) {
+      done = true;
+      if (get_impl_options.value) {
+        get_impl_options.value->PinSelf();
+      }
+      RecordTick(stats_, MEMTABLE_HIT);
     }
-    RecordTick(stats_, MEMTABLE_HIT);
-  } else if ((s.ok() || s.IsMergeInProgress()) &&
-             super_version->imm->Get(
-                 lkey,
-                 get_impl_options.value ? get_impl_options.value->GetSelf()
-                                        : nullptr,
-                 get_impl_options.columns, ts, &s, &merge_context,
-                 &max_covering_tombstone_seq, read_options, &read_cb)) {
-    done = true;
-    if (get_impl_options.value) {
-      get_impl_options.value->PinSelf();
+  } else {
+    // GetMergeOperands
+    if (super_version->mem->Get(
+            lkey,
+            get_impl_options.value ? get_impl_options.value->GetSelf()
+                                   : nullptr,
+            get_impl_options.columns, ts, &s, &merge_context,
+            &max_covering_tombstone_seq, read_options,
+            false /* immutable_memtable */, &read_cb,
+            /*is_blob_index=*/nullptr, /*do_merge=*/false)) {
+      done = true;
+      RecordTick(stats_, MEMTABLE_HIT);
+    } else if ((s.ok() || s.IsMergeInProgress()) &&
+               super_version->imm->GetMergeOperands(lkey, &s, &merge_context,
+                                                    &max_covering_tombstone_seq,
+                                                    read_options)) {
+      done = true;
+      RecordTick(stats_, MEMTABLE_HIT);
     }
-    RecordTick(stats_, MEMTABLE_HIT);
   }
-  if (!done && !s.ok() && !s.IsMergeInProgress()) {
+  if (!s.ok() && !s.IsMergeInProgress() && !s.IsNotFound()) {
+    assert(done);
     ReturnAndCleanupSuperVersion(cfd, super_version);
     return s;
   }
@@ -453,6 +479,14 @@ Status DBImplSecondary::GetImpl(const ReadOptions& read_options,
       size = get_impl_options.value->size();
     } else if (get_impl_options.columns) {
       size = get_impl_options.columns->serialized_size();
+    } else if (get_impl_options.merge_operands) {
+      *get_impl_options.number_of_operands =
+          static_cast<int>(merge_context.GetNumOperands());
+      for (const Slice& sl : merge_context.GetOperands()) {
+        size += sl.size();
+        get_impl_options.merge_operands->PinSelf(sl);
+        get_impl_options.merge_operands++;
+      }
     }
     RecordTick(stats_, BYTES_READ, size);
     RecordTimeToHistogram(stats_, BYTES_PER_READ, size);
@@ -533,7 +567,7 @@ ArenaWrappedDBIter* DBImplSecondary::NewIteratorImpl(
   snapshot = versions_->LastSequence();
   assert(snapshot != kMaxSequenceNumber);
   auto db_iter = NewArenaWrappedDbIterator(
-      env_, read_options, *cfh->cfd()->ioptions(),
+      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,
@@ -737,7 +771,8 @@ Status DBImplSecondary::TryCatchUpWithPrimary() {
 }
 
 Status DB::OpenAsSecondary(const Options& options, const std::string& dbname,
-                           const std::string& secondary_path, DB** dbptr) {
+                           const std::string& secondary_path,
+                           std::unique_ptr<DB>* dbptr) {
   *dbptr = nullptr;
 
   DBOptions db_options(options);
@@ -759,7 +794,7 @@ Status DB::OpenAsSecondary(
     const DBOptions& db_options, const std::string& dbname,
     const std::string& secondary_path,
     const std::vector<ColumnFamilyDescriptor>& column_families,
-    std::vector<ColumnFamilyHandle*>* handles, DB** dbptr) {
+    std::vector<ColumnFamilyHandle*>* handles, std::unique_ptr<DB>* dbptr) {
   *dbptr = nullptr;
 
   DBOptions tmp_opts(db_options);
@@ -826,7 +861,7 @@ Status DB::OpenAsSecondary(
   impl->mutex_.Unlock();
   sv_context.Clean();
   if (s.ok()) {
-    *dbptr = impl;
+    dbptr->reset(impl);
     for (auto h : *handles) {
       impl->NewThreadStatusCfInfo(
           static_cast_with_check<ColumnFamilyHandleImpl>(h)->cfd());
@@ -864,21 +899,24 @@ Status DBImplSecondary::CompactWithoutInstallation(
   ColumnFamilyMetaData cf_meta;
   version->GetColumnFamilyMetaData(&cf_meta);
 
-  const MutableCFOptions* mutable_cf_options = cfd->GetLatestMutableCFOptions();
-  ColumnFamilyOptions cf_options = cfd->GetLatestCFOptions();
   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(
-      *mutable_cf_options, input.output_level, cf_options.compaction_style,
-      vstorage->base_level(), cf_options.level_compaction_dynamic_level_bytes);
+      cfd->GetLatestMutableCFOptions(), input.output_level,
+      cfd->ioptions().compaction_style, vstorage->base_level(),
+      cfd->ioptions().level_compaction_dynamic_level_bytes);
 
   std::vector<CompactionInputFiles> input_files;
   Status s = cfd->compaction_picker()->GetCompactionInputsFromFileNumbers(
       &input_files, &input_set, vstorage, comp_options);
   if (!s.ok()) {
+    ROCKS_LOG_ERROR(
+        immutable_db_options_.info_log,
+        "GetCompactionInputsFromFileNumbers() failed - %s.\n DebugString: %s",
+        s.ToString().c_str(), version->DebugString().c_str());
     return s;
   }
 
@@ -886,7 +924,7 @@ Status DBImplSecondary::CompactWithoutInstallation(
   assert(cfd->compaction_picker());
   c.reset(cfd->compaction_picker()->CompactFiles(
       comp_options, input_files, input.output_level, vstorage,
-      *mutable_cf_options, mutable_db_options_, 0));
+      cfd->GetLatestMutableCFOptions(), mutable_db_options_, 0));
   assert(c != nullptr);
 
   c->FinalizeInputInfo(version);
@@ -917,6 +955,8 @@ Status DBImplSecondary::CompactWithoutInstallation(
       options.canceled ? *options.canceled : kManualCompactionCanceledFalse_,
       input.db_id, db_session_id_, secondary_path_, input, result);
 
+  compaction_job.Prepare();
+
   mutex_.Unlock();
   s = compaction_job.Run();
   mutex_.Lock();
@@ -954,6 +994,7 @@ Status DB::OpenAndCompact(
   DBOptions db_options;
   ConfigOptions config_options;
   config_options.env = override_options.env;
+  config_options.ignore_unknown_options = true;
   std::vector<ColumnFamilyDescriptor> all_column_families;
 
   TEST_SYNC_POINT_CALLBACK(
@@ -1057,5 +1098,4 @@ Status DB::OpenAndCompact(
                         output, override_options);
 }
 
-
 }  // namespace ROCKSDB_NAMESPACE

db/db_impl/db_impl_secondary.h

@@ -5,7 +5,6 @@
 
 #pragma once
 
-
 #include <string>
 #include <vector>
 
@@ -53,7 +52,8 @@ class LogReaderContainer {
     Logger* info_log;
     std::string fname;
     Status* status;  // nullptr if immutable_db_options_.paranoid_checks==false
-    void Corruption(size_t bytes, const Status& s) override {
+    void Corruption(size_t bytes, const Status& s,
+                    uint64_t /*log_number*/ = kMaxSequenceNumber) override {
       ROCKS_LOG_WARN(info_log, "%s%s: dropping %d bytes; %s",
                      (this->status == nullptr ? "(ignoring error) " : ""),
                      fname.c_str(), static_cast<int>(bytes),

db/db_impl/db_impl_write.cc

@@ -12,6 +12,7 @@
 #include "db/error_handler.h"
 #include "db/event_helpers.h"
 #include "logging/logging.h"
+#include "memtable/wbwi_memtable.h"
 #include "monitoring/perf_context_imp.h"
 #include "options/options_helper.h"
 #include "test_util/sync_point.h"
@@ -66,7 +67,7 @@ Status DBImpl::Merge(const WriteOptions& o, ColumnFamilyHandle* column_family,
     return s;
   }
   auto cfh = static_cast_with_check<ColumnFamilyHandleImpl>(column_family);
-  if (!cfh->cfd()->ioptions()->merge_operator) {
+  if (!cfh->cfd()->ioptions().merge_operator) {
     return Status::NotSupported("Provide a merge_operator when opening DB");
   } else {
     return DB::Merge(o, column_family, key, val);
@@ -189,16 +190,137 @@ Status DBImpl::WriteWithCallback(const WriteOptions& write_options,
   return s;
 }
 
-// The main write queue. This is the only write queue that updates LastSequence.
-// When using one write queue, the same sequence also indicates the last
-// published sequence.
+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) {
+  // 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.
+  assert(assigned_seqno.lower_bound > versions_->LastSequence());
+  autovector<ReadOnlyMemTable*> memtables;
+  autovector<ColumnFamilyData*> cfds;
+  InstrumentedMutexLock lock(&mutex_);
+  ColumnFamilySet* cf_set = versions_->GetColumnFamilySet();
+
+  // Create WBWIMemTables
+  for (const auto& [cf_id, stat] : wbwi->GetCFStats()) {
+    ColumnFamilyData* cfd = cf_set->GetColumnFamily(cf_id);
+    if (!cfd) {
+      if (ignore_missing_cf) {
+        continue;
+      }
+      for (auto mem : memtables) {
+        mem->Unref();
+        delete mem;
+      }
+      for (auto cfd_ptr : cfds) {
+        cfd_ptr->UnrefAndTryDelete();
+      }
+      Status s = Status::InvalidArgument(
+          "Invalid column family id from WriteBatchWithIndex: " +
+          std::to_string(cf_id));
+      if (memtable_updated) {
+        s = Status::Corruption(
+            "Part of the write batch is applied. Memtable is in a inconsistent "
+            "state. " +
+            s.ToString());
+        error_handler_.SetBGError(s, BackgroundErrorReason::kMemTable);
+      }
+
+      return s;
+    }
+    WBWIMemTable* wbwi_memtable =
+        new WBWIMemTable(wbwi, cfd->user_comparator(), cf_id, &cfd->ioptions(),
+                         &cfd->GetLatestMutableCFOptions(), stat);
+    wbwi_memtable->Ref();
+    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);
+    memtables.push_back(wbwi_memtable);
+    cfd->Ref();
+    cfds.push_back(cfd);
+  }
+
+  // Stop writes to the DB by entering both write threads
+  WriteThread::Writer nonmem_w;
+  if (two_write_queues_) {
+    nonmem_write_thread_.EnterUnbatched(&nonmem_w, &mutex_);
+  }
+  WaitForPendingWrites();
+
+  // 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()
+    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 (!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
+      // be empty and no new WAL is created for the rest of the calls.
+      assert(i == 0);
+      if (i != 0 || memtable_updated) {
+        // escalate error to non-recoverable
+        s = Status::Corruption(
+            "Part of the write batch is applied. Memtable is in a inconsistent "
+            "state. " +
+            s.ToString());
+        error_handler_.SetBGError(s, BackgroundErrorReason::kMemTable);
+      } else {
+        // SwitchMemtable() already sets appropriate bg error
+      }
+      for (size_t j = i; j < memtables.size(); j++) {
+        memtables[j]->Unref();
+        delete memtables[j];
+      }
+      break;
+    }
+  }
+  for (size_t i = 0; i < cfds.size(); ++i) {
+    if (cfds[i]->UnrefAndTryDelete()) {
+      cfds[i] = nullptr;
+    }
+  }
+
+  // exit the second queue before returning
+  if (two_write_queues_) {
+    nonmem_write_thread_.ExitUnbatched(&nonmem_w);
+  }
+  if (s.ok()) {
+    // Trigger flushes for the new immutable memtables.
+    for (const auto cfd : cfds) {
+      if (cfd == nullptr) {
+        continue;
+      }
+      cfd->imm()->FlushRequested();
+      FlushRequest flush_req;
+      // TODO: a new flush reason for ingesting memtable
+      GenerateFlushRequest({cfd}, FlushReason::kExternalFileIngestion,
+                           &flush_req);
+      EnqueuePendingFlush(flush_req);
+    }
+    MaybeScheduleFlushOrCompaction();
+  }
+  return s;
+}
+
 Status DBImpl::WriteImpl(const WriteOptions& write_options,
                          WriteBatch* my_batch, WriteCallback* callback,
                          UserWriteCallback* user_write_cb, uint64_t* log_used,
                          uint64_t log_ref, bool disable_memtable,
                          uint64_t* seq_used, size_t batch_cnt,
                          PreReleaseCallback* pre_release_callback,
-                         PostMemTableCallback* post_memtable_callback) {
+                         PostMemTableCallback* post_memtable_callback,
+                         std::shared_ptr<WriteBatchWithIndex> wbwi,
+                         uint64_t prep_log) {
   assert(!seq_per_batch_ || batch_cnt != 0);
   assert(my_batch == nullptr || my_batch->Count() == 0 ||
          write_options.protection_bytes_per_key == 0 ||
@@ -287,6 +409,23 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
     return Status::NotSupported(
         "DeleteRange is not compatible with row cache.");
   }
+  if (wbwi) {
+    assert(prep_log > 0);
+    // Used only in WriteCommittedTxn::CommitInternal() with no `callback`.
+    assert(!callback);
+    if (immutable_db_options_.unordered_write) {
+      return Status::NotSupported(
+          "Ingesting WriteBatch does not support unordered_write");
+    }
+    if (immutable_db_options_.enable_pipelined_write) {
+      return Status::NotSupported(
+          "Ingesting WriteBatch does not support pipelined_write");
+    }
+    if (immutable_db_options_.atomic_flush) {
+      return Status::NotSupported(
+          "Ingesting WriteBatch does not support atomic_flush");
+    }
+  }
   // Otherwise IsLatestPersistentState optimization does not make sense
   assert(!WriteBatchInternal::IsLatestPersistentState(my_batch) ||
          disable_memtable);
@@ -344,7 +483,8 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
   PERF_TIMER_GUARD(write_pre_and_post_process_time);
   WriteThread::Writer w(write_options, my_batch, callback, user_write_cb,
                         log_ref, disable_memtable, batch_cnt,
-                        pre_release_callback, post_memtable_callback);
+                        pre_release_callback, post_memtable_callback,
+                        /*_ingest_wbwi=*/wbwi != nullptr);
   StopWatch write_sw(immutable_db_options_.clock, stats_, DB_WRITE);
 
   write_thread_.JoinBatchGroup(&w);
@@ -441,9 +581,13 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
   TEST_SYNC_POINT("DBImpl::WriteImpl:BeforeLeaderEnters");
   last_batch_group_size_ =
       write_thread_.EnterAsBatchGroupLeader(&w, &write_group);
+  if (wbwi) {
+    assert(write_group.size == 1);
+  }
 
   IOStatus io_s;
   Status pre_release_cb_status;
+  size_t seq_inc = 0;
   if (status.ok()) {
     // Rules for when we can update the memtable concurrently
     // 1. supported by memtable
@@ -494,10 +638,25 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
     // Note about seq_per_batch_: either disableWAL is set for the entire write
     // group or not. In either case we inc seq for each write batch with no
     // failed callback. This means that there could be a batch with
-    // disalbe_memtable in between; although we do not write this batch to
+    // disable_memtable in between; although we do not write this batch to
     // memtable it still consumes a seq. Otherwise, if !seq_per_batch_, we inc
     // the seq per valid written key to mem.
-    size_t seq_inc = seq_per_batch_ ? valid_batches : total_count;
+    seq_inc = seq_per_batch_ ? valid_batches : total_count;
+    if (wbwi) {
+      // Reserve sequence numbers for the ingested memtable. We need to reserve
+      // at lease this amount for recovery. During recovery,
+      // transactions do not commit by ingesting WBWI. The sequence number
+      // associated with the commit entry in WAL is used as the starting
+      // sequence number for inserting into memtable. We need to reserve
+      // enough sequence numbers here (at least the number of operations
+      // in write batch) to assign to memtable entries for this transaction.
+      // This prevents updates in different transactions from using out-of-order
+      // sequence numbers or the same key+seqno.
+      //
+      // WBWI ingestion requires not grouping writes, so we don't need to
+      // consider incrementing sequence number for WBWI from other writers.
+      seq_inc += wbwi->GetWriteBatch()->Count();
+    }
 
     const bool concurrent_update = two_write_queues_;
     // Update stats while we are an exclusive group leader, so we know
@@ -555,6 +714,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
     assert(last_sequence != kMaxSequenceNumber);
     const SequenceNumber current_sequence = last_sequence + 1;
     last_sequence += seq_inc;
+    // Seqno assigned to this write are [current_sequence, last_sequence]
 
     if (log_context.need_log_sync) {
       VersionEdit synced_wals;
@@ -674,6 +834,31 @@ 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()) {
+      // 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.
+      // wbwi gets the rest `wbwi_count` sequence numbers.
+      assert(seq_inc == memtable_update_count + wbwi_count);
+      assert(wbwi_count > 0);
+      assert(last_sequence != kMaxSequenceNumber);
+      SequenceNumber lb = last_sequence + 1 - wbwi_count;
+      SequenceNumber ub = last_sequence;
+      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);
+    }
+  }
+
   if (should_exit_batch_group) {
     if (status.ok()) {
       for (auto* tmp_w : write_group) {
@@ -1210,7 +1395,6 @@ void DBImpl::MemTableInsertStatusCheck(const Status& status) {
   if (!status.ok()) {
     mutex_.Lock();
     assert(!error_handler_.IsBGWorkStopped());
-    // Maybe change the return status to void?
     error_handler_.SetBGError(status, BackgroundErrorReason::kMemTable);
     mutex_.Unlock();
   }
@@ -1380,7 +1564,8 @@ IOStatus DBImpl::WriteToWAL(const WriteBatch& merged_batch,
                             const WriteOptions& write_options,
                             log::Writer* log_writer, uint64_t* log_used,
                             uint64_t* log_size,
-                            LogFileNumberSize& log_file_number_size) {
+                            LogFileNumberSize& log_file_number_size,
+                            SequenceNumber sequence) {
   assert(log_size != nullptr);
 
   Slice log_entry = WriteBatchInternal::Contents(&merged_batch);
@@ -1406,7 +1591,7 @@ IOStatus DBImpl::WriteToWAL(const WriteBatch& merged_batch,
   if (!io_s.ok()) {
     return io_s;
   }
-  io_s = log_writer->AddRecord(write_options, log_entry);
+  io_s = log_writer->AddRecord(write_options, log_entry, sequence);
 
   if (UNLIKELY(needs_locking)) {
     log_write_mutex_.Unlock();
@@ -1456,7 +1641,7 @@ IOStatus DBImpl::WriteToWAL(const WriteThread::WriteGroup& write_group,
   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);
+                    &log_size, log_file_number_size, sequence);
   if (to_be_cached_state) {
     cached_recoverable_state_ = *to_be_cached_state;
     cached_recoverable_state_empty_ = false;
@@ -1582,7 +1767,7 @@ IOStatus DBImpl::ConcurrentWriteToWAL(
   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);
+                    &log_size, log_file_number_size, sequence);
   if (to_be_cached_state) {
     cached_recoverable_state_ = *to_be_cached_state;
     cached_recoverable_state_empty_ = false;
@@ -2204,16 +2389,13 @@ void DBImpl::NotifyOnMemTableSealed(ColumnFamilyData* /*cfd*/,
   mutex_.Lock();
 }
 
-// REQUIRES: mutex_ is held
-// REQUIRES: this thread is currently at the front of the writer queue
-// REQUIRES: this thread is currently at the front of the 2nd writer queue if
-// two_write_queues_ is true (This is to simplify the reasoning.)
-Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
+Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context,
+                              ReadOnlyMemTable* new_imm,
+                              SequenceNumber last_seqno) {
   mutex_.AssertHeld();
   assert(lock_wal_count_ == 0);
 
   // TODO: plumb Env::IOActivity, Env::IOPriority
-  const ReadOptions read_options;
   const WriteOptions write_options;
 
   log::Writer* new_log = nullptr;
@@ -2246,16 +2428,19 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
   }
   uint64_t new_log_number =
       creating_new_log ? versions_->NewFileNumber() : logfile_number_;
-  const MutableCFOptions mutable_cf_options = *cfd->GetLatestMutableCFOptions();
+  // For use outside of holding DB mutex
+  const MutableCFOptions mutable_cf_options_copy =
+      cfd->GetLatestMutableCFOptions();
 
   // Set memtable_info for memtable sealed callback
+  // TODO: memtable_info for `new_imm`
   MemTableInfo memtable_info;
   memtable_info.cf_name = cfd->GetName();
   memtable_info.first_seqno = cfd->mem()->GetFirstSequenceNumber();
   memtable_info.earliest_seqno = cfd->mem()->GetEarliestSequenceNumber();
   memtable_info.num_entries = cfd->mem()->NumEntries();
   memtable_info.num_deletes = cfd->mem()->NumDeletion();
-  if (!cfd->ioptions()->persist_user_defined_timestamps &&
+  if (!cfd->ioptions().persist_user_defined_timestamps &&
       cfd->user_comparator()->timestamp_size() > 0) {
     const Slice& newest_udt = cfd->mem()->GetNewestUDT();
     memtable_info.newest_udt.assign(newest_udt.data(), newest_udt.size());
@@ -2264,20 +2449,41 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
   // flush happens before logging, but that should be ok.
   int num_imm_unflushed = cfd->imm()->NumNotFlushed();
   const auto preallocate_block_size =
-      GetWalPreallocateBlockSize(mutable_cf_options.write_buffer_size);
+      GetWalPreallocateBlockSize(mutable_cf_options_copy.write_buffer_size);
   mutex_.Unlock();
   if (creating_new_log) {
+    PredecessorWALInfo info;
+    log_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();
     // 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,
-                     preallocate_block_size, &new_log);
+                     preallocate_block_size, info, &new_log);
     if (s.ok()) {
       s = io_s;
     }
   }
   if (s.ok()) {
-    SequenceNumber seq = versions_->LastSequence();
-    new_mem = cfd->ConstructNewMemtable(mutable_cf_options, seq);
+    // FIXME: from the comment for GetEarliestSequenceNumber(), any key with
+    //  seqno >= earliest_seqno should be in this or later memtable. This means
+    //  we should use LastSequence() + 1 or last_seqno + 1 here. And it needs to
+    //  be incremented with file ingestion and other operations that consumes
+    //  sequence number.
+    SequenceNumber seq;
+    if (new_imm) {
+      assert(last_seqno > versions_->LastSequence());
+      seq = last_seqno;
+    } else {
+      seq = versions_->LastSequence();
+    }
+    new_mem = cfd->ConstructNewMemtable(mutable_cf_options_copy,
+                                        /*earliest_seq=*/seq);
     context->superversion_context.NewSuperVersion();
 
     ROCKS_LOG_INFO(immutable_db_options_.info_log,
@@ -2359,6 +2565,8 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
         versions_->PreComputeMinLogNumberWithUnflushedData(logfile_number_);
     if (min_wal_number_to_keep >
         versions_->GetWalSet().GetMinWalNumberToKeep()) {
+      // TODO: plumb Env::IOActivity, Env::IOPriority
+      const ReadOptions read_options;
       // Get a snapshot of the empty column families.
       // LogAndApply may release and reacquire db
       // mutex, during that period, column family may become empty (e.g. its
@@ -2416,10 +2624,21 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
   cfd->mem()->SetNextLogNumber(logfile_number_);
   assert(new_mem != nullptr);
   cfd->imm()->Add(cfd->mem(), &context->memtables_to_free_);
+  if (new_imm) {
+    // Need to assign memtable id here before SetMemtable() below assigns id to
+    // the new live memtable
+    cfd->AssignMemtableID(new_imm);
+    // NOTE: new_imm and cfd->mem() references the same WAL and has the same
+    // NextLogNumber(). They should be flushed together. For non-atomic-flush,
+    // 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_);
+    cfd->imm()->Add(new_imm, &context->memtables_to_free_);
+  }
   new_mem->Ref();
   cfd->SetMemtable(new_mem);
-  InstallSuperVersionAndScheduleWork(cfd, &context->superversion_context,
-                                     mutable_cf_options);
+  InstallSuperVersionAndScheduleWork(cfd, &context->superversion_context);
 
   // Notify client that memtable is sealed, now that we have successfully
   // installed a new memtable
@@ -2428,6 +2647,9 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
   // that is okay.  If we did, it most likely means that s was already an error.
   // In any case, ignore any unchecked error for i_os here.
   io_s.PermitUncheckedError();
+  // We guarantee that if a non-ok status is returned, `new_imm` was not added
+  // to the db.
+  assert(s.ok());
   return s;
 }
 

db/db_io_failure_test.cc

@@ -97,12 +97,20 @@ class CorruptionFS : public FileSystemWrapper {
         for (size_t i = 0; i < num_reqs; ++i) {
           FSReadRequest& req = reqs[i];
           if (fs_.fs_buffer_) {
-            FSAllocationPtr buffer(new char[req.len], [](void* ptr) {
-              delete[] static_cast<char*>(ptr);
-            });
-            req.fs_scratch = std::move(buffer);
+            // See https://github.com/facebook/rocksdb/pull/13195 for why we
+            // want to set up our test implementation for FSAllocationPtr this
+            // way.
+            char* internalData = new char[req.len];
             req.status = Read(req.offset, req.len, options, &req.result,
-                              static_cast<char*>(req.fs_scratch.get()), dbg);
+                              internalData, dbg);
+
+            Slice* internalSlice = new Slice(internalData, req.len);
+            FSAllocationPtr internalPtr(internalSlice, [](void* ptr) {
+              delete[] static_cast<const char*>(
+                  static_cast<Slice*>(ptr)->data_);
+              delete static_cast<Slice*>(ptr);
+            });
+            req.fs_scratch = std::move(internalPtr);
           } else {
             req.status = Read(req.offset, req.len, options, &req.result,
                               req.scratch, dbg);

db/db_iter.cc

@@ -581,8 +581,14 @@ bool DBIter::FindNextUserEntryInternal(bool skipping_saved_key,
     } else {
       iter_.Next();
     }
+
     // This could be a long-running operation due to tombstones, etc.
-    ROCKSDB_THREAD_YIELD_HOOK();
+    bool aborted = ROCKSDB_THREAD_YIELD_CHECK_ABORT();
+    if (aborted) {
+      valid_ = false;
+      status_ = Status::Aborted("Query abort.");
+      return false;
+    }
   } while (iter_.Valid());
 
   valid_ = false;

db/db_iterator_test.cc

@@ -2544,6 +2544,145 @@ TEST_P(DBIteratorTest, RefreshWithSnapshot) {
   ASSERT_OK(db_->Close());
 }
 
+TEST_P(DBIteratorTest, AutoRefreshIterator) {
+  constexpr int kNumKeys = 1000;
+  Options options = CurrentOptions();
+  options.disable_auto_compactions = true;
+  for (const DBIter::Direction direction :
+       {DBIter::kForward, DBIter::kReverse}) {
+    for (const bool auto_refresh_enabled : {false, true}) {
+      for (const bool explicit_snapshot : {false, true}) {
+        DestroyAndReopen(options);
+        // Multi dimensional iterator:
+        //
+        // L0 (level iterator): [key000000]
+        // L1 (table iterator): [key000001]
+        // Memtable           : [key000000, key000999]
+        for (int i = 0; i < kNumKeys + 2; i++) {
+          ASSERT_OK(Put(Key(i % kNumKeys), "val" + std::to_string(i)));
+          if (i <= 1) {
+            ASSERT_OK(Flush());
+          }
+          if (i == 0) {
+            MoveFilesToLevel(1);
+          }
+        }
+
+        ReadOptions read_options;
+        std::unique_ptr<ManagedSnapshot> snapshot = nullptr;
+        if (explicit_snapshot) {
+          snapshot = std::make_unique<ManagedSnapshot>(db_);
+        }
+        read_options.snapshot =
+            explicit_snapshot ? snapshot->snapshot() : nullptr;
+        read_options.auto_refresh_iterator_with_snapshot = auto_refresh_enabled;
+        std::unique_ptr<Iterator> iter(NewIterator(read_options));
+
+        int trigger_compact_on_it = kNumKeys / 2;
+
+        // This update should NOT be visible from the iterator.
+        ASSERT_OK(Put(Key(trigger_compact_on_it + 1), "new val"));
+
+        ASSERT_EQ(1, NumTableFilesAtLevel(1));
+        ASSERT_EQ(1, NumTableFilesAtLevel(0));
+
+        uint64_t all_memtables_size_before_refresh;
+        uint64_t all_memtables_size_after_refresh;
+
+        std::string prop_value;
+        ASSERT_OK(iter->GetProperty("rocksdb.iterator.super-version-number",
+                                    &prop_value));
+        int superversion_number = std::stoi(prop_value);
+
+        std::vector<LiveFileMetaData> old_files;
+        db_->GetLiveFilesMetaData(&old_files);
+
+        int expected_next_key_int;
+        if (direction == DBIter::kForward) {
+          expected_next_key_int = 0;
+          iter->SeekToFirst();
+        } else {  // DBIter::kReverse
+          expected_next_key_int = kNumKeys - 1;
+          iter->SeekToLast();
+        }
+
+        int it_num = 0;
+        std::unordered_map<std::string, std::string> kvs;
+        while (iter->Valid()) {
+          ASSERT_OK(iter->status());
+          it_num++;
+          if (it_num == trigger_compact_on_it) {
+            // Bump the superversion by manually scheduling flush + compaction.
+            ASSERT_OK(Flush());
+            ASSERT_OK(dbfull()->CompactRange(CompactRangeOptions(), nullptr,
+                                             nullptr));
+            ASSERT_OK(dbfull()->TEST_WaitForBackgroundWork());
+
+            // For accuracy, capture the memtables size right before consecutive
+            // iterator call to Next() will update its' stale superversion ref.
+            dbfull()->GetIntProperty("rocksdb.size-all-mem-tables",
+                                     &all_memtables_size_before_refresh);
+          }
+
+          if (it_num == trigger_compact_on_it + 1) {
+            dbfull()->GetIntProperty("rocksdb.size-all-mem-tables",
+                                     &all_memtables_size_after_refresh);
+            ASSERT_OK(iter->GetProperty("rocksdb.iterator.super-version-number",
+                                        &prop_value));
+            uint64_t new_superversion_number = std::stoi(prop_value);
+            Status expected_status_for_preexisting_files;
+            if (auto_refresh_enabled && explicit_snapshot) {
+              // Iterator is expected to detect its' superversion staleness.
+              ASSERT_LT(superversion_number, new_superversion_number);
+              // ... and since our iterator was the only reference to that very
+              // superversion, we expect most of the active memory to be
+              // returned upon automatical iterator refresh.
+              ASSERT_GT(all_memtables_size_before_refresh,
+                        all_memtables_size_after_refresh);
+              expected_status_for_preexisting_files = Status::NotFound();
+            } else {
+              ASSERT_EQ(superversion_number, new_superversion_number);
+              ASSERT_EQ(all_memtables_size_after_refresh,
+                        all_memtables_size_before_refresh);
+              expected_status_for_preexisting_files = Status::OK();
+            }
+
+            for (const auto& file : old_files) {
+              ASSERT_EQ(env_->FileExists(file.db_path + "/" + file.name),
+                        expected_status_for_preexisting_files);
+            }
+          }
+
+          // Ensure we're visiting the keys in desired order and at most once!
+          ASSERT_EQ(IdFromKey(iter->key().ToString()), expected_next_key_int);
+          kvs[iter->key().ToString()] = iter->value().ToString();
+
+          if (direction == DBIter::kForward) {
+            iter->Next();
+            expected_next_key_int++;
+          } else {
+            iter->Prev();
+            expected_next_key_int--;
+          }
+        }
+        ASSERT_OK(iter->status());
+
+        // Data validation.
+        ASSERT_EQ(kvs.size(), kNumKeys);
+        for (int i = 0; i < kNumKeys; i++) {
+          auto kv = kvs.find(Key(i));
+          ASSERT_TRUE(kv != kvs.end());
+          int val = i;
+          if (i <= 1) {
+            val += kNumKeys;
+          }
+          ASSERT_EQ(kv->second, "val" + std::to_string(val));
+        }
+      }
+    }
+  }
+}
+
 TEST_P(DBIteratorTest, CreationFailure) {
   SyncPoint::GetInstance()->SetCallBack(
       "DBImpl::NewInternalIterator:StatusCallback", [](void* arg) {

db/db_log_iter_test.cc

@@ -332,7 +332,6 @@ TEST_F(DBTestXactLogIterator, TransactionLogIteratorBlobs) {
 }
 }  // namespace ROCKSDB_NAMESPACE
 
-
 int main(int argc, char** argv) {
   ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
   ::testing::InitGoogleTest(&argc, argv);

db/db_merge_operand_test.cc

@@ -5,11 +5,11 @@
 
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
+#include "rocksdb/merge_operator.h"
 #include "rocksdb/perf_context.h"
 #include "rocksdb/utilities/debug.h"
 #include "table/block_based/block_builder.h"
 #include "test_util/sync_point.h"
-#include "rocksdb/merge_operator.h"
 #include "utilities/fault_injection_env.h"
 #include "utilities/merge_operators.h"
 #include "utilities/merge_operators/sortlist.h"
@@ -120,8 +120,11 @@ TEST_F(DBMergeOperandTest, FlushedMergeOperandReadAfterFreeBug) {
 
 TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {
   Options options = CurrentOptions();
+
+  int limit = 2;
   // Use only the latest two merge operands.
-  options.merge_operator = std::make_shared<LimitedStringAppendMergeOp>(2, ',');
+  options.merge_operator =
+      std::make_shared<LimitedStringAppendMergeOp>(limit, ',');
   Reopen(options);
   int num_records = 4;
   int number_of_operands = 0;
@@ -263,6 +266,7 @@ TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {
   ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
                                   "k3.2", values.data(), &merge_operands_info,
                                   &number_of_operands));
+  ASSERT_EQ(number_of_operands, 2);
   ASSERT_EQ(values[0], "cd");
   ASSERT_EQ(values[1], "de");
 
@@ -280,6 +284,7 @@ TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {
   ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
                                   "k4", values.data(), &merge_operands_info,
                                   &number_of_operands));
+  ASSERT_EQ(number_of_operands, 4);
   ASSERT_EQ(values[0], "ba");
   ASSERT_EQ(values[1], "cb");
   ASSERT_EQ(values[2], "dc");
@@ -299,9 +304,26 @@ TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {
   ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
                                   "k5", values.data(), &merge_operands_info,
                                   &number_of_operands));
+  ASSERT_EQ(number_of_operands, 4);
   ASSERT_EQ(values[0], "remember");
   ASSERT_EQ(values[1], "i");
   ASSERT_EQ(values[2], "am");
+  ASSERT_EQ(values[3], "rocks");
+
+  // GetMergeOperands() in ReadOnly DB
+  ASSERT_OK(Merge("k6", "better"));
+  ASSERT_OK(Merge("k6", "call"));
+  ASSERT_OK(Merge("k6", "saul"));
+
+  ASSERT_OK(ReadOnlyReopen(options));
+  std::vector<PinnableSlice> readonly_values(num_records);
+  ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
+                                  "k6", readonly_values.data(),
+                                  &merge_operands_info, &number_of_operands));
+  ASSERT_EQ(number_of_operands, 3);
+  ASSERT_EQ(readonly_values[0], "better");
+  ASSERT_EQ(readonly_values[1], "call");
+  ASSERT_EQ(readonly_values[2], "saul");
 }
 
 TEST_F(DBMergeOperandTest, BlobDBGetMergeOperandsBasic) {
@@ -577,6 +599,42 @@ TEST_F(DBMergeOperandTest, GetMergeOperandsBaseDeletionInImmMem) {
   }
 }
 
+TEST_F(DBMergeOperandTest, GetMergeOperandCallbackStopAtImm) {
+  Options options = CurrentOptions();
+  options.max_write_buffer_number = 10;
+  options.merge_operator = MergeOperators::CreateStringAppendOperator();
+  DestroyAndReopen(options);
+
+  Random rnd(301);
+  ASSERT_OK(db_->PauseBackgroundWork());
+  ASSERT_OK(Merge("key", "v1"));
+  ASSERT_OK(dbfull()->TEST_SwitchMemtable());
+  // Keep this merge in an immutable memtable
+  uint64_t num_imm = 0;
+  ASSERT_TRUE(
+      db_->GetIntProperty(DB::Properties::kNumImmutableMemTable, &num_imm));
+  ASSERT_EQ(num_imm, 1);
+  ASSERT_OK(Merge("key", "v2"));
+
+  std::vector<PinnableSlice> merge_operands(2);
+  GetMergeOperandsOptions merge_operands_info;
+  merge_operands_info.expected_max_number_of_operands = 2;
+  int num_fetched = 0;
+  merge_operands_info.continue_cb = [&num_fetched](Slice /* value */) {
+    num_fetched++;
+    // Stop in the first immutable memtable.
+    return num_fetched < 2;
+  };
+  int num_merge_operands = 0;
+  ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
+                                  "key", merge_operands.data(),
+                                  &merge_operands_info, &num_merge_operands));
+  ASSERT_EQ(2, num_merge_operands);
+  ASSERT_EQ(2, num_fetched);
+
+  ASSERT_EQ("v1", merge_operands[0]);
+  ASSERT_EQ("v2", merge_operands[1]);
+}
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/db_merge_operator_test.cc

@@ -120,6 +120,10 @@ TEST_F(DBMergeOperatorTest, LimitMergeOperands) {
   ASSERT_OK(Merge("k3", "de"));
   ASSERT_OK(db_->Get(ReadOptions(), "k3", &value));
   ASSERT_EQ(value, "cd,de");
+  // Tests that merge operands reach exact limit at memtable.
+  ASSERT_OK(Merge("k3", "fg"));
+  ASSERT_OK(db_->Get(ReadOptions(), "k3", &value));
+  ASSERT_EQ(value, "de,fg");
 
   // All K4 values are in different levels
   ASSERT_OK(Merge("k4", "ab"));

db/db_options_test.cc

@@ -1451,7 +1451,6 @@ TEST_F(DBOptionsTest, ChangeCompression) {
   SyncPoint::GetInstance()->DisableProcessing();
 }
 
-
 TEST_F(DBOptionsTest, BottommostCompressionOptsWithFallbackType) {
   // Verify the bottommost compression options still take effect even when the
   // bottommost compression type is left at its default value. Verify for both
@@ -1596,6 +1595,69 @@ TEST_F(DBOptionsTest, TempOptionsFailTest) {
   ASSERT_FALSE(found_temp_file);
 }
 
+TEST_F(DBOptionsTest, SetOptionsNoManifestWrite) {
+  ASSERT_OK(Put("x", "x"));
+  ASSERT_OK(Flush());
+
+  // In addition to checking manifest file, we want to ensure that SetOptions
+  // is essentially atomic, without releasing the DB mutex between applying
+  // the options to the cfd and installing new Version and SuperVersion. We
+  // probabilistically verify that by attempting to catch an inconsistency.
+  auto* const cfd =
+      static_cast<ColumnFamilyHandleImpl*>(db_->DefaultColumnFamily())->cfd();
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+  std::optional<std::thread> t;
+  SyncPoint::GetInstance()->SetCallBack(
+      "VersionSet::LogAndApply:WakeUpAndNotDone", [&](void* arg) {
+        auto* mu = static_cast<InstrumentedMutex*>(arg);
+        // Option not yet modified
+        ASSERT_FALSE(cfd->GetLatestMutableCFOptions().disable_auto_compactions);
+        ASSERT_FALSE(
+            cfd->current()->GetMutableCFOptions().disable_auto_compactions);
+        ASSERT_FALSE(
+            cfd->GetCurrentMutableCFOptions().disable_auto_compactions);
+        t = std::thread([mu, cfd]() {
+          InstrumentedMutexLock l(mu);
+          // Assuming above correctness, we can only acquire the mutex after
+          // options fully installed.
+          ASSERT_TRUE(
+              cfd->GetLatestMutableCFOptions().disable_auto_compactions);
+          ASSERT_TRUE(
+              cfd->current()->GetMutableCFOptions().disable_auto_compactions);
+          ASSERT_TRUE(
+              cfd->GetCurrentMutableCFOptions().disable_auto_compactions);
+        });
+      });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  // Baseline manifest file info
+  std::vector<std::string> live_files;
+  uint64_t orig_manifest_file_size;
+  ASSERT_OK(dbfull()->GetLiveFiles(live_files, &orig_manifest_file_size));
+  uint64_t orig_manifest_file_num = dbfull()->TEST_Current_Manifest_FileNo();
+
+  // Although this test mostly concerns SetOptions, we also include SetDBOptions
+  // just for the added scope
+  ASSERT_OK(db_->SetDBOptions({{"max_open_files", "100"}}));
+  ASSERT_OK(db_->SetOptions({{"disable_auto_compactions", "true"}}));
+
+  // Verify that our above check was activated and completed
+  ASSERT_TRUE(t.has_value());
+  t->join();
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+
+  // Verify manifest was not written to
+  uint64_t new_manifest_file_size;
+  ASSERT_OK(dbfull()->GetLiveFiles(live_files, &new_manifest_file_size));
+  uint64_t new_manifest_file_num = dbfull()->TEST_Current_Manifest_FileNo();
+  ASSERT_EQ(orig_manifest_file_num, new_manifest_file_num);
+  ASSERT_EQ(orig_manifest_file_size, new_manifest_file_size);
+
+  ASSERT_EQ(Get("x"), "x");
+}
+
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/db_rate_limiter_test.cc

@@ -226,7 +226,6 @@ TEST_P(DBRateLimiterOnReadTest, Iterator) {
   ASSERT_EQ(expected, options_.rate_limiter->GetTotalRequests(Env::IO_USER));
 }
 
-
 TEST_P(DBRateLimiterOnReadTest, VerifyChecksum) {
   if (use_direct_io_ && !IsDirectIOSupported()) {
     return;
@@ -271,7 +270,6 @@ TEST_P(DBRateLimiterOnReadTest, VerifyFileChecksums) {
   ASSERT_EQ(expected, options_.rate_limiter->GetTotalRequests(Env::IO_USER));
 }
 
-
 class DBRateLimiterOnWriteTest : public DBTestBase {
  public:
   explicit DBRateLimiterOnWriteTest()

db/db_secondary_test.cc

@@ -14,10 +14,9 @@
 #include "rocksdb/utilities/transaction_db.h"
 #include "test_util/sync_point.h"
 #include "test_util/testutil.h"
-#include "utilities/fault_injection_env.h"
+#include "utilities/merge_operators/string_append/stringappend2.h"
 
 namespace ROCKSDB_NAMESPACE {
-
 class DBSecondaryTestBase : public DBBasicTestWithTimestampBase {
  public:
   explicit DBSecondaryTestBase(const std::string& dbname)
@@ -331,6 +330,43 @@ TEST_F(DBSecondaryTest, InternalCompactionMultiLevels) {
   //  cfh, input1, &result));
 }
 
+TEST_F(DBSecondaryTest, GetMergeOperands) {
+  Options options;
+  options.merge_operator = MergeOperators::CreateStringAppendOperator();
+  options.env = env_;
+  Reopen(options);
+
+  ASSERT_OK(Merge("k1", "v1"));
+  ASSERT_OK(Merge("k1", "v2"));
+  ASSERT_OK(Merge("k1", "v3"));
+  ASSERT_OK(Merge("k1", "v4"));
+
+  options.max_open_files = -1;
+  OpenSecondary(options);
+
+  ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
+
+  int num_records = 4;
+  int number_of_operands = 0;
+  std::vector<PinnableSlice> values(num_records);
+  GetMergeOperandsOptions merge_operands_info;
+  merge_operands_info.expected_max_number_of_operands = num_records;
+
+  auto cfh = db_secondary_->DefaultColumnFamily();
+
+  const Status s = db_secondary_->GetMergeOperands(
+      ReadOptions(), cfh, "k1", values.data(), &merge_operands_info,
+      &number_of_operands);
+  ASSERT_NOK(s);
+  ASSERT_TRUE(s.IsMergeInProgress());
+
+  ASSERT_EQ(number_of_operands, 4);
+  ASSERT_EQ(values[0].ToString(), "v1");
+  ASSERT_EQ(values[1].ToString(), "v2");
+  ASSERT_EQ(values[2].ToString(), "v3");
+  ASSERT_EQ(values[3].ToString(), "v4");
+}
+
 TEST_F(DBSecondaryTest, InternalCompactionCompactedFiles) {
   Options options;
   options.env = env_;

db/db_sst_test.cc

@@ -1991,7 +1991,6 @@ TEST_F(DBSSTTest, DBWithSFMForBlobFilesAtomicFlush) {
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
 }
 
-
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/db_statistics_test.cc

@@ -198,7 +198,6 @@ TEST_F(DBStatisticsTest, ExcludeTickers) {
   ASSERT_GT(options.statistics->getTickerCount(BYTES_READ), 0);
 }
 
-
 TEST_F(DBStatisticsTest, VerifyChecksumReadStat) {
   Options options = CurrentOptions();
   options.file_checksum_gen_factory = GetFileChecksumGenCrc32cFactory();

db/db_table_properties_test.cc

@@ -25,7 +25,6 @@
 #include "util/atomic.h"
 #include "util/random.h"
 
-
 namespace ROCKSDB_NAMESPACE {
 
 // A helper function that ensures the table properties returned in
@@ -744,7 +743,6 @@ INSTANTIATE_TEST_CASE_P(DBTablePropertiesTest, DBTablePropertiesTest,
 
 }  // namespace ROCKSDB_NAMESPACE
 
-
 int main(int argc, char** argv) {
   ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
   ::testing::InitGoogleTest(&argc, argv);

db/db_tailing_iter_test.cc

@@ -703,7 +703,6 @@ TEST_P(DBTestTailingIterator, SeekToFirstWithUpperBoundBug) {
 
 }  // namespace ROCKSDB_NAMESPACE
 
-
 int main(int argc, char** argv) {
   ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
   ::testing::InitGoogleTest(&argc, argv);

db/db_test.cc

@@ -1161,7 +1161,6 @@ class DelayFilterFactory : public CompactionFilterFactory {
 };
 }  // anonymous namespace
 
-
 static std::string CompressibleString(Random* rnd, int len) {
   std::string r;
   test::CompressibleString(rnd, 0.8, len, &r);
@@ -3410,8 +3409,6 @@ class ModelDB : public DB {
     return Status::NotSupported();
   }
 
-  Status DeleteFile(std::string /*name*/) override { return Status::OK(); }
-
   Status GetUpdatesSince(
       ROCKSDB_NAMESPACE::SequenceNumber,
       std::unique_ptr<ROCKSDB_NAMESPACE::TransactionLogIterator>*,
@@ -3715,6 +3712,139 @@ TEST_F(DBTest, BlockBasedTablePrefixIndexTest) {
   ASSERT_EQ("v2", Get("k2"));
 }
 
+TEST_F(DBTest, SetOptionsEffectiveInSuperVersions) {
+  // Basically, to test the SetOptions take effect with (and only with)
+  // new SuperVersions, and remain in effect through some things like
+  // flush and compaction, we use some queries that depend on the current
+  // prefix extractor.
+  //
+  // Making the semantics of read options dependent on the current state of
+  // mutable options is kind of an anti-pattern that prefix_seek_opt_in_only
+  // is helping to phase out. However, this is useful for rather directly
+  // testing the expected behavior of mutable options handling.
+  ReadOptions ropts;
+  ropts.prefix_same_as_start = true;
+  Options options = CurrentOptions();
+  options.prefix_extractor.reset(NewFixedPrefixTransform(5));
+  options.prefix_seek_opt_in_only = false;
+  Reopen(options);
+
+  ASSERT_OK(Put("goat1", "g1"));
+  ASSERT_OK(Put("goat2", "g2"));
+
+  std::unique_ptr<Iterator> iter(db_->NewIterator(ropts));
+
+  auto VerifyTransform4 = [&](int caller_line) {
+    SCOPED_TRACE("Called from " + std::to_string(caller_line));
+    // Nothing with this prefix
+    iter->Seek("game1");
+    ASSERT_OK(iter->status());
+    ASSERT_FALSE(iter->Valid());
+
+    iter->Seek("goat1");
+    ASSERT_OK(iter->status());
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ("goat1", iter->key());
+    iter->Next();
+    ASSERT_OK(iter->status());
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ("goat2", iter->key());
+  };
+
+  auto VerifyTransform5 = [&](int caller_line) {
+    SCOPED_TRACE("Called from " + std::to_string(caller_line));
+    // Nothing with this prefix
+    iter->Seek("game1");
+    ASSERT_OK(iter->status());
+    ASSERT_FALSE(iter->Valid());
+
+    iter->Seek("goat1");
+    ASSERT_OK(iter->status());
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ("goat1", iter->key());
+    iter->Next();
+    ASSERT_OK(iter->status());
+    ASSERT_FALSE(iter->Valid());
+
+    iter->Seek("goat2");
+    ASSERT_OK(iter->status());
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ("goat2", iter->key());
+    iter->Next();
+    ASSERT_OK(iter->status());
+    ASSERT_FALSE(iter->Valid());
+  };
+
+  for (int i = 0;; ++i) {
+    SCOPED_TRACE("Iteration " + std::to_string(i));
+    // Baseline
+    VerifyTransform5(__LINE__);
+
+    if (i == 0) {
+      // Test a "normal" change with nothing happening in parallel
+      ASSERT_OK(db_->SetOptions({{"prefix_extractor", "fixed:4"}}));
+
+      // Iterator still uses old superversion
+      VerifyTransform5(__LINE__);
+
+      // Refresh updates the SuperVersion
+      ASSERT_OK(iter->Refresh());
+    } else if (i == 1) {
+      // Test a setting change in parallel with flush
+      iter = nullptr;
+      SyncPoint::GetInstance()->DisableProcessing();
+      SyncPoint::GetInstance()->ClearAllCallBacks();
+      SyncPoint::GetInstance()->SetCallBack(
+          "FlushJob::WriteLevel0Table:num_memtables", [&](void*) {
+            // During flush, without DB mutex held
+            ASSERT_OK(db_->SetOptions({{"prefix_extractor", "fixed:4"}}));
+            iter.reset(db_->NewIterator(ropts));
+            VerifyTransform4(__LINE__);
+          });
+      SyncPoint::GetInstance()->EnableProcessing();
+      ASSERT_OK(Flush());
+      SyncPoint::GetInstance()->DisableProcessing();
+      SyncPoint::GetInstance()->ClearAllCallBacks();
+      // Callback was called
+      ASSERT_NE(iter, nullptr);
+    } else if (i == 2) {
+      // Test a setting change in parallel with compaction
+      iter = nullptr;
+      SyncPoint::GetInstance()->DisableProcessing();
+      SyncPoint::GetInstance()->ClearAllCallBacks();
+      SyncPoint::GetInstance()->SetCallBack(
+          "CompactionJob::Run():EndStatusSet", [&](void*) {
+            // During compaction, without DB mutex held
+            ASSERT_OK(db_->SetOptions({{"prefix_extractor", "fixed:4"}}));
+            iter.reset(db_->NewIterator(ropts));
+            VerifyTransform4(__LINE__);
+          });
+      SyncPoint::GetInstance()->EnableProcessing();
+      // Need data overlapping that L0 file to prevent trivial move
+      ASSERT_OK(Put("aaaaa", "a"));
+      ASSERT_OK(Put("zzzzz", "a"));
+      ASSERT_OK(CompactRange({}, {}, {}));
+      SyncPoint::GetInstance()->DisableProcessing();
+      SyncPoint::GetInstance()->ClearAllCallBacks();
+      // Callback was called
+      ASSERT_NE(iter, nullptr);
+    } else {
+      break;
+    }
+    // Change has taken effect
+    VerifyTransform4(__LINE__);
+    // Same after a new iterator (in case a new SuperVersion reverted the
+    // setting)
+    iter.reset(db_->NewIterator(ropts));
+    VerifyTransform4(__LINE__);
+
+    // Back to baseline setting
+    ASSERT_OK(db_->SetOptions({{"prefix_extractor", "fixed:5"}}));
+    // New iterator uses latest SuperVersion
+    iter.reset(db_->NewIterator(ropts));
+  }
+}
+
 TEST_F(DBTest, BlockBasedTablePrefixHashIndexTest) {
   // create a DB with block prefix index
   BlockBasedTableOptions table_options;
@@ -4378,7 +4508,6 @@ TEST_F(DBTest, ConcurrentMemtableNotSupported) {
   ASSERT_NOK(db_->CreateColumnFamily(cf_options, "name", &handle));
 }
 
-
 TEST_F(DBTest, SanitizeNumThreads) {
   for (int attempt = 0; attempt < 2; attempt++) {
     const size_t kTotalTasks = 8;
@@ -5161,7 +5290,6 @@ TEST_F(DBTest, DynamicLevelCompressionPerLevel) {
   for (int i = 0; i < kNKeys; i++) {
     keys[i] = i;
   }
-  RandomShuffle(std::begin(keys), std::end(keys));
 
   Random rnd(301);
   Options options;
@@ -5186,7 +5314,7 @@ TEST_F(DBTest, DynamicLevelCompressionPerLevel) {
   options.compression_per_level[0] = kNoCompression;
   // No compression for the Ln whre L0 is compacted to
   options.compression_per_level[1] = kNoCompression;
-  // Snpapy compression for Ln+1
+  // Snappy compression for Ln+1
   options.compression_per_level[2] = kSnappyCompression;
 
   OnFileDeletionListener* listener = new OnFileDeletionListener();
@@ -5198,6 +5326,7 @@ TEST_F(DBTest, DynamicLevelCompressionPerLevel) {
   // 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());
@@ -5216,8 +5345,9 @@ TEST_F(DBTest, DynamicLevelCompressionPerLevel) {
   // 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 = 21; i < 120; i++) {
+  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());
@@ -5238,9 +5368,45 @@ TEST_F(DBTest, DynamicLevelCompressionPerLevel) {
   }));
   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);
-    ASSERT_OK(dbfull()->DeleteFile(file.name));
+    Slice start(file.smallestkey), limit(file.largestkey);
+    const RangePtr ranges(&start, &limit);
+    // 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());
 
@@ -5788,7 +5954,6 @@ TEST_F(DBTest, FileCreationRandomFailure) {
   }
 }
 
-
 TEST_F(DBTest, DynamicMiscOptions) {
   // Test max_sequential_skip_in_iterations
   Options options;
@@ -7230,7 +7395,6 @@ TEST_F(DBTest, ReusePinnableSlice) {
             1);
 }
 
-
 TEST_F(DBTest, DeletingOldWalAfterDrop) {
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
       {{"Test:AllowFlushes", "DBImpl::BGWorkFlush"},
@@ -7355,7 +7519,6 @@ TEST_F(DBTest, LargeBlockSizeTest) {
   ASSERT_NOK(TryReopenWithColumnFamilies({"default", "pikachu"}, options));
 }
 
-
 TEST_F(DBTest, CreationTimeOfOldestFile) {
   const int kNumKeysPerFile = 32;
   const int kNumLevelFiles = 2;

db/db_test2.cc

@@ -16,6 +16,7 @@
 #include "db/db_test_util.h"
 #include "db/read_callback.h"
 #include "db/version_edit.h"
+#include "env/fs_readonly.h"
 #include "options/options_helper.h"
 #include "port/port.h"
 #include "port/stack_trace.h"
@@ -134,7 +135,6 @@ TEST_F(DBTest2, PartitionedIndexUserToInternalKey) {
   }
 }
 
-
 class PrefixFullBloomWithReverseComparator
     : public DBTestBase,
       public ::testing::WithParamInterface<bool> {
@@ -1475,8 +1475,7 @@ TEST_P(PresetCompressionDictTest, Flush) {
     // 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 &&
-        compression_type_ != kZSTDNotFinalCompression) {
+    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.
@@ -1553,8 +1552,7 @@ TEST_P(PresetCompressionDictTest, CompactNonBottommost) {
     // 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 &&
-        compression_type_ != kZSTDNotFinalCompression) {
+    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.
@@ -1615,8 +1613,7 @@ TEST_P(PresetCompressionDictTest, CompactBottommost) {
   // 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 &&
-      compression_type_ != kZSTDNotFinalCompression) {
+  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.
@@ -1976,7 +1973,6 @@ TEST_F(DBTest2, CompactionStall) {
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
 }
 
-
 TEST_F(DBTest2, FirstSnapshotTest) {
   Options options;
   options.write_buffer_size = 100000;  // Small write buffer
@@ -2103,16 +2099,15 @@ TEST_P(PinL0IndexAndFilterBlocksTest,
   ASSERT_EQ(2, TestGetTickerCount(options, BLOCK_CACHE_ADD));
   ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_DATA_MISS));
 
-  std::string value;
   // Miss and hit count should remain the same, they're all pinned.
-  ASSERT_TRUE(db_->KeyMayExist(ReadOptions(), handles_[1], "key", &value));
+  ASSERT_TRUE(db_->KeyMayExist(ReadOptions(), handles_[1], "key", nullptr));
   ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
   ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
   ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
   ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT));
 
   // Miss and hit count should remain the same, they're all pinned.
-  value = Get(1, "key");
+  std::string value = Get(1, "key");
   ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
   ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
   ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
@@ -3600,7 +3595,6 @@ TEST_F(DBTest2, OptimizeForSmallDB) {
   value.Reset();
 }
 
-
 TEST_F(DBTest2, IterRaceFlush1) {
   ASSERT_OK(Put("foo", "v1"));
 
@@ -4104,7 +4098,6 @@ TEST_F(DBTest2, ReadCallbackTest) {
   }
 }
 
-
 TEST_F(DBTest2, LiveFilesOmitObsoleteFiles) {
   // Regression test for race condition where an obsolete file is returned to
   // user as a "live file" but then deleted, all while file deletions are
@@ -4432,7 +4425,7 @@ TEST_F(DBTest2, TraceAndReplay) {
       db2->NewDefaultReplayer(handles, std::move(trace_reader), &replayer));
 
   TraceExecutionResultHandler res_handler;
-  std::function<void(Status, std::unique_ptr<TraceRecordResult> &&)> res_cb =
+  std::function<void(Status, std::unique_ptr<TraceRecordResult>&&)> res_cb =
       [&res_handler](Status exec_s, std::unique_ptr<TraceRecordResult>&& res) {
         ASSERT_TRUE(exec_s.ok() || exec_s.IsNotSupported());
         if (res != nullptr) {
@@ -5164,7 +5157,6 @@ TEST_F(DBTest2, TraceWithFilter) {
   ASSERT_EQ(count, 6);
 }
 
-
 TEST_F(DBTest2, PinnableSliceAndMmapReads) {
   Options options = CurrentOptions();
   options.env = env_;
@@ -7693,7 +7685,6 @@ TEST_F(DBTest2, RecoverEpochNumber) {
   }
 }
 
-
 TEST_F(DBTest2, RenameDirectory) {
   Options options = CurrentOptions();
   DestroyAndReopen(options);
@@ -8006,7 +7997,7 @@ TEST_F(DBTest2, GetLatestSeqAndTsForKey) {
   ASSERT_EQ(0, options.statistics->getTickerCount(GET_HIT_L0));
 }
 
-#if defined(ZSTD_ADVANCED)
+#if defined(ZSTD)
 TEST_F(DBTest2, ZSTDChecksum) {
   // Verify that corruption during decompression is caught.
   Options options = CurrentOptions();
@@ -8062,6 +8053,63 @@ TEST_F(DBTest2, TableCacheMissDuringReadFromBlockCacheTier) {
   ASSERT_EQ(orig_num_file_opens, TestGetTickerCount(options, NO_FILE_OPENS));
 }
 
+TEST_F(DBTest2, GetFileChecksumsFromCurrentManifest_CRC32) {
+  Options opts = CurrentOptions();
+  opts.create_if_missing = true;
+  opts.file_checksum_gen_factory = GetFileChecksumGenCrc32cFactory();
+  opts.level0_file_num_compaction_trigger = 10;
+
+  // Bootstrap the test database.
+  DB* db = nullptr;
+  std::string dbname = test::PerThreadDBPath("file_chksum");
+  ASSERT_OK(DB::Open(opts, dbname, &db));
+
+  WriteOptions wopts;
+  FlushOptions fopts;
+  fopts.wait = true;
+  Random rnd(test::RandomSeed());
+  for (int i = 0; i < 4; i++) {
+    ASSERT_OK(db->Put(wopts, Key(i), rnd.RandomString(100)));
+    ASSERT_OK(db->Flush(fopts));
+  }
+
+  // Obtain rich files metadata for source of truth.
+  std::vector<LiveFileMetaData> live_files;
+  db->GetLiveFilesMetaData(&live_files);
+
+  ASSERT_OK(db->Close());
+  delete db;
+  db = nullptr;
+
+  // Process current MANIFEST file and build internal file checksum mappings.
+  std::unique_ptr<FileChecksumList> checksum_list(NewFileChecksumList());
+  auto read_only_fs =
+      std::make_shared<ReadOnlyFileSystem>(env_->GetFileSystem());
+  ASSERT_OK(experimental::GetFileChecksumsFromCurrentManifest(
+      read_only_fs.get(), dbname, checksum_list.get()));
+
+  ASSERT_TRUE(checksum_list != nullptr);
+
+  // Retrieve files, related checksums and checksum functions.
+  std::vector<uint64_t> file_numbers;
+  std::vector<std::string> checksums;
+  std::vector<std::string> checksum_func_names;
+  ASSERT_OK(checksum_list->GetAllFileChecksums(&file_numbers, &checksums,
+                                               &checksum_func_names));
+
+  // Compare results.
+  ASSERT_EQ(live_files.size(), checksum_list->size());
+  for (size_t i = 0; i < live_files.size(); i++) {
+    std::string stored_checksum;
+    std::string stored_func_name;
+    ASSERT_OK(checksum_list->SearchOneFileChecksum(
+        live_files[i].file_number, &stored_checksum, &stored_func_name));
+
+    ASSERT_EQ(live_files[i].file_checksum, stored_checksum);
+    ASSERT_EQ(live_files[i].file_checksum_func_name, stored_func_name);
+  }
+}
+
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/db_test_util.cc

@@ -122,7 +122,6 @@ DBTestBase::~DBTestBase() {
 }
 
 bool DBTestBase::ShouldSkipOptions(int option_config, int skip_mask) {
-
   if ((skip_mask & kSkipUniversalCompaction) &&
       (option_config == kUniversalCompaction ||
        option_config == kUniversalCompactionMultiLevel ||
@@ -1208,7 +1207,6 @@ std::string DBTestBase::FilesPerLevel(int cf) {
   return result;
 }
 
-
 std::vector<uint64_t> DBTestBase::GetBlobFileNumbers() {
   VersionSet* const versions = dbfull()->GetVersionSet();
   assert(versions);
@@ -1600,42 +1598,74 @@ std::vector<std::uint64_t> DBTestBase::ListTableFiles(Env* env,
   return file_numbers;
 }
 
-void DBTestBase::VerifyDBFromMap(std::map<std::string, std::string> true_data,
-                                 size_t* total_reads_res, bool tailing_iter,
-                                 std::map<std::string, Status> status) {
-  size_t total_reads = 0;
+void DBTestBase::VerifyDBFromMap(
+    std::map<std::string, std::string> true_data, size_t* total_reads_res,
+    bool tailing_iter, ReadOptions* ro, ColumnFamilyHandle* cf,
+    std::unordered_set<std::string>* not_found) const {
+  ReadOptions temp_ro;
+  if (!ro) {
+    ro = &temp_ro;
+    ro->verify_checksums = true;
+  }
+  if (!cf) {
+    cf = db_->DefaultColumnFamily();
+  }
 
-  for (auto& kv : true_data) {
-    Status s = status[kv.first];
-    if (s.ok()) {
-      ASSERT_EQ(Get(kv.first), kv.second);
-    } else {
-      std::string value;
-      ASSERT_EQ(s, db_->Get(ReadOptions(), kv.first, &value));
-    }
+  // Get
+  size_t total_reads = 0;
+  std::string result;
+  for (auto& [k, v] : true_data) {
+    ASSERT_OK(db_->Get(*ro, cf, k, &result)) << "key is " << k;
+    ASSERT_EQ(v, result);
     total_reads++;
   }
+  if (not_found) {
+    for (const auto& k : *not_found) {
+      ASSERT_TRUE(db_->Get(*ro, cf, k, &result).IsNotFound())
+          << "key is " << k << " val is " << result;
+    }
+  }
+
+  // MultiGet
+  std::vector<Slice> key_slice;
+  for (const auto& [k, _] : true_data) {
+    key_slice.emplace_back(k);
+  }
+  std::vector<std::string> values;
+  std::vector<ColumnFamilyHandle*> cfs(key_slice.size(), cf);
+  std::vector<Status> status = db_->MultiGet(*ro, cfs, key_slice, &values);
+  total_reads += key_slice.size();
+  auto data_iter = true_data.begin();
+  for (size_t i = 0; i < key_slice.size(); ++i, ++data_iter) {
+    ASSERT_OK(status[i]);
+    ASSERT_EQ(values[i], data_iter->second);
+  }
+  // MultiGet - not found
+  if (not_found) {
+    key_slice.clear();
+    for (const auto& k : *not_found) {
+      key_slice.emplace_back(k);
+    }
+    cfs = std::vector<ColumnFamilyHandle*>(key_slice.size(), cf);
+    values.clear();
+    status = db_->MultiGet(*ro, cfs, key_slice, &values);
+    for (const auto& s : status) {
+      ASSERT_TRUE(s.IsNotFound());
+    }
+  }
 
   // Normal Iterator
   {
     int iter_cnt = 0;
-    ReadOptions ro;
-    ro.total_order_seek = true;
-    Iterator* iter = db_->NewIterator(ro);
+    ReadOptions ro_ = *ro;
+    ro_.total_order_seek = true;
+    Iterator* iter = db_->NewIterator(ro_, cf);
     // Verify Iterator::Next()
     iter_cnt = 0;
-    auto data_iter = true_data.begin();
-    Status s;
-    for (iter->SeekToFirst(); iter->Valid(); iter->Next(), data_iter++) {
+    data_iter = true_data.begin();
+    for (iter->SeekToFirst(); iter->Valid(); iter->Next(), ++data_iter) {
       ASSERT_EQ(iter->key().ToString(), data_iter->first);
-      Status current_status = status[data_iter->first];
-      if (!current_status.ok()) {
-        s = current_status;
-      }
-      ASSERT_EQ(iter->status(), s);
-      if (current_status.ok()) {
-        ASSERT_EQ(iter->value().ToString(), data_iter->second);
-      }
+      ASSERT_EQ(iter->value().ToString(), data_iter->second);
       iter_cnt++;
       total_reads++;
     }
@@ -1646,20 +1676,12 @@ void DBTestBase::VerifyDBFromMap(std::map<std::string, std::string> true_data,
 
     // Verify Iterator::Prev()
     // Use a new iterator to make sure its status is clean.
-    iter = db_->NewIterator(ro);
+    iter = db_->NewIterator(ro_, cf);
     iter_cnt = 0;
-    s = Status::OK();
     auto data_rev = true_data.rbegin();
     for (iter->SeekToLast(); iter->Valid(); iter->Prev(), data_rev++) {
       ASSERT_EQ(iter->key().ToString(), data_rev->first);
-      Status current_status = status[data_rev->first];
-      if (!current_status.ok()) {
-        s = current_status;
-      }
-      ASSERT_EQ(iter->status(), s);
-      if (current_status.ok()) {
-        ASSERT_EQ(iter->value().ToString(), data_rev->second);
-      }
+      ASSERT_EQ(iter->value().ToString(), data_rev->second);
       iter_cnt++;
       total_reads++;
     }
@@ -1667,12 +1689,20 @@ void DBTestBase::VerifyDBFromMap(std::map<std::string, std::string> true_data,
     ASSERT_EQ(data_rev, true_data.rend())
         << iter_cnt << " / " << true_data.size();
 
-    // Verify Iterator::Seek()
-    for (const auto& kv : true_data) {
-      iter->Seek(kv.first);
-      ASSERT_EQ(kv.first, iter->key().ToString());
-      ASSERT_EQ(kv.second, iter->value().ToString());
-      total_reads++;
+    // Verify Iterator::Seek() and SeekForPrev()
+    for (const auto& [k, v] : true_data) {
+      for (bool prev : {false, true}) {
+        if (prev) {
+          iter->SeekForPrev(k);
+        } else {
+          iter->Seek(k);
+        }
+        ASSERT_TRUE(iter->Valid());
+        ASSERT_OK(iter->status());
+        ASSERT_EQ(iter->key(), k);
+        ASSERT_EQ(iter->value(), v);
+        ++total_reads;
+      }
     }
     delete iter;
   }
@@ -1680,14 +1710,14 @@ void DBTestBase::VerifyDBFromMap(std::map<std::string, std::string> true_data,
   if (tailing_iter) {
     // Tailing iterator
     int iter_cnt = 0;
-    ReadOptions ro;
-    ro.tailing = true;
-    ro.total_order_seek = true;
-    Iterator* iter = db_->NewIterator(ro);
+    ReadOptions ro_ = *ro;
+    ro_.tailing = true;
+    ro_.total_order_seek = true;
+    Iterator* iter = db_->NewIterator(ro_, cf);
 
     // Verify ForwardIterator::Next()
     iter_cnt = 0;
-    auto data_iter = true_data.begin();
+    data_iter = true_data.begin();
     for (iter->SeekToFirst(); iter->Valid(); iter->Next(), data_iter++) {
       ASSERT_EQ(iter->key().ToString(), data_iter->first);
       ASSERT_EQ(iter->value().ToString(), data_iter->second);
@@ -1733,7 +1763,6 @@ void DBTestBase::VerifyDBInternal(
   iter->~InternalIterator();
 }
 
-
 uint64_t DBTestBase::GetNumberOfSstFilesForColumnFamily(
     DB* db, std::string column_family_name) {
   std::vector<LiveFileMetaData> metadata;

db/db_test_util.h

@@ -1106,6 +1106,11 @@ class DBTestBase : public testing::Test {
     return std::string(buf);
   }
 
+  // Expects valid key created by Key().
+  static int IdFromKey(const std::string& key) {
+    return std::stoi(key.substr(3));
+  }
+
   static bool ShouldSkipOptions(int option_config, int skip_mask = kNoSkip);
 
   // Switch to a fresh database with the next option configuration to
@@ -1367,7 +1372,8 @@ class DBTestBase : public testing::Test {
   void VerifyDBFromMap(
       std::map<std::string, std::string> true_data,
       size_t* total_reads_res = nullptr, bool tailing_iter = false,
-      std::map<std::string, Status> status = std::map<std::string, Status>());
+      ReadOptions* ro = nullptr, ColumnFamilyHandle* cf = nullptr,
+      std::unordered_set<std::string>* not_found = nullptr) const;
 
   void VerifyDBInternal(
       std::vector<std::pair<std::string, std::string>> true_data);

db/db_universal_compaction_test.cc

@@ -2369,7 +2369,6 @@ TEST_F(DBTestUniversalCompaction2, PeriodicCompactionOffpeak) {
 
 }  // namespace ROCKSDB_NAMESPACE
 
-
 int main(int argc, char** argv) {
   ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
   ::testing::InitGoogleTest(&argc, argv);

db/db_wal_test.cc

@@ -1153,6 +1153,7 @@ TEST_F(DBWALTest, FullPurgePreservesRecycledLog) {
     dbfull()->FindObsoleteFiles(&job_context, true /* force */);
     dbfull()->TEST_UnlockMutex();
     dbfull()->PurgeObsoleteFiles(job_context);
+    job_context.Clean();
 
     if (i == 0) {
       ASSERT_OK(
@@ -1838,9 +1839,194 @@ class RecoveryTestHelper {
   }
 };
 
-class DBWALTestWithParams : public DBWALTestBase,
-                            public ::testing::WithParamInterface<
-                                std::tuple<bool, int, int, CompressionType>> {
+TEST_F(DBWALTest, TrackAndVerifyWALsRecycleWAL) {
+  Options options = CurrentOptions();
+  options.avoid_flush_during_shutdown = true;
+  options.wal_recovery_mode = WALRecoveryMode::kPointInTimeRecovery;
+  options.recycle_log_file_num = 1;
+  options.track_and_verify_wals = true;
+
+  DestroyAndReopen(options);
+
+  ASSERT_OK(Put("key_ignore", "wal_to_recycle"));
+  ASSERT_OK(Put("key_ignore1", "wal_to_recycle"));
+  ASSERT_OK(Put("key_ignore2", "wal_to_recycle"));
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(Put("key_ignore", "wal_to_recycle"));
+  ASSERT_OK(Put("key_ignore1", "wal_to_recycle"));
+  ASSERT_OK(Put("key_ignore2", "wal_to_recycle"));
+  ASSERT_OK(Flush());
+
+  // Stop background flush to avoid deleting any WAL
+  options.env->SetBackgroundThreads(1, Env::HIGH);
+  test::SleepingBackgroundTask sleeping_task;
+  options.env->Schedule(&test::SleepingBackgroundTask::DoSleepTask,
+                        &sleeping_task, Env::Priority::HIGH);
+
+  // Recycle the first WAL
+  ASSERT_OK(Put("key1", "old_value"));
+  // Recycle the second WAL
+  ASSERT_OK(dbfull()->TEST_SwitchWAL());
+  ASSERT_OK(Put("key1", "new_value"));
+
+  // Create a WAL hole on sequence number by truncating the first WAL to 0 byte
+  VectorWalPtr log_files;
+  ASSERT_OK(db_->GetSortedWalFiles(log_files));
+  ASSERT_EQ(log_files.size(), 2);
+  std::string log_name = LogFileName(dbname_, log_files.front()->LogNumber());
+  Close();
+  // Drop `Put("key1", "old_value")` in the first WAL
+  ASSERT_OK(test::TruncateFile(options.env, log_name, 0 /* new_length */));
+
+  Status s = DB::Open(options, dbname_, &db_);
+
+  ASSERT_OK(s);
+
+  ASSERT_EQ("wal_to_recycle", Get("key_ignore2"));
+  ASSERT_EQ("NOT_FOUND", Get("key1"));
+
+  Close();
+}
+
+class DBWALTrackAndVerifyWALsWithParamsTest
+    : public DBWALTestBase,
+      public ::testing::WithParamInterface<WALRecoveryMode> {
+ public:
+  DBWALTrackAndVerifyWALsWithParamsTest()
+      : DBWALTestBase("/db_wal_track_and_verify_wals_with_params_test") {}
+};
+
+INSTANTIATE_TEST_CASE_P(
+    DBWALTrackAndVerifyWALsWithParamsTest,
+    DBWALTrackAndVerifyWALsWithParamsTest,
+    ::testing::Values(WALRecoveryMode::kTolerateCorruptedTailRecords,
+                      WALRecoveryMode::kAbsoluteConsistency,
+                      WALRecoveryMode::kPointInTimeRecovery,
+                      WALRecoveryMode::kSkipAnyCorruptedRecords));
+
+TEST_P(DBWALTrackAndVerifyWALsWithParamsTest, Basic) {
+  Options options = CurrentOptions();
+  options.avoid_flush_during_shutdown = true;
+  options.track_and_verify_wals = true;
+  options.wal_recovery_mode = GetParam();
+
+  // Stop background flush to avoid deleting any WAL
+  options.env->SetBackgroundThreads(1, Env::HIGH);
+  test::SleepingBackgroundTask sleeping_task;
+  options.env->Schedule(&test::SleepingBackgroundTask::DoSleepTask,
+                        &sleeping_task, Env::Priority::HIGH);
+
+  for (int i = 0; i < 5; i++) {
+    DestroyAndReopen(options);
+
+    ASSERT_OK(Put("key1", "old_value"));
+    SequenceNumber last_seqno_recorded_in_fist_wal =
+        dbfull()->GetLatestSequenceNumber();
+
+    ASSERT_OK(dbfull()->TEST_SwitchWAL());
+    ASSERT_OK(Put("key1", "new_value"));
+
+    VectorWalPtr log_files;
+    ASSERT_OK(db_->GetSortedWalFiles(log_files));
+    ASSERT_EQ(log_files.size(), 2);
+    uint64_t first_log_number = log_files.front()->LogNumber();
+    std::string first_log_name = LogFileName(dbname_, first_log_number);
+    std::string second_log_name =
+        LogFileName(dbname_, log_files.back()->LogNumber());
+
+    if (i == 0) {
+      // Delete the obsolete WAL and verify it will not be seen as a WAL hole
+      sleeping_task.WakeUp();
+      sleeping_task.WaitUntilDone();
+      ASSERT_OK(dbfull()->TEST_WaitForBackgroundWork());
+      // Stop background flush to avoid deleting any WAL
+      sleeping_task.Reset();
+      options.env->Schedule(&test::SleepingBackgroundTask::DoSleepTask,
+                            &sleeping_task, Env::Priority::HIGH);
+      Close();
+    } else if (i == 1) {
+      // Create a WAL hole on WAL number by deleting the first WAL and verify
+      // the hole will be detected
+      Close();
+      ASSERT_OK(options.env->DeleteFile(first_log_name));
+    } else if (i == 2) {
+      // Create a WAL hole on sequence number by truncating the first WAL and
+      // verify the hole will be detected
+      Close();
+      ASSERT_OK(
+          test::TruncateFile(options.env, first_log_name, 0 /* new_length */));
+    } else if (i == 3) {
+      // Create a WAL hole on size difference by truncating the first WAL and
+      // mocking a correct sequence number to force triggering corruption based
+      // on size instead of sequence number and verify the hole will be detected
+      Close();
+      ASSERT_OK(
+          test::TruncateFile(options.env, first_log_name, 0 /* new_length */));
+      SyncPoint::GetInstance()->SetCallBack(
+          "DBImpl::UpdatePredecessorWALInfo", [&](void* arg) {
+            std::pair<uint64_t, SequenceNumber*>* pair =
+                static_cast<std::pair<uint64_t, SequenceNumber*>*>(arg);
+            if (pair->first == first_log_number) {
+              *(pair->second) = last_seqno_recorded_in_fist_wal;
+            }
+          });
+      SyncPoint::GetInstance()->EnableProcessing();
+    } else if (i == 4) {
+      // Delete all wals and verify opening a DB with no WAL will be detected
+      Close();
+      ASSERT_OK(options.env->DeleteFile(first_log_name));
+      ASSERT_OK(options.env->DeleteFile(second_log_name));
+    }
+
+    Status s = DB::Open(options, dbname_, &db_);
+
+    if (i == 0) {
+      ASSERT_OK(s);
+      ASSERT_EQ("new_value", Get("key1"));
+      continue;
+    } else if (i == 3) {
+      SyncPoint::GetInstance()->ClearAllCallBacks();
+      SyncPoint::GetInstance()->DisableProcessing();
+    } else if (i == 4) {
+      ASSERT_TRUE(s.IsCorruption());
+      ASSERT_TRUE(
+          s.ToString().find("Opening an existing DB with no WAL files") !=
+          std::string::npos);
+      Close();
+      continue;
+    }
+
+    if (options.wal_recovery_mode == WALRecoveryMode::kPointInTimeRecovery) {
+      ASSERT_OK(s);
+      ASSERT_EQ("NOT_FOUND", Get("key1"));
+    } else if (options.wal_recovery_mode ==
+                   WALRecoveryMode::kAbsoluteConsistency ||
+               options.wal_recovery_mode ==
+                   WALRecoveryMode::kTolerateCorruptedTailRecords) {
+      ASSERT_TRUE(s.IsCorruption());
+      std::string msg;
+      if (i == 1) {
+        msg = "Missing WAL";
+      } else if (i == 2) {
+        msg = "Mismatched last sequence number recorded in the WAL";
+      } else if (i == 3) {
+        msg = "Mismatched size of the WAL";
+      }
+      ASSERT_TRUE(s.ToString().find(msg) != std::string::npos);
+    } else {
+      ASSERT_OK(s);
+      ASSERT_EQ("new_value", Get("key1"));
+    }
+
+    Close();
+  }
+}
+
+class DBWALTestWithParams
+    : public DBWALTestBase,
+      public ::testing::WithParamInterface<
+          std::tuple<bool, int, int, CompressionType, bool>> {
  public:
   DBWALTestWithParams() : DBWALTestBase("/db_wal_test_with_params") {}
 };
@@ -1853,7 +2039,8 @@ INSTANTIATE_TEST_CASE_P(
                                             RecoveryTestHelper::kWALFilesCount,
                                         1),
                        ::testing::Values(CompressionType::kNoCompression,
-                                         CompressionType::kZSTD)));
+                                         CompressionType::kZSTD),
+                       ::testing::Bool()));
 
 class DBWALTestWithParamsVaryingRecoveryMode
     : public DBWALTestBase,
@@ -1891,6 +2078,7 @@ TEST_P(DBWALTestWithParams, kTolerateCorruptedTailRecords) {
 
   // Fill data for testing
   Options options = CurrentOptions();
+  options.track_and_verify_wals = std::get<4>(GetParam());
   const size_t row_count = RecoveryTestHelper::FillData(this, &options);
   // test checksum failure or parsing
   RecoveryTestHelper::CorruptWAL(this, options, corrupt_offset * .3,
@@ -1914,6 +2102,7 @@ TEST_P(DBWALTestWithParams, kTolerateCorruptedTailRecords) {
 TEST_P(DBWALTestWithParams, kAbsoluteConsistency) {
   // Verify clean slate behavior
   Options options = CurrentOptions();
+  options.track_and_verify_wals = std::get<4>(GetParam());
   const size_t row_count = RecoveryTestHelper::FillData(this, &options);
   options.create_if_missing = false;
   ASSERT_OK(TryReopen(options));
@@ -2124,7 +2313,6 @@ TEST_F(DBWALTest, FixSyncWalOnObseletedWalWithNewManifestCausingMissingWAL) {
         wal_synced = true;
       });
 
-
   SyncPoint::GetInstance()->EnableProcessing();
 
   ASSERT_OK(Flush());
@@ -2165,6 +2353,7 @@ TEST_P(DBWALTestWithParams, kPointInTimeRecovery) {
 
   // Fill data for testing
   Options options = CurrentOptions();
+  options.track_and_verify_wals = std::get<4>(GetParam());
   options.wal_compression = compression_type;
   const size_t row_count = RecoveryTestHelper::FillData(this, &options);
 
@@ -2222,6 +2411,7 @@ TEST_P(DBWALTestWithParams, kSkipAnyCorruptedRecords) {
 
   // Fill data for testing
   Options options = CurrentOptions();
+  options.track_and_verify_wals = std::get<4>(GetParam());
   options.wal_compression = compression_type;
   const size_t row_count = RecoveryTestHelper::FillData(this, &options);
 
@@ -2699,12 +2889,12 @@ TEST_F(DBWALTest, ReadOnlyRecoveryNoTruncate) {
   // create DB and close with file truncate disabled
   std::atomic_bool enable_truncate{false};
 
-  SyncPoint::GetInstance()->SetCallBack(
-      "PosixWritableFile::Close", [&](void* arg) {
-        if (!enable_truncate) {
-          *(static_cast<size_t*>(arg)) = 0;
-        }
-      });
+  SyncPoint::GetInstance()->SetCallBack("PosixWritableFile::Close",
+                                        [&](void* arg) {
+                                          if (!enable_truncate) {
+                                            *(static_cast<size_t*>(arg)) = 0;
+                                          }
+                                        });
   SyncPoint::GetInstance()->EnableProcessing();
 
   DestroyAndReopen(options);
@@ -2793,7 +2983,6 @@ TEST_F(DBWALTest, WalInManifestButNotInSortedWals) {
   Close();
 }
 
-
 TEST_F(DBWALTest, WalTermTest) {
   Options options = CurrentOptions();
   options.env = env_;

db/db_with_timestamp_basic_test.cc

@@ -1480,6 +1480,55 @@ TEST_F(DBBasicTestWithTimestamp, ReseekToUserKeyBeforeSavedKey) {
   Close();
 }
 
+TEST_F(DBBasicTestWithTimestamp,
+       FIXME_ReverseIterationWithBlobAndUnpreparedValue) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.env = env_;
+  options.enable_blob_files = true;
+  options.max_sequential_skip_in_iterations = 0;
+
+  const size_t kTimestampSize = Timestamp(0, 0).size();
+  TestComparator test_cmp(kTimestampSize);
+  options.comparator = &test_cmp;
+
+  DestroyAndReopen(options);
+
+  constexpr uint64_t kMaxKey = 1024;
+
+  const std::vector<std::string> write_timestamps = {Timestamp(1, 0),
+                                                     Timestamp(3, 0)};
+  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)));
+    }
+  }
+
+  ASSERT_OK(Flush());
+
+  {
+    const std::string read_timestamp_str = Timestamp(4, 0);
+    const Slice read_timestamp(read_timestamp_str);
+
+    ReadOptions read_opts;
+    read_opts.timestamp = &read_timestamp;
+    read_opts.allow_unprepared_value = true;
+
+    std::unique_ptr<Iterator> it(db_->NewIterator(read_opts));
+
+    it->SeekForPrev(Key1(kMaxKey));
+    ASSERT_TRUE(it->Valid());
+    ASSERT_OK(it->status());
+
+    // FIXME: PrepareValue() should succeed and status() should remain OK
+    ASSERT_FALSE(it->PrepareValue());
+    ASSERT_TRUE(it->status().IsCorruption());
+  }
+
+  Close();
+}
+
 TEST_F(DBBasicTestWithTimestamp, MultiGetWithFastLocalBloom) {
   Options options = CurrentOptions();
   options.env = env_;
@@ -3257,7 +3306,6 @@ TEST_F(DBBasicTestWithTimestamp, MultiGetNoReturnTs) {
   Close();
 }
 
-
 INSTANTIATE_TEST_CASE_P(
     Timestamp, DBBasicTestWithTimestampCompressionSettings,
     ::testing::Combine(

db/db_write_buffer_manager_test.cc

@@ -780,7 +780,6 @@ TEST_P(DBWriteBufferManagerTest, MixedSlowDownOptionsMultipleDB) {
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
 }
 
-
 // Tests a `WriteBufferManager` constructed with `allow_stall == false` does not
 // thrash memtable switching when full and a CF receives multiple writes.
 // Instead, we expect to switch a CF's memtable for flush only when that CF does

db/dbformat.cc

@@ -178,11 +178,9 @@ std::string ParsedInternalKey::DebugString(bool log_err_key, bool hex,
     result += "<redacted>";
   }
 
-  char buf[50];
-  snprintf(buf, sizeof(buf), "' seq:%" PRIu64 ", type:%d", sequence,
-           static_cast<int>(type));
+  result += "' seq:" + std::to_string(sequence);
+  result += ", type:" + std::to_string(type);
 
-  result += buf;
   return result;
 }
 

db/dbformat.h

@@ -498,6 +498,8 @@ class InternalKey {
   // Intended only to be used together with ConvertFromUserKey().
   std::string* rep() { return &rep_; }
 
+  const std::string* const_rep() const { return &rep_; }
+
   // Assuming that *rep() contains a user key, this method makes internal key
   // out of it in-place. This saves a memcpy compared to Set()/SetFrom().
   void ConvertFromUserKey(SequenceNumber s, ValueType t) {
@@ -1179,4 +1181,64 @@ struct ParsedInternalKeyComparator {
   const InternalKeyComparator* cmp;
 };
 
+class PredecessorWALInfo {
+ public:
+  PredecessorWALInfo()
+      : log_number_(0),
+        size_bytes_(0),
+        last_seqno_recorded_(0),
+        initialized_(false) {}
+
+  explicit PredecessorWALInfo(uint64_t log_number, uint64_t size_bytes,
+                              SequenceNumber last_seqno_recorded)
+      : log_number_(log_number),
+        size_bytes_(size_bytes),
+        last_seqno_recorded_(last_seqno_recorded),
+        initialized_(true) {}
+
+  uint64_t GetLogNumber() const {
+    assert(initialized_);
+    return log_number_;
+  }
+
+  uint64_t GetSizeBytes() const {
+    assert(initialized_);
+    return size_bytes_;
+  }
+
+  SequenceNumber GetLastSeqnoRecorded() const {
+    assert(initialized_);
+    return last_seqno_recorded_;
+  }
+
+  bool IsInitialized() const { return initialized_; }
+
+  inline void EncodeTo(std::string* dst) const {
+    assert(dst != nullptr);
+    assert(initialized_);
+    PutFixed64(dst, log_number_);
+    PutFixed64(dst, size_bytes_);
+    PutFixed64(dst, last_seqno_recorded_);
+  }
+
+  inline Status DecodeFrom(Slice* src) {
+    if (!GetFixed64(src, &log_number_)) {
+      return Status::Corruption("Error decoding log number");
+    }
+    if (!GetFixed64(src, &size_bytes_)) {
+      return Status::Corruption("Error decoding size bytes");
+    }
+    if (!GetFixed64(src, &last_seqno_recorded_)) {
+      return Status::Corruption("Error decoding last seqno recorded");
+    }
+    initialized_ = true;
+    return Status::OK();
+  }
+
+ private:
+  uint64_t log_number_;
+  uint64_t size_bytes_;
+  SequenceNumber last_seqno_recorded_;
+  bool initialized_;
+};
 }  // namespace ROCKSDB_NAMESPACE

db/deletefile_test.cc

@@ -135,57 +135,6 @@ class DeleteFileTest : public DBTestBase {
   }
 };
 
-TEST_F(DeleteFileTest, AddKeysAndQueryLevels) {
-  Options options = CurrentOptions();
-  SetOptions(&options);
-  Destroy(options);
-  options.create_if_missing = true;
-  Reopen(options);
-
-  CreateTwoLevels();
-  std::vector<LiveFileMetaData> metadata;
-  db_->GetLiveFilesMetaData(&metadata);
-
-  std::string level1file;
-  int level1keycount = 0;
-  std::string level2file;
-  int level2keycount = 0;
-  int level1index = 0;
-  int level2index = 1;
-
-  ASSERT_EQ((int)metadata.size(), 2);
-  if (metadata[0].level == 2) {
-    level1index = 1;
-    level2index = 0;
-  }
-
-  level1file = metadata[level1index].name;
-  int startkey = atoi(metadata[level1index].smallestkey.c_str());
-  int endkey = atoi(metadata[level1index].largestkey.c_str());
-  level1keycount = (endkey - startkey + 1);
-  level2file = metadata[level2index].name;
-  startkey = atoi(metadata[level2index].smallestkey.c_str());
-  endkey = atoi(metadata[level2index].largestkey.c_str());
-  level2keycount = (endkey - startkey + 1);
-
-  // COntrolled setup. Levels 1 and 2 should both have 50K files.
-  // This is a little fragile as it depends on the current
-  // compaction heuristics.
-  ASSERT_EQ(level1keycount, 50000);
-  ASSERT_EQ(level2keycount, 50000);
-
-  Status status = db_->DeleteFile("0.sst");
-  ASSERT_TRUE(status.IsInvalidArgument());
-
-  // intermediate level files cannot be deleted.
-  status = db_->DeleteFile(level1file);
-  ASSERT_TRUE(status.IsInvalidArgument());
-
-  // Lowest level file deletion should succeed.
-  status = db_->DeleteFile(level2file);
-  ASSERT_OK(status);
-}
-
 TEST_F(DeleteFileTest, PurgeObsoleteFilesTest) {
   Options options = CurrentOptions();
   SetOptions(&options);
@@ -496,145 +445,6 @@ TEST_F(DeleteFileTest, BackgroundPurgeTestMultipleJobs) {
   CheckFileTypeCounts(dbname_, 0, 1, 1);
 }
 
-TEST_F(DeleteFileTest, DeleteFileWithIterator) {
-  Options options = CurrentOptions();
-  SetOptions(&options);
-  Destroy(options);
-  options.create_if_missing = true;
-  Reopen(options);
-
-  CreateTwoLevels();
-  ReadOptions read_options;
-  Iterator* it = db_->NewIterator(read_options);
-  ASSERT_OK(it->status());
-  std::vector<LiveFileMetaData> metadata;
-  db_->GetLiveFilesMetaData(&metadata);
-
-  std::string level2file;
-
-  ASSERT_EQ(metadata.size(), static_cast<size_t>(2));
-  if (metadata[0].level == 1) {
-    level2file = metadata[1].name;
-  } else {
-    level2file = metadata[0].name;
-  }
-
-  Status status = db_->DeleteFile(level2file);
-  fprintf(stdout, "Deletion status %s: %s\n", level2file.c_str(),
-          status.ToString().c_str());
-  ASSERT_OK(status);
-  it->SeekToFirst();
-  int numKeysIterated = 0;
-  while (it->Valid()) {
-    numKeysIterated++;
-    it->Next();
-  }
-  ASSERT_EQ(numKeysIterated, 50000);
-  delete it;
-}
-
-TEST_F(DeleteFileTest, DeleteLogFiles) {
-  Options options = CurrentOptions();
-  SetOptions(&options);
-  Destroy(options);
-  options.create_if_missing = true;
-  Reopen(options);
-
-  AddKeys(10, 0);
-  VectorLogPtr logfiles;
-  ASSERT_OK(db_->GetSortedWalFiles(logfiles));
-  ASSERT_GT(logfiles.size(), 0UL);
-  // Take the last log file which is expected to be alive and try to delete it
-  // Should not succeed because live logs are not allowed to be deleted
-  std::unique_ptr<LogFile> alive_log = std::move(logfiles.back());
-  ASSERT_EQ(alive_log->Type(), kAliveLogFile);
-  ASSERT_OK(env_->FileExists(wal_dir_ + "/" + alive_log->PathName()));
-  fprintf(stdout, "Deleting alive log file %s\n",
-          alive_log->PathName().c_str());
-  ASSERT_NOK(db_->DeleteFile(alive_log->PathName()));
-  ASSERT_OK(env_->FileExists(wal_dir_ + "/" + alive_log->PathName()));
-  logfiles.clear();
-
-  // Call Flush to bring about a new working log file and add more keys
-  // Call Flush again to flush out memtable and move alive log to archived log
-  // and try to delete the archived log file
-  FlushOptions fopts;
-  ASSERT_OK(db_->Flush(fopts));
-  AddKeys(10, 0);
-  ASSERT_OK(db_->Flush(fopts));
-  ASSERT_OK(db_->GetSortedWalFiles(logfiles));
-  ASSERT_GT(logfiles.size(), 0UL);
-  std::unique_ptr<LogFile> archived_log = std::move(logfiles.front());
-  ASSERT_EQ(archived_log->Type(), kArchivedLogFile);
-  ASSERT_OK(env_->FileExists(wal_dir_ + "/" + archived_log->PathName()));
-  fprintf(stdout, "Deleting archived log file %s\n",
-          archived_log->PathName().c_str());
-  ASSERT_OK(db_->DeleteFile(archived_log->PathName()));
-  ASSERT_TRUE(
-      env_->FileExists(wal_dir_ + "/" + archived_log->PathName()).IsNotFound());
-}
-
-TEST_F(DeleteFileTest, DeleteNonDefaultColumnFamily) {
-  Options options = CurrentOptions();
-  SetOptions(&options);
-  Destroy(options);
-  options.create_if_missing = true;
-  Reopen(options);
-  CreateAndReopenWithCF({"new_cf"}, options);
-
-  Random rnd(5);
-  for (int i = 0; i < 1000; ++i) {
-    ASSERT_OK(db_->Put(WriteOptions(), handles_[1], test::RandomKey(&rnd, 10),
-                       test::RandomKey(&rnd, 10)));
-  }
-  ASSERT_OK(db_->Flush(FlushOptions(), handles_[1]));
-  for (int i = 0; i < 1000; ++i) {
-    ASSERT_OK(db_->Put(WriteOptions(), handles_[1], test::RandomKey(&rnd, 10),
-                       test::RandomKey(&rnd, 10)));
-  }
-  ASSERT_OK(db_->Flush(FlushOptions(), handles_[1]));
-
-  std::vector<LiveFileMetaData> metadata;
-  db_->GetLiveFilesMetaData(&metadata);
-  ASSERT_EQ(2U, metadata.size());
-  ASSERT_EQ("new_cf", metadata[0].column_family_name);
-  ASSERT_EQ("new_cf", metadata[1].column_family_name);
-  auto old_file = metadata[0].smallest_seqno < metadata[1].smallest_seqno
-                      ? metadata[0].name
-                      : metadata[1].name;
-  auto new_file = metadata[0].smallest_seqno > metadata[1].smallest_seqno
-                      ? metadata[0].name
-                      : metadata[1].name;
-  ASSERT_TRUE(db_->DeleteFile(new_file).IsInvalidArgument());
-  ASSERT_OK(db_->DeleteFile(old_file));
-
-  {
-    std::unique_ptr<Iterator> itr(db_->NewIterator(ReadOptions(), handles_[1]));
-    ASSERT_OK(itr->status());
-    int count = 0;
-    for (itr->SeekToFirst(); itr->Valid(); itr->Next()) {
-      ASSERT_OK(itr->status());
-      ++count;
-    }
-    ASSERT_OK(itr->status());
-    ASSERT_EQ(count, 1000);
-  }
-
-  Close();
-  ReopenWithColumnFamilies({kDefaultColumnFamilyName, "new_cf"}, options);
-
-  {
-    std::unique_ptr<Iterator> itr(db_->NewIterator(ReadOptions(), handles_[1]));
-    int count = 0;
-    for (itr->SeekToFirst(); itr->Valid(); itr->Next()) {
-      ASSERT_OK(itr->status());
-      ++count;
-    }
-    ASSERT_OK(itr->status());
-    ASSERT_EQ(count, 1000);
-  }
-}
-
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/error_handler.cc

@@ -227,7 +227,7 @@ std::map<std::tuple<BackgroundErrorReason, bool>, Status::Severity>
          Status::Severity::kFatalError},
 };
 
-void ErrorHandler::CancelErrorRecovery() {
+void ErrorHandler::CancelErrorRecoveryForShutDown() {
   db_mutex_->AssertHeld();
 
   // We'll release the lock before calling sfm, so make sure no new
@@ -573,6 +573,8 @@ Status ErrorHandler::ClearBGError() {
 
   // Signal that recovery succeeded
   if (recovery_error_.ok()) {
+    // If this assertion fails, it means likely bg error is not set after a
+    // file is quarantined during MANIFEST write.
     assert(files_to_quarantine_.empty());
     Status old_bg_error = bg_error_;
     // old_bg_error is only for notifying listeners, so may not be checked
@@ -585,8 +587,15 @@ Status ErrorHandler::ClearBGError() {
     recovery_error_.PermitUncheckedError();
     recovery_in_prog_ = false;
     soft_error_no_bg_work_ = false;
-    EventHelpers::NotifyOnErrorRecoveryEnd(db_options_.listeners, old_bg_error,
-                                           bg_error_, db_mutex_);
+    if (!db_->shutdown_initiated_) {
+      // NotifyOnErrorRecoveryEnd() may release and re-acquire db_mutex_.
+      // Prevent DB from being closed while we notify listeners. DB close will
+      // wait until allow_db_shutdown_ = true, see ReadyForShutdown().
+      allow_db_shutdown_ = false;
+      EventHelpers::NotifyOnErrorRecoveryEnd(
+          db_options_.listeners, old_bg_error, bg_error_, db_mutex_);
+      allow_db_shutdown_ = true;
+    }
   }
   return recovery_error_;
 }

db/error_handler.h

@@ -44,6 +44,7 @@ class ErrorHandler {
         auto_recovery_(false),
         recovery_in_prog_(false),
         soft_error_no_bg_work_(false),
+        allow_db_shutdown_(true),
         is_db_stopped_(false),
         bg_error_stats_(db_options.statistics) {
     // Clear the checked flag for uninitialized errors
@@ -63,6 +64,12 @@ class ErrorHandler {
 
   Status GetRecoveryError() const { return recovery_error_; }
 
+  // REQUIREs: db mutex held
+  //
+  // Returns non-OK status if encountered error during recovery.
+  // Returns OK if bg error is successfully cleared. May releases and
+  // re-acquire db mutex to notify listeners. However, DB close (if initiated)
+  // will be blocked until db mutex is released after return.
   Status ClearBGError();
 
   bool IsDBStopped() { return is_db_stopped_.load(std::memory_order_acquire); }
@@ -79,8 +86,14 @@ class ErrorHandler {
 
   bool IsRecoveryInProgress() { return recovery_in_prog_; }
 
+  // REQUIRES: db mutex held
+  bool ReadyForShutdown() {
+    db_mutex_->AssertHeld();
+    return !recovery_in_prog_ && allow_db_shutdown_;
+  }
+
   Status RecoverFromBGError(bool is_manual = false);
-  void CancelErrorRecovery();
+  void CancelErrorRecoveryForShutDown();
 
   void EndAutoRecovery();
 
@@ -121,6 +134,8 @@ class ErrorHandler {
   // A flag to indicate that for the soft error, we should not allow any
   // background work except the work is from recovery.
   bool soft_error_no_bg_work_;
+  // Used in ClearBGError() to prevent DB from being closed.
+  bool allow_db_shutdown_;
 
   // Used to store the context for recover, such as flush reason.
   DBRecoverContext recover_context_;

db/event_helpers.cc

@@ -81,9 +81,8 @@ void EventHelpers::LogAndNotifyTableFileCreationFinished(
     JSONWriter jwriter;
     AppendCurrentTime(&jwriter);
     jwriter << "cf_name" << cf_name << "job" << job_id << "event"
-            << "table_file_creation"
-            << "file_number" << fd.GetNumber() << "file_size"
-            << fd.GetFileSize() << "file_checksum"
+            << "table_file_creation" << "file_number" << fd.GetNumber()
+            << "file_size" << fd.GetFileSize() << "file_checksum"
             << Slice(file_checksum).ToString(true) << "file_checksum_func_name"
             << file_checksum_func_name << "smallest_seqno" << fd.smallest_seqno
             << "largest_seqno" << fd.largest_seqno;
@@ -199,8 +198,7 @@ void EventHelpers::LogAndNotifyTableFileDeletion(
   JSONWriter jwriter;
   AppendCurrentTime(&jwriter);
 
-  jwriter << "job" << job_id << "event"
-          << "table_file_deletion"
+  jwriter << "job" << job_id << "event" << "table_file_deletion"
           << "file_number" << file_number;
   if (!status.ok()) {
     jwriter << "status" << status.ToString();
@@ -280,11 +278,11 @@ void EventHelpers::LogAndNotifyBlobFileCreationFinished(
     JSONWriter jwriter;
     AppendCurrentTime(&jwriter);
     jwriter << "cf_name" << cf_name << "job" << job_id << "event"
-            << "blob_file_creation"
-            << "file_number" << file_number << "total_blob_count"
-            << total_blob_count << "total_blob_bytes" << total_blob_bytes
-            << "file_checksum" << file_checksum << "file_checksum_func_name"
-            << file_checksum_func_name << "status" << s.ToString();
+            << "blob_file_creation" << "file_number" << file_number
+            << "total_blob_count" << total_blob_count << "total_blob_bytes"
+            << total_blob_bytes << "file_checksum" << file_checksum
+            << "file_checksum_func_name" << file_checksum_func_name << "status"
+            << s.ToString();
 
     jwriter.EndObject();
     event_logger->Log(jwriter);
@@ -311,8 +309,7 @@ void EventHelpers::LogAndNotifyBlobFileDeletion(
     JSONWriter jwriter;
     AppendCurrentTime(&jwriter);
 
-    jwriter << "job" << job_id << "event"
-            << "blob_file_deletion"
+    jwriter << "job" << job_id << "event" << "blob_file_deletion"
             << "file_number" << file_number;
     if (!status.ok()) {
       jwriter << "status" << status.ToString();