Skip CancelAllBackgroundWork if DBImpl is already closed (#6268)

Summary:
WritePreparedTxnDB calls CancelAllBackgroundWork in its destructor to avoid dangling references to it from background job's SnapshotChecker callback. However, if the DBImpl is already closed, the info log might be closed with it, which causes memory leak when CancelAllBackgroundWork tries to print to the info log. The patch fixes that by calling CancelAllBackgroundWork only if the db is not closed already.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6268

Differential Revision: D19303439

Pulled By: maysamyabandeh

fbshipit-source-id: 4228a6be7e78d43c90630347baa89b008200bd15

.gitignore

@@ -79,3 +79,4 @@ tp2/
 fbcode/
 fbcode
 buckifier/*.pyc
+buckifier/__pycache__

.travis.yml

@@ -6,25 +6,34 @@ os:
 compiler:
   - clang
   - gcc
-osx_image: xcode8.3
+osx_image: xcode9.4
 jdk:
   - openjdk7
 cache:
   - ccache
-  - apt
 
 addons:
   apt:
     sources:
       - ubuntu-toolchain-r-test
     packages:
-      - curl
-      - g++-8
-      - libbz2-dev
       - libgflags-dev
+      - libbz2-dev
+      - liblz4-dev
       - libsnappy-dev
-      - mingw-w64
+      - liblzma-dev  # xv
+      - libzstd-dev
       - zlib1g-dev
+  homebrew:
+    update: true
+    packages:
+      - ccache
+      - gflags
+      - lz4
+      - snappy
+      - xz
+      - zstd
+
 env:
   - TEST_GROUP=platform_dependent # 16-18 minutes
   - TEST_GROUP=1 # 33-35 minutes
@@ -60,15 +69,17 @@ matrix:
   - os : osx
     compiler: gcc
 
-# https://docs.travis-ci.com/user/caching/#ccache-cache
 install:
   - if [ "${TRAVIS_OS_NAME}" == osx ]; then
-      brew install ccache zstd lz4 snappy xz;
       PATH=$PATH:/usr/local/opt/ccache/libexec;
     fi
   - if [ "${JOB_NAME}" == cmake-gcc8 ]; then
+      sudo apt-get install -y g++-8;
       CC=gcc-8 && CXX=g++-8;
     fi
+  - if [ "${JOB_NAME}" == cmake-mingw ]; then
+      sudo apt-get install -y mingw-w64 ;
+    fi
   - if [[ "${JOB_NAME}" == cmake* ]] && [ "${TRAVIS_OS_NAME}" == linux ]; then
       mkdir cmake-dist && curl -sfSL https://github.com/Kitware/CMake/releases/download/v3.14.5/cmake-3.14.5-Linux-x86_64.tar.gz | tar --strip-components=1 -C cmake-dist -xz && export PATH=$PWD/cmake-dist/bin:$PATH;
     fi
@@ -89,13 +100,13 @@ script:
       OPT=-DTRAVIS V=1 ROCKSDBTESTS_END=db_block_cache_test make -j4 all_but_some_tests check_some
       ;;
     1)
-      OPT=-DTRAVIS V=1 ROCKSDBTESTS_START=db_block_cache_test ROCKSDBTESTS_END=full_filter_block_test make -j4 check_some
+      OPT=-DTRAVIS V=1 ROCKSDBTESTS_START=db_block_cache_test ROCKSDBTESTS_END=db_iter_test make -j4 check_some
       ;;
     2)
-      OPT=-DTRAVIS V=1 make -j4 tools && OPT=-DTRAVIS V=1 ROCKSDBTESTS_START=full_filter_block_test ROCKSDBTESTS_END=write_batch_with_index_test make -j4 check_some
+      OPT=-DTRAVIS V=1 make -j4 tools && OPT=-DTRAVIS V=1 ROCKSDBTESTS_START=db_iter_test ROCKSDBTESTS_END=options_file_test make -j4 check_some
       ;;
     3)
-      OPT=-DTRAVIS V=1 ROCKSDBTESTS_START=write_batch_with_index_test ROCKSDBTESTS_END=write_prepared_transaction_test make -j4 check_some
+      OPT=-DTRAVIS V=1 ROCKSDBTESTS_START=options_file_test ROCKSDBTESTS_END=write_prepared_transaction_test make -j4 check_some
       ;;
     4)
       OPT=-DTRAVIS V=1 ROCKSDBTESTS_START=write_prepared_transaction_test make -j4 check_some
@@ -113,7 +124,7 @@ script:
       ;;
     cmake-mingw)
       sudo update-alternatives --set x86_64-w64-mingw32-g++ /usr/bin/x86_64-w64-mingw32-g++-posix;
-      mkdir build && cd build && cmake -DJNI=1 .. -DCMAKE_C_COMPILER=x86_64-w64-mingw32-gcc -DCMAKE_CXX_COMPILER=x86_64-w64-mingw32-g++ -DCMAKE_SYSTEM_NAME=Windows && make -j4 rocksdb rocksdbjni
+      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
       ;;
     cmake*)
       mkdir build && cd build && cmake -DJNI=1 .. -DCMAKE_BUILD_TYPE=Release && make -j4 rocksdb rocksdbjni
@@ -122,5 +133,3 @@ script:
 notifications:
     email:
       - leveldb@fb.com
-    webhooks:
-      - https://buildtimetrend.herokuapp.com/travis

CMakeLists.txt

@@ -14,7 +14,7 @@
 #        cd build
 # 3. Run cmake to generate project files for Windows, add more options to enable required third-party libraries.
 #    See thirdparty.inc for more information.
-#        sample command: cmake -G "Visual Studio 15 Win64" -DWITH_GFLAGS=1 -DWITH_SNAPPY=1 -DWITH_JEMALLOC=1 -DWITH_JNI=1 ..
+#        sample command: cmake -G "Visual Studio 15 Win64" -DCMAKE_BUILD_TYPE=Release -DWITH_GFLAGS=1 -DWITH_SNAPPY=1 -DWITH_JEMALLOC=1 -DWITH_JNI=1 ..
 # 4. Then build the project in debug mode (you may want to add /m[:<N>] flag to run msbuild in <N> parallel threads
 #                                          or simply /m to use all avail cores)
 #        msbuild rocksdb.sln
@@ -45,6 +45,16 @@ if(POLICY CMP0042)
   cmake_policy(SET CMP0042 NEW)
 endif()
 
+if(NOT CMAKE_BUILD_TYPE)
+  if(EXISTS "${CMAKE_SOURCE_DIR}/.git")
+    set(default_build_type "Debug")
+  else()
+    set(default_build_type "RelWithDebInfo")
+  endif()
+  set(CMAKE_BUILD_TYPE "${default_build_type}" CACHE STRING
+    "Default BUILD_TYPE is ${default_build_type}" FORCE)
+endif()
+
 find_program(CCACHE_FOUND ccache)
 if(CCACHE_FOUND)
   set_property(GLOBAL PROPERTY RULE_LAUNCH_COMPILE ccache)
@@ -62,15 +72,22 @@ if (WITH_WINDOWS_UTF8_FILENAMES)
 endif()
 # third-party/folly is only validated to work on Linux and Windows for now.
 # So only turn it on there by default.
-if(CMAKE_SYSTEM_NAME MATCHES "Linux" OR CMAKE_SYSTEM_NAME MATCHES "Windows")
-  option(WITH_FOLLY_DISTRIBUTED_MUTEX "build with folly::DistributedMutex" ON)
+if(CMAKE_SYSTEM_NAME MATCHES "Linux|Windows")
+  if(MSVC AND MSVC_VERSION LESS 1910)
+    # Folly does not compile with MSVC older than VS2017
+    option(WITH_FOLLY_DISTRIBUTED_MUTEX "build with folly::DistributedMutex" OFF)
+  else()
+    option(WITH_FOLLY_DISTRIBUTED_MUTEX "build with folly::DistributedMutex" ON)
+  endif()
 else()
   option(WITH_FOLLY_DISTRIBUTED_MUTEX "build with folly::DistributedMutex" OFF)
 endif()
+
+include(CMakeDependentOption)
+CMAKE_DEPENDENT_OPTION(WITH_GFLAGS "build with GFlags" ON
+  "NOT MSVC;NOT MINGW" OFF)
+
 if(MSVC)
-  # Defaults currently different for GFLAGS.
-  #  We will address find_package work a little later
-  option(WITH_GFLAGS "build with GFlags" OFF)
   option(WITH_XPRESS "build with windows built in compression" OFF)
   include(${CMAKE_CURRENT_SOURCE_DIR}/thirdparty.inc)
 else()
@@ -87,14 +104,11 @@ else()
   endif()
 
   # No config file for this
-  option(WITH_GFLAGS "build with GFlags" ON)
   if(WITH_GFLAGS)
-    find_package(gflags)
-    if(gflags_FOUND)
-      add_definitions(-DGFLAGS=1)
-      include_directories(${gflags_INCLUDE_DIR})
-      list(APPEND THIRDPARTY_LIBS ${gflags_LIBRARIES})
-    endif()
+    find_package(gflags REQUIRED)
+    add_definitions(-DGFLAGS=1)
+    include_directories(${gflags_INCLUDE_DIR})
+    list(APPEND THIRDPARTY_LIBS gflags::gflags)
   endif()
 
   if(WITH_SNAPPY)
@@ -174,7 +188,7 @@ else()
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -W -Wextra -Wall")
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wsign-compare -Wshadow -Wno-unused-parameter -Wno-unused-variable -Woverloaded-virtual -Wnon-virtual-dtor -Wno-missing-field-initializers -Wno-strict-aliasing")
   if(MINGW)
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-format")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-format -fno-asynchronous-unwind-tables")
     add_definitions(-D_POSIX_C_SOURCE=1)
   endif()
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
@@ -545,6 +559,7 @@ set(SOURCES
         env/env_chroot.cc
         env/env_encryption.cc
         env/env_hdfs.cc
+        env/file_system.cc
         env/mock_env.cc
         file/delete_scheduler.cc
         file/file_prefetch_buffer.cc
@@ -599,9 +614,11 @@ set(SOURCES
         table/block_based/data_block_hash_index.cc
         table/block_based/data_block_footer.cc
         table/block_based/filter_block_reader_common.cc
+        table/block_based/filter_policy.cc
         table/block_based/flush_block_policy.cc
         table/block_based/full_filter_block.cc
         table/block_based/index_builder.cc
+        table/block_based/parsed_full_filter_block.cc
         table/block_based/partitioned_filter_block.cc
         table/block_based/uncompression_dict_reader.cc
         table/block_fetcher.cc
@@ -629,7 +646,6 @@ set(SOURCES
         test_util/testutil.cc
         test_util/transaction_test_util.cc
         tools/block_cache_analyzer/block_cache_trace_analyzer.cc
-        tools/db_bench_tool.cc
         tools/dump/db_dump_tool.cc
         tools/ldb_cmd.cc
         tools/ldb_tool.cc
@@ -637,7 +653,6 @@ set(SOURCES
         tools/trace_analyzer_tool.cc
         trace_replay/trace_replay.cc
         trace_replay/block_cache_tracer.cc
-        util/bloom.cc
         util/coding.cc
         util/compaction_job_stats_impl.cc
         util/comparator.cc
@@ -645,7 +660,6 @@ set(SOURCES
         util/concurrent_task_limiter_impl.cc
         util/crc32c.cc
         util/dynamic_bloom.cc
-        util/filter_policy.cc
         util/hash.cc
         util/murmurhash.cc
         util/random.cc
@@ -753,6 +767,7 @@ else()
   list(APPEND SOURCES
     port/port_posix.cc
     env/env_posix.cc
+    env/fs_posix.cc
     env/io_posix.cc)
 endif()
 
@@ -767,7 +782,8 @@ endif()
 
 set(ROCKSDB_STATIC_LIB rocksdb${ARTIFACT_SUFFIX})
 set(ROCKSDB_SHARED_LIB rocksdb-shared${ARTIFACT_SUFFIX})
-set(ROCKSDB_IMPORT_LIB ${ROCKSDB_SHARED_LIB})
+
+option(ROCKSDB_BUILD_SHARED "Build shared versions of the RocksDB libraries" ON)
 
 option(WITH_LIBRADOS "Build with librados" OFF)
 if(WITH_LIBRADOS)
@@ -777,41 +793,45 @@ if(WITH_LIBRADOS)
 endif()
 
 if(WIN32)
-  set(SYSTEM_LIBS ${SYSTEM_LIBS} Shlwapi.lib Rpcrt4.lib)
-  set(LIBS ${ROCKSDB_STATIC_LIB} ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})
+  set(SYSTEM_LIBS ${SYSTEM_LIBS} shlwapi.lib rpcrt4.lib)
 else()
   set(SYSTEM_LIBS ${CMAKE_THREAD_LIBS_INIT})
-  set(LIBS ${ROCKSDB_SHARED_LIB} ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})
-
-  add_library(${ROCKSDB_SHARED_LIB} SHARED ${SOURCES})
-  target_link_libraries(${ROCKSDB_SHARED_LIB}
-    ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})
-  set_target_properties(${ROCKSDB_SHARED_LIB} PROPERTIES
-                        LINKER_LANGUAGE CXX
-                        VERSION ${rocksdb_VERSION}
-                        SOVERSION ${rocksdb_VERSION_MAJOR}
-                        CXX_STANDARD 11
-                        OUTPUT_NAME "rocksdb")
 endif()
 
 add_library(${ROCKSDB_STATIC_LIB} STATIC ${SOURCES})
 target_link_libraries(${ROCKSDB_STATIC_LIB}
   ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})
 
-if(WIN32)
-  add_library(${ROCKSDB_IMPORT_LIB} SHARED ${SOURCES})
-  target_link_libraries(${ROCKSDB_IMPORT_LIB}
+if(ROCKSDB_BUILD_SHARED)
+  add_library(${ROCKSDB_SHARED_LIB} SHARED ${SOURCES})
+  target_link_libraries(${ROCKSDB_SHARED_LIB}
     ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})
-  set_target_properties(${ROCKSDB_IMPORT_LIB} PROPERTIES
-    COMPILE_DEFINITIONS "ROCKSDB_DLL;ROCKSDB_LIBRARY_EXPORTS")
-  if(MSVC)
-    set_target_properties(${ROCKSDB_STATIC_LIB} PROPERTIES
-      COMPILE_FLAGS "/Fd${CMAKE_CFG_INTDIR}/${ROCKSDB_STATIC_LIB}.pdb")
-    set_target_properties(${ROCKSDB_IMPORT_LIB} PROPERTIES
-      COMPILE_FLAGS "/Fd${CMAKE_CFG_INTDIR}/${ROCKSDB_IMPORT_LIB}.pdb")
+
+  if(WIN32)
+    set_target_properties(${ROCKSDB_SHARED_LIB} PROPERTIES
+      COMPILE_DEFINITIONS "ROCKSDB_DLL;ROCKSDB_LIBRARY_EXPORTS")
+    if(MSVC)
+      set_target_properties(${ROCKSDB_STATIC_LIB} PROPERTIES
+        COMPILE_FLAGS "/Fd${CMAKE_CFG_INTDIR}/${ROCKSDB_STATIC_LIB}.pdb")
+      set_target_properties(${ROCKSDB_SHARED_LIB} PROPERTIES
+        COMPILE_FLAGS "/Fd${CMAKE_CFG_INTDIR}/${ROCKSDB_SHARED_LIB}.pdb")
+    endif()
+  else()
+    set_target_properties(${ROCKSDB_SHARED_LIB} PROPERTIES
+                          LINKER_LANGUAGE CXX
+                          VERSION ${rocksdb_VERSION}
+                          SOVERSION ${rocksdb_VERSION_MAJOR}
+                          CXX_STANDARD 11
+                          OUTPUT_NAME "rocksdb")
   endif()
 endif()
 
+if(ROCKSDB_BUILD_SHARED AND NOT WIN32)
+  set(ROCKSDB_LIB ${ROCKSDB_SHARED_LIB})
+else()
+  set(ROCKSDB_LIB ${ROCKSDB_STATIC_LIB})
+endif()
+
 option(WITH_JNI "build with JNI" OFF)
 if(WITH_JNI OR JNI)
   message(STATUS "JNI library is enabled")
@@ -858,15 +878,17 @@ if(NOT WIN32 OR ROCKSDB_INSTALL_ON_WINDOWS)
     INCLUDES DESTINATION "${CMAKE_INSTALL_INCLUDEDIR}"
   )
 
-  install(
-    TARGETS ${ROCKSDB_SHARED_LIB}
-    EXPORT RocksDBTargets
-    COMPONENT runtime
-    ARCHIVE DESTINATION "${CMAKE_INSTALL_LIBDIR}"
-    RUNTIME DESTINATION "${CMAKE_INSTALL_BINDIR}"
-    LIBRARY DESTINATION "${CMAKE_INSTALL_LIBDIR}"
-    INCLUDES DESTINATION "${CMAKE_INSTALL_INCLUDEDIR}"
-  )
+  if(ROCKSDB_BUILD_SHARED)
+    install(
+      TARGETS ${ROCKSDB_SHARED_LIB}
+      EXPORT RocksDBTargets
+      COMPONENT runtime
+      ARCHIVE DESTINATION "${CMAKE_INSTALL_LIBDIR}"
+      RUNTIME DESTINATION "${CMAKE_INSTALL_BINDIR}"
+      LIBRARY DESTINATION "${CMAKE_INSTALL_LIBDIR}"
+      INCLUDES DESTINATION "${CMAKE_INSTALL_INCLUDEDIR}"
+    )
+  endif()
 
   install(
     EXPORT RocksDBTargets
@@ -884,9 +906,15 @@ if(NOT WIN32 OR ROCKSDB_INSTALL_ON_WINDOWS)
   )
 endif()
 
-option(WITH_TESTS "build with tests" ON)
+add_subdirectory(third-party/gtest-1.8.1/fused-src/gtest)
+add_library(testharness STATIC
+  test_util/testharness.cc)
+target_link_libraries(testharness gtest)
+
+# Tests are excluded from Release builds
+CMAKE_DEPENDENT_OPTION(WITH_TESTS "build with tests" ON
+  "CMAKE_BUILD_TYPE STREQUAL Debug" OFF)
 if(WITH_TESTS)
-  add_subdirectory(third-party/gtest-1.8.1/fused-src/gtest)
   set(TESTS
         cache/cache_test.cc
         cache/lru_cache_test.cc
@@ -1002,6 +1030,7 @@ if(WITH_TESTS)
         util/filelock_test.cc
         util/hash_test.cc
         util/heap_test.cc
+        util/random_test.cc
         util/rate_limiter_test.cc
         util/repeatable_thread_test.cc
         util/slice_transform_test.cc
@@ -1040,23 +1069,6 @@ if(WITH_TESTS)
     list(APPEND TESTS third-party/folly/folly/synchronization/test/DistributedMutexTest.cpp)
   endif()
 
-  set(BENCHMARKS
-    cache/cache_bench.cc
-    memtable/memtablerep_bench.cc
-    db/range_del_aggregator_bench.cc
-    tools/db_bench.cc
-    table/table_reader_bench.cc
-    utilities/persistent_cache/hash_table_bench.cc)
-  add_library(testharness OBJECT test_util/testharness.cc)
-  foreach(sourcefile ${BENCHMARKS})
-    get_filename_component(exename ${sourcefile} NAME_WE)
-    add_executable(${exename}${ARTIFACT_SUFFIX} ${sourcefile}
-      $<TARGET_OBJECTS:testharness>)
-    target_link_libraries(${exename}${ARTIFACT_SUFFIX} gtest ${LIBS})
-  endforeach(sourcefile ${BENCHMARKS})
-
-  # For test util library that is build only in DEBUG mode
-  # and linked to tests. Add test only code that is not #ifdefed for Release here.
   set(TESTUTIL_SOURCE
       db/db_test_util.cc
       monitoring/thread_status_updater_debug.cc
@@ -1064,11 +1076,11 @@ if(WITH_TESTS)
       test_util/fault_injection_test_env.cc
       utilities/cassandra/test_utils.cc
   )
-  # test utilities are only build in debug
   enable_testing()
   add_custom_target(check COMMAND ${CMAKE_CTEST_COMMAND})
   set(TESTUTILLIB testutillib${ARTIFACT_SUFFIX})
   add_library(${TESTUTILLIB} STATIC ${TESTUTIL_SOURCE})
+  target_link_libraries(${TESTUTILLIB} ${ROCKSDB_LIB})
   if(MSVC)
     set_target_properties(${TESTUTILLIB} PROPERTIES COMPILE_FLAGS "/Fd${CMAKE_CFG_INTDIR}/testutillib${ARTIFACT_SUFFIX}.pdb")
   endif()
@@ -1078,46 +1090,86 @@ if(WITH_TESTS)
         EXCLUDE_FROM_DEFAULT_BUILD_RELWITHDEBINFO 1
         )
 
-  # Tests are excluded from Release builds
-  set(TEST_EXES ${TESTS})
-
-  foreach(sourcefile ${TEST_EXES})
+  foreach(sourcefile ${TESTS})
       get_filename_component(exename ${sourcefile} NAME_WE)
-      add_executable(${CMAKE_PROJECT_NAME}_${exename}${ARTIFACT_SUFFIX} ${sourcefile}
-        $<TARGET_OBJECTS:testharness>)
+      add_executable(${CMAKE_PROJECT_NAME}_${exename}${ARTIFACT_SUFFIX} ${sourcefile})
       set_target_properties(${CMAKE_PROJECT_NAME}_${exename}${ARTIFACT_SUFFIX}
         PROPERTIES EXCLUDE_FROM_DEFAULT_BUILD_RELEASE 1
         EXCLUDE_FROM_DEFAULT_BUILD_MINRELEASE 1
         EXCLUDE_FROM_DEFAULT_BUILD_RELWITHDEBINFO 1
         OUTPUT_NAME ${exename}${ARTIFACT_SUFFIX}
         )
-      target_link_libraries(${CMAKE_PROJECT_NAME}_${exename}${ARTIFACT_SUFFIX} testutillib${ARTIFACT_SUFFIX} gtest ${LIBS})
+      target_link_libraries(${CMAKE_PROJECT_NAME}_${exename}${ARTIFACT_SUFFIX} testutillib${ARTIFACT_SUFFIX} testharness gtest ${ROCKSDB_LIB})
       if(NOT "${exename}" MATCHES "db_sanity_test")
         add_test(NAME ${exename} COMMAND ${exename}${ARTIFACT_SUFFIX})
         add_dependencies(check ${CMAKE_PROJECT_NAME}_${exename}${ARTIFACT_SUFFIX})
       endif()
-  endforeach(sourcefile ${TEST_EXES})
+  endforeach(sourcefile ${TESTS})
 
-  # C executables must link to a shared object
-  set(C_TESTS db/c_test.c)
-  set(C_TEST_EXES ${C_TESTS})
+  if(WIN32)
+    # C executables must link to a shared object
+    if(ROCKSDB_BUILD_SHARED)
+      set(ROCKSDB_LIB_FOR_C ${ROCKSDB_SHARED_LIB})
+    else()
+      set(ROCKSDB_LIB_FOR_C OFF)
+    endif()
+  else()
+    set(ROCKSDB_LIB_FOR_C ${ROCKSDB_LIB})
+  endif()
 
-  foreach(sourcefile ${C_TEST_EXES})
-      string(REPLACE ".c" "" exename ${sourcefile})
-      string(REGEX REPLACE "^((.+)/)+" "" exename ${exename})
-      add_executable(${exename}${ARTIFACT_SUFFIX} ${sourcefile})
-      set_target_properties(${exename}${ARTIFACT_SUFFIX}
-        PROPERTIES EXCLUDE_FROM_DEFAULT_BUILD_RELEASE 1
-        EXCLUDE_FROM_DEFAULT_BUILD_MINRELEASE 1
-        EXCLUDE_FROM_DEFAULT_BUILD_RELWITHDEBINFO 1
-        )
-      target_link_libraries(${exename}${ARTIFACT_SUFFIX} ${ROCKSDB_IMPORT_LIB} testutillib${ARTIFACT_SUFFIX})
-      add_test(NAME ${exename} COMMAND ${exename}${ARTIFACT_SUFFIX})
-      add_dependencies(check ${exename}${ARTIFACT_SUFFIX})
-  endforeach(sourcefile ${C_TEST_EXES})
+  if(ROCKSDB_LIB_FOR_C)
+    set(C_TESTS db/c_test.c)
+    # C executables must link to a shared object
+    add_executable(c_test db/c_test.c)
+    target_link_libraries(c_test ${ROCKSDB_SHARED_LIB} testharness)
+    add_test(NAME c_test COMMAND c_test${ARTIFACT_SUFFIX})
+    add_dependencies(check c_test)
+  endif()
+endif()
+
+option(WITH_BENCHMARK_TOOLS "build with benchmarks" ON)
+if(WITH_BENCHMARK_TOOLS)
+  add_executable(db_bench
+    tools/db_bench.cc
+    tools/db_bench_tool.cc)
+  target_link_libraries(db_bench
+    ${ROCKSDB_LIB})
+
+  add_executable(cache_bench
+    cache/cache_bench.cc)
+  target_link_libraries(cache_bench
+    ${ROCKSDB_LIB})
+
+  add_executable(memtablerep_bench
+    memtable/memtablerep_bench.cc)
+  target_link_libraries(memtablerep_bench
+    ${ROCKSDB_LIB})
+
+  add_executable(range_del_aggregator_bench
+    db/range_del_aggregator_bench.cc)
+  target_link_libraries(range_del_aggregator_bench
+    ${ROCKSDB_LIB})
+
+  add_executable(table_reader_bench
+    table/table_reader_bench.cc)
+  target_link_libraries(table_reader_bench
+    ${ROCKSDB_LIB} testharness)
+
+  add_executable(filter_bench
+    util/filter_bench.cc)
+  target_link_libraries(filter_bench
+    ${ROCKSDB_LIB})
+
+  add_executable(hash_table_bench
+    utilities/persistent_cache/hash_table_bench.cc)
+  target_link_libraries(hash_table_bench
+    ${ROCKSDB_LIB})
 endif()
 
 option(WITH_TOOLS "build with tools" ON)
 if(WITH_TOOLS)
   add_subdirectory(tools)
+  add_subdirectory(db_stress_tool)
+  add_custom_target(tools
+    DEPENDS ${tool_deps})
 endif()

HISTORY.md

@@ -1,32 +1,115 @@
 # Rocksdb Change Log
 ## Unreleased
+### Public API Change
+* Added a rocksdb::FileSystem class in include/rocksdb/file_system.h to encapsulate file creation/read/write operations, and an option DBOptions::file_system to allow a user to pass in an instance of rocksdb::FileSystem. If its a non-null value, this will take precendence over DBOptions::env for file operations. A new API rocksdb::FileSystem::Default() returns a platform default object. The DBOptions::env option and Env::Default() API will continue to be used for threading and other OS related functions, and where DBOptions::file_system is not specified, for file operations. For storage developers who are accustomed to rocksdb::Env, the interface in rocksdb::FileSystem is new and will probably undergo some changes as more storage systems are ported to it from rocksdb::Env. As of now, no env other than Posix has been ported to the new interface.
+* A new rocksdb::NewSstFileManager() API that allows the caller to pass in separate Env and FileSystem objects.
+* Changed Java API for RocksDB.keyMayExist functions to use Holder<byte[]> instead of StringBuilder, so that retrieved values need not decode to Strings.
+* A new `OptimisticTransactionDBOptions` Option that allows users to configure occ validation policy. The default policy changes from kValidateSerial to kValidateParallel to reduce mutex contention.
+
+### Bug Fixes
+* Fix a bug that can cause unnecessary bg thread to be scheduled(#6104).
+* Fix a bug in which a snapshot read could be affected by a DeleteRange after the snapshot (#6062).
+* Fix crash caused by concurrent CF iterations and drops(#6147).
+* Fixed two performance issues related to memtable history trimming. First, a new SuperVersion is now created only if some memtables were actually trimmed. Second, trimming is only scheduled if there is at least one flushed memtable that is kept in memory for the purposes of transaction conflict checking.
+* Fix a bug in WriteBatchWithIndex::MultiGetFromBatchAndDB, which is called by Transaction::MultiGet, that causes due to stale pointer access when the number of keys is > 32
+* BlobDB no longer updates the SST to blob file mapping upon failed compactions.
+* Fixed a bug where BlobDB was comparing the `ColumnFamilyHandle` pointers themselves instead of only the column family IDs when checking whether an API call uses the default column family or not.
+* Fix a race condition for cfd->log_number_ between manifest switch and memtable switch (PR 6249) when number of column families is greater than 1.
+
+### New Features
+* It is now possible to enable periodic compactions for the base DB when using BlobDB.
+* BlobDB now garbage collects non-TTL blobs when `enable_garbage_collection` is set to `true` in `BlobDBOptions`. Garbage collection is performed during compaction: any valid blobs located in the oldest N files (where N is the number of non-TTL blob files multiplied by the value of `BlobDBOptions::garbage_collection_cutoff`) encountered during compaction get relocated to new blob files, and old blob files are dropped once they are no longer needed. Note: we recommend enabling periodic compactions for the base DB when using this feature to deal with the case when some old blob files are kept alive by SSTs that otherwise do not get picked for compaction.
+* `db_bench` now supports the `garbage_collection_cutoff` option for BlobDB.
+
+## 6.6.0 (11/25/2019)
+### Bug Fixes
+* Fix data corruption caused by output of intra-L0 compaction on ingested file not being placed in correct order in L0.
+* Fix a data race between Version::GetColumnFamilyMetaData() and Compaction::MarkFilesBeingCompacted() for access to being_compacted (#6056). The current fix acquires the db mutex during Version::GetColumnFamilyMetaData(), which may cause regression.
+* Fix a bug in DBIter that is_blob_ state isn't updated when iterating backward using seek.
+* Fix a bug when format_version=3, partitioned filters, and prefix search are used in conjunction. The bug could result into Seek::(prefix) returning NotFound for an existing prefix.
+* Revert the feature "Merging iterator to avoid child iterator reseek for some cases (#5286)" since it might cause strong results when reseek happens with a different iterator upper bound.
+* Fix a bug causing a crash during ingest external file when background compaction cause severe error (file not found).
+* Fix a bug when partitioned filters and prefix search are used in conjunction, ::SeekForPrev could return invalid for an existing prefix. ::SeekForPrev might be called by the user, or internally on ::Prev, or within ::Seek if the return value involves Delete or a Merge operand.
+* Fix OnFlushCompleted fired before flush result persisted in MANIFEST when there's concurrent flush job. The bug exists since OnFlushCompleted was introduced in rocksdb 3.8.
+* Fixed an sst_dump crash on some plain table SST files.
+* Fixed a memory leak in some error cases of opening plain table SST files.
+* Fix a bug when a crash happens while calling WriteLevel0TableForRecovery for multiple column families, leading to a column family's log number greater than the first corrutped log number when the DB is being opened in PointInTime recovery mode during next recovery attempt (#5856).
+
+### New Features
+* Universal compaction to support options.periodic_compaction_seconds. A full compaction will be triggered if any file is over the threshold.
+* `GetLiveFilesMetaData` and `GetColumnFamilyMetaData` now expose the file number of SST files as well as the oldest blob file referenced by each SST.
+* A batched MultiGet API (DB::MultiGet()) that supports retrieving keys from multiple column families.
+* Full and partitioned filters in the block-based table use an improved Bloom filter implementation, enabled with format_version 5 (or above) because previous releases cannot read this filter. This replacement is faster and more accurate, especially for high bits per key or millions of keys in a single (full) filter. For example, the new Bloom filter has the same false positive rate at 9.55 bits per key as the old one at 10 bits per key, and a lower false positive rate at 16 bits per key than the old one at 100 bits per key.
+* Added AVX2 instructions to USE_SSE builds to accelerate the new Bloom filter and XXH3-based hash function on compatible x86_64 platforms (Haswell and later, ~2014).
+* Support options.ttl or options.periodic_compaction_seconds with options.max_open_files = -1. File's oldest ancester time and file creation time will be written to manifest. If it is availalbe, this information will be used instead of creation_time and file_creation_time in table properties.
+* Setting options.ttl for universal compaction now has the same meaning as setting periodic_compaction_seconds.
+* SstFileMetaData also returns file creation time and oldest ancester time.
+* The `sst_dump` command line tool `recompress` command now displays how many blocks were compressed and how many were not, in particular how many were not compressed because the compression ratio was not met (12.5% threshold for GoodCompressionRatio), as seen in the `number.block.not_compressed` counter stat since version 6.0.0.
+* The block cache usage is now takes into account the overhead of metadata per each entry. This results into more accurate management of memory. A side-effect of this feature is that less items are fit into the block cache of the same size, which would result to higher cache miss rates. This can be remedied by increasing the block cache size or passing kDontChargeCacheMetadata to its constuctor to restore the old behavior.
+* When using BlobDB, a mapping is maintained and persisted in the MANIFEST between each SST file and the oldest non-TTL blob file it references.
+* `db_bench` now supports and by default issues non-TTL Puts to BlobDB. TTL Puts can be enabled by specifying a non-zero value for the `blob_db_max_ttl_range` command line parameter explicitly.
+* `sst_dump` now supports printing BlobDB blob indexes in a human-readable format. This can be enabled by specifying the `decode_blob_index` flag on the command line.
+* A number of new information elements are now exposed through the EventListener interface. For flushes, the file numbers of the new SST file and the oldest blob file referenced by the SST are propagated. For compactions, the level, file number, and the oldest blob file referenced are passed to the client for each compaction input and output file.
+
+### Public API Change
+* RocksDB release 4.1 or older will not be able to open DB generated by the new release. 4.2 was released on Feb 23, 2016.
+* TTL Compactions in Level compaction style now initiate successive cascading compactions on a key range so that it reaches the bottom level quickly on TTL expiry. `creation_time` table property for compaction output files is now set to the minimum of the creation times of all compaction inputs.
+* With FIFO compaction style, options.periodic_compaction_seconds will have the same meaning as options.ttl. Whichever stricter will be used. With the default options.periodic_compaction_seconds value with options.ttl's default of 0, RocksDB will give a default of 30 days.
+* Added an API GetCreationTimeOfOldestFile(uint64_t* creation_time) to get the file_creation_time of the oldest SST file in the DB.
+* FilterPolicy now exposes additional API to make it possible to choose filter configurations based on context, such as table level and compaction style. See `LevelAndStyleCustomFilterPolicy` in db_bloom_filter_test.cc. While most existing custom implementations of FilterPolicy should continue to work as before, those wrapping the return of NewBloomFilterPolicy will require overriding new function `GetBuilderWithContext()`, because calling `GetFilterBitsBuilder()` on the FilterPolicy returned by NewBloomFilterPolicy is no longer supported.
+* An unlikely usage of FilterPolicy is no longer supported. Calling GetFilterBitsBuilder() on the FilterPolicy returned by NewBloomFilterPolicy will now cause an assertion violation in debug builds, because RocksDB has internally migrated to a more elaborate interface that is expected to evolve further. Custom implementations of FilterPolicy should work as before, except those wrapping the return of NewBloomFilterPolicy, which will require a new override of a protected function in FilterPolicy.
+* NewBloomFilterPolicy now takes bits_per_key as a double instead of an int. This permits finer control over the memory vs. accuracy trade-off in the new Bloom filter implementation and should not change source code compatibility.
+* The option BackupableDBOptions::max_valid_backups_to_open is now only used when opening BackupEngineReadOnly. When opening a read/write BackupEngine, anything but the default value logs a warning and is treated as the default. This change ensures that backup deletion has proper accounting of shared files to ensure they are deleted when no longer referenced by a backup.
+* Deprecate `snap_refresh_nanos` option.
+* Added DisableManualCompaction/EnableManualCompaction to stop and resume manual compaction.
+* Add TryCatchUpWithPrimary() to StackableDB in non-LITE mode.
+* Add a new Env::LoadEnv() overloaded function to return a shared_ptr to Env.
+* Flush sets file name to "(nil)" for OnTableFileCreationCompleted() if the flush does not produce any L0. This can happen if the file is empty thus delete by RocksDB.
+
+### Default Option Changes
+* Changed the default value of periodic_compaction_seconds to `UINT64_MAX - 1` which allows RocksDB to auto-tune periodic compaction scheduling. When using the default value, periodic compactions are now auto-enabled if a compaction filter is used. A value of `0` will turn off the feature completely.
+* Changed the default value of ttl to `UINT64_MAX - 1` which allows RocksDB to auto-tune ttl value. When using the default value, TTL will be auto-enabled to 30 days, when the feature is supported. To revert the old behavior, you can explicitly set it to 0.
+
+### Performance Improvements
+* For 64-bit hashing, RocksDB is standardizing on a slightly modified preview version of XXH3. This function is now used for many non-persisted hashes, along with fastrange64() in place of the modulus operator, and some benchmarks show a slight improvement.
+* Level iterator to invlidate the iterator more often in prefix seek and the level is filtered out by prefix bloom.
+
+## 6.5.2 (11/15/2019)
+### Bug Fixes
+* Fix a assertion failure in MultiGet() when BlockBasedTableOptions::no_block_cache is true and there is no compressed block cache
+* Fix a buffer overrun problem in BlockBasedTable::MultiGet() when compression is enabled and no compressed block cache is configured.
+* If a call to BackupEngine::PurgeOldBackups or BackupEngine::DeleteBackup suffered a crash, power failure, or I/O error, files could be left over from old backups that could only be purged with a call to GarbageCollect. Any call to PurgeOldBackups, DeleteBackup, or GarbageCollect should now suffice to purge such files.
+
+## 6.5.1 (10/16/2019)
+### Bug Fixes
+* Revert the feature "Merging iterator to avoid child iterator reseek for some cases (#5286)" since it might cause strange results when reseek happens with a different iterator upper bound.
+* Fix a bug in BlockBasedTableIterator that might return incorrect results when reseek happens with a different iterator upper bound.
+* Fix a bug when partitioned filters and prefix search are used in conjunction, ::SeekForPrev could return invalid for an existing prefix. ::SeekForPrev might be called by the user, or internally on ::Prev, or within ::Seek if the return value involves Delete or a Merge operand.
+
+## 6.5.0 (9/13/2019)
 ### Bug Fixes
 * Fixed a number of data races in BlobDB.
 * Fix a bug where the compaction snapshot refresh feature is not disabled as advertised when `snap_refresh_nanos` is set to 0..
 * Fix bloom filter lookups by the MultiGet batching API when BlockBasedTableOptions::whole_key_filtering is false, by checking that a key is in the perfix_extractor domain and extracting the prefix before looking up.
 * Fix a bug in file ingestion caused by incorrect file number allocation when the number of column families involved in the ingestion exceeds 2.
-* Fix a bug when format_version=3, partitioned fitlers, and prefix search are used in conjunction. The bug could result into Seek::(prefix) returning NotFound for an existing prefix.
-* Revert the feature "Merging iterator to avoid child iterator reseek for some cases (#5286)" since it might cause strong results when reseek happens with a different iterator upper bound.
+
 ### New Features
 * Introduced DBOptions::max_write_batch_group_size_bytes to configure maximum limit on number of bytes that are written in a single batch of WAL or memtable write. It is followed when the leader write size is larger than 1/8 of this limit.
 * VerifyChecksum() by default will issue readahead. Allow ReadOptions to be passed in to those functions to override the readhead size. For checksum verifying before external SST file ingestion, a new option IngestExternalFileOptions.verify_checksums_readahead_size, is added for this readahead setting.
 * When user uses options.force_consistency_check in RocksDb, instead of crashing the process, we now pass the error back to the users without killing the process.
 * Add an option `memtable_insert_hint_per_batch` to WriteOptions. If it is true, each WriteBatch will maintain its own insert hints for each memtable in concurrent write. See include/rocksdb/options.h for more details.
-* The `sst_dump` command line tool `recompress` command now displays how many blocks were compressed and how many were not, in particular how many were not compressed because the compression ratio was not met (12.5% threshold for GoodCompressionRatio), as seen in the `number.block.not_compressed` counter stat since version 6.0.0.
-* The block cache usage is now takes into account the overhead of metadata per each entry. This results into more accurate managment of memory. A side-effect of this feature is that less items are fit into the block cache of the same size, which would result to higher cache miss rates. This can be remedied by increasing the block cache size or passing kDontChargeCacheMetadata to its constuctor to restore the old behavior.
+
 ### Public API Change
 * Added max_write_buffer_size_to_maintain option to better control memory usage of immutable memtables.
 * Added a lightweight API GetCurrentWalFile() to get last live WAL filename and size. Meant to be used as a helper for backup/restore tooling in a larger ecosystem such as MySQL with a MyRocks storage engine.
 * The MemTable Bloom filter, when enabled, now always uses cache locality. Options::bloom_locality now only affects the PlainTable SST format.
-* Deprecate `snap_refresh_nanos` option.
-* Added DisableManualCompaction/EnableManualCompaction to stop and resume manual compaction.
-* Add TryCatchUpWithPrimary() to StackableDB in non-LITE mode.
+
 ### Performance Improvements
 * Improve the speed of the MemTable Bloom filter, reducing the write overhead of enabling it by 1/3 to 1/2, with similar benefit to read performance.
 
 ## 6.4.0 (7/30/2019)
 ### Default Option Change
-* LRUCacheOptions.high_pri_pool_ratio is set to 0.5 (previously 0.0) by default, which means that by default midpoint insertion is enabled. The same change is made for the default value of high_pri_pool_ratio argument in NewLRUCache(). When block cache is not explictly created, the small block cache created by BlockBasedTable will still has this option to be 0.0.
+* LRUCacheOptions.high_pri_pool_ratio is set to 0.5 (previously 0.0) by default, which means that by default midpoint insertion is enabled. The same change is made for the default value of high_pri_pool_ratio argument in NewLRUCache(). When block cache is not explicitly created, the small block cache created by BlockBasedTable will still has this option to be 0.0.
 * Change BlockBasedTableOptions.cache_index_and_filter_blocks_with_high_priority's default value from false to true.
 
 ### Public API Change
@@ -50,7 +133,7 @@
 * Support loading custom objects in unit tests. In the affected unit tests, RocksDB will create custom Env objects based on environment variable TEST_ENV_URI. Users need to make sure custom object types are properly registered. For example, a static library should expose a `RegisterCustomObjects` function. By linking the unit test binary with the static library, the unit test can execute this function.
 
 ### Performance Improvements
-* Reduce iterator key comparision for upper/lower bound check.
+* Reduce iterator key comparison for upper/lower bound check.
 * Improve performance of row_cache: make reads with newer snapshots than data in an SST file share the same cache key, except in some transaction cases.
 * The compression dictionary is no longer copied to a new object upon retrieval.
 
@@ -146,7 +229,7 @@
 * Introduce two more stats levels, kExceptHistogramOrTimers and kExceptTimers.
 * Added a feature to perform data-block sampling for compressibility, and report stats to user.
 * Add support for trace filtering.
-* Add DBOptions.avoid_unnecessary_blocking_io. If true, we avoid file deletion when destorying ColumnFamilyHandle and Iterator. Instead, a job is scheduled to delete the files in background.
+* Add DBOptions.avoid_unnecessary_blocking_io. If true, we avoid file deletion when destroying ColumnFamilyHandle and Iterator. Instead, a job is scheduled to delete the files in background.
 
 ### Public API Change
 * Remove bundled fbson library.
@@ -525,7 +608,7 @@ if set to something > 0 user will see 2 changes in iterators behavior 1) only ke
 
 ## 5.2.0 (02/08/2017)
 ### Public API Change
-* NewLRUCache() will determine number of shard bits automatically based on capacity, if the user doesn't pass one. This also impacts the default block cache when the user doesn't explict provide one.
+* NewLRUCache() will determine number of shard bits automatically based on capacity, if the user doesn't pass one. This also impacts the default block cache when the user doesn't explicit provide one.
 * Change the default of delayed slowdown value to 16MB/s and further increase the L0 stop condition to 36 files.
 * Options::use_direct_writes and Options::use_direct_reads are now ready to use.
 * (Experimental) Two-level indexing that partition the index and creates a 2nd level index on the partitions. The feature can be enabled by setting kTwoLevelIndexSearch as IndexType and configuring index_per_partition.
@@ -596,7 +679,7 @@ if set to something > 0 user will see 2 changes in iterators behavior 1) only ke
 
 ### New Features
 * A tool to migrate DB after options change. See include/rocksdb/utilities/option_change_migration.h.
-* Add ReadOptions.background_purge_on_iterator_cleanup. If true, we avoid file deletion when destorying iterators.
+* Add ReadOptions.background_purge_on_iterator_cleanup. If true, we avoid file deletion when destroying iterators.
 
 ## 4.10.0 (7/5/2016)
 ### Public API Change

Makefile

@@ -439,6 +439,8 @@ BENCHTOOLOBJECTS = $(BENCH_LIB_SOURCES:.cc=.o) $(LIBOBJECTS) $(TESTUTIL)
 
 ANALYZETOOLOBJECTS = $(ANALYZER_LIB_SOURCES:.cc=.o)
 
+STRESSTOOLOBJECTS = $(STRESS_LIB_SOURCES:.cc=.o) $(LIBOBJECTS) $(TESTUTIL)
+
 EXPOBJECTS = $(LIBOBJECTS) $(TESTUTIL)
 
 TESTS = \
@@ -458,6 +460,7 @@ TESTS = \
 	env_test \
 	env_logger_test \
 	hash_test \
+	random_test \
 	thread_local_test \
 	rate_limiter_test \
 	perf_context_test \
@@ -649,7 +652,7 @@ TEST_LIBS = \
 	librocksdb_env_basic_test.a
 
 # TODO: add back forward_iterator_bench, after making it build in all environemnts.
-BENCHMARKS = db_bench table_reader_bench cache_bench memtablerep_bench persistent_cache_bench range_del_aggregator_bench
+BENCHMARKS = db_bench table_reader_bench cache_bench memtablerep_bench filter_bench persistent_cache_bench range_del_aggregator_bench
 
 # if user didn't config LIBNAME, set the default
 ifeq ($(LIBNAME),)
@@ -662,6 +665,7 @@ endif
 endif
 LIBRARY = ${LIBNAME}.a
 TOOLS_LIBRARY = ${LIBNAME}_tools.a
+STRESS_LIBRARY = ${LIBNAME}_stress.a
 
 ROCKSDB_MAJOR = $(shell egrep "ROCKSDB_MAJOR.[0-9]" include/rocksdb/version.h | cut -d ' ' -f 3)
 ROCKSDB_MINOR = $(shell egrep "ROCKSDB_MINOR.[0-9]" include/rocksdb/version.h | cut -d ' ' -f 3)
@@ -740,7 +744,8 @@ endif  # PLATFORM_SHARED_EXT
 	release tags tags0 valgrind_check whitebox_crash_test format static_lib shared_lib all \
 	dbg rocksdbjavastatic rocksdbjava install install-static install-shared uninstall \
 	analyze tools tools_lib \
-	blackbox_crash_test_with_atomic_flush whitebox_crash_test_with_atomic_flush
+	blackbox_crash_test_with_atomic_flush whitebox_crash_test_with_atomic_flush  \
+	blackbox_crash_test_with_txn whitebox_crash_test_with_txn
 
 
 all: $(LIBRARY) $(BENCHMARKS) tools tools_lib test_libs $(TESTS)
@@ -751,6 +756,8 @@ static_lib: $(LIBRARY)
 
 shared_lib: $(SHARED)
 
+stress_lib: $(STRESS_LIBRARY)
+
 tools: $(TOOLS)
 
 tools_lib: $(TOOLS_LIBRARY)
@@ -932,6 +939,7 @@ check: all
 	fi
 	rm -rf $(TMPD)
 ifneq ($(PLATFORM), OS_AIX)
+	python tools/check_all_python.py
 ifeq ($(filter -DROCKSDB_LITE,$(OPT)),)
 	python tools/ldb_test.py
 	sh tools/rocksdb_dump_test.sh
@@ -950,6 +958,8 @@ crash_test: whitebox_crash_test blackbox_crash_test
 
 crash_test_with_atomic_flush: whitebox_crash_test_with_atomic_flush blackbox_crash_test_with_atomic_flush
 
+crash_test_with_txn: whitebox_crash_test_with_txn blackbox_crash_test_with_txn
+
 blackbox_crash_test: db_stress
 	python -u tools/db_crashtest.py --simple blackbox $(CRASH_TEST_EXT_ARGS)
 	python -u tools/db_crashtest.py blackbox $(CRASH_TEST_EXT_ARGS)
@@ -957,6 +967,9 @@ blackbox_crash_test: db_stress
 blackbox_crash_test_with_atomic_flush: db_stress
 	python -u tools/db_crashtest.py --cf_consistency blackbox $(CRASH_TEST_EXT_ARGS)
 
+blackbox_crash_test_with_txn: db_stress
+	python -u tools/db_crashtest.py --txn blackbox $(CRASH_TEST_EXT_ARGS)
+
 ifeq ($(CRASH_TEST_KILL_ODD),)
   CRASH_TEST_KILL_ODD=888887
 endif
@@ -971,6 +984,10 @@ whitebox_crash_test_with_atomic_flush: db_stress
 	python -u tools/db_crashtest.py --cf_consistency whitebox  --random_kill_odd \
       $(CRASH_TEST_KILL_ODD) $(CRASH_TEST_EXT_ARGS)
 
+whitebox_crash_test_with_txn: db_stress
+	python -u tools/db_crashtest.py --txn whitebox --random_kill_odd \
+      $(CRASH_TEST_KILL_ODD) $(CRASH_TEST_EXT_ARGS)
+
 asan_check:
 	$(MAKE) clean
 	COMPILE_WITH_ASAN=1 $(MAKE) check -j32
@@ -986,6 +1003,11 @@ asan_crash_test_with_atomic_flush:
 	COMPILE_WITH_ASAN=1 $(MAKE) crash_test_with_atomic_flush
 	$(MAKE) clean
 
+asan_crash_test_with_txn:
+	$(MAKE) clean
+	COMPILE_WITH_ASAN=1 $(MAKE) crash_test_with_txn
+	$(MAKE) clean
+
 ubsan_check:
 	$(MAKE) clean
 	COMPILE_WITH_UBSAN=1 $(MAKE) check -j32
@@ -1001,6 +1023,11 @@ ubsan_crash_test_with_atomic_flush:
 	COMPILE_WITH_UBSAN=1 $(MAKE) crash_test_with_atomic_flush
 	$(MAKE) clean
 
+ubsan_crash_test_with_txn:
+	$(MAKE) clean
+	COMPILE_WITH_UBSAN=1 $(MAKE) crash_test_with_txn
+	$(MAKE) clean
+
 valgrind_test:
 	ROCKSDB_VALGRIND_RUN=1 DISABLE_JEMALLOC=1 $(MAKE) valgrind_check
 
@@ -1070,6 +1097,9 @@ parallel_check: $(TESTS)
 		TMPD=$(TMPD) J=$(J) db_test=0 parloop;
 
 analyze: clean
+	USE_CLANG=1 $(MAKE) analyze_incremental
+
+analyze_incremental:
 	$(CLANG_SCAN_BUILD) --use-analyzer=$(CLANG_ANALYZER) \
 		--use-c++=$(CXX) --use-cc=$(CC) --status-bugs \
 		-o $(CURDIR)/scan_build_report \
@@ -1098,14 +1128,23 @@ unity_test: db/db_test.o db/db_test_util.o $(TESTHARNESS) $(TOOLLIBOBJECTS) unit
 rocksdb.h rocksdb.cc: build_tools/amalgamate.py Makefile $(LIB_SOURCES) unity.cc
 	build_tools/amalgamate.py -I. -i./include unity.cc -x include/rocksdb/c.h -H rocksdb.h -o rocksdb.cc
 
-clean:
+clean: clean-ext-libraries-all clean-rocks
+
+clean-not-downloaded: clean-ext-libraries-bin clean-rocks
+
+clean-rocks:
 	rm -f $(BENCHMARKS) $(TOOLS) $(TESTS) $(PARALLEL_TEST) $(LIBRARY) $(SHARED)
 	rm -rf $(CLEAN_FILES) ios-x86 ios-arm scan_build_report
 	$(FIND) . -name "*.[oda]" -exec rm -f {} \;
 	$(FIND) . -type f -regex ".*\.\(\(gcda\)\|\(gcno\)\)" -exec rm {} \;
-	rm -rf bzip2* snappy* zlib* lz4* zstd*
 	cd java; $(MAKE) clean
 
+clean-ext-libraries-all:
+	rm -rf bzip2* snappy* zlib* lz4* zstd*
+
+clean-ext-libraries-bin:
+	find . -maxdepth 1 -type d \( -name bzip2\* -or -name snappy\* -or -name zlib\* -or -name lz4\* -or -name zstd\* \) -prune -exec rm -rf {} \;
+
 tags:
 	ctags -R .
 	cscope -b `$(FIND) . -name '*.cc'` `$(FIND) . -name '*.h'` `$(FIND) . -name '*.c'`
@@ -1135,6 +1174,10 @@ $(TOOLS_LIBRARY): $(BENCH_LIB_SOURCES:.cc=.o) $(TOOL_LIB_SOURCES:.cc=.o) $(LIB_S
 	$(AM_V_AR)rm -f $@
 	$(AM_V_at)$(AR) $(ARFLAGS) $@ $^
 
+$(STRESS_LIBRARY): $(LIB_SOURCES:.cc=.o) $(TESTUTIL) $(ANALYZER_LIB_SOURCES:.cc=.o) $(STRESS_LIB_SOURCES:.cc=.o)
+	$(AM_V_AR)rm -f $@
+	$(AM_V_at)$(AR) $(ARFLAGS) $@ $^
+
 librocksdb_env_basic_test.a: env/env_basic_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_V_AR)rm -f $@
 	$(AM_V_at)$(AR) $(ARFLAGS) $@ $^
@@ -1162,7 +1205,10 @@ persistent_cache_bench: utilities/persistent_cache/persistent_cache_bench.o $(LI
 memtablerep_bench: memtable/memtablerep_bench.o $(LIBOBJECTS) $(TESTUTIL)
 	$(AM_LINK)
 
-db_stress: tools/db_stress.o $(LIBOBJECTS) $(TESTUTIL)
+filter_bench: util/filter_bench.o $(LIBOBJECTS) $(TESTUTIL)
+	$(AM_LINK)
+
+db_stress: db_stress_tool/db_stress.o $(STRESSTOOLOBJECTS)
 	$(AM_LINK)
 
 write_stress: tools/write_stress.o $(LIBOBJECTS) $(TESTUTIL)
@@ -1204,6 +1250,9 @@ coding_test: util/coding_test.o $(LIBOBJECTS) $(TESTHARNESS)
 hash_test: util/hash_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
+random_test: util/random_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 option_change_migration_test: utilities/option_change_migration/option_change_migration_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
@@ -1701,7 +1750,7 @@ JAVA_INCLUDE = -I$(JAVA_HOME)/include/ -I$(JAVA_HOME)/include/linux
 ifeq ($(PLATFORM), OS_SOLARIS)
 	ARCH := $(shell isainfo -b)
 else ifeq ($(PLATFORM), OS_OPENBSD)
-	ifneq (,$(filter $(MACHINE), amd64 arm64 aarch64 sparc64))
+	ifneq (,$(filter amd64 ppc64 ppc64le arm64 aarch64 sparc64, $(MACHINE)))
 		ARCH := 64
 	else
 		ARCH := 32
@@ -1710,12 +1759,23 @@ else
 	ARCH := $(shell getconf LONG_BIT)
 endif
 
-ifeq (,$(filter $(MACHINE), ppc arm64 aarch64 sparc64))
-        ROCKSDBJNILIB = librocksdbjni-linux$(ARCH).so
-else
-        ROCKSDBJNILIB = librocksdbjni-linux-$(MACHINE).so
+ifeq ($(shell ldd /usr/bin/env 2>/dev/null | grep -q musl; echo $$?),0)
+        JNI_LIBC = musl
+# GNU LibC (or glibc) is so pervasive we can assume it is the default
+# else
+#        JNI_LIBC = glibc
 endif
-ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-linux$(ARCH).jar
+
+ifneq ($(origin JNI_LIBC), undefined)
+  JNI_LIBC_POSTFIX = -$(JNI_LIBC)
+endif
+
+ifneq (,$(filter ppc% arm64 aarch64 sparc64, $(MACHINE)))
+	ROCKSDBJNILIB = librocksdbjni-linux-$(MACHINE)$(JNI_LIBC_POSTFIX).so
+else
+	ROCKSDBJNILIB = librocksdbjni-linux$(ARCH)$(JNI_LIBC_POSTFIX).so
+endif
+ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-linux$(ARCH)$(JNI_LIBC_POSTFIX).jar
 ROCKSDB_JAR_ALL = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH).jar
 ROCKSDB_JAVADOCS_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-javadoc.jar
 ROCKSDB_SOURCES_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-sources.jar
@@ -1726,15 +1786,15 @@ ZLIB_SHA256 ?= c3e5e9fdd5004dcb542feda5ee4f0ff0744628baf8ed2dd5d66f8ca1197cb1a1
 ZLIB_DOWNLOAD_BASE ?= http://zlib.net
 BZIP2_VER ?= 1.0.6
 BZIP2_SHA256 ?= a2848f34fcd5d6cf47def00461fcb528a0484d8edef8208d6d2e2909dc61d9cd
-BZIP2_DOWNLOAD_BASE ?= https://web.archive.org/web/20180624184835/http://www.bzip.org
+BZIP2_DOWNLOAD_BASE ?= https://downloads.sourceforge.net/project/bzip2
 SNAPPY_VER ?= 1.1.7
 SNAPPY_SHA256 ?= 3dfa02e873ff51a11ee02b9ca391807f0c8ea0529a4924afa645fbf97163f9d4
 SNAPPY_DOWNLOAD_BASE ?= https://github.com/google/snappy/archive
-LZ4_VER ?= 1.8.3
-LZ4_SHA256 ?= 33af5936ac06536805f9745e0b6d61da606a1f8b4cc5c04dd3cbaca3b9b4fc43
+LZ4_VER ?= 1.9.2
+LZ4_SHA256 ?= 658ba6191fa44c92280d4aa2c271b0f4fbc0e34d249578dd05e50e76d0e5efcc
 LZ4_DOWNLOAD_BASE ?= https://github.com/lz4/lz4/archive
-ZSTD_VER ?= 1.4.0
-ZSTD_SHA256 ?= 63be339137d2b683c6d19a9e34f4fb684790e864fee13c7dd40e197a64c705c1
+ZSTD_VER ?= 1.4.4
+ZSTD_SHA256 ?= a364f5162c7d1a455cc915e8e3cf5f4bd8b75d09bc0f53965b0c9ca1383c52c8
 ZSTD_DOWNLOAD_BASE ?= https://github.com/facebook/zstd/archive
 CURL_SSL_OPTS ?= --tlsv1
 
@@ -1788,7 +1848,7 @@ endif
 libbz2.a:
 	-rm -rf bzip2-$(BZIP2_VER)
 ifeq (,$(wildcard ./bzip2-$(BZIP2_VER).tar.gz))
-	curl --output bzip2-$(BZIP2_VER).tar.gz -L ${BZIP2_DOWNLOAD_BASE}/$(BZIP2_VER)/bzip2-$(BZIP2_VER).tar.gz
+	curl --output bzip2-$(BZIP2_VER).tar.gz -L ${CURL_SSL_OPTS} ${BZIP2_DOWNLOAD_BASE}/bzip2-$(BZIP2_VER).tar.gz
 endif
 	BZIP2_SHA256_ACTUAL=`$(SHA256_CMD) bzip2-$(BZIP2_VER).tar.gz | cut -d ' ' -f 1`; \
 	if [ "$(BZIP2_SHA256)" != "$$BZIP2_SHA256_ACTUAL" ]; then \
@@ -1889,35 +1949,47 @@ rocksdbjavastatic: $(java_static_all_libobjects)
 	cd java/src/main/java;jar -cf ../../../target/$(ROCKSDB_SOURCES_JAR) org
 
 rocksdbjavastaticrelease: rocksdbjavastatic
-	cd java/crossbuild && vagrant destroy -f && vagrant up linux32 && vagrant halt linux32 && vagrant up linux64 && vagrant halt linux64
+	cd java/crossbuild && (vagrant destroy -f || true) && vagrant up linux32 && vagrant halt linux32 && vagrant up linux64 && vagrant halt linux64 && vagrant up linux64-musl && vagrant halt linux64-musl
 	cd java;jar -cf target/$(ROCKSDB_JAR_ALL) HISTORY*.md
 	cd java/target;jar -uf $(ROCKSDB_JAR_ALL) librocksdbjni-*.so librocksdbjni-*.jnilib
 	cd java/target/classes;jar -uf ../$(ROCKSDB_JAR_ALL) org/rocksdb/*.class org/rocksdb/util/*.class
 
-rocksdbjavastaticreleasedocker: rocksdbjavastatic rocksdbjavastaticdockerx86 rocksdbjavastaticdockerx86_64
+rocksdbjavastaticreleasedocker: rocksdbjavastatic rocksdbjavastaticdockerx86 rocksdbjavastaticdockerx86_64 rocksdbjavastaticdockerx86musl rocksdbjavastaticdockerx86_64musl
 	cd java;jar -cf target/$(ROCKSDB_JAR_ALL) HISTORY*.md
 	cd java/target;jar -uf $(ROCKSDB_JAR_ALL) librocksdbjni-*.so librocksdbjni-*.jnilib
 	cd java/target/classes;jar -uf ../$(ROCKSDB_JAR_ALL) org/rocksdb/*.class org/rocksdb/util/*.class
 
 rocksdbjavastaticdockerx86:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_x86-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos6_x86-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+	docker run --rm --name rocksdb_linux_x86-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos6_x86-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
 
 rocksdbjavastaticdockerx86_64:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_x64-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos6_x64-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+	docker run --rm --name rocksdb_linux_x64-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos6_x64-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
 
 rocksdbjavastaticdockerppc64le:
 	mkdir -p java/target
-	docker run --rm --name rocksdb_linux_ppc64le-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos7_ppc64le-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+	docker run --rm --name rocksdb_linux_ppc64le-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos7_ppc64le-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
 
 rocksdbjavastaticdockerarm64v8:
 	mkdir -p java/target
-	DOCKER_LINUX_ARM64V8_CONTAINER=`docker ps -aqf name=rocksdb_linux_arm64v8-be`; \
-	if [ -z "$$DOCKER_LINUX_ARM64V8_CONTAINER" ]; then \
-		docker container create --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host --name rocksdb_linux_arm64v8-be evolvedbinary/rocksjava:centos7_arm64v8-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh; \
-	fi
-	docker start -a rocksdb_linux_arm64v8-be
+	docker run --rm --name rocksdb_linux_arm64v8-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos7_arm64v8-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+
+rocksdbjavastaticdockerx86musl:
+	mkdir -p java/target
+	docker run --rm --name rocksdb_linux_x86-musl-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:alpine3_x86-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+
+rocksdbjavastaticdockerx86_64musl:
+	mkdir -p java/target
+	docker run --rm --name rocksdb_linux_x64-musl-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:alpine3_x64-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+
+rocksdbjavastaticdockerppc64lemusl:
+	mkdir -p java/target
+	docker run --rm --name rocksdb_linux_ppc64le-musl-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:alpine3_ppc64le-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+
+rocksdbjavastaticdockerarm64v8musl:
+	mkdir -p java/target
+	docker run --rm --name rocksdb_linux_arm64v8-musl-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host:ro --volume /rocksdb-local-build --volume `pwd`/java/target:/rocksdb-java-target --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:alpine3_arm64v8-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
 
 rocksdbjavastaticpublish: rocksdbjavastaticrelease rocksdbjavastaticpublishcentral
 
@@ -1928,6 +2000,8 @@ rocksdbjavastaticpublishcentral:
 	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/target/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-sources.jar -Dclassifier=sources
 	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/target/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-linux64.jar -Dclassifier=linux64
 	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/target/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-linux32.jar -Dclassifier=linux32
+	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/target/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-linux64-musl.jar -Dclassifier=linux64-musl
+	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/target/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-linux32-musl.jar -Dclassifier=linux32-musl
 	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/target/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-osx.jar -Dclassifier=osx
 	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/target/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-win64.jar -Dclassifier=win64
 	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/target/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH).jar
@@ -1981,6 +2055,7 @@ jtest_run:
 
 jtest: rocksdbjava
 	cd java;$(MAKE) sample;$(MAKE) test;
+	python tools/check_all_python.py # TODO peterd: find a better place for this check in CI targets
 
 jdb_bench:
 	cd java;$(MAKE) db_bench;
@@ -2028,6 +2103,9 @@ endif
 .cc.o:
 	$(AM_V_CC)$(CXX) $(CXXFLAGS) -c $< -o $@ $(COVERAGEFLAGS)
 
+.cpp.o:
+	$(AM_V_CC)$(CXX) $(CXXFLAGS) -c $< -o $@ $(COVERAGEFLAGS)
+
 .c.o:
 	$(AM_V_CC)$(CC) $(CFLAGS) -c $< -o $@
 endif
@@ -2035,9 +2113,13 @@ endif
 #  	Source files dependencies detection
 # ---------------------------------------------------------------------------
 
-all_sources = $(LIB_SOURCES) $(MAIN_SOURCES) $(MOCK_LIB_SOURCES) $(TOOL_LIB_SOURCES) $(BENCH_LIB_SOURCES) $(TEST_LIB_SOURCES) $(ANALYZER_LIB_SOURCES)
+all_sources = $(LIB_SOURCES) $(MAIN_SOURCES) $(MOCK_LIB_SOURCES) $(TOOL_LIB_SOURCES) $(BENCH_LIB_SOURCES) $(TEST_LIB_SOURCES) $(ANALYZER_LIB_SOURCES) $(STRESS_LIB_SOURCES)
 DEPFILES = $(all_sources:.cc=.cc.d)
 
+ifeq ($(USE_FOLLY_DISTRIBUTED_MUTEX),1)
+  DEPFILES += $(FOLLY_SOURCES:.cpp=.cpp.d)
+endif
+
 # Add proper dependency support so changing a .h file forces a .cc file to
 # rebuild.
 
@@ -2047,6 +2129,10 @@ DEPFILES = $(all_sources:.cc=.cc.d)
 	@$(CXX) $(CXXFLAGS) $(PLATFORM_SHARED_CFLAGS) \
 	  -MM -MT'$@' -MT'$(<:.cc=.o)' "$<" -o '$@'
 
+%.cpp.d: %.cpp
+	@$(CXX) $(CXXFLAGS) $(PLATFORM_SHARED_CFLAGS) \
+	  -MM -MT'$@' -MT'$(<:.cpp=.o)' "$<" -o '$@'
+
 ifeq ($(HAVE_POWER8),1)
 DEPFILES_C = $(LIB_SOURCES_C:.c=.c.d)
 DEPFILES_ASM = $(LIB_SOURCES_ASM:.S=.S.d)

README.md

@@ -9,11 +9,11 @@ It is built on earlier work on [LevelDB](https://github.com/google/leveldb) by S
 and Jeff Dean (jeff@google.com)
 
 This code is a library that forms the core building block for a fast
-key value server, especially suited for storing data on flash drives.
+key-value server, especially suited for storing data on flash drives.
 It has a Log-Structured-Merge-Database (LSM) design with flexible tradeoffs
 between Write-Amplification-Factor (WAF), Read-Amplification-Factor (RAF)
 and Space-Amplification-Factor (SAF). It has multi-threaded compactions,
-making it specially suitable for storing multiple terabytes of data in a
+making it especially suitable for storing multiple terabytes of data in a
 single database.
 
 Start with example usage here: https://github.com/facebook/rocksdb/tree/master/examples

TARGETS

@@ -1,3 +1,8 @@
+# This file @generated by `python buckifier/buckify_rocksdb.py`
+# --> DO NOT EDIT MANUALLY <--
+# This file is a Facebook-specific integration for buck builds, so can
+# only be validated by Facebook employees.
+#
 load("@fbcode_macros//build_defs:auto_headers.bzl", "AutoHeaders")
 load("@fbcode_macros//build_defs:cpp_library.bzl", "cpp_library")
 load(":defs.bzl", "test_binary")
@@ -21,6 +26,7 @@ ROCKSDB_EXTERNAL_DEPS = [
     ("lz4", None, "lz4"),
     ("zstd", None),
     ("tbb", None),
+    ("liburing", None, "uring"),
     ("googletest", None, "gtest"),
 ]
 
@@ -41,6 +47,7 @@ ROCKSDB_OS_PREPROCESSOR_FLAGS = [
             "-DROCKSDB_PTHREAD_ADAPTIVE_MUTEX",
             "-DROCKSDB_RANGESYNC_PRESENT",
             "-DROCKSDB_SCHED_GETCPU_PRESENT",
+            "-DROCKSDB_IOURING_PRESENT",
             "-DHAVE_SSE42",
             "-DNUMA",
         ],
@@ -65,6 +72,7 @@ ROCKSDB_PREPROCESSOR_FLAGS = [
     "-DZSTD_STATIC_LINKING_ONLY",
     "-DGFLAGS=gflags",
     "-DTBB",
+    "-DLIBURING",
 
     # Added missing flags from output of build_detect_platform
     "-DROCKSDB_BACKTRACE",
@@ -173,6 +181,8 @@ cpp_library(
         "env/env_encryption.cc",
         "env/env_hdfs.cc",
         "env/env_posix.cc",
+        "env/file_system.cc",
+        "env/fs_posix.cc",
         "env/io_posix.cc",
         "env/mock_env.cc",
         "file/delete_scheduler.cc",
@@ -230,9 +240,11 @@ cpp_library(
         "table/block_based/data_block_footer.cc",
         "table/block_based/data_block_hash_index.cc",
         "table/block_based/filter_block_reader_common.cc",
+        "table/block_based/filter_policy.cc",
         "table/block_based/flush_block_policy.cc",
         "table/block_based/full_filter_block.cc",
         "table/block_based/index_builder.cc",
+        "table/block_based/parsed_full_filter_block.cc",
         "table/block_based/partitioned_filter_block.cc",
         "table/block_based/uncompression_dict_reader.cc",
         "table/block_fetcher.cc",
@@ -264,7 +276,6 @@ cpp_library(
         "tools/sst_dump_tool.cc",
         "trace_replay/block_cache_tracer.cc",
         "trace_replay/trace_replay.cc",
-        "util/bloom.cc",
         "util/build_version.cc",
         "util/coding.cc",
         "util/compaction_job_stats_impl.cc",
@@ -273,7 +284,6 @@ cpp_library(
         "util/concurrent_task_limiter_impl.cc",
         "util/crc32c.cc",
         "util/dynamic_bloom.cc",
-        "util/filter_policy.cc",
         "util/hash.cc",
         "util/murmurhash.cc",
         "util/random.cc",
@@ -391,6 +401,32 @@ cpp_library(
     external_deps = ROCKSDB_EXTERNAL_DEPS,
 )
 
+cpp_library(
+    name = "rocksdb_stress_lib",
+    srcs = [
+        "db_stress_tool/batched_ops_stress.cc",
+        "db_stress_tool/cf_consistency_stress.cc",
+        "db_stress_tool/db_stress_common.cc",
+        "db_stress_tool/db_stress_driver.cc",
+        "db_stress_tool/db_stress_gflags.cc",
+        "db_stress_tool/db_stress_shared_state.cc",
+        "db_stress_tool/db_stress_test_base.cc",
+        "db_stress_tool/db_stress_tool.cc",
+        "db_stress_tool/no_batched_ops_stress.cc",
+        "test_util/testutil.cc",
+        "tools/block_cache_analyzer/block_cache_trace_analyzer.cc",
+        "tools/trace_analyzer_tool.cc",
+    ],
+    auto_headers = AutoHeaders.RECURSIVE_GLOB,
+    arch_preprocessor_flags = ROCKSDB_ARCH_PREPROCESSOR_FLAGS,
+    compiler_flags = ROCKSDB_COMPILER_FLAGS,
+    os_deps = ROCKSDB_OS_DEPS,
+    os_preprocessor_flags = ROCKSDB_OS_PREPROCESSOR_FLAGS,
+    preprocessor_flags = ROCKSDB_PREPROCESSOR_FLAGS,
+    deps = [":rocksdb_lib"],
+    external_deps = ROCKSDB_EXTERNAL_DEPS,
+)
+
 cpp_library(
     name = "env_basic_test_lib",
     srcs = ["env/env_basic_test.cc"],
@@ -757,15 +793,15 @@ ROCKS_TESTS = [
         [],
     ],
     [
-        "db_merge_operator_test",
-        "db/db_merge_operator_test.cc",
-        "parallel",
+        "db_merge_operand_test",
+        "db/db_merge_operand_test.cc",
+        "serial",
         [],
         [],
     ],
     [
-        "db_merge_operand_test",
-        "db/db_merge_operand_test.cc",
+        "db_merge_operator_test",
+        "db/db_merge_operator_test.cc",
         "parallel",
         [],
         [],
@@ -1211,6 +1247,13 @@ ROCKS_TESTS = [
         [],
         [],
     ],
+    [
+        "random_test",
+        "util/random_test.cc",
+        "serial",
+        [],
+        [],
+    ],
     [
         "range_del_aggregator_test",
         "db/range_del_aggregator_test.cc",

appveyor.yml

@@ -17,6 +17,13 @@ environment:
   ZSTD_INCLUDE: $(ZSTD_HOME)\lib;$(ZSTD_HOME)\lib\dictBuilder
   ZSTD_LIB_DEBUG: $(ZSTD_HOME)\build\VS2010\bin\x64_Debug\libzstd_static.lib
   ZSTD_LIB_RELEASE: $(ZSTD_HOME)\build\VS2010\bin\x64_Release\libzstd_static.lib
+  matrix:
+    - APPVEYOR_BUILD_WORKER_IMAGE: Visual Studio 2015
+      CMAKE_GENERATOR: Visual Studio 14 Win64
+      DEV_ENV: C:\Program Files (x86)\Microsoft Visual Studio 14.0\Common7\IDE\devenv.com
+    - APPVEYOR_BUILD_WORKER_IMAGE: Visual Studio 2017
+      CMAKE_GENERATOR: Visual Studio 15 Win64
+      DEV_ENV: C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\Common7\IDE\devenv.com
 
 install:
   - md %THIRDPARTY_HOME%
@@ -27,7 +34,7 @@ install:
   - cd snappy-1.1.7
   - mkdir build
   - cd build
-  - cmake -DCMAKE_GENERATOR_PLATFORM=x64 ..
+  - cmake -G "%CMAKE_GENERATOR%" ..
   - msbuild Snappy.sln /p:Configuration=Debug /p:Platform=x64
   - msbuild Snappy.sln /p:Configuration=Release /p:Platform=x64
   - echo "Building LZ4 dependency..."
@@ -35,7 +42,7 @@ install:
   - curl -fsSL -o lz4-1.8.3.zip https://github.com/lz4/lz4/archive/v1.8.3.zip
   - unzip lz4-1.8.3.zip
   - cd lz4-1.8.3\visual\VS2010
-  - ps: $CMD="C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\Common7\IDE\devenv.com"; & $CMD lz4.sln /upgrade
+  - ps: $CMD="$Env:DEV_ENV"; & $CMD lz4.sln /upgrade
   - msbuild lz4.sln /p:Configuration=Debug /p:Platform=x64
   - msbuild lz4.sln /p:Configuration=Release /p:Platform=x64
   - echo "Building ZStd dependency..."
@@ -43,15 +50,16 @@ install:
   - curl -fsSL -o zstd-1.4.0.zip https://github.com/facebook/zstd/archive/v1.4.0.zip
   - unzip zstd-1.4.0.zip
   - cd zstd-1.4.0\build\VS2010
-  - ps: $CMD="C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\Common7\IDE\devenv.com"; & $CMD zstd.sln /upgrade
+  - ps: $CMD="$Env:DEV_ENV"; & $CMD zstd.sln /upgrade
   - msbuild zstd.sln /p:Configuration=Debug /p:Platform=x64
   - msbuild zstd.sln /p:Configuration=Release /p:Platform=x64
 
 before_build:
   - md %APPVEYOR_BUILD_FOLDER%\build
   - cd %APPVEYOR_BUILD_FOLDER%\build
-  - cmake -G "Visual Studio 15 Win64" -DOPTDBG=1 -DPORTABLE=1 -DSNAPPY=1 -DLZ4=1 -DZSTD=1 -DXPRESS=1 -DJNI=1 ..
+  - cmake -G "%CMAKE_GENERATOR%" -DCMAKE_BUILD_TYPE=Debug -DOPTDBG=1 -DPORTABLE=1 -DSNAPPY=1 -DLZ4=1 -DZSTD=1 -DXPRESS=1 -DJNI=1 ..
   - cd ..
+
 build:
   project: build\rocksdb.sln
   parallel: true

buckifier/buckify_rocksdb.py

@@ -3,6 +3,10 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 from __future__ import unicode_literals
+try:
+    from builtins import str
+except ImportError:
+    from __builtin__ import str
 from targets_builder import TARGETSBuilder
 import json
 import os
@@ -108,9 +112,9 @@ def get_tests(repo_path):
 # Parse extra dependencies passed by user from command line
 def get_dependencies():
     deps_map = {
-        ''.encode('ascii'): {
-            'extra_deps'.encode('ascii'): [],
-            'extra_compiler_flags'.encode('ascii'): []
+        '': {
+            'extra_deps': [],
+            'extra_compiler_flags': []
         }
     }
     if len(sys.argv) < 2:
@@ -119,13 +123,7 @@ def get_dependencies():
     def encode_dict(data):
         rv = {}
         for k, v in data.items():
-            if isinstance(k, unicode):
-                k = k.encode('ascii')
-            if isinstance(v, unicode):
-                v = v.encode('ascii')
-            elif isinstance(v, list):
-                v = [x.encode('ascii') for x in v]
-            elif isinstance(v, dict):
+            if isinstance(v, dict):
                 v = encode_dict(v)
             rv[k] = v
         return rv
@@ -169,6 +167,13 @@ def generate_targets(repo_path, deps_map):
         src_mk.get("ANALYZER_LIB_SOURCES", []) +
         ["test_util/testutil.cc"],
         [":rocksdb_lib"])
+    # rocksdb_stress_lib
+    TARGETS.add_library(
+        "rocksdb_stress_lib",
+        src_mk.get("ANALYZER_LIB_SOURCES", [])
+        + src_mk.get('STRESS_LIB_SOURCES', [])
+        + ["test_util/testutil.cc"],
+        [":rocksdb_lib"])
 
     print("Extra dependencies:\n{0}".format(str(deps_map)))
     # test for every test we found in the Makefile

buckifier/targets_builder.py

@@ -3,6 +3,12 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 from __future__ import unicode_literals
+try:
+    from builtins import object
+    from builtins import str
+except ImportError:
+    from __builtin__ import object
+    from __builtin__ import str
 import targets_cfg
 
 def pretty_list(lst, indent=8):
@@ -18,7 +24,7 @@ def pretty_list(lst, indent=8):
     return res
 
 
-class TARGETSBuilder:
+class TARGETSBuilder(object):
     def __init__(self, path):
         self.path = path
         self.targets_file = open(path, 'w')

buckifier/targets_cfg.py

@@ -3,7 +3,13 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 from __future__ import unicode_literals
-rocksdb_target_header = """load("@fbcode_macros//build_defs:auto_headers.bzl", "AutoHeaders")
+
+rocksdb_target_header = """# This file \100generated by `python buckifier/buckify_rocksdb.py`
+# --> DO NOT EDIT MANUALLY <--
+# This file is a Facebook-specific integration for buck builds, so can
+# only be validated by Facebook employees.
+#
+load("@fbcode_macros//build_defs:auto_headers.bzl", "AutoHeaders")
 load("@fbcode_macros//build_defs:cpp_library.bzl", "cpp_library")
 load(":defs.bzl", "test_binary")
 
@@ -26,6 +32,7 @@ ROCKSDB_EXTERNAL_DEPS = [
     ("lz4", None, "lz4"),
     ("zstd", None),
     ("tbb", None),
+    ("liburing", None, "uring"),
     ("googletest", None, "gtest"),
 ]
 
@@ -46,6 +53,7 @@ ROCKSDB_OS_PREPROCESSOR_FLAGS = [
             "-DROCKSDB_PTHREAD_ADAPTIVE_MUTEX",
             "-DROCKSDB_RANGESYNC_PRESENT",
             "-DROCKSDB_SCHED_GETCPU_PRESENT",
+            "-DROCKSDB_IOURING_PRESENT",
             "-DHAVE_SSE42",
             "-DNUMA",
         ],
@@ -70,6 +78,7 @@ ROCKSDB_PREPROCESSOR_FLAGS = [
     "-DZSTD_STATIC_LINKING_ONLY",
     "-DGFLAGS=gflags",
     "-DTBB",
+    "-DLIBURING",
 
     # Added missing flags from output of build_detect_platform
     "-DROCKSDB_BACKTRACE",

buckifier/util.py

@@ -6,11 +6,16 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 from __future__ import unicode_literals
+try:
+    from builtins import object
+except ImportError:
+    from __builtin__ import object
 import subprocess
+import sys
 import os
 import time
 
-class ColorString:
+class ColorString(object):
     """ Generate colorful strings on terminal """
     HEADER = '\033[95m'
     BLUE = '\033[94m'
@@ -21,7 +26,13 @@ class ColorString:
 
     @staticmethod
     def _make_color_str(text, color):
-        return "".join([color, text.encode('utf-8'), ColorString.ENDC])
+        # In Python2, default encoding for unicode string is ASCII
+        if sys.version_info.major <= 2:
+            return "".join(
+                [color, text.encode('utf-8'), ColorString.ENDC])
+        # From Python3, default encoding for unicode string is UTF-8
+        return "".join(
+            [color, text, ColorString.ENDC])
 
     @staticmethod
     def ok(text):

build_tools/RocksDBCommonHelper.php

@@ -1,377 +0,0 @@
-<?php
-// Copyright 2004-present Facebook. All Rights Reserved.
-//  This source code is licensed under both the GPLv2 (found in the
-//  COPYING file in the root directory) and Apache 2.0 License
-//  (found in the LICENSE.Apache file in the root directory).
-
-// Name of the environment variables which need to be set by the entity which
-// triggers continuous runs so that code at the end of the file gets executed
-// and Sandcastle run starts.
-const ENV_POST_RECEIVE_HOOK = "POST_RECEIVE_HOOK";
-const ENV_HTTPS_APP_VALUE = "HTTPS_APP_VALUE";
-const ENV_HTTPS_TOKEN_VALUE = "HTTPS_TOKEN_VALUE";
-
-const PRIMARY_TOKEN_FILE = '/home/krad/.sandcastle';
-const CONT_RUN_ALIAS = "leveldb";
-
-//////////////////////////////////////////////////////////////////////
-/*  Run tests in sandcastle */
-function postURL($diffID, $url) {
-  assert(strlen($diffID) > 0);
-  assert(is_numeric($diffID));
-  assert(strlen($url) > 0);
-
-  $cmd_args = array(
-    'diff_id' => (int)$diffID,
-    'name' => sprintf(
-      'click here for sandcastle tests for D%d',
-      (int)$diffID
-    ),
-    'link' => $url
-  );
-  $cmd = 'echo ' . escapeshellarg(json_encode($cmd_args))
-         . ' | arc call-conduit differential.updateunitresults';
-
-  shell_exec($cmd);
-}
-
-function buildUpdateTestStatusCmd($diffID, $test, $status) {
-  assert(strlen($diffID) > 0);
-  assert(is_numeric($diffID));
-  assert(strlen($test) > 0);
-  assert(strlen($status) > 0);
-
-  $cmd_args = array(
-    'diff_id' => (int)$diffID,
-    'name' => $test,
-    'result' => $status
-  );
-
-  $cmd = 'echo ' . escapeshellarg(json_encode($cmd_args))
-         . ' | arc call-conduit differential.updateunitresults';
-
-  return $cmd;
-}
-
-function updateTestStatus($diffID, $test) {
-  assert(strlen($diffID) > 0);
-  assert(is_numeric($diffID));
-  assert(strlen($test) > 0);
-
-  shell_exec(buildUpdateTestStatusCmd($diffID, $test, "waiting"));
-}
-
-function getSteps($applyDiff, $diffID, $username, $test) {
-  assert(strlen($username) > 0);
-  assert(strlen($test) > 0);
-
-  if ($applyDiff) {
-    assert(strlen($diffID) > 0);
-    assert(is_numeric($diffID));
-
-    $arcrc_content = (PHP_OS == "Darwin" ?
-        exec("cat ~/.arcrc | gzip -f | base64") :
-            exec("cat ~/.arcrc | gzip -f | base64 -w0"));
-    assert(strlen($arcrc_content) > 0);
-
-    // Sandcastle machines don't have arc setup. We copy the user certificate
-    // and authenticate using that in Sandcastle.
-    $setup = array(
-      "name" => "Setup arcrc",
-      "shell" => "echo " . escapeshellarg($arcrc_content) . " | base64 --decode"
-                 . " | gzip -d > ~/.arcrc",
-      "user" => "root"
-    );
-
-    // arc demands certain permission on its config.
-    // also fix the sticky bit issue in sandcastle
-    $fix_permission = array(
-      "name" => "Fix environment",
-      "shell" => "chmod 600 ~/.arcrc && chmod +t /dev/shm",
-      "user" => "root"
-    );
-
-    // Construct the steps in the order of execution.
-    $steps[] = $setup;
-    $steps[] = $fix_permission;
-  }
-
-  // fbcode is a sub-repo. We cannot patch until we add it to ignore otherwise
-  // Git thinks it is an uncommitted change.
-  $fix_git_ignore = array(
-    "name" => "Fix git ignore",
-    "shell" => "echo fbcode >> .git/info/exclude",
-    "user" => "root"
-  );
-
-  // This fixes "FATAL: ThreadSanitizer can not mmap the shadow memory"
-  // Source:
-  // https://github.com/google/sanitizers/wiki/ThreadSanitizerCppManual#FAQ
-  $fix_kernel_issue = array(
-    "name" => "Fix kernel issue with tsan",
-    "shell" => "echo 2 >/proc/sys/kernel/randomize_va_space",
-    "user" => "root"
-  );
-
-  $steps[] = $fix_git_ignore;
-  $steps[] = $fix_kernel_issue;
-
-  // This will be the command used to execute particular type of tests.
-  $cmd = "";
-
-  if ($applyDiff) {
-    // Patch the code (keep your fingures crossed).
-    $patch = array(
-      "name" => "Patch " . $diffID,
-      "shell" => "arc --arcrc-file ~/.arcrc "
-                  . "patch --nocommit --diff " . escapeshellarg($diffID),
-      "user" => "root"
-    );
-
-    $steps[] = $patch;
-
-    updateTestStatus($diffID, $test);
-    $cmd = buildUpdateTestStatusCmd($diffID, $test, "running") . "; ";
-  }
-
-  // Run the actual command.
-  $cmd = $cmd . "J=$(nproc) ./build_tools/precommit_checker.py " .
-           escapeshellarg($test) . "; exit_code=$?; ";
-
-  if ($applyDiff) {
-    $cmd = $cmd . "([[ \$exit_code -eq 0 ]] &&"
-                . buildUpdateTestStatusCmd($diffID, $test, "pass") . ")"
-                . "||" . buildUpdateTestStatusCmd($diffID, $test, "fail")
-                . "; ";
-  }
-
-  // shell command to sort the tests based on exit code and print
-  // the output of the log files.
-  $cat_sorted_logs = "
-    while read code log_file;
-      do echo \"################ cat \$log_file [exit_code : \$code] ################\";
-      cat \$log_file;
-    done < <(tail -n +2 LOG | sort -k7,7n -k4,4gr | awk '{print \$7,\$NF}')";
-
-  // Shell command to cat all log files
-  $cat_all_logs = "for f in `ls t/!(run-*)`; do echo \$f;cat \$f; done";
-
-  // If LOG file exist use it to cat log files sorted by exit code, otherwise
-  // cat everything
-  $logs_cmd = "if [ -f LOG ]; then {$cat_sorted_logs}; else {$cat_all_logs}; fi";
-
-  $cmd = $cmd . " cat /tmp/precommit-check.log"
-              . "; shopt -s extglob; {$logs_cmd}"
-              . "; shopt -u extglob; [[ \$exit_code -eq 0 ]]";
-  assert(strlen($cmd) > 0);
-
-  $run_test = array(
-    "name" => "Run " . $test,
-    "shell" => $cmd,
-    "user" => "root",
-    "parser" => "python build_tools/error_filter.py " . escapeshellarg($test),
-  );
-
-  $steps[] = $run_test;
-
-  if ($applyDiff) {
-    // Clean up the user arc config we are using.
-    $cleanup = array(
-      "name" => "Arc cleanup",
-      "shell" => "rm -f ~/.arcrc",
-      "user" => "root"
-    );
-
-    $steps[] = $cleanup;
-  }
-
-  assert(count($steps) > 0);
-  return $steps;
-}
-
-function getSandcastleConfig() {
-  $sandcastle_config = array();
-
-  $cwd = getcwd();
-  $cwd_token_file = "{$cwd}/.sandcastle";
-  // This is a case when we're executed from a continuous run. Fetch the values
-  // from the environment.
-  if (getenv(ENV_POST_RECEIVE_HOOK)) {
-    $sandcastle_config[0] = getenv(ENV_HTTPS_APP_VALUE);
-    $sandcastle_config[1] = getenv(ENV_HTTPS_TOKEN_VALUE);
-  } else {
-    // This is a typical `[p]arc diff` case. Fetch the values from the specific
-    // configuration files.
-    for ($i = 0; $i < 50; $i++) {
-      if (file_exists(PRIMARY_TOKEN_FILE) ||
-          file_exists($cwd_token_file)) {
-        break;
-      }
-      // If we failed to fetch the tokens, sleep for 0.2 second and try again
-      usleep(200000);
-    }
-    assert(file_exists(PRIMARY_TOKEN_FILE) ||
-           file_exists($cwd_token_file));
-
-    // Try the primary location first, followed by a secondary.
-    if (file_exists(PRIMARY_TOKEN_FILE)) {
-      $cmd = 'cat ' . PRIMARY_TOKEN_FILE;
-    } else {
-      $cmd = 'cat ' . escapeshellarg($cwd_token_file);
-    }
-
-    assert(strlen($cmd) > 0);
-    $sandcastle_config = explode(':', rtrim(shell_exec($cmd)));
-  }
-
-  // In this case be very explicit about the implications.
-  if (count($sandcastle_config) != 2) {
-    echo "Sandcastle configuration files don't contain valid information " .
-         "or the necessary environment variables aren't defined. Unable " .
-         "to validate the code changes.";
-    exit(1);
-  }
-
-  assert(strlen($sandcastle_config[0]) > 0);
-  assert(strlen($sandcastle_config[1]) > 0);
-  assert(count($sandcastle_config) > 0);
-
-  return $sandcastle_config;
-}
-
-// This function can be called either from `[p]arc diff` command or during
-// the Git post-receive hook.
- function startTestsInSandcastle($applyDiff, $workflow, $diffID) {
-  // Default options don't terminate on failure, but that's what we want. In
-  // the current case we use assertions intentionally as "terminate on failure
-  // invariants".
-  assert_options(ASSERT_BAIL, true);
-
-  // In case of a diff we'll send notificatios to the author. Else it'll go to
-  // the entire team because failures indicate that build quality has regressed.
-  $username = $applyDiff ? exec("whoami") : CONT_RUN_ALIAS;
-  assert(strlen($username) > 0);
-
-  if ($applyDiff) {
-    assert($workflow);
-    assert(strlen($diffID) > 0);
-    assert(is_numeric($diffID));
-  }
-
-  // List of tests we want to run in Sandcastle.
-  $tests = array("unit", "unit_non_shm", "unit_481", "clang_unit", "tsan",
-                 "asan", "lite_test", "valgrind", "release", "release_481",
-                 "clang_release", "clang_analyze", "code_cov",
-                 "java_build", "no_compression", "unity", "ubsan");
-
-  $send_email_template = array(
-    'type' => 'email',
-    'triggers' => array('fail'),
-    'emails' => array($username . '@fb.com'),
-  );
-
-  // Construct a job definition for each test and add it to the master plan.
-  foreach ($tests as $test) {
-    $stepName = "RocksDB diff " . $diffID . " test " . $test;
-
-    if (!$applyDiff) {
-      $stepName = "RocksDB continuous integration test " . $test;
-    }
-
-    $arg[] = array(
-      "name" => $stepName,
-      "report" => array($send_email_template),
-      "steps" => getSteps($applyDiff, $diffID, $username, $test)
-    );
-  }
-
-  // We cannot submit the parallel execution master plan to Sandcastle and
-  // need supply the job plan as a determinator. So we construct a small job
-  // that will spit out the master job plan which Sandcastle will parse and
-  // execute. Why compress the job definitions? Otherwise we run over the max
-  // string size.
-  $cmd = "echo " . base64_encode(json_encode($arg))
-         . (PHP_OS == "Darwin" ?
-             " | gzip -f | base64" :
-                 " | gzip -f | base64 -w0");
-  assert(strlen($cmd) > 0);
-
-  $arg_encoded = shell_exec($cmd);
-  assert(strlen($arg_encoded) > 0);
-
-  $runName = "Run diff " . $diffID . "for user " . $username;
-
-  if (!$applyDiff) {
-    $runName = "RocksDB continuous integration build and test run";
-  }
-
-  $command = array(
-    "name" => $runName,
-    "steps" => array()
-  );
-
-  $command["steps"][] = array(
-    "name" => "Generate determinator",
-    "shell" => "echo " . $arg_encoded . " | base64 --decode | gzip -d"
-               . " | base64 --decode",
-    "determinator" => true,
-    "user" => "root"
-  );
-
-  // Submit to Sandcastle.
-  $url = 'https://interngraph.intern.facebook.com/sandcastle/create';
-
-  $job = array(
-    'command' => 'SandcastleUniversalCommand',
-    'args' => $command,
-    'capabilities' => array(
-      'vcs' => 'rocksdb-int-git',
-      'type' => 'lego',
-    ),
-    'hash' => 'origin/master',
-    'user' => $username,
-    'alias' => 'rocksdb-precommit',
-    'tags' => array('rocksdb'),
-    'description' => 'Rocksdb precommit job',
-  );
-
-  // Fetch the configuration necessary to submit a successful HTTPS request.
-  $sandcastle_config = getSandcastleConfig();
-
-  $app = $sandcastle_config[0];
-  $token = $sandcastle_config[1];
-
-  $cmd = 'curl -s -k '
-          . ' -F app=' . escapeshellarg($app)
-          . ' -F token=' . escapeshellarg($token)
-          . ' -F job=' . escapeshellarg(json_encode($job))
-          .' ' . escapeshellarg($url);
-
-  $output = shell_exec($cmd);
-  assert(strlen($output) > 0);
-
-  // Extract Sandcastle URL from the response.
-  preg_match('/url": "(.+)"/', $output, $sandcastle_url);
-
-  assert(count($sandcastle_url) > 0, "Unable to submit Sandcastle request.");
-  assert(strlen($sandcastle_url[1]) > 0, "Unable to extract Sandcastle URL.");
-
-  if ($applyDiff) {
-    echo "\nSandcastle URL: " . $sandcastle_url[1] . "\n";
-    // Ask Phabricator to display it on the diff UI.
-    postURL($diffID, $sandcastle_url[1]);
-  } else {
-    echo "Continuous integration started Sandcastle tests. You can look at ";
-    echo "the progress at:\n" . $sandcastle_url[1] . "\n";
-  }
-}
-
-// Continuous run cript will set the environment variable and based on that
-// we'll trigger the execution of tests in Sandcastle. In that case we don't
-// need to apply any diffs and there's no associated workflow either.
-if (getenv(ENV_POST_RECEIVE_HOOK)) {
-  startTestsInSandcastle(
-    false /* $applyDiff */,
-    NULL /* $workflow */,
-    NULL /* $diffID */);
-}

build_tools/build_detect_platform

@@ -56,10 +56,10 @@ if [ -z "$ROCKSDB_NO_FBCODE" -a -d /mnt/gvfs/third-party ]; then
     if [ -n "$ROCKSDB_FBCODE_BUILD_WITH_481" ]; then
       # we need this to build with MySQL. Don't use for other purposes.
       source "$PWD/build_tools/fbcode_config4.8.1.sh"
-    elif [ -n "$ROCKSDB_FBCODE_BUILD_WITH_PLATFORM007" ]; then
-      source "$PWD/build_tools/fbcode_config_platform007.sh"
-    else
+    elif [ -n "$ROCKSDB_FBCODE_BUILD_WITH_5xx" ]; then
       source "$PWD/build_tools/fbcode_config.sh"
+    else
+      source "$PWD/build_tools/fbcode_config_platform007.sh"
     fi
 fi
 
@@ -150,6 +150,19 @@ case "$TARGET_OS" in
             PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -latomic"
         fi
         PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -lpthread -lrt"
+        # check for liburing
+        $CXX $CFLAGS -x c++ - -luring -o /dev/null 2>/dev/null  <<EOF
+          #include <liburing.h>
+          int main() {
+            struct io_uring ring;
+            io_uring_queue_init(1, &ring, 0);
+            return 0;
+          }
+EOF
+        if [ "$?" = 0 ]; then
+          PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -luring"
+          COMMON_FLAGS="$COMMON_FLAGS -DROCKSDB_IOURING_PRESENT"
+        fi
         if test -z "$USE_FOLLY_DISTRIBUTED_MUTEX"; then
           USE_FOLLY_DISTRIBUTED_MUTEX=1
         fi
@@ -572,6 +585,8 @@ if test "$TRY_SSE_ETC"; then
   TRY_SSE42="-msse4.2"
   # PCLMUL available since westmere, ca. 2010-2011
   TRY_PCLMUL="-mpclmul"
+  # AVX2 available since haswell, ca. 2013-2015
+  TRY_AVX2="-mavx2"
 fi
 
 $CXX $PLATFORM_CXXFLAGS $COMMON_FLAGS $TRY_SSE42 -x c++ - -o /dev/null 2>/dev/null <<EOF
@@ -605,6 +620,34 @@ elif test "$USE_SSE"; then
   echo "warning: USE_SSE specified but compiler could not use PCLMUL intrinsics, disabling" >&2
 fi
 
+$CXX $PLATFORM_CXXFLAGS $COMMON_FLAGS $TRY_AVX2 -x c++ - -o /dev/null 2>/dev/null <<EOF
+  #include <cstdint>
+  #include <immintrin.h>
+  int main() {
+    const auto a = _mm256_setr_epi32(0, 1, 2, 3, 4, 7, 6, 5);
+    const auto b = _mm256_permutevar8x32_epi32(a, a);
+    (void)b;
+  }
+EOF
+if [ "$?" = 0 ]; then
+  COMMON_FLAGS="$COMMON_FLAGS $TRY_AVX2 -DHAVE_AVX2"
+elif test "$USE_SSE"; then
+  echo "warning: USE_SSE specified but compiler could not use AVX2 intrinsics, disabling" >&2
+fi
+
+$CXX $PLATFORM_CXXFLAGS $COMMON_FLAGS -x c++ - -o /dev/null 2>/dev/null <<EOF
+  #include <cstdint>
+  int main() {
+    uint64_t a = 0xffffFFFFffffFFFF;
+    __uint128_t b = __uint128_t(a) * a;
+    a = static_cast<uint64_t>(b >> 64);
+    (void)a;
+  }
+EOF
+if [ "$?" = 0 ]; then
+  COMMON_FLAGS="$COMMON_FLAGS -DHAVE_UINT128_EXTENSION"
+fi
+
 # iOS doesn't support thread-local storage, but this check would erroneously
 # succeed because the cross-compiler flags are added by the Makefile, not this
 # script.

build_tools/dependencies_platform007.sh

@@ -1,6 +1,6 @@
 # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
 GCC_BASE=/mnt/gvfs/third-party2/gcc/7331085db891a2ef4a88a48a751d834e8d68f4cb/7.x/centos7-native/b2ef2b6
-CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/1bd23f9917738974ad0ff305aa23eb5f93f18305/9.0.0/centos7-native/c9f9104
+CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/963d9aeda70cc4779885b1277484fe7544a04e3e/9.0.0/platform007/9e92d53/
 LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/6ace84e956873d53638c738b6f65f3f469cca74c/7.x/platform007/5620abc
 GLIBC_BASE=/mnt/gvfs/third-party2/glibc/192b0f42d63dcf6210d6ceae387b49af049e6e0c/2.26/platform007/f259413
 SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/7f9bdaada18f59bc27ec2b0871eb8a6144343aef/1.1.3/platform007/ca4da3d
@@ -13,7 +13,8 @@ JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/c26f08f47ac35fc31da2633b7da92d6b86
 NUMA_BASE=/mnt/gvfs/third-party2/numa/3f3fb57a5ccc5fd21c66416c0b83e0aa76a05376/2.0.11/platform007/ca4da3d
 LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/40c73d874898b386a71847f1b99115d93822d11f/1.4/platform007/6f3e0a9
 TBB_BASE=/mnt/gvfs/third-party2/tbb/4ce8e8dba77cdbd81b75d6f0c32fd7a1b76a11ec/2018_U5/platform007/ca4da3d
+LIBURING_BASE=/mnt/gvfs/third-party2/liburing/79427253fd0d42677255aacfe6d13bfe63f752eb/20190828/platform007/ca4da3d
 KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/fb251ecd2f5ae16f8671f7014c246e52a748fe0b/fb/platform007/da39a3e
-BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/2e3cb7d119b3cea5f1e738cc13a1ac69f49eb875/2.29.1/centos7-native/da39a3e
+BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/ab9f09bba370e7066cafd4eb59752db93f2e8312/2.29.1/platform007/15a3614
 VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/d42d152a15636529b0861ec493927200ebebca8e/3.15.0/platform007/ca4da3d
 LUA_BASE=/mnt/gvfs/third-party2/lua/f0cd714433206d5139df61659eb7b28b1dea6683/5.3.4/platform007/5007832

build_tools/error_filter.py

@@ -133,13 +133,16 @@ _TEST_NAME_TO_PARSERS = {
     'lite_test': [CompilerErrorParser, GTestErrorParser],
     'stress_crash': [CompilerErrorParser, DbCrashErrorParser],
     'stress_crash_with_atomic_flush': [CompilerErrorParser, DbCrashErrorParser],
+    'stress_crash_with_txn': [CompilerErrorParser, DbCrashErrorParser],
     'write_stress': [CompilerErrorParser, WriteStressErrorParser],
     'asan': [CompilerErrorParser, GTestErrorParser, AsanErrorParser],
     'asan_crash': [CompilerErrorParser, AsanErrorParser, DbCrashErrorParser],
     'asan_crash_with_atomic_flush': [CompilerErrorParser, AsanErrorParser, DbCrashErrorParser],
+    'asan_crash_with_txn': [CompilerErrorParser, AsanErrorParser, DbCrashErrorParser],
     'ubsan': [CompilerErrorParser, GTestErrorParser, UbsanErrorParser],
     'ubsan_crash': [CompilerErrorParser, UbsanErrorParser, DbCrashErrorParser],
     'ubsan_crash_with_atomic_flush': [CompilerErrorParser, UbsanErrorParser, DbCrashErrorParser],
+    'ubsan_crash_with_txn': [CompilerErrorParser, UbsanErrorParser, DbCrashErrorParser],
     'valgrind': [CompilerErrorParser, GTestErrorParser, ValgrindErrorParser],
     'tsan': [CompilerErrorParser, GTestErrorParser, TsanErrorParser],
     'format_compatible': [CompilerErrorParser, CompatErrorParser],

build_tools/fbcode_config_platform007.sh

@@ -86,6 +86,15 @@ else
 fi
 CFLAGS+=" -DTBB"
 
+# location of LIBURING
+LIBURING_INCLUDE=" -isystem $LIBURING_BASE/include/"
+if test -z $PIC_BUILD; then
+  LIBURING_LIBS="$LIBURING_BASE/lib/liburing.a"
+else
+  LIBURING_LIBS="$LIBURING_BASE/lib/liburing_pic.a"
+fi
+CFLAGS+=" -DLIBURING"
+
 test "$USE_SSE" || USE_SSE=1
 export USE_SSE
 test "$PORTABLE" || PORTABLE=1
@@ -94,7 +103,7 @@ export PORTABLE
 BINUTILS="$BINUTILS_BASE/bin"
 AR="$BINUTILS/ar"
 
-DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE $TBB_INCLUDE"
+DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE $TBB_INCLUDE $LIBURING_INCLUDE"
 
 STDLIBS="-L $GCC_BASE/lib64"
 
@@ -135,10 +144,10 @@ else
 fi
 
 CFLAGS+=" $DEPS_INCLUDE"
-CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DROCKSDB_SUPPORT_THREAD_LOCAL -DHAVE_SSE42"
+CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DROCKSDB_SUPPORT_THREAD_LOCAL -DHAVE_SSE42 -DROCKSDB_IOURING_PRESENT"
 CXXFLAGS+=" $CFLAGS"
 
-EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS"
+EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS $LIBURING_LIBS"
 EXEC_LDFLAGS+=" -B$BINUTILS/gold"
 EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/platform007/lib/ld.so"
 EXEC_LDFLAGS+=" $LIBUNWIND"
@@ -148,7 +157,7 @@ EXEC_LDFLAGS+=" -ldl"
 
 PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"
 
-EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $TBB_LIBS"
+EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $TBB_LIBS $LIBURING_LIBS"
 
 VALGRIND_VER="$VALGRIND_BASE/bin/"
 

build_tools/format-diff.sh

@@ -13,9 +13,14 @@ then
   echo "You didn't have clang-format-diff.py and/or clang-format available in your computer!"
   echo "You can download clang-format-diff.py by running: "
   echo "    curl --location http://goo.gl/iUW1u2 -o ${CLANG_FORMAT_DIFF}"
-  echo "You can download clang-format by running: "
+  echo "You can download clang-format by running:"
   echo "    brew install clang-format"
+  echo "  Or"
+  echo "    apt install clang-format"
+  echo "  This might work too:"
+  echo "    yum install git-clang-format"
   echo "Then, move both files (i.e. ${CLANG_FORMAT_DIFF} and clang-format) to some directory within PATH=${PATH}"
+  echo "and make sure ${CLANG_FORMAT_DIFF} is executable."
   exit 128
 fi
 

build_tools/precommit_checker.py

@@ -1,4 +1,4 @@
-#!/usr/local/fbcode/gcc-4.9-glibc-2.20-fb/bin/python2.7
+#!/usr/bin/env python2.7
 # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
 
 from __future__ import absolute_import

build_tools/rocksdb-lego-determinator

@@ -450,6 +450,36 @@ STRESS_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS="[
     }
 ]"
 
+#
+# RocksDB stress/crash test with txn
+#
+STRESS_CRASH_TEST_WITH_TXN_COMMANDS="[
+    {
+        'name':'Rocksdb Stress and Crash Test with txn',
+        'oncall':'$ONCALL',
+        'executeLocal': 'true',
+        'timeout': 86400,
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Build and run RocksDB debug stress tests',
+                'shell':'$SHM $DEBUG $NON_TSAN_CRASH make J=1 db_stress || $CONTRUN_NAME=db_stress $TASK_CREATION_TOOL',
+                'user':'root',
+                $PARSER
+            },
+            {
+                'name':'Build and run RocksDB debug crash tests with txn',
+                'timeout': 86400,
+                'shell':'$SHM $DEBUG $NON_TSAN_CRASH make J=1 crash_test_with_txn || $CONTRUN_NAME=crash_test_with_txn $TASK_CREATION_TOOL',
+                'user':'root',
+                $PARSER
+            },
+            $UPLOAD_DB_DIR,
+        ],
+        $REPORT
+    }
+]"
+
 # RocksDB write stress test.
 # We run on disk device on purpose (i.e. no $SHM)
 # because we want to add some randomness to fsync commands
@@ -541,6 +571,30 @@ ASAN_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS="[
     }
 ]"
 
+#
+# RocksDB crash testing with txn under address sanitizer
+#
+ASAN_CRASH_TEST_WITH_TXN_COMMANDS="[
+    {
+        'name':'Rocksdb crash test with txn under ASAN',
+        'oncall':'$ONCALL',
+        'executeLocal': 'true',
+        'timeout': 86400,
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Build and run RocksDB debug asan_crash_test_with_txn',
+                'timeout': 86400,
+                'shell':'$SHM $DEBUG $NON_TSAN_CRASH make J=1 asan_crash_test_with_txn || $CONTRUN_NAME=asan_crash_test_with_txn $TASK_CREATION_TOOL',
+                'user':'root',
+                $PARSER
+            },
+            $UPLOAD_DB_DIR,
+        ],
+        $REPORT
+    }
+]"
+
 #
 # RocksDB test under undefined behavior sanitizer
 #
@@ -553,7 +607,7 @@ UBSAN_TEST_COMMANDS="[
             $CLEANUP_ENV,
             {
                 'name':'Test RocksDB debug under UBSAN',
-                'shell':'set -o pipefail && $SHM $UBSAN $DEBUG make $PARALLELISM ubsan_check || $CONTRUN_NAME=ubsan_check $TASK_CREATION_TOOL',
+                'shell':'set -o pipefail && $SHM $UBSAN $CLANG $DEBUG make $PARALLELISM ubsan_check || $CONTRUN_NAME=ubsan_check $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             }
@@ -576,7 +630,7 @@ UBSAN_CRASH_TEST_COMMANDS="[
             {
                 'name':'Build and run RocksDB debug ubsan_crash_test',
                 'timeout': 86400,
-                'shell':'$SHM $DEBUG $NON_TSAN_CRASH make J=1 ubsan_crash_test || $CONTRUN_NAME=ubsan_crash_test $TASK_CREATION_TOOL',
+                'shell':'$SHM $DEBUG $NON_TSAN_CRASH $CLANG make J=1 ubsan_crash_test || $CONTRUN_NAME=ubsan_crash_test $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -599,7 +653,31 @@ UBSAN_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS="[
             {
                 'name':'Build and run RocksDB debug ubsan_crash_test_with_atomic_flush',
                 'timeout': 86400,
-                'shell':'$SHM $DEBUG $NON_TSAN_CRASH make J=1 ubsan_crash_test_with_atomic_flush || $CONTRUN_NAME=ubsan_crash_test_with_atomic_flush $TASK_CREATION_TOOL',
+                'shell':'$SHM $DEBUG $NON_TSAN_CRASH $CLANG make J=1 ubsan_crash_test_with_atomic_flush || $CONTRUN_NAME=ubsan_crash_test_with_atomic_flush $TASK_CREATION_TOOL',
+                'user':'root',
+                $PARSER
+            },
+            $UPLOAD_DB_DIR,
+        ],
+        $REPORT
+    }
+]"
+
+#
+# RocksDB crash testing with txn under undefined behavior sanitizer
+#
+UBSAN_CRASH_TEST_WITH_TXN_COMMANDS="[
+    {
+        'name':'Rocksdb crash test with txn under UBSAN',
+        'oncall':'$ONCALL',
+        'executeLocal': 'true',
+        'timeout': 86400,
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Build and run RocksDB debug ubsan_crash_test_with_txn',
+                'timeout': 86400,
+                'shell':'$SHM $DEBUG $NON_TSAN_CRASH $CLANG make J=1 ubsan_crash_test_with_txn || $CONTRUN_NAME=ubsan_crash_test_with_txn $TASK_CREATION_TOOL',
                 'user':'root',
                 $PARSER
             },
@@ -702,6 +780,30 @@ TSAN_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS="[
     }
 ]"
 
+#
+# RocksDB crash test with txn under TSAN
+#
+TSAN_CRASH_TEST_WITH_TXN_COMMANDS="[
+    {
+        'name':'Rocksdb Crash Test with txn under TSAN',
+        'oncall':'$ONCALL',
+        'executeLocal': 'true',
+        'timeout': 86400,
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Compile and run',
+                'timeout': 86400,
+                'shell':'set -o pipefail && $SHM $DEBUG $TSAN $TSAN_CRASH CRASH_TEST_KILL_ODD=1887 make J=1 crash_test_with_txn || $CONTRUN_NAME=tsan_crash_test_with_txn $TASK_CREATION_TOOL',
+                'user':'root',
+                $PARSER
+            },
+            $UPLOAD_DB_DIR,
+        ],
+        $REPORT
+    }
+]"
+
 #
 # RocksDB format compatible
 #
@@ -889,6 +991,9 @@ case $1 in
   stress_crash_with_atomic_flush)
     echo $STRESS_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS
     ;;
+  stress_crash_with_txn)
+    echo $STRESS_CRASH_TEST_WITH_TXN_COMMANDS
+    ;;
   write_stress)
     echo $WRITE_STRESS_COMMANDS
     ;;
@@ -901,6 +1006,9 @@ case $1 in
   asan_crash_with_atomic_flush)
     echo $ASAN_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS
     ;;
+  asan_crash_with_txn)
+    echo $ASAN_CRASH_TEST_WITH_TXN_COMMANDS
+    ;;
   ubsan)
     echo $UBSAN_TEST_COMMANDS
     ;;
@@ -910,6 +1018,9 @@ case $1 in
   ubsan_crash_with_atomic_flush)
     echo $UBSAN_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS
     ;;
+  ubsan_crash_with_txn)
+    echo $UBSAN_CRASH_TEST_WITH_TXN_COMMANDS
+    ;;
   valgrind)
     echo $VALGRIND_TEST_COMMANDS
     ;;
@@ -922,6 +1033,9 @@ case $1 in
   tsan_crash_with_atomic_flush)
     echo $TSAN_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS
     ;;
+  tsan_crash_with_txn)
+    echo $TSAN_CRASH_TEST_WITH_TXN_COMMANDS
+    ;;
   format_compatible)
     echo $FORMAT_COMPATIBLE_COMMANDS
     ;;

build_tools/update_dependencies.sh

@@ -92,6 +92,7 @@ get_lib_base jemalloc   LATEST  platform007
 get_lib_base numa       LATEST  platform007
 get_lib_base libunwind  LATEST  platform007
 get_lib_base tbb        LATEST  platform007
+get_lib_base liburing   LATEST  platform007
 
 get_lib_base kernel-headers fb platform007
 get_lib_base binutils   LATEST centos7-native

cache/cache_test.cc

@@ -365,7 +365,7 @@ TEST_P(CacheTest, EvictionPolicy) {
   Insert(200, 201);
 
   // Frequently used entry must be kept around
-  for (int i = 0; i < kCacheSize + 200; i++) {
+  for (int i = 0; i < kCacheSize * 2; i++) {
     Insert(1000+i, 2000+i);
     ASSERT_EQ(101, Lookup(100));
   }
@@ -418,7 +418,7 @@ TEST_P(CacheTest, EvictionPolicyRef) {
   Insert(303, 104);
 
   // Insert entries much more than Cache capacity
-  for (int i = 0; i < kCacheSize + 200; i++) {
+  for (int i = 0; i < kCacheSize * 2; i++) {
     Insert(1000 + i, 2000 + i);
   }
 

cache/lru_cache.h

@@ -133,7 +133,6 @@ struct LRUHandle {
   // Caclculate the memory usage by metadata
   inline size_t CalcTotalCharge(
       CacheMetadataChargePolicy metadata_charge_policy) {
-    assert(key_length);
     size_t meta_charge = 0;
     if (metadata_charge_policy == kFullChargeCacheMetadata) {
 #ifdef ROCKSDB_MALLOC_USABLE_SIZE

cmake/modules/Findgflags.cmake

@@ -0,0 +1,29 @@
+# - Find gflags library
+# Find the gflags includes and library
+#
+# gflags_INCLUDE_DIR - where to find gflags.h.
+# gflags_LIBRARIES - List of libraries when using gflags.
+# gflags_FOUND - True if gflags found.
+
+find_path(GFLAGS_INCLUDE_DIR
+  NAMES gflags/gflags.h)
+
+find_library(GFLAGS_LIBRARIES
+  NAMES gflags)
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(gflags
+  DEFAULT_MSG GFLAGS_LIBRARIES GFLAGS_INCLUDE_DIR)
+
+mark_as_advanced(
+  GFLAGS_LIBRARIES
+  GFLAGS_INCLUDE_DIR)
+
+if(gflags_FOUND AND NOT (TARGET gflags::gflags))
+  add_library(gflags::gflags UNKNOWN IMPORTED)
+  set_target_properties(gflags::gflags
+    PROPERTIES
+      IMPORTED_LOCATION ${GFLAGS_LIBRARIES}
+      INTERFACE_INCLUDE_DIRECTORIES ${GFLAGS_INCLUDE_DIR}
+      IMPORTED_LINK_INTERFACE_LANGUAGES "CXX")
+endif()

cmake/modules/ReadVersion.cmake

@@ -1,7 +1,7 @@
 # Read rocksdb version from version.h header file.
 
 function(get_rocksdb_version version_var)
-  file(READ "${CMAKE_SOURCE_DIR}/include/rocksdb/version.h" version_header_file)
+  file(READ "${CMAKE_CURRENT_SOURCE_DIR}/include/rocksdb/version.h" version_header_file)
   foreach(component MAJOR MINOR PATCH)
     string(REGEX MATCH "#define ROCKSDB_${component} ([0-9]+)" _ ${version_header_file})
     set(ROCKSDB_VERSION_${component} ${CMAKE_MATCH_1})

db/arena_wrapped_db_iter.cc

@@ -64,7 +64,7 @@ Status ArenaWrappedDBIter::Refresh() {
     arena_.~Arena();
     new (&arena_) Arena();
 
-    SuperVersion* sv = cfd_->GetReferencedSuperVersion(db_impl_->mutex());
+    SuperVersion* sv = cfd_->GetReferencedSuperVersion(db_impl_);
     if (read_callback_) {
       read_callback_->Refresh(latest_seq);
     }

db/blob_index.h

@@ -5,12 +5,14 @@
 #pragma once
 #ifndef ROCKSDB_LITE
 
+#include <sstream>
+#include <string>
+
 #include "rocksdb/options.h"
 #include "util/coding.h"
 #include "util/string_util.h"
 
 namespace rocksdb {
-namespace blob_db {
 
 // BlobIndex is a pointer to the blob and metadata of the blob. The index is
 // stored in base DB as ValueType::kTypeBlobIndex.
@@ -109,6 +111,23 @@ class BlobIndex {
     return Status::OK();
   }
 
+  std::string DebugString(bool output_hex) const {
+    std::ostringstream oss;
+
+    if (IsInlined()) {
+      oss << "[inlined blob] value:" << value_.ToString(output_hex);
+    } else {
+      oss << "[blob ref] file:" << file_number_ << " offset:" << offset_
+          << " size:" << size_;
+    }
+
+    if (HasTTL()) {
+      oss << " exp:" << expiration_;
+    }
+
+    return oss.str();
+  }
+
   static void EncodeInlinedTTL(std::string* dst, uint64_t expiration,
                                const Slice& value) {
     assert(dst != nullptr);
@@ -156,6 +175,5 @@ class BlobIndex {
   CompressionType compression_ = kNoCompression;
 };
 
-}  // namespace blob_db
 }  // namespace rocksdb
 #endif  // ROCKSDB_LITE

db/builder.cc

@@ -66,8 +66,9 @@ TableBuilder* NewTableBuilder(
 }
 
 Status BuildTable(
-    const std::string& dbname, Env* env, const ImmutableCFOptions& ioptions,
-    const MutableCFOptions& mutable_cf_options, const EnvOptions& env_options,
+    const std::string& dbname, Env* env, FileSystem* fs,
+    const ImmutableCFOptions& ioptions,
+    const MutableCFOptions& mutable_cf_options, const FileOptions& file_options,
     TableCache* table_cache, InternalIterator* iter,
     std::vector<std::unique_ptr<FragmentedRangeTombstoneIterator>>
         range_del_iters,
@@ -115,23 +116,23 @@ Status BuildTable(
     compression_opts_for_flush.max_dict_bytes = 0;
     compression_opts_for_flush.zstd_max_train_bytes = 0;
     {
-      std::unique_ptr<WritableFile> file;
+      std::unique_ptr<FSWritableFile> file;
 #ifndef NDEBUG
-      bool use_direct_writes = env_options.use_direct_writes;
+      bool use_direct_writes = file_options.use_direct_writes;
       TEST_SYNC_POINT_CALLBACK("BuildTable:create_file", &use_direct_writes);
 #endif  // !NDEBUG
-      s = NewWritableFile(env, fname, &file, env_options);
+      s = NewWritableFile(fs, fname, &file, file_options);
       if (!s.ok()) {
         EventHelpers::LogAndNotifyTableFileCreationFinished(
             event_logger, ioptions.listeners, dbname, column_family_name, fname,
-            job_id, meta->fd, tp, reason, s);
+            job_id, meta->fd, kInvalidBlobFileNumber, tp, reason, s);
         return s;
       }
       file->SetIOPriority(io_priority);
       file->SetWriteLifeTimeHint(write_hint);
 
       file_writer.reset(
-          new WritableFileWriter(std::move(file), fname, env_options, env,
+          new WritableFileWriter(std::move(file), fname, file_options, env,
                                  ioptions.statistics, ioptions.listeners));
       builder = NewTableBuilder(
           ioptions, mutable_cf_options, internal_comparator,
@@ -157,8 +158,9 @@ Status BuildTable(
     for (; c_iter.Valid(); c_iter.Next()) {
       const Slice& key = c_iter.key();
       const Slice& value = c_iter.value();
+      const ParsedInternalKey& ikey = c_iter.ikey();
       builder->Add(key, value);
-      meta->UpdateBoundaries(key, c_iter.ikey().sequence);
+      meta->UpdateBoundaries(key, value, ikey.sequence, ikey.type);
 
       // TODO(noetzli): Update stats after flush, too.
       if (io_priority == Env::IO_HIGH &&
@@ -217,7 +219,7 @@ Status BuildTable(
       // we will regrad this verification as user reads since the goal is
       // to cache it here for further user reads
       std::unique_ptr<InternalIterator> it(table_cache->NewIterator(
-          ReadOptions(), env_options, internal_comparator, *meta,
+          ReadOptions(), file_options, internal_comparator, *meta,
           nullptr /* range_del_agg */,
           mutable_cf_options.prefix_extractor.get(), nullptr,
           (internal_stats == nullptr) ? nullptr
@@ -240,13 +242,16 @@ Status BuildTable(
   }
 
   if (!s.ok() || meta->fd.GetFileSize() == 0) {
-    env->DeleteFile(fname);
+    fs->DeleteFile(fname, IOOptions(), nullptr);
   }
 
+  if (meta->fd.GetFileSize() == 0) {
+    fname = "(nil)";
+  }
   // Output to event logger and fire events.
   EventHelpers::LogAndNotifyTableFileCreationFinished(
       event_logger, ioptions.listeners, dbname, column_family_name, fname,
-      job_id, meta->fd, tp, reason, s);
+      job_id, meta->fd, meta->oldest_blob_file_number, tp, reason, s);
 
   return s;
 }

db/builder.h

@@ -62,8 +62,9 @@ TableBuilder* NewTableBuilder(
 // @param column_family_name Name of the column family that is also identified
 //    by column_family_id, or empty string if unknown.
 extern Status BuildTable(
-    const std::string& dbname, Env* env, const ImmutableCFOptions& options,
-    const MutableCFOptions& mutable_cf_options, const EnvOptions& env_options,
+    const std::string& dbname, Env* env, FileSystem* fs,
+    const ImmutableCFOptions& options,
+    const MutableCFOptions& mutable_cf_options, const FileOptions& file_options,
     TableCache* table_cache, InternalIterator* iter,
     std::vector<std::unique_ptr<FragmentedRangeTombstoneIterator>>
         range_del_iters,

db/c.cc

@@ -852,6 +852,17 @@ void rocksdb_delete_cf(
         Slice(key, keylen)));
 }
 
+void rocksdb_delete_range_cf(rocksdb_t* db,
+                             const rocksdb_writeoptions_t* options,
+                             rocksdb_column_family_handle_t* column_family,
+                             const char* start_key, size_t start_key_len,
+                             const char* end_key, size_t end_key_len,
+                             char** errptr) {
+  SaveError(errptr, db->rep->DeleteRange(options->rep, column_family->rep,
+                                         Slice(start_key, start_key_len),
+                                         Slice(end_key, end_key_len)));
+}
+
 void rocksdb_merge(
     rocksdb_t* db,
     const rocksdb_writeoptions_t* options,
@@ -2032,6 +2043,17 @@ void rocksdb_block_based_options_set_index_type(
   options->rep.index_type = static_cast<BlockBasedTableOptions::IndexType>(v);
 }
 
+void rocksdb_block_based_options_set_data_block_index_type(
+    rocksdb_block_based_table_options_t* options, int v) {
+  options->rep.data_block_index_type =
+          static_cast<BlockBasedTableOptions::DataBlockIndexType>(v);
+}
+
+void rocksdb_block_based_options_set_data_block_hash_ratio(
+    rocksdb_block_based_table_options_t* options, double v) {
+  options->rep.data_block_hash_table_util_ratio = v;
+}
+
 void rocksdb_block_based_options_set_hash_index_allow_collision(
     rocksdb_block_based_table_options_t* options, unsigned char v) {
   options->rep.hash_index_allow_collision = v;
@@ -3037,7 +3059,16 @@ rocksdb_filterpolicy_t* rocksdb_filterpolicy_create_bloom_format(int bits_per_ke
     bool KeyMayMatch(const Slice& key, const Slice& filter) const override {
       return rep_->KeyMayMatch(key, filter);
     }
-    static void DoNothing(void*) { }
+    // No need to override GetFilterBitsBuilder if this one is overridden
+    rocksdb::FilterBitsBuilder* GetBuilderWithContext(
+        const rocksdb::FilterBuildingContext& context) const override {
+      return rep_->GetBuilderWithContext(context);
+    }
+    rocksdb::FilterBitsReader* GetFilterBitsReader(
+        const Slice& contents) const override {
+      return rep_->GetFilterBitsReader(contents);
+    }
+    static void DoNothing(void*) {}
   };
   Wrapper* wrapper = new Wrapper;
   wrapper->rep_ = NewBloomFilterPolicy(bits_per_key, original_format);

db/c_test.c

@@ -487,8 +487,11 @@ int main(int argc, char** argv) {
   rocksdb_options_set_paranoid_checks(options, 1);
   rocksdb_options_set_max_open_files(options, 10);
   rocksdb_options_set_base_background_compactions(options, 1);
+
   table_options = rocksdb_block_based_options_create();
   rocksdb_block_based_options_set_block_cache(table_options, cache);
+  rocksdb_block_based_options_set_data_block_index_type(table_options, 1);
+  rocksdb_block_based_options_set_data_block_hash_ratio(table_options, 0.75);
   rocksdb_options_set_block_based_table_factory(options, table_options);
 
   rocksdb_options_set_compression(options, rocksdb_no_compression);
@@ -1048,17 +1051,20 @@ int main(int argc, char** argv) {
   }
 
   StartPhase("filter");
-  for (run = 0; run < 2; run++) {
-    // First run uses custom filter, second run uses bloom filter
+  for (run = 0; run <= 2; run++) {
+    // First run uses custom filter
+    // Second run uses old block-based bloom filter
+    // Third run uses full bloom filter
     CheckNoError(err);
     rocksdb_filterpolicy_t* policy;
     if (run == 0) {
-      policy = rocksdb_filterpolicy_create(
-          NULL, FilterDestroy, FilterCreate, FilterKeyMatch, NULL, FilterName);
+      policy = rocksdb_filterpolicy_create(NULL, FilterDestroy, FilterCreate,
+                                           FilterKeyMatch, NULL, FilterName);
+    } else if (run == 1) {
+      policy = rocksdb_filterpolicy_create_bloom(8);
     } else {
-      policy = rocksdb_filterpolicy_create_bloom(10);
+      policy = rocksdb_filterpolicy_create_bloom_full(8);
     }
-
     rocksdb_block_based_options_set_filter_policy(table_options, policy);
 
     // Create new database
@@ -1071,12 +1077,24 @@ int main(int argc, char** argv) {
     CheckNoError(err);
     rocksdb_put(db, woptions, "bar", 3, "barvalue", 8, &err);
     CheckNoError(err);
+
+    {
+      // Add enough keys to get just one reasonably populated Bloom filter
+      const int keys_to_add = 1500;
+      int i;
+      char keybuf[100];
+      for (i = 0; i < keys_to_add; i++) {
+        snprintf(keybuf, sizeof(keybuf), "yes%020d", i);
+        rocksdb_put(db, woptions, keybuf, strlen(keybuf), "val", 3, &err);
+        CheckNoError(err);
+      }
+    }
     rocksdb_compact_range(db, NULL, 0, NULL, 0);
 
     fake_filter_result = 1;
     CheckGet(db, roptions, "foo", "foovalue");
     CheckGet(db, roptions, "bar", "barvalue");
-    if (phase == 0) {
+    if (run == 0) {
       // Must not find value when custom filter returns false
       fake_filter_result = 0;
       CheckGet(db, roptions, "foo", NULL);
@@ -1086,6 +1104,43 @@ int main(int argc, char** argv) {
       CheckGet(db, roptions, "foo", "foovalue");
       CheckGet(db, roptions, "bar", "barvalue");
     }
+
+    {
+      // Query some keys not added to identify Bloom filter implementation
+      // from false positive queries, using perfcontext to detect Bloom
+      // filter behavior
+      rocksdb_perfcontext_t* perf = rocksdb_perfcontext_create();
+      rocksdb_perfcontext_reset(perf);
+
+      const int keys_to_query = 10000;
+      int i;
+      char keybuf[100];
+      for (i = 0; i < keys_to_query; i++) {
+        fake_filter_result = i % 2;
+        snprintf(keybuf, sizeof(keybuf), "no%020d", i);
+        CheckGet(db, roptions, keybuf, NULL);
+      }
+
+      const int hits =
+          (int)rocksdb_perfcontext_metric(perf, rocksdb_bloom_sst_hit_count);
+      if (run == 0) {
+        // Due to half true, half false with fake filter result
+        CheckCondition(hits == keys_to_query / 2);
+      } else if (run == 1) {
+        // Essentially a fingerprint of the block-based Bloom schema
+        CheckCondition(hits == 241);
+      } else {
+        // Essentially a fingerprint of the full Bloom schema(s),
+        // format_version < 5, which vary for three different CACHE_LINE_SIZEs
+        CheckCondition(hits == 224 || hits == 180 || hits == 125);
+      }
+      CheckCondition(
+          (keys_to_query - hits) ==
+          (int)rocksdb_perfcontext_metric(perf, rocksdb_bloom_sst_miss_count));
+
+      rocksdb_perfcontext_destroy(perf);
+    }
+
     // Reset the policy
     rocksdb_block_based_options_set_filter_policy(table_options, NULL);
     rocksdb_options_set_block_based_table_factory(options, table_options);
@@ -1189,6 +1244,15 @@ int main(int argc, char** argv) {
     rocksdb_put_cf(db, woptions, handles[1], "foo", 3, "hello", 5, &err);
     CheckNoError(err);
 
+    rocksdb_put_cf(db, woptions, handles[1], "foobar1", 7, "hello1", 6, &err);
+    CheckNoError(err);
+    rocksdb_put_cf(db, woptions, handles[1], "foobar2", 7, "hello2", 6, &err);
+    CheckNoError(err);
+    rocksdb_put_cf(db, woptions, handles[1], "foobar3", 7, "hello3", 6, &err);
+    CheckNoError(err);
+    rocksdb_put_cf(db, woptions, handles[1], "foobar4", 7, "hello4", 6, &err);
+    CheckNoError(err);
+
     rocksdb_flushoptions_t *flush_options = rocksdb_flushoptions_create();
     rocksdb_flushoptions_set_wait(flush_options, 1);
     rocksdb_flush_cf(db, flush_options, handles[1], &err);
@@ -1201,6 +1265,10 @@ int main(int argc, char** argv) {
     rocksdb_delete_cf(db, woptions, handles[1], "foo", 3, &err);
     CheckNoError(err);
 
+    rocksdb_delete_range_cf(db, woptions, handles[1], "foobar2", 7, "foobar4",
+                            7, &err);
+    CheckNoError(err);
+
     CheckGetCF(db, roptions, handles[1], "foo", NULL);
     CheckPinGetCF(db, roptions, handles[1], "foo", NULL);
 
@@ -1253,7 +1321,7 @@ int main(int argc, char** argv) {
     for (i = 0; rocksdb_iter_valid(iter) != 0; rocksdb_iter_next(iter)) {
       i++;
     }
-    CheckCondition(i == 1);
+    CheckCondition(i == 3);
     rocksdb_iter_get_error(iter, &err);
     CheckNoError(err);
     rocksdb_iter_destroy(iter);
@@ -1277,7 +1345,7 @@ int main(int argc, char** argv) {
     for (i = 0; rocksdb_iter_valid(iter) != 0; rocksdb_iter_next(iter)) {
       i++;
     }
-    CheckCondition(i == 1);
+    CheckCondition(i == 3);
     rocksdb_iter_get_error(iter, &err);
     CheckNoError(err);
     rocksdb_iter_destroy(iter);
@@ -1467,6 +1535,7 @@ int main(int argc, char** argv) {
        CheckCondition(!rocksdb_iter_valid(iter));
 
        rocksdb_iter_destroy(iter);
+       rocksdb_readoptions_set_iterate_upper_bound(roptions, NULL, 0);
     }
   }
 

db/column_family.cc

@@ -30,6 +30,7 @@
 #include "memtable/hash_skiplist_rep.h"
 #include "monitoring/thread_status_util.h"
 #include "options/options_helper.h"
+#include "port/port.h"
 #include "table/block_based/block_based_table_factory.h"
 #include "table/merging_iterator.h"
 #include "util/autovector.h"
@@ -59,11 +60,8 @@ ColumnFamilyHandleImpl::~ColumnFamilyHandleImpl() {
     ColumnFamilyOptions initial_cf_options_copy = cfd_->initial_cf_options();
     JobContext job_context(0);
     mutex_->Lock();
-    if (cfd_->Unref()) {
-      bool dropped = cfd_->IsDropped();
-
-      delete cfd_;
-
+    bool dropped = cfd_->IsDropped();
+    if (cfd_->UnrefAndTryDelete()) {
       if (dropped) {
         db_->FindObsoleteFiles(&job_context, false, true);
       }
@@ -162,6 +160,11 @@ Status CheckConcurrentWritesSupported(const ColumnFamilyOptions& cf_options) {
     return Status::InvalidArgument(
         "Memtable doesn't concurrent writes (allow_concurrent_memtable_write)");
   }
+  if (cf_options.max_successive_merges != 0) {
+    return Status::InvalidArgument(
+        "max_successive_merges > 0 is incompatible "
+        "with concurrent writes (allow_concurrent_memtable_write)");
+  }
   return Status::OK();
 }
 
@@ -187,6 +190,11 @@ Status CheckCFPathsSupported(const DBOptions& db_options,
   return Status::OK();
 }
 
+namespace {
+const uint64_t kDefaultTtl = 0xfffffffffffffffe;
+const uint64_t kDefaultPeriodicCompSecs = 0xfffffffffffffffe;
+};  // namespace
+
 ColumnFamilyOptions SanitizeOptions(const ImmutableDBOptions& db_options,
                                     const ColumnFamilyOptions& src) {
   ColumnFamilyOptions result = src;
@@ -342,6 +350,61 @@ ColumnFamilyOptions SanitizeOptions(const ImmutableDBOptions& db_options,
     result.max_compaction_bytes = result.target_file_size_base * 25;
   }
 
+  bool is_block_based_table =
+      (result.table_factory->Name() == BlockBasedTableFactory().Name());
+
+  const uint64_t kAdjustedTtl = 30 * 24 * 60 * 60;
+  if (result.ttl == kDefaultTtl) {
+    if (is_block_based_table &&
+        result.compaction_style != kCompactionStyleFIFO) {
+      result.ttl = kAdjustedTtl;
+    } else {
+      result.ttl = 0;
+    }
+  }
+
+  const uint64_t kAdjustedPeriodicCompSecs = 30 * 24 * 60 * 60;
+
+  // Turn on periodic compactions and set them to occur once every 30 days if
+  // compaction filters are used and periodic_compaction_seconds is set to the
+  // default value.
+  if (result.compaction_style != kCompactionStyleFIFO) {
+    if ((result.compaction_filter != nullptr ||
+         result.compaction_filter_factory != nullptr) &&
+        result.periodic_compaction_seconds == kDefaultPeriodicCompSecs &&
+        is_block_based_table) {
+      result.periodic_compaction_seconds = kAdjustedPeriodicCompSecs;
+    }
+  } else {
+    // result.compaction_style == kCompactionStyleFIFO
+    if (result.ttl == 0) {
+      if (is_block_based_table) {
+        if (result.periodic_compaction_seconds == kDefaultPeriodicCompSecs) {
+          result.periodic_compaction_seconds = kAdjustedPeriodicCompSecs;
+        }
+        result.ttl = result.periodic_compaction_seconds;
+      }
+    } else if (result.periodic_compaction_seconds != 0) {
+      result.ttl = std::min(result.ttl, result.periodic_compaction_seconds);
+    }
+  }
+
+  // TTL compactions would work similar to Periodic Compactions in Universal in
+  // most of the cases. So, if ttl is set, execute the periodic compaction
+  // codepath.
+  if (result.compaction_style == kCompactionStyleUniversal && result.ttl != 0) {
+    if (result.periodic_compaction_seconds != 0) {
+      result.periodic_compaction_seconds =
+          std::min(result.ttl, result.periodic_compaction_seconds);
+    } else {
+      result.periodic_compaction_seconds = result.ttl;
+    }
+  }
+
+  if (result.periodic_compaction_seconds == kDefaultPeriodicCompSecs) {
+    result.periodic_compaction_seconds = 0;
+  }
+
   return result;
 }
 
@@ -378,13 +441,18 @@ void SuperVersion::Cleanup() {
     to_delete.push_back(m);
   }
   current->Unref();
+  if (cfd->Unref()) {
+    delete cfd;
+  }
 }
 
-void SuperVersion::Init(MemTable* new_mem, MemTableListVersion* new_imm,
-                        Version* new_current) {
+void SuperVersion::Init(ColumnFamilyData* new_cfd, MemTable* new_mem,
+                        MemTableListVersion* new_imm, Version* new_current) {
+  cfd = new_cfd;
   mem = new_mem;
   imm = new_imm;
   current = new_current;
+  cfd->Ref();
   mem->Ref();
   imm->Ref();
   current->Ref();
@@ -412,7 +480,7 @@ ColumnFamilyData::ColumnFamilyData(
     uint32_t id, const std::string& name, Version* _dummy_versions,
     Cache* _table_cache, WriteBufferManager* write_buffer_manager,
     const ColumnFamilyOptions& cf_options, const ImmutableDBOptions& db_options,
-    const EnvOptions& env_options, ColumnFamilySet* column_family_set,
+    const FileOptions& file_options, ColumnFamilySet* column_family_set,
     BlockCacheTracer* const block_cache_tracer)
     : id_(id),
       name_(name),
@@ -454,7 +522,7 @@ ColumnFamilyData::ColumnFamilyData(
   if (_dummy_versions != nullptr) {
     internal_stats_.reset(
         new InternalStats(ioptions_.num_levels, db_options.env, this));
-    table_cache_.reset(new TableCache(ioptions_, env_options, _table_cache,
+    table_cache_.reset(new TableCache(ioptions_, file_options, _table_cache,
                                       block_cache_tracer));
     if (ioptions_.compaction_style == kCompactionStyleLevel) {
       compaction_picker_.reset(
@@ -520,21 +588,7 @@ ColumnFamilyData::~ColumnFamilyData() {
   // compaction_queue_ and we destroyed it
   assert(!queued_for_flush_);
   assert(!queued_for_compaction_);
-
-  if (super_version_ != nullptr) {
-    // Release SuperVersion reference kept in ThreadLocalPtr.
-    // This must be done outside of mutex_ since unref handler can lock mutex.
-    super_version_->db_mutex->Unlock();
-    local_sv_.reset();
-    super_version_->db_mutex->Lock();
-
-    bool is_last_reference __attribute__((__unused__));
-    is_last_reference = super_version_->Unref();
-    assert(is_last_reference);
-    super_version_->Cleanup();
-    delete super_version_;
-    super_version_ = nullptr;
-  }
+  assert(super_version_ == nullptr);
 
   if (dummy_versions_ != nullptr) {
     // List must be empty
@@ -554,6 +608,36 @@ ColumnFamilyData::~ColumnFamilyData() {
   }
 }
 
+bool ColumnFamilyData::UnrefAndTryDelete() {
+  int old_refs = refs_.fetch_sub(1);
+  assert(old_refs > 0);
+
+  if (old_refs == 1) {
+    assert(super_version_ == nullptr);
+    delete this;
+    return true;
+  }
+
+  if (old_refs == 2 && super_version_ != nullptr) {
+    // Only the super_version_ holds me
+    SuperVersion* sv = super_version_;
+    super_version_ = nullptr;
+    // Release SuperVersion reference kept in ThreadLocalPtr.
+    // This must be done outside of mutex_ since unref handler can lock mutex.
+    sv->db_mutex->Unlock();
+    local_sv_.reset();
+    sv->db_mutex->Lock();
+
+    if (sv->Unref()) {
+      // May delete this ColumnFamilyData after calling Cleanup()
+      sv->Cleanup();
+      delete sv;
+      return true;
+    }
+  }
+  return false;
+}
+
 void ColumnFamilyData::SetDropped() {
   // can't drop default CF
   assert(id_ != 0);
@@ -888,8 +972,8 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
   return write_stall_condition;
 }
 
-const EnvOptions* ColumnFamilyData::soptions() const {
-  return &(column_family_set_->env_options_);
+const FileOptions* ColumnFamilyData::soptions() const {
+  return &(column_family_set_->file_options_);
 }
 
 void ColumnFamilyData::SetCurrent(Version* current_version) {
@@ -929,8 +1013,12 @@ bool ColumnFamilyData::NeedsCompaction() const {
 
 Compaction* ColumnFamilyData::PickCompaction(
     const MutableCFOptions& mutable_options, LogBuffer* log_buffer) {
+  SequenceNumber earliest_mem_seqno =
+      std::min(mem_->GetEarliestSequenceNumber(),
+               imm_.current()->GetEarliestSequenceNumber(false));
   auto* result = compaction_picker_->PickCompaction(
-      GetName(), mutable_options, current_->storage_info(), log_buffer);
+      GetName(), mutable_options, current_->storage_info(), log_buffer,
+      earliest_mem_seqno);
   if (result != nullptr) {
     result->SetInputVersion(current_);
   }
@@ -1015,9 +1103,8 @@ Compaction* ColumnFamilyData::CompactRange(
   return result;
 }
 
-SuperVersion* ColumnFamilyData::GetReferencedSuperVersion(
-    InstrumentedMutex* db_mutex) {
-  SuperVersion* sv = GetThreadLocalSuperVersion(db_mutex);
+SuperVersion* ColumnFamilyData::GetReferencedSuperVersion(DBImpl* db) {
+  SuperVersion* sv = GetThreadLocalSuperVersion(db);
   sv->Ref();
   if (!ReturnThreadLocalSuperVersion(sv)) {
     // This Unref() corresponds to the Ref() in GetThreadLocalSuperVersion()
@@ -1029,8 +1116,7 @@ SuperVersion* ColumnFamilyData::GetReferencedSuperVersion(
   return sv;
 }
 
-SuperVersion* ColumnFamilyData::GetThreadLocalSuperVersion(
-    InstrumentedMutex* db_mutex) {
+SuperVersion* ColumnFamilyData::GetThreadLocalSuperVersion(DBImpl* db) {
   // The SuperVersion is cached in thread local storage to avoid acquiring
   // mutex when SuperVersion does not change since the last use. When a new
   // SuperVersion is installed, the compaction or flush thread cleans up
@@ -1057,16 +1143,21 @@ SuperVersion* ColumnFamilyData::GetThreadLocalSuperVersion(
 
     if (sv && sv->Unref()) {
       RecordTick(ioptions_.statistics, NUMBER_SUPERVERSION_CLEANUPS);
-      db_mutex->Lock();
+      db->mutex()->Lock();
       // NOTE: underlying resources held by superversion (sst files) might
       // not be released until the next background job.
       sv->Cleanup();
-      sv_to_delete = sv;
+      if (db->immutable_db_options().avoid_unnecessary_blocking_io) {
+        db->AddSuperVersionsToFreeQueue(sv);
+        db->SchedulePurge();
+      } else {
+        sv_to_delete = sv;
+      }
     } else {
-      db_mutex->Lock();
+      db->mutex()->Lock();
     }
     sv = super_version_->Ref();
-    db_mutex->Unlock();
+    db->mutex()->Unlock();
 
     delete sv_to_delete;
   }
@@ -1104,7 +1195,7 @@ void ColumnFamilyData::InstallSuperVersion(
   SuperVersion* new_superversion = sv_context->new_superversion.release();
   new_superversion->db_mutex = db_mutex;
   new_superversion->mutable_cf_options = mutable_cf_options;
-  new_superversion->Init(mem_, imm_.current(), current_);
+  new_superversion->Init(this, mem_, imm_.current(), current_);
   SuperVersion* old_superversion = super_version_;
   super_version_ = new_superversion;
   ++super_version_number_;
@@ -1168,24 +1259,15 @@ Status ColumnFamilyData::ValidateOptions(
     return s;
   }
 
-  if (cf_options.ttl > 0) {
-    if (db_options.max_open_files != -1) {
-      return Status::NotSupported(
-          "TTL is only supported when files are always "
-          "kept open (set max_open_files = -1). ");
-    }
+  if (cf_options.ttl > 0 && cf_options.ttl != kDefaultTtl) {
     if (cf_options.table_factory->Name() != BlockBasedTableFactory().Name()) {
       return Status::NotSupported(
           "TTL is only supported in Block-Based Table format. ");
     }
   }
 
-  if (cf_options.periodic_compaction_seconds > 0) {
-    if (db_options.max_open_files != -1) {
-      return Status::NotSupported(
-          "Periodic Compaction is only supported when files are always "
-          "kept open (set max_open_files = -1). ");
-    }
+  if (cf_options.periodic_compaction_seconds > 0 &&
+      cf_options.periodic_compaction_seconds != kDefaultPeriodicCompSecs) {
     if (cf_options.table_factory->Name() != BlockBasedTableFactory().Name()) {
       return Status::NotSupported(
           "Periodic Compaction is only supported in "
@@ -1229,6 +1311,10 @@ Env::WriteLifeTimeHint ColumnFamilyData::CalculateSSTWriteHint(int level) {
   // L1: medium, L2: long, ...
   if (level - base_level >= 2) {
     return Env::WLTH_EXTREME;
+  } else if (level < base_level) {
+    // There is no restriction which prevents level passed in to be smaller
+    // than base_level.
+    return Env::WLTH_MEDIUM;
   }
   return static_cast<Env::WriteLifeTimeHint>(level - base_level +
                             static_cast<int>(Env::WLTH_MEDIUM));
@@ -1261,7 +1347,7 @@ Directory* ColumnFamilyData::GetDataDir(size_t path_id) const {
 
 ColumnFamilySet::ColumnFamilySet(const std::string& dbname,
                                  const ImmutableDBOptions* db_options,
-                                 const EnvOptions& env_options,
+                                 const FileOptions& file_options,
                                  Cache* table_cache,
                                  WriteBufferManager* write_buffer_manager,
                                  WriteController* write_controller,
@@ -1269,11 +1355,11 @@ ColumnFamilySet::ColumnFamilySet(const std::string& dbname,
     : max_column_family_(0),
       dummy_cfd_(new ColumnFamilyData(
           0, "", nullptr, nullptr, nullptr, ColumnFamilyOptions(), *db_options,
-          env_options, nullptr, block_cache_tracer)),
+          file_options, nullptr, block_cache_tracer)),
       default_cfd_cache_(nullptr),
       db_name_(dbname),
       db_options_(db_options),
-      env_options_(env_options),
+      file_options_(file_options),
       table_cache_(table_cache),
       write_buffer_manager_(write_buffer_manager),
       write_controller_(write_controller),
@@ -1288,14 +1374,12 @@ ColumnFamilySet::~ColumnFamilySet() {
     // cfd destructor will delete itself from column_family_data_
     auto cfd = column_family_data_.begin()->second;
     bool last_ref __attribute__((__unused__));
-    last_ref = cfd->Unref();
+    last_ref = cfd->UnrefAndTryDelete();
     assert(last_ref);
-    delete cfd;
   }
   bool dummy_last_ref __attribute__((__unused__));
-  dummy_last_ref = dummy_cfd_->Unref();
+  dummy_last_ref = dummy_cfd_->UnrefAndTryDelete();
   assert(dummy_last_ref);
-  delete dummy_cfd_;
 }
 
 ColumnFamilyData* ColumnFamilySet::GetDefault() const {
@@ -1345,7 +1429,7 @@ ColumnFamilyData* ColumnFamilySet::CreateColumnFamily(
   assert(column_families_.find(name) == column_families_.end());
   ColumnFamilyData* new_cfd = new ColumnFamilyData(
       id, name, dummy_versions, table_cache_, write_buffer_manager_, options,
-      *db_options_, env_options_, this, block_cache_tracer_);
+      *db_options_, file_options_, this, block_cache_tracer_);
   column_families_.insert({name, id});
   column_family_data_.insert({id, new_cfd});
   max_column_family_ = std::max(max_column_family_, id);

db/column_family.h

@@ -198,6 +198,7 @@ class ColumnFamilyHandleInternal : public ColumnFamilyHandleImpl {
 struct SuperVersion {
   // Accessing members of this class is not thread-safe and requires external
   // synchronization (ie db mutex held or on write thread).
+  ColumnFamilyData* cfd;
   MemTable* mem;
   MemTableListVersion* imm;
   Version* current;
@@ -221,8 +222,8 @@ struct SuperVersion {
   // that needs to be deleted in to_delete vector. Unrefing those
   // objects needs to be done in the mutex
   void Cleanup();
-  void Init(MemTable* new_mem, MemTableListVersion* new_imm,
-            Version* new_current);
+  void Init(ColumnFamilyData* new_cfd, MemTable* new_mem,
+            MemTableListVersion* new_imm, Version* new_current);
 
   // The value of dummy is not actually used. kSVInUse takes its address as a
   // mark in the thread local storage to indicate the SuperVersion is in use
@@ -288,6 +289,11 @@ class ColumnFamilyData {
     return old_refs == 1;
   }
 
+  // UnrefAndTryDelete() decreases the reference count and do free if needed,
+  // return true if this is freed else false, UnrefAndTryDelete() can only
+  // be called while holding a DB mutex, or during single-threaded recovery.
+  bool UnrefAndTryDelete();
+
   // SetDropped() can only be called under following conditions:
   // 1) Holding a DB mutex,
   // 2) from single-threaded write thread, AND
@@ -318,7 +324,7 @@ class ColumnFamilyData {
   }
   FlushReason GetFlushReason() const { return flush_reason_; }
   // thread-safe
-  const EnvOptions* soptions() const;
+  const FileOptions* soptions() const;
   const ImmutableCFOptions* ioptions() const { return &ioptions_; }
   // REQUIRES: DB mutex held
   // This returns the MutableCFOptions used by current SuperVersion
@@ -430,11 +436,11 @@ class ColumnFamilyData {
   SuperVersion* GetSuperVersion() { return super_version_; }
   // thread-safe
   // Return a already referenced SuperVersion to be used safely.
-  SuperVersion* GetReferencedSuperVersion(InstrumentedMutex* db_mutex);
+  SuperVersion* GetReferencedSuperVersion(DBImpl* db);
   // thread-safe
   // Get SuperVersion stored in thread local storage. If it does not exist,
   // get a reference from a current SuperVersion.
-  SuperVersion* GetThreadLocalSuperVersion(InstrumentedMutex* db_mutex);
+  SuperVersion* GetThreadLocalSuperVersion(DBImpl* db);
   // Try to return SuperVersion back to thread local storage. Retrun true on
   // success and false on failure. It fails when the thread local storage
   // contains anything other than SuperVersion::kSVInUse flag.
@@ -504,7 +510,7 @@ class ColumnFamilyData {
                    WriteBufferManager* write_buffer_manager,
                    const ColumnFamilyOptions& options,
                    const ImmutableDBOptions& db_options,
-                   const EnvOptions& env_options,
+                   const FileOptions& file_options,
                    ColumnFamilySet* column_family_set,
                    BlockCacheTracer* const block_cache_tracer);
 
@@ -632,7 +638,7 @@ class ColumnFamilySet {
 
   ColumnFamilySet(const std::string& dbname,
                   const ImmutableDBOptions* db_options,
-                  const EnvOptions& env_options, Cache* table_cache,
+                  const FileOptions& file_options, Cache* table_cache,
                   WriteBufferManager* write_buffer_manager,
                   WriteController* write_controller,
                   BlockCacheTracer* const block_cache_tracer);
@@ -690,7 +696,7 @@ class ColumnFamilySet {
 
   const std::string db_name_;
   const ImmutableDBOptions* const db_options_;
-  const EnvOptions env_options_;
+  const FileOptions file_options_;
   Cache* table_cache_;
   WriteBufferManager* write_buffer_manager_;
   WriteController* write_controller_;

db/column_family_test.cc

@@ -67,9 +67,9 @@ class ColumnFamilyTestBase : public testing::Test {
 #ifndef ROCKSDB_LITE
     const char* test_env_uri = getenv("TEST_ENV_URI");
     if (test_env_uri) {
-      Status s = ObjectRegistry::NewInstance()->NewSharedObject(test_env_uri,
-                                                                &env_guard_);
-      base_env = env_guard_.get();
+      Env* test_env = nullptr;
+      Status s = Env::LoadEnv(test_env_uri, &test_env, &env_guard_);
+      base_env = test_env;
       EXPECT_OK(s);
       EXPECT_NE(Env::Default(), base_env);
     }
@@ -2365,6 +2365,45 @@ TEST_P(ColumnFamilyTest, ReadDroppedColumnFamily) {
   }
 }
 
+TEST_P(ColumnFamilyTest, LiveIteratorWithDroppedColumnFamily) {
+  db_options_.create_missing_column_families = true;
+  db_options_.max_open_files = 20;
+  // delete obsolete files always
+  db_options_.delete_obsolete_files_period_micros = 0;
+  Open({"default", "one", "two"});
+  ColumnFamilyOptions options;
+  options.level0_file_num_compaction_trigger = 100;
+  options.level0_slowdown_writes_trigger = 200;
+  options.level0_stop_writes_trigger = 200;
+  options.write_buffer_size = 100000;  // small write buffer size
+  Reopen({options, options, options});
+
+  // 1MB should create ~10 files for each CF
+  int kKeysNum = 10000;
+  PutRandomData(1, kKeysNum, 100);
+
+  {
+    std::unique_ptr<Iterator> iterator(
+        db_->NewIterator(ReadOptions(), handles_[1]));
+    iterator->SeekToFirst();
+
+    DropColumnFamilies({1});
+
+    // Make sure iterator created can still be used.
+    int count = 0;
+    for (; iterator->Valid(); iterator->Next()) {
+      ASSERT_OK(iterator->status());
+      ++count;
+    }
+    ASSERT_OK(iterator->status());
+    ASSERT_EQ(count, kKeysNum);
+  }
+
+  Reopen();
+  Close();
+  Destroy();
+}
+
 TEST_P(ColumnFamilyTest, FlushAndDropRaceCondition) {
   db_options_.create_missing_column_families = true;
   Open({"default", "one"});

db/compact_files_test.cc

@@ -387,7 +387,7 @@ TEST_F(CompactFilesTest, GetCompactionJobInfo) {
   auto l0_files_1 = collector->GetFlushedFiles();
   CompactionOptions co;
   co.compression = CompressionType::kLZ4Compression;
-  CompactionJobInfo compaction_job_info;
+  CompactionJobInfo compaction_job_info{};
   ASSERT_OK(
       db->CompactFiles(co, l0_files_1, 0, -1, nullptr, &compaction_job_info));
   ASSERT_EQ(compaction_job_info.base_input_level, 0);

db/compaction/compaction.cc

@@ -275,9 +275,7 @@ Compaction::~Compaction() {
     input_version_->Unref();
   }
   if (cfd_ != nullptr) {
-    if (cfd_->Unref()) {
-      delete cfd_;
-    }
+    cfd_->UnrefAndTryDelete();
   }
 }
 
@@ -545,17 +543,16 @@ bool Compaction::ShouldFormSubcompactions() const {
   }
 }
 
-uint64_t Compaction::MaxInputFileCreationTime() const {
-  uint64_t max_creation_time = 0;
+uint64_t Compaction::MinInputFileOldestAncesterTime() const {
+  uint64_t min_oldest_ancester_time = port::kMaxUint64;
   for (const auto& file : inputs_[0].files) {
-    if (file->fd.table_reader != nullptr &&
-        file->fd.table_reader->GetTableProperties() != nullptr) {
-      uint64_t creation_time =
-          file->fd.table_reader->GetTableProperties()->creation_time;
-      max_creation_time = std::max(max_creation_time, creation_time);
+    uint64_t oldest_ancester_time = file->TryGetOldestAncesterTime();
+    if (oldest_ancester_time != 0) {
+      min_oldest_ancester_time =
+          std::min(min_oldest_ancester_time, oldest_ancester_time);
     }
   }
-  return max_creation_time;
+  return min_oldest_ancester_time;
 }
 
 int Compaction::GetInputBaseLevel() const {

db/compaction/compaction.h

@@ -291,7 +291,7 @@ class Compaction {
 
   uint32_t max_subcompactions() const { return max_subcompactions_; }
 
-  uint64_t MaxInputFileCreationTime() const;
+  uint64_t MinInputFileOldestAncesterTime() const;
 
  private:
   // mark (or clear) all files that are being compacted

db/compaction/compaction_iterator.cc

@@ -3,6 +3,8 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).
 
+#include <cinttypes>
+
 #include "db/compaction/compaction_iterator.h"
 #include "db/snapshot_checker.h"
 #include "port/likely.h"
@@ -38,7 +40,8 @@ CompactionIterator::CompactionIterator(
     const CompactionFilter* compaction_filter,
     const std::atomic<bool>* shutting_down,
     const SequenceNumber preserve_deletes_seqnum,
-    const std::atomic<bool>* manual_compaction_paused)
+    const std::atomic<bool>* manual_compaction_paused,
+    const std::shared_ptr<Logger> info_log)
     : CompactionIterator(
           input, cmp, merge_helper, last_sequence, snapshots,
           earliest_write_conflict_snapshot, snapshot_checker, env,
@@ -46,7 +49,7 @@ CompactionIterator::CompactionIterator(
           std::unique_ptr<CompactionProxy>(
               compaction ? new CompactionProxy(compaction) : nullptr),
           compaction_filter, shutting_down, preserve_deletes_seqnum,
-          manual_compaction_paused) {}
+          manual_compaction_paused, info_log) {}
 
 CompactionIterator::CompactionIterator(
     InternalIterator* input, const Comparator* cmp, MergeHelper* merge_helper,
@@ -59,7 +62,8 @@ CompactionIterator::CompactionIterator(
     const CompactionFilter* compaction_filter,
     const std::atomic<bool>* shutting_down,
     const SequenceNumber preserve_deletes_seqnum,
-    const std::atomic<bool>* manual_compaction_paused)
+    const std::atomic<bool>* manual_compaction_paused,
+    const std::shared_ptr<Logger> info_log)
     : input_(input),
       cmp_(cmp),
       merge_helper_(merge_helper),
@@ -78,7 +82,8 @@ CompactionIterator::CompactionIterator(
       current_user_key_sequence_(0),
       current_user_key_snapshot_(0),
       merge_out_iter_(merge_helper_),
-      current_key_committed_(false) {
+      current_key_committed_(false),
+      info_log_(info_log) {
   assert(compaction_filter_ == nullptr || compaction_ != nullptr);
   assert(snapshots_ != nullptr);
   bottommost_level_ =
@@ -142,6 +147,11 @@ void CompactionIterator::Next() {
       // MergeUntil stops when it encounters a corrupt key and does not
       // include them in the result, so we expect the keys here to be valid.
       assert(valid_key);
+      if (!valid_key) {
+        ROCKS_LOG_FATAL(info_log_, "Invalid key (%s) in compaction",
+                        key_.ToString(true).c_str());
+      }
+
       // Keep current_key_ in sync.
       current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
       key_ = current_key_.GetInternalKey();
@@ -338,7 +348,18 @@ void CompactionIterator::NextFromInput() {
       // not compact out.  We will keep this Put, but can drop it's data.
       // (See Optimization 3, below.)
       assert(ikey_.type == kTypeValue);
+      if (ikey_.type != kTypeValue) {
+        ROCKS_LOG_FATAL(info_log_,
+                        "Unexpected key type %d for compaction output",
+                        ikey_.type);
+      }
       assert(current_user_key_snapshot_ == last_snapshot);
+      if (current_user_key_snapshot_ != last_snapshot) {
+        ROCKS_LOG_FATAL(info_log_,
+                        "current_user_key_snapshot_ (%" PRIu64
+                        ") != last_snapshot (%" PRIu64 ")",
+                        current_user_key_snapshot_, last_snapshot);
+      }
 
       value_.clear();
       valid_ = true;
@@ -480,6 +501,12 @@ void CompactionIterator::NextFromInput() {
       // checking since there has already been a record returned for this key
       // in this snapshot.
       assert(last_sequence >= current_user_key_sequence_);
+      if (last_sequence < current_user_key_sequence_) {
+        ROCKS_LOG_FATAL(info_log_,
+                        "last_sequence (%" PRIu64
+                        ") < current_user_key_sequence_ (%" PRIu64 ")",
+                        last_sequence, current_user_key_sequence_);
+      }
 
       ++iter_stats_.num_record_drop_hidden;  // (A)
       input_->Next();
@@ -563,6 +590,10 @@ void CompactionIterator::NextFromInput() {
         // MergeUntil stops when it encounters a corrupt key and does not
         // include them in the result, so we expect the keys here to valid.
         assert(valid_key);
+        if (!valid_key) {
+          ROCKS_LOG_FATAL(info_log_, "Invalid key (%s) in compaction",
+                          key_.ToString(true).c_str());
+        }
         // Keep current_key_ in sync.
         current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
         key_ = current_key_.GetInternalKey();
@@ -608,29 +639,58 @@ void CompactionIterator::NextFromInput() {
 }
 
 void CompactionIterator::PrepareOutput() {
-  // Zeroing out the sequence number leads to better compression.
-  // If this is the bottommost level (no files in lower levels)
-  // and the earliest snapshot is larger than this seqno
-  // and the userkey differs from the last userkey in compaction
-  // then we can squash the seqno to zero.
-  //
-  // This is safe for TransactionDB write-conflict checking since transactions
-  // only care about sequence number larger than any active snapshots.
-  //
-  // Can we do the same for levels above bottom level as long as
-  // KeyNotExistsBeyondOutputLevel() return true?
-  if ((compaction_ != nullptr && !compaction_->allow_ingest_behind()) &&
-      ikeyNotNeededForIncrementalSnapshot() && bottommost_level_ && valid_ &&
-      IN_EARLIEST_SNAPSHOT(ikey_.sequence) && ikey_.type != kTypeMerge) {
-    assert(ikey_.type != kTypeDeletion && ikey_.type != kTypeSingleDeletion);
-    ikey_.sequence = 0;
-    current_key_.UpdateInternalKey(0, ikey_.type);
+  if (valid_) {
+    if (compaction_filter_ && ikey_.type == kTypeBlobIndex) {
+      const auto blob_decision = compaction_filter_->PrepareBlobOutput(
+          user_key(), value_, &compaction_filter_value_);
+
+      if (blob_decision == CompactionFilter::BlobDecision::kCorruption) {
+        status_ = Status::Corruption(
+            "Corrupted blob reference encountered during GC");
+        valid_ = false;
+      } else if (blob_decision == CompactionFilter::BlobDecision::kIOError) {
+        status_ = Status::IOError("Could not relocate blob during GC");
+        valid_ = false;
+      } else if (blob_decision ==
+                 CompactionFilter::BlobDecision::kChangeValue) {
+        value_ = compaction_filter_value_;
+      }
+    }
+
+    // Zeroing out the sequence number leads to better compression.
+    // If this is the bottommost level (no files in lower levels)
+    // and the earliest snapshot is larger than this seqno
+    // and the userkey differs from the last userkey in compaction
+    // then we can squash the seqno to zero.
+    //
+    // This is safe for TransactionDB write-conflict checking since transactions
+    // only care about sequence number larger than any active snapshots.
+    //
+    // Can we do the same for levels above bottom level as long as
+    // KeyNotExistsBeyondOutputLevel() return true?
+    if (valid_ && compaction_ != nullptr &&
+        !compaction_->allow_ingest_behind() &&
+        ikeyNotNeededForIncrementalSnapshot() && bottommost_level_ &&
+        IN_EARLIEST_SNAPSHOT(ikey_.sequence) && ikey_.type != kTypeMerge) {
+      assert(ikey_.type != kTypeDeletion && ikey_.type != kTypeSingleDeletion);
+      if (ikey_.type == kTypeDeletion || ikey_.type == kTypeSingleDeletion) {
+        ROCKS_LOG_FATAL(info_log_,
+                        "Unexpected key type %d for seq-zero optimization",
+                        ikey_.type);
+      }
+      ikey_.sequence = 0;
+      current_key_.UpdateInternalKey(0, ikey_.type);
+    }
   }
 }
 
 inline SequenceNumber CompactionIterator::findEarliestVisibleSnapshot(
     SequenceNumber in, SequenceNumber* prev_snapshot) {
   assert(snapshots_->size());
+  if (snapshots_->size() == 0) {
+    ROCKS_LOG_FATAL(info_log_,
+                    "No snapshot left in findEarliestVisibleSnapshot");
+  }
   auto snapshots_iter = std::lower_bound(
       snapshots_->begin(), snapshots_->end(), in);
   if (snapshots_iter == snapshots_->begin()) {
@@ -638,6 +698,10 @@ inline SequenceNumber CompactionIterator::findEarliestVisibleSnapshot(
   } else {
     *prev_snapshot = *std::prev(snapshots_iter);
     assert(*prev_snapshot < in);
+    if (*prev_snapshot >= in) {
+      ROCKS_LOG_FATAL(info_log_,
+                      "*prev_snapshot >= in in findEarliestVisibleSnapshot");
+    }
   }
   if (snapshot_checker_ == nullptr) {
     return snapshots_iter != snapshots_->end()
@@ -647,6 +711,9 @@ inline SequenceNumber CompactionIterator::findEarliestVisibleSnapshot(
   for (; snapshots_iter != snapshots_->end(); ++snapshots_iter) {
     auto cur = *snapshots_iter;
     assert(in <= cur);
+    if (in > cur) {
+      ROCKS_LOG_FATAL(info_log_, "in > cur in findEarliestVisibleSnapshot");
+    }
     // Skip if cur is in released_snapshots.
     if (has_released_snapshot && released_snapshots_.count(cur) > 0) {
       continue;
@@ -671,9 +738,14 @@ inline bool CompactionIterator::ikeyNotNeededForIncrementalSnapshot() {
 
 bool CompactionIterator::IsInEarliestSnapshot(SequenceNumber sequence) {
   assert(snapshot_checker_ != nullptr);
-  assert(earliest_snapshot_ == kMaxSequenceNumber ||
-         (earliest_snapshot_iter_ != snapshots_->end() &&
-          *earliest_snapshot_iter_ == earliest_snapshot_));
+  bool pre_condition = (earliest_snapshot_ == kMaxSequenceNumber ||
+                        (earliest_snapshot_iter_ != snapshots_->end() &&
+                         *earliest_snapshot_iter_ == earliest_snapshot_));
+  assert(pre_condition);
+  if (!pre_condition) {
+    ROCKS_LOG_FATAL(info_log_,
+                    "Pre-Condition is not hold in IsInEarliestSnapshot");
+  }
   auto in_snapshot =
       snapshot_checker_->CheckInSnapshot(sequence, earliest_snapshot_);
   while (UNLIKELY(in_snapshot == SnapshotCheckerResult::kSnapshotReleased)) {
@@ -692,6 +764,10 @@ bool CompactionIterator::IsInEarliestSnapshot(SequenceNumber sequence) {
         snapshot_checker_->CheckInSnapshot(sequence, earliest_snapshot_);
   }
   assert(in_snapshot != SnapshotCheckerResult::kSnapshotReleased);
+  if (in_snapshot == SnapshotCheckerResult::kSnapshotReleased) {
+    ROCKS_LOG_FATAL(info_log_,
+                    "Unexpected released snapshot in IsInEarliestSnapshot");
+  }
   return in_snapshot == SnapshotCheckerResult::kInSnapshot;
 }
 

db/compaction/compaction_iterator.h

@@ -70,7 +70,8 @@ class CompactionIterator {
       const CompactionFilter* compaction_filter = nullptr,
       const std::atomic<bool>* shutting_down = nullptr,
       const SequenceNumber preserve_deletes_seqnum = 0,
-      const std::atomic<bool>* manual_compaction_paused = nullptr);
+      const std::atomic<bool>* manual_compaction_paused = nullptr,
+      const std::shared_ptr<Logger> info_log = nullptr);
 
   // Constructor with custom CompactionProxy, used for tests.
   CompactionIterator(
@@ -84,7 +85,8 @@ class CompactionIterator {
       const CompactionFilter* compaction_filter = nullptr,
       const std::atomic<bool>* shutting_down = nullptr,
       const SequenceNumber preserve_deletes_seqnum = 0,
-      const std::atomic<bool>* manual_compaction_paused = nullptr);
+      const std::atomic<bool>* manual_compaction_paused = nullptr,
+      const std::shared_ptr<Logger> info_log = nullptr);
 
   ~CompactionIterator();
 
@@ -222,6 +224,7 @@ class CompactionIterator {
   // Used to avoid purging uncommitted values. The application can specify
   // uncommitted values by providing a SnapshotChecker object.
   bool current_key_committed_;
+  std::shared_ptr<Logger> info_log_;
 
   bool IsShuttingDown() {
     // This is a best-effort facility, so memory_order_relaxed is sufficient.

db/compaction/compaction_iterator_test.cc

@@ -184,7 +184,7 @@ class TestSnapshotChecker : public SnapshotChecker {
  public:
   explicit TestSnapshotChecker(
       SequenceNumber last_committed_sequence,
-      const std::unordered_map<SequenceNumber, SequenceNumber>& snapshots = {})
+      const std::unordered_map<SequenceNumber, SequenceNumber>& snapshots = {{}})
       : last_committed_sequence_(last_committed_sequence),
         snapshots_(snapshots) {}
 

db/compaction/compaction_job.cc

@@ -301,7 +301,7 @@ void CompactionJob::AggregateStatistics() {
 
 CompactionJob::CompactionJob(
     int job_id, Compaction* compaction, const ImmutableDBOptions& db_options,
-    const EnvOptions env_options, VersionSet* versions,
+    const FileOptions& file_options, VersionSet* versions,
     const std::atomic<bool>* shutting_down,
     const SequenceNumber preserve_deletes_seqnum, LogBuffer* log_buffer,
     Directory* db_directory, Directory* output_directory, Statistics* stats,
@@ -318,10 +318,11 @@ CompactionJob::CompactionJob(
       compaction_stats_(compaction->compaction_reason(), 1),
       dbname_(dbname),
       db_options_(db_options),
-      env_options_(env_options),
+      file_options_(file_options),
       env_(db_options.env),
-      env_options_for_read_(
-          env_->OptimizeForCompactionTableRead(env_options, db_options_)),
+      fs_(db_options.fs.get()),
+      file_options_for_read_(
+          fs_->OptimizeForCompactionTableRead(file_options, db_options_)),
       versions_(versions),
       shutting_down_(shutting_down),
       manual_compaction_paused_(manual_compaction_paused),
@@ -647,7 +648,7 @@ Status CompactionJob::Run() {
         // we will regard this verification as user reads since the goal is
         // to cache it here for further user reads
         InternalIterator* iter = cfd->table_cache()->NewIterator(
-            ReadOptions(), env_options_, cfd->internal_comparator(),
+            ReadOptions(), file_options_, cfd->internal_comparator(),
             *files_meta[file_idx], /*range_del_agg=*/nullptr, prefix_extractor,
             /*table_reader_ptr=*/nullptr,
             cfd->internal_stats()->GetFileReadHist(
@@ -836,7 +837,7 @@ 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> input(versions_->MakeInputIterator(
-      sub_compact->compaction, &range_del_agg, env_options_for_read_));
+      sub_compact->compaction, &range_del_agg, file_options_for_read_));
 
   AutoThreadOperationStageUpdater stage_updater(
       ThreadStatus::STAGE_COMPACTION_PROCESS_KV);
@@ -891,7 +892,8 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
       &existing_snapshots_, earliest_write_conflict_snapshot_,
       snapshot_checker_, env_, ShouldReportDetailedTime(env_, stats_), false,
       &range_del_agg, sub_compact->compaction, compaction_filter,
-      shutting_down_, preserve_deletes_seqnum_, manual_compaction_paused_));
+      shutting_down_, preserve_deletes_seqnum_, manual_compaction_paused_,
+      db_options_.info_log));
   auto c_iter = sub_compact->c_iter.get();
   c_iter->SeekToFirst();
   if (c_iter->Valid() && sub_compact->compaction->output_level() != 0) {
@@ -933,8 +935,9 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
     assert(sub_compact->current_output() != nullptr);
     sub_compact->builder->Add(key, value);
     sub_compact->current_output_file_size = sub_compact->builder->FileSize();
+    const ParsedInternalKey& ikey = c_iter->ikey();
     sub_compact->current_output()->meta.UpdateBoundaries(
-        key, c_iter->ikey().sequence);
+        key, value, ikey.sequence, ikey.type);
     sub_compact->num_output_records++;
 
     // Close output file if it is big enough. Two possibilities determine it's
@@ -1349,17 +1352,20 @@ Status CompactionJob::FinishCompactionOutputFile(
   }
   std::string fname;
   FileDescriptor output_fd;
+  uint64_t oldest_blob_file_number = kInvalidBlobFileNumber;
   if (meta != nullptr) {
     fname =
         TableFileName(sub_compact->compaction->immutable_cf_options()->cf_paths,
                       meta->fd.GetNumber(), meta->fd.GetPathId());
     output_fd = meta->fd;
+    oldest_blob_file_number = meta->oldest_blob_file_number;
   } else {
     fname = "(nil)";
   }
   EventHelpers::LogAndNotifyTableFileCreationFinished(
       event_logger_, cfd->ioptions()->listeners, dbname_, cfd->GetName(), fname,
-      job_id_, output_fd, tp, TableFileCreationReason::kCompaction, s);
+      job_id_, output_fd, oldest_blob_file_number, tp,
+      TableFileCreationReason::kCompaction, s);
 
 #ifndef ROCKSDB_LITE
   // Report new file to SstFileManagerImpl
@@ -1452,13 +1458,13 @@ Status CompactionJob::OpenCompactionOutputFile(
       TableFileCreationReason::kCompaction);
 #endif  // !ROCKSDB_LITE
   // Make the output file
-  std::unique_ptr<WritableFile> writable_file;
+  std::unique_ptr<FSWritableFile> writable_file;
 #ifndef NDEBUG
-  bool syncpoint_arg = env_options_.use_direct_writes;
+  bool syncpoint_arg = file_options_.use_direct_writes;
   TEST_SYNC_POINT_CALLBACK("CompactionJob::OpenCompactionOutputFile",
                            &syncpoint_arg);
 #endif
-  Status s = NewWritableFile(env_, fname, &writable_file, env_options_);
+  Status s = NewWritableFile(fs_, fname, &writable_file, file_options_);
   if (!s.ok()) {
     ROCKS_LOG_ERROR(
         db_options_.info_log,
@@ -1469,33 +1475,12 @@ Status CompactionJob::OpenCompactionOutputFile(
     LogFlush(db_options_.info_log);
     EventHelpers::LogAndNotifyTableFileCreationFinished(
         event_logger_, cfd->ioptions()->listeners, dbname_, cfd->GetName(),
-        fname, job_id_, FileDescriptor(), TableProperties(),
-        TableFileCreationReason::kCompaction, s);
+        fname, job_id_, FileDescriptor(), kInvalidBlobFileNumber,
+        TableProperties(), TableFileCreationReason::kCompaction, s);
     return s;
   }
 
-  SubcompactionState::Output out;
-  out.meta.fd =
-      FileDescriptor(file_number, sub_compact->compaction->output_path_id(), 0);
-  out.finished = false;
-
-  sub_compact->outputs.push_back(out);
-  writable_file->SetIOPriority(Env::IO_LOW);
-  writable_file->SetWriteLifeTimeHint(write_hint_);
-  writable_file->SetPreallocationBlockSize(static_cast<size_t>(
-      sub_compact->compaction->OutputFilePreallocationSize()));
-  const auto& listeners =
-      sub_compact->compaction->immutable_cf_options()->listeners;
-  sub_compact->outfile.reset(
-      new WritableFileWriter(std::move(writable_file), fname, env_options_,
-                             env_, db_options_.statistics.get(), listeners));
-
-  // If the Column family flag is to only optimize filters for hits,
-  // we can skip creating filters if this is the bottommost_level where
-  // data is going to be found
-  bool skip_filters =
-      cfd->ioptions()->optimize_filters_for_hits && bottommost_level_;
-
+  // Try to figure out the output file's oldest ancester time.
   int64_t temp_current_time = 0;
   auto get_time_status = env_->GetCurrentTime(&temp_current_time);
   // Safe to proceed even if GetCurrentTime fails. So, log and proceed.
@@ -1505,13 +1490,39 @@ Status CompactionJob::OpenCompactionOutputFile(
                    get_time_status.ToString().c_str());
   }
   uint64_t current_time = static_cast<uint64_t>(temp_current_time);
-
-  uint64_t latest_key_time =
-      sub_compact->compaction->MaxInputFileCreationTime();
-  if (latest_key_time == 0) {
-    latest_key_time = current_time;
+  uint64_t oldest_ancester_time =
+      sub_compact->compaction->MinInputFileOldestAncesterTime();
+  if (oldest_ancester_time == port::kMaxUint64) {
+    oldest_ancester_time = current_time;
   }
 
+  // Initialize a SubcompactionState::Output and add it to sub_compact->outputs
+  {
+    SubcompactionState::Output out;
+    out.meta.fd = FileDescriptor(file_number,
+                                 sub_compact->compaction->output_path_id(), 0);
+    out.meta.oldest_ancester_time = oldest_ancester_time;
+    out.meta.file_creation_time = current_time;
+    out.finished = false;
+    sub_compact->outputs.push_back(out);
+  }
+
+  writable_file->SetIOPriority(Env::IOPriority::IO_LOW);
+  writable_file->SetWriteLifeTimeHint(write_hint_);
+  writable_file->SetPreallocationBlockSize(static_cast<size_t>(
+      sub_compact->compaction->OutputFilePreallocationSize()));
+  const auto& listeners =
+      sub_compact->compaction->immutable_cf_options()->listeners;
+  sub_compact->outfile.reset(
+      new WritableFileWriter(std::move(writable_file), fname, file_options_,
+                             env_, db_options_.statistics.get(), listeners));
+
+  // If the Column family flag is to only optimize filters for hits,
+  // we can skip creating filters if this is the bottommost_level where
+  // data is going to be found
+  bool skip_filters =
+      cfd->ioptions()->optimize_filters_for_hits && bottommost_level_;
+
   sub_compact->builder.reset(NewTableBuilder(
       *cfd->ioptions(), *(sub_compact->compaction->mutable_cf_options()),
       cfd->internal_comparator(), cfd->int_tbl_prop_collector_factories(),
@@ -1519,9 +1530,9 @@ Status CompactionJob::OpenCompactionOutputFile(
       sub_compact->compaction->output_compression(),
       0 /*sample_for_compression */,
       sub_compact->compaction->output_compression_opts(),
-      sub_compact->compaction->output_level(), skip_filters, latest_key_time,
-      0 /* oldest_key_time */, sub_compact->compaction->max_output_file_size(),
-      current_time));
+      sub_compact->compaction->output_level(), skip_filters,
+      oldest_ancester_time, 0 /* oldest_key_time */,
+      sub_compact->compaction->max_output_file_size(), current_time));
   LogFlush(db_options_.info_log);
   return s;
 }

db/compaction/compaction_job.h

@@ -64,7 +64,7 @@ class CompactionJob {
  public:
   CompactionJob(int job_id, Compaction* compaction,
                 const ImmutableDBOptions& db_options,
-                const EnvOptions env_options, VersionSet* versions,
+                const FileOptions& file_options, VersionSet* versions,
                 const std::atomic<bool>* shutting_down,
                 const SequenceNumber preserve_deletes_seqnum,
                 LogBuffer* log_buffer, Directory* db_directory,
@@ -150,11 +150,12 @@ class CompactionJob {
   // DBImpl state
   const std::string& dbname_;
   const ImmutableDBOptions& db_options_;
-  const EnvOptions env_options_;
+  const FileOptions file_options_;
 
   Env* env_;
+  FileSystem* fs_;
   // env_option optimized for compaction table reads
-  EnvOptions env_options_for_read_;
+  FileOptions file_options_for_read_;
   VersionSet* versions_;
   const std::atomic<bool>* shutting_down_;
   const std::atomic<bool>* manual_compaction_paused_;

db/compaction/compaction_job_test.cc

@@ -12,6 +12,7 @@
 #include <string>
 #include <tuple>
 
+#include "db/blob_index.h"
 #include "db/column_family.h"
 #include "db/compaction/compaction_job.h"
 #include "db/db_impl/db_impl.h"
@@ -71,6 +72,7 @@ class CompactionJobTest : public testing::Test {
  public:
   CompactionJobTest()
       : env_(Env::Default()),
+        fs_(std::make_shared<LegacyFileSystemWrapper>(env_)),
         dbname_(test::PerThreadDBPath("compaction_job_test")),
         db_options_(),
         mutable_cf_options_(cf_options_),
@@ -85,6 +87,8 @@ class CompactionJobTest : public testing::Test {
         mock_table_factory_(new mock::MockTableFactory()),
         error_handler_(nullptr, db_options_, &mutex_) {
     EXPECT_OK(env_->CreateDirIfMissing(dbname_));
+    db_options_.env = env_;
+    db_options_.fs = fs_;
     db_options_.db_paths.emplace_back(dbname_,
                                       std::numeric_limits<uint64_t>::max());
   }
@@ -97,11 +101,34 @@ class CompactionJobTest : public testing::Test {
     return TableFileName(db_paths, meta.fd.GetNumber(), meta.fd.GetPathId());
   }
 
-  std::string KeyStr(const std::string& user_key, const SequenceNumber seq_num,
-      const ValueType t) {
+  static std::string KeyStr(const std::string& user_key,
+                            const SequenceNumber seq_num, const ValueType t) {
     return InternalKey(user_key, seq_num, t).Encode().ToString();
   }
 
+  static std::string BlobStr(uint64_t blob_file_number, uint64_t offset,
+                             uint64_t size) {
+    std::string blob_index;
+    BlobIndex::EncodeBlob(&blob_index, blob_file_number, offset, size,
+                          kNoCompression);
+    return blob_index;
+  }
+
+  static std::string BlobStrTTL(uint64_t blob_file_number, uint64_t offset,
+                                uint64_t size, uint64_t expiration) {
+    std::string blob_index;
+    BlobIndex::EncodeBlobTTL(&blob_index, expiration, blob_file_number, offset,
+                             size, kNoCompression);
+    return blob_index;
+  }
+
+  static std::string BlobStrInlinedTTL(const Slice& value,
+                                       uint64_t expiration) {
+    std::string blob_index;
+    BlobIndex::EncodeInlinedTTL(&blob_index, expiration, value);
+    return blob_index;
+  }
+
   void AddMockFile(const stl_wrappers::KVMap& contents, int level = 0) {
     assert(contents.size() > 0);
 
@@ -110,12 +137,13 @@ class CompactionJobTest : public testing::Test {
     InternalKey smallest_key, largest_key;
     SequenceNumber smallest_seqno = kMaxSequenceNumber;
     SequenceNumber largest_seqno = 0;
+    uint64_t oldest_blob_file_number = kInvalidBlobFileNumber;
     for (auto kv : contents) {
       ParsedInternalKey key;
       std::string skey;
       std::string value;
       std::tie(skey, value) = kv;
-      ParseInternalKey(skey, &key);
+      bool parsed = ParseInternalKey(skey, &key);
 
       smallest_seqno = std::min(smallest_seqno, key.sequence);
       largest_seqno = std::max(largest_seqno, key.sequence);
@@ -132,6 +160,24 @@ class CompactionJobTest : public testing::Test {
       }
 
       first_key = false;
+
+      if (parsed && key.type == kTypeBlobIndex) {
+        BlobIndex blob_index;
+        const Status s = blob_index.DecodeFrom(value);
+        if (!s.ok()) {
+          continue;
+        }
+
+        if (blob_index.IsInlined() || blob_index.HasTTL() ||
+            blob_index.file_number() == kInvalidBlobFileNumber) {
+          continue;
+        }
+
+        if (oldest_blob_file_number == kInvalidBlobFileNumber ||
+            oldest_blob_file_number > blob_index.file_number()) {
+          oldest_blob_file_number = blob_index.file_number();
+        }
+      }
     }
 
     uint64_t file_number = versions_->NewFileNumber();
@@ -140,7 +186,8 @@ class CompactionJobTest : public testing::Test {
 
     VersionEdit edit;
     edit.AddFile(level, file_number, 0, 10, smallest_key, largest_key,
-        smallest_seqno, largest_seqno, false);
+                 smallest_seqno, largest_seqno, false, oldest_blob_file_number,
+                 kUnknownOldestAncesterTime, kUnknownFileCreationTime);
 
     mutex_.Lock();
     versions_->LogAndApply(versions_->GetColumnFamilySet()->GetDefault(),
@@ -223,8 +270,8 @@ class CompactionJobTest : public testing::Test {
     Status s = env_->NewWritableFile(
         manifest, &file, env_->OptimizeForManifestWrite(env_options_));
     ASSERT_OK(s);
-    std::unique_ptr<WritableFileWriter> file_writer(
-        new WritableFileWriter(std::move(file), manifest, env_options_));
+    std::unique_ptr<WritableFileWriter> file_writer(new WritableFileWriter(
+        NewLegacyWritableFileWrapper(std::move(file)), manifest, env_options_));
     {
       log::Writer log(std::move(file_writer), 0, false);
       std::string record;
@@ -250,7 +297,8 @@ class CompactionJobTest : public testing::Test {
       const stl_wrappers::KVMap& expected_results,
       const std::vector<SequenceNumber>& snapshots = {},
       SequenceNumber earliest_write_conflict_snapshot = kMaxSequenceNumber,
-      int output_level = 1, bool verify = true) {
+      int output_level = 1, bool verify = true,
+      uint64_t expected_oldest_blob_file_number = kInvalidBlobFileNumber) {
     auto cfd = versions_->GetColumnFamilySet()->GetDefault();
 
     size_t num_input_files = 0;
@@ -296,20 +344,26 @@ class CompactionJobTest : public testing::Test {
     mutex_.Unlock();
 
     if (verify) {
-      if (expected_results.size() == 0) {
-        ASSERT_GE(compaction_job_stats_.elapsed_micros, 0U);
-        ASSERT_EQ(compaction_job_stats_.num_input_files, num_input_files);
+      ASSERT_GE(compaction_job_stats_.elapsed_micros, 0U);
+      ASSERT_EQ(compaction_job_stats_.num_input_files, num_input_files);
+
+      if (expected_results.empty()) {
         ASSERT_EQ(compaction_job_stats_.num_output_files, 0U);
       } else {
-        ASSERT_GE(compaction_job_stats_.elapsed_micros, 0U);
-        ASSERT_EQ(compaction_job_stats_.num_input_files, num_input_files);
         ASSERT_EQ(compaction_job_stats_.num_output_files, 1U);
         mock_table_factory_->AssertLatestFile(expected_results);
+
+        auto output_files =
+            cfd->current()->storage_info()->LevelFiles(output_level);
+        ASSERT_EQ(output_files.size(), 1);
+        ASSERT_EQ(output_files[0]->oldest_blob_file_number,
+                  expected_oldest_blob_file_number);
       }
     }
   }
 
   Env* env_;
+  std::shared_ptr<FileSystem> fs_;
   std::string dbname_;
   EnvOptions env_options_;
   ImmutableDBOptions db_options_;
@@ -960,6 +1014,54 @@ TEST_F(CompactionJobTest, CorruptionAfterDeletion) {
   RunCompaction({files}, expected_results);
 }
 
+TEST_F(CompactionJobTest, OldestBlobFileNumber) {
+  NewDB();
+
+  // Note: blob1 is inlined TTL, so it will not be considered for the purposes
+  // of identifying the oldest referenced blob file. Similarly, blob6 will be
+  // ignored because it has TTL and hence refers to a TTL blob file.
+  const stl_wrappers::KVMap::value_type blob1(
+      KeyStr("a", 1U, kTypeBlobIndex), BlobStrInlinedTTL("foo", 1234567890ULL));
+  const stl_wrappers::KVMap::value_type blob2(KeyStr("b", 2U, kTypeBlobIndex),
+                                              BlobStr(59, 123456, 999));
+  const stl_wrappers::KVMap::value_type blob3(KeyStr("c", 3U, kTypeBlobIndex),
+                                              BlobStr(138, 1000, 1 << 8));
+  auto file1 = mock::MakeMockFile({blob1, blob2, blob3});
+  AddMockFile(file1);
+
+  const stl_wrappers::KVMap::value_type blob4(KeyStr("d", 4U, kTypeBlobIndex),
+                                              BlobStr(199, 3 << 10, 1 << 20));
+  const stl_wrappers::KVMap::value_type blob5(KeyStr("e", 5U, kTypeBlobIndex),
+                                              BlobStr(19, 6789, 333));
+  const stl_wrappers::KVMap::value_type blob6(
+      KeyStr("f", 6U, kTypeBlobIndex),
+      BlobStrTTL(5, 2048, 1 << 7, 1234567890ULL));
+  auto file2 = mock::MakeMockFile({blob4, blob5, blob6});
+  AddMockFile(file2);
+
+  const stl_wrappers::KVMap::value_type expected_blob1(
+      KeyStr("a", 0U, kTypeBlobIndex), blob1.second);
+  const stl_wrappers::KVMap::value_type expected_blob2(
+      KeyStr("b", 0U, kTypeBlobIndex), blob2.second);
+  const stl_wrappers::KVMap::value_type expected_blob3(
+      KeyStr("c", 0U, kTypeBlobIndex), blob3.second);
+  const stl_wrappers::KVMap::value_type expected_blob4(
+      KeyStr("d", 0U, kTypeBlobIndex), blob4.second);
+  const stl_wrappers::KVMap::value_type expected_blob5(
+      KeyStr("e", 0U, kTypeBlobIndex), blob5.second);
+  const stl_wrappers::KVMap::value_type expected_blob6(
+      KeyStr("f", 0U, kTypeBlobIndex), blob6.second);
+  auto expected_results =
+      mock::MakeMockFile({expected_blob1, expected_blob2, expected_blob3,
+                          expected_blob4, expected_blob5, expected_blob6});
+
+  SetLastSequence(6U);
+  auto files = cfd_->current()->storage_info()->LevelFiles(0);
+  RunCompaction({files}, expected_results, std::vector<SequenceNumber>(),
+                kMaxSequenceNumber, /* output_level */ 1, /* verify */ true,
+                /* expected_oldest_blob_file_number */ 19);
+}
+
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

db/compaction/compaction_picker.cc

@@ -39,20 +39,41 @@ bool FindIntraL0Compaction(const std::vector<FileMetaData*>& level_files,
                            size_t min_files_to_compact,
                            uint64_t max_compact_bytes_per_del_file,
                            uint64_t max_compaction_bytes,
-                           CompactionInputFiles* comp_inputs) {
-  size_t compact_bytes = static_cast<size_t>(level_files[0]->fd.file_size);
-  uint64_t compensated_compact_bytes = level_files[0]->compensated_file_size;
+                           CompactionInputFiles* comp_inputs,
+                           SequenceNumber earliest_mem_seqno) {
+  // Do not pick ingested file when there is at least one memtable not flushed
+  // which of seqno is overlap with the sst.
+  TEST_SYNC_POINT("FindIntraL0Compaction");
+  size_t start = 0;
+  for (; start < level_files.size(); start++) {
+    if (level_files[start]->being_compacted) {
+      return false;
+    }
+    // If there is no data in memtable, the earliest sequence number would the
+    // largest sequence number in last memtable.
+    // Because all files are sorted in descending order by largest_seqno, so we
+    // only need to check the first one.
+    if (level_files[start]->fd.largest_seqno <= earliest_mem_seqno) {
+      break;
+    }
+  }
+  if (start >= level_files.size()) {
+    return false;
+  }
+  size_t compact_bytes = static_cast<size_t>(level_files[start]->fd.file_size);
+  uint64_t compensated_compact_bytes =
+      level_files[start]->compensated_file_size;
   size_t compact_bytes_per_del_file = port::kMaxSizet;
-  // Compaction range will be [0, span_len).
-  size_t span_len;
+  // Compaction range will be [start, limit).
+  size_t limit;
   // Pull in files until the amount of compaction work per deleted file begins
   // increasing or maximum total compaction size is reached.
   size_t new_compact_bytes_per_del_file = 0;
-  for (span_len = 1; span_len < level_files.size(); ++span_len) {
-    compact_bytes += static_cast<size_t>(level_files[span_len]->fd.file_size);
-    compensated_compact_bytes += level_files[span_len]->compensated_file_size;
-    new_compact_bytes_per_del_file = compact_bytes / span_len;
-    if (level_files[span_len]->being_compacted ||
+  for (limit = start + 1; limit < level_files.size(); ++limit) {
+    compact_bytes += static_cast<size_t>(level_files[limit]->fd.file_size);
+    compensated_compact_bytes += level_files[limit]->compensated_file_size;
+    new_compact_bytes_per_del_file = compact_bytes / (limit - start);
+    if (level_files[limit]->being_compacted ||
         new_compact_bytes_per_del_file > compact_bytes_per_del_file ||
         compensated_compact_bytes > max_compaction_bytes) {
       break;
@@ -60,11 +81,11 @@ bool FindIntraL0Compaction(const std::vector<FileMetaData*>& level_files,
     compact_bytes_per_del_file = new_compact_bytes_per_del_file;
   }
 
-  if (span_len >= min_files_to_compact &&
+  if ((limit - start) >= min_files_to_compact &&
       compact_bytes_per_del_file < max_compact_bytes_per_del_file) {
     assert(comp_inputs != nullptr);
     comp_inputs->level = 0;
-    for (size_t i = 0; i < span_len; ++i) {
+    for (size_t i = start; i < limit; ++i) {
       comp_inputs->files.push_back(level_files[i]);
     }
     return true;

db/compaction/compaction_picker.h

@@ -54,10 +54,10 @@ class CompactionPicker {
   // Returns nullptr if there is no compaction to be done.
   // Otherwise returns a pointer to a heap-allocated object that
   // describes the compaction.  Caller should delete the result.
-  virtual Compaction* PickCompaction(const std::string& cf_name,
-                                     const MutableCFOptions& mutable_cf_options,
-                                     VersionStorageInfo* vstorage,
-                                     LogBuffer* log_buffer) = 0;
+  virtual Compaction* PickCompaction(
+      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
+      VersionStorageInfo* vstorage, LogBuffer* log_buffer,
+      SequenceNumber earliest_memtable_seqno = kMaxSequenceNumber) = 0;
 
   // Return a compaction object for compacting the range [begin,end] in
   // the specified level.  Returns nullptr if there is nothing in that
@@ -247,10 +247,11 @@ class NullCompactionPicker : public CompactionPicker {
   virtual ~NullCompactionPicker() {}
 
   // Always return "nullptr"
-  Compaction* PickCompaction(const std::string& /*cf_name*/,
-                             const MutableCFOptions& /*mutable_cf_options*/,
-                             VersionStorageInfo* /*vstorage*/,
-                             LogBuffer* /*log_buffer*/) override {
+  Compaction* PickCompaction(
+      const std::string& /*cf_name*/,
+      const MutableCFOptions& /*mutable_cf_options*/,
+      VersionStorageInfo* /*vstorage*/, LogBuffer* /* log_buffer */,
+      SequenceNumber /* earliest_memtable_seqno */) override {
     return nullptr;
   }
 
@@ -292,11 +293,11 @@ class NullCompactionPicker : public CompactionPicker {
 //                                        files. Cannot be nullptr.
 //
 // @return                                true iff compaction was found.
-bool FindIntraL0Compaction(const std::vector<FileMetaData*>& level_files,
-                           size_t min_files_to_compact,
-                           uint64_t max_compact_bytes_per_del_file,
-                           uint64_t max_compaction_bytes,
-                           CompactionInputFiles* comp_inputs);
+bool FindIntraL0Compaction(
+    const std::vector<FileMetaData*>& level_files, size_t min_files_to_compact,
+    uint64_t max_compact_bytes_per_del_file, uint64_t max_compaction_bytes,
+    CompactionInputFiles* comp_inputs,
+    SequenceNumber earliest_mem_seqno = kMaxSequenceNumber);
 
 CompressionType GetCompressionType(const ImmutableCFOptions& ioptions,
                                    const VersionStorageInfo* vstorage,

db/compaction/compaction_picker_fifo.cc

@@ -202,7 +202,8 @@ Compaction* FIFOCompactionPicker::PickSizeCompaction(
 
 Compaction* FIFOCompactionPicker::PickCompaction(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
-    VersionStorageInfo* vstorage, LogBuffer* log_buffer) {
+    VersionStorageInfo* vstorage, LogBuffer* log_buffer,
+    SequenceNumber /*earliest_memtable_seqno*/) {
   assert(vstorage->num_levels() == 1);
 
   Compaction* c = nullptr;

db/compaction/compaction_picker_fifo.h

@@ -19,10 +19,10 @@ class FIFOCompactionPicker : public CompactionPicker {
                        const InternalKeyComparator* icmp)
       : CompactionPicker(ioptions, icmp) {}
 
-  virtual Compaction* PickCompaction(const std::string& cf_name,
-                                     const MutableCFOptions& mutable_cf_options,
-                                     VersionStorageInfo* version,
-                                     LogBuffer* log_buffer) override;
+  virtual Compaction* PickCompaction(
+      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
+      VersionStorageInfo* version, LogBuffer* log_buffer,
+      SequenceNumber earliest_memtable_seqno = kMaxSequenceNumber) override;
 
   virtual Compaction* CompactRange(
       const std::string& cf_name, const MutableCFOptions& mutable_cf_options,

db/compaction/compaction_picker_level.cc

@@ -45,12 +45,14 @@ class LevelCompactionBuilder {
  public:
   LevelCompactionBuilder(const std::string& cf_name,
                          VersionStorageInfo* vstorage,
+                         SequenceNumber earliest_mem_seqno,
                          CompactionPicker* compaction_picker,
                          LogBuffer* log_buffer,
                          const MutableCFOptions& mutable_cf_options,
                          const ImmutableCFOptions& ioptions)
       : cf_name_(cf_name),
         vstorage_(vstorage),
+        earliest_mem_seqno_(earliest_mem_seqno),
         compaction_picker_(compaction_picker),
         log_buffer_(log_buffer),
         mutable_cf_options_(mutable_cf_options),
@@ -97,6 +99,7 @@ class LevelCompactionBuilder {
 
   const std::string& cf_name_;
   VersionStorageInfo* vstorage_;
+  SequenceNumber earliest_mem_seqno_;
   CompactionPicker* compaction_picker_;
   LogBuffer* log_buffer_;
   int start_level_ = -1;
@@ -537,17 +540,19 @@ bool LevelCompactionBuilder::PickIntraL0Compaction() {
     // resort to L0->L0 compaction yet.
     return false;
   }
-  return FindIntraL0Compaction(
-      level_files, kMinFilesForIntraL0Compaction, port::kMaxUint64,
-      mutable_cf_options_.max_compaction_bytes, &start_level_inputs_);
+  return FindIntraL0Compaction(level_files, kMinFilesForIntraL0Compaction,
+                               port::kMaxUint64,
+                               mutable_cf_options_.max_compaction_bytes,
+                               &start_level_inputs_, earliest_mem_seqno_);
 }
 }  // namespace
 
 Compaction* LevelCompactionPicker::PickCompaction(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
-    VersionStorageInfo* vstorage, LogBuffer* log_buffer) {
-  LevelCompactionBuilder builder(cf_name, vstorage, this, log_buffer,
-                                 mutable_cf_options, ioptions_);
+    VersionStorageInfo* vstorage, LogBuffer* log_buffer,
+    SequenceNumber earliest_mem_seqno) {
+  LevelCompactionBuilder builder(cf_name, vstorage, earliest_mem_seqno, this,
+                                 log_buffer, mutable_cf_options, ioptions_);
   return builder.PickCompaction();
 }
 }  // namespace rocksdb

db/compaction/compaction_picker_level.h

@@ -20,10 +20,10 @@ class LevelCompactionPicker : public CompactionPicker {
   LevelCompactionPicker(const ImmutableCFOptions& ioptions,
                         const InternalKeyComparator* icmp)
       : CompactionPicker(ioptions, icmp) {}
-  virtual Compaction* PickCompaction(const std::string& cf_name,
-                                     const MutableCFOptions& mutable_cf_options,
-                                     VersionStorageInfo* vstorage,
-                                     LogBuffer* log_buffer) override;
+  virtual Compaction* PickCompaction(
+      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
+      VersionStorageInfo* vstorage, LogBuffer* log_buffer,
+      SequenceNumber earliest_memtable_seqno = kMaxSequenceNumber) override;
 
   virtual bool NeedsCompaction(
       const VersionStorageInfo* vstorage) const override;

db/compaction/compaction_picker_test.cc

@@ -57,6 +57,8 @@ class CompactionPickerTest : public testing::Test {
         log_buffer_(InfoLogLevel::INFO_LEVEL, &logger_),
         file_num_(1),
         vstorage_(nullptr) {
+    mutable_cf_options_.ttl = 0;
+    mutable_cf_options_.periodic_compaction_seconds = 0;
     // ioptions_.compaction_pri = kMinOverlappingRatio has its own set of
     // tests to cover.
     ioptions_.compaction_pri = kByCompensatedSize;
@@ -88,15 +90,14 @@ class CompactionPickerTest : public testing::Test {
            SequenceNumber smallest_seq = 100, SequenceNumber largest_seq = 100,
            size_t compensated_file_size = 0) {
     assert(level < vstorage_->num_levels());
-    FileMetaData* f = new FileMetaData;
-    f->fd = FileDescriptor(file_number, path_id, file_size);
-    f->smallest = InternalKey(smallest, smallest_seq, kTypeValue);
-    f->largest = InternalKey(largest, largest_seq, kTypeValue);
-    f->fd.smallest_seqno = smallest_seq;
-    f->fd.largest_seqno = largest_seq;
+    FileMetaData* f = new FileMetaData(
+        file_number, path_id, file_size,
+        InternalKey(smallest, smallest_seq, kTypeValue),
+        InternalKey(largest, largest_seq, kTypeValue), smallest_seq,
+        largest_seq, /* marked_for_compact */ false, kInvalidBlobFileNumber,
+        kUnknownOldestAncesterTime, kUnknownFileCreationTime);
     f->compensated_file_size =
         (compensated_file_size != 0) ? compensated_file_size : file_size;
-    f->refs = 0;
     vstorage_->AddFile(level, f);
     files_.emplace_back(f);
     file_map_.insert({file_number, {f, level}});
@@ -501,6 +502,168 @@ TEST_F(CompactionPickerTest, AllowsTrivialMoveUniversal) {
   ASSERT_TRUE(compaction->is_trivial_move());
 }
 
+TEST_F(CompactionPickerTest, UniversalPeriodicCompaction1) {
+  // The case where universal periodic compaction can be picked
+  // with some newer files being compacted.
+  const uint64_t kFileSize = 100000;
+
+  mutable_cf_options_.periodic_compaction_seconds = 1000;
+  UniversalCompactionPicker universal_compaction_picker(ioptions_, &icmp_);
+
+  NewVersionStorage(5, kCompactionStyleUniversal);
+
+  Add(0, 1U, "150", "200", kFileSize, 0, 500, 550);
+  Add(0, 2U, "201", "250", kFileSize, 0, 401, 450);
+  Add(0, 4U, "260", "300", kFileSize, 0, 260, 300);
+  Add(3, 5U, "010", "080", kFileSize, 0, 200, 251);
+  Add(4, 3U, "301", "350", kFileSize, 0, 101, 150);
+  Add(4, 6U, "501", "750", kFileSize, 0, 101, 150);
+
+  file_map_[2].first->being_compacted = true;
+  UpdateVersionStorageInfo();
+  vstorage_->TEST_AddFileMarkedForPeriodicCompaction(4, file_map_[3].first);
+
+  std::unique_ptr<Compaction> compaction(
+      universal_compaction_picker.PickCompaction(
+          cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
+
+  ASSERT_TRUE(compaction);
+  ASSERT_EQ(4, compaction->output_level());
+  ASSERT_EQ(0, compaction->start_level());
+  ASSERT_EQ(1U, compaction->num_input_files(0));
+}
+
+TEST_F(CompactionPickerTest, UniversalPeriodicCompaction2) {
+  // The case where universal periodic compaction does not
+  // pick up only level to compact if it doesn't cover
+  // any file marked as periodic compaction.
+  const uint64_t kFileSize = 100000;
+
+  mutable_cf_options_.periodic_compaction_seconds = 1000;
+  UniversalCompactionPicker universal_compaction_picker(ioptions_, &icmp_);
+
+  NewVersionStorage(5, kCompactionStyleUniversal);
+
+  Add(0, 1U, "150", "200", kFileSize, 0, 500, 550);
+  Add(3, 5U, "010", "080", kFileSize, 0, 200, 251);
+  Add(4, 3U, "301", "350", kFileSize, 0, 101, 150);
+  Add(4, 6U, "501", "750", kFileSize, 0, 101, 150);
+
+  file_map_[5].first->being_compacted = true;
+  UpdateVersionStorageInfo();
+  vstorage_->TEST_AddFileMarkedForPeriodicCompaction(0, file_map_[1].first);
+
+  std::unique_ptr<Compaction> compaction(
+      universal_compaction_picker.PickCompaction(
+          cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
+
+  ASSERT_FALSE(compaction);
+}
+
+TEST_F(CompactionPickerTest, UniversalPeriodicCompaction3) {
+  // The case where universal periodic compaction does not
+  // pick up only the last sorted run which is an L0 file if it isn't
+  // marked as periodic compaction.
+  const uint64_t kFileSize = 100000;
+
+  mutable_cf_options_.periodic_compaction_seconds = 1000;
+  UniversalCompactionPicker universal_compaction_picker(ioptions_, &icmp_);
+
+  NewVersionStorage(5, kCompactionStyleUniversal);
+
+  Add(0, 1U, "150", "200", kFileSize, 0, 500, 550);
+  Add(0, 5U, "010", "080", kFileSize, 0, 200, 251);
+  Add(0, 6U, "501", "750", kFileSize, 0, 101, 150);
+
+  file_map_[5].first->being_compacted = true;
+  UpdateVersionStorageInfo();
+  vstorage_->TEST_AddFileMarkedForPeriodicCompaction(0, file_map_[1].first);
+
+  std::unique_ptr<Compaction> compaction(
+      universal_compaction_picker.PickCompaction(
+          cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
+
+  ASSERT_FALSE(compaction);
+}
+
+TEST_F(CompactionPickerTest, UniversalPeriodicCompaction4) {
+  // The case where universal periodic compaction couldn't form
+  // a compaction that inlcudes any file marked for periodic compaction.
+  // Right now we form the compaction anyway if it is more than one
+  // sorted run. Just put the case here to validate that it doesn't
+  // crash.
+  const uint64_t kFileSize = 100000;
+
+  mutable_cf_options_.periodic_compaction_seconds = 1000;
+  UniversalCompactionPicker universal_compaction_picker(ioptions_, &icmp_);
+
+  NewVersionStorage(5, kCompactionStyleUniversal);
+
+  Add(0, 1U, "150", "200", kFileSize, 0, 500, 550);
+  Add(2, 2U, "010", "080", kFileSize, 0, 200, 251);
+  Add(3, 5U, "010", "080", kFileSize, 0, 200, 251);
+  Add(4, 3U, "301", "350", kFileSize, 0, 101, 150);
+  Add(4, 6U, "501", "750", kFileSize, 0, 101, 150);
+
+  file_map_[2].first->being_compacted = true;
+  UpdateVersionStorageInfo();
+  vstorage_->TEST_AddFileMarkedForPeriodicCompaction(0, file_map_[2].first);
+
+  std::unique_ptr<Compaction> compaction(
+      universal_compaction_picker.PickCompaction(
+          cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
+  ASSERT_TRUE(!compaction ||
+              compaction->start_level() != compaction->output_level());
+}
+
+TEST_F(CompactionPickerTest, UniversalPeriodicCompaction5) {
+  // Test single L0 file periodic compaction triggering.
+  const uint64_t kFileSize = 100000;
+
+  mutable_cf_options_.periodic_compaction_seconds = 1000;
+  UniversalCompactionPicker universal_compaction_picker(ioptions_, &icmp_);
+
+  NewVersionStorage(5, kCompactionStyleUniversal);
+
+  Add(0, 6U, "150", "200", kFileSize, 0, 500, 550);
+  UpdateVersionStorageInfo();
+  vstorage_->TEST_AddFileMarkedForPeriodicCompaction(0, file_map_[6].first);
+
+  std::unique_ptr<Compaction> compaction(
+      universal_compaction_picker.PickCompaction(
+          cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
+  ASSERT_TRUE(compaction);
+  ASSERT_EQ(0, compaction->start_level());
+  ASSERT_EQ(1U, compaction->num_input_files(0));
+  ASSERT_EQ(6U, compaction->input(0, 0)->fd.GetNumber());
+  ASSERT_EQ(4, compaction->output_level());
+}
+
+TEST_F(CompactionPickerTest, UniversalPeriodicCompaction6) {
+  // Test single sorted run non-L0 periodic compaction
+  const uint64_t kFileSize = 100000;
+
+  mutable_cf_options_.periodic_compaction_seconds = 1000;
+  UniversalCompactionPicker universal_compaction_picker(ioptions_, &icmp_);
+
+  NewVersionStorage(5, kCompactionStyleUniversal);
+
+  Add(4, 5U, "150", "200", kFileSize, 0, 500, 550);
+  Add(4, 6U, "350", "400", kFileSize, 0, 500, 550);
+  UpdateVersionStorageInfo();
+  vstorage_->TEST_AddFileMarkedForPeriodicCompaction(4, file_map_[6].first);
+
+  std::unique_ptr<Compaction> compaction(
+      universal_compaction_picker.PickCompaction(
+          cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
+  ASSERT_TRUE(compaction);
+  ASSERT_EQ(4, compaction->start_level());
+  ASSERT_EQ(2U, compaction->num_input_files(0));
+  ASSERT_EQ(5U, compaction->input(0, 0)->fd.GetNumber());
+  ASSERT_EQ(6U, compaction->input(0, 1)->fd.GetNumber());
+  ASSERT_EQ(4, compaction->output_level());
+}
+
 TEST_F(CompactionPickerTest, NeedsCompactionFIFO) {
   NewVersionStorage(1, kCompactionStyleFIFO);
   const int kFileCount =
@@ -1488,12 +1651,12 @@ TEST_F(CompactionPickerTest, IntraL0MaxCompactionBytesNotHit) {
   // All 5 L0 files will be picked for intra L0 compaction. The one L1 file
   // spans entire L0 key range and is marked as being compacted to avoid
   // L0->L1 compaction.
-  Add(0, 1U, "100", "150", 200000U);
-  Add(0, 2U, "151", "200", 200000U);
-  Add(0, 3U, "201", "250", 200000U);
-  Add(0, 4U, "251", "300", 200000U);
-  Add(0, 5U, "301", "350", 200000U);
-  Add(1, 6U, "100", "350", 200000U);
+  Add(0, 1U, "100", "150", 200000U, 0, 100, 101);
+  Add(0, 2U, "151", "200", 200000U, 0, 102, 103);
+  Add(0, 3U, "201", "250", 200000U, 0, 104, 105);
+  Add(0, 4U, "251", "300", 200000U, 0, 106, 107);
+  Add(0, 5U, "301", "350", 200000U, 0, 108, 109);
+  Add(1, 6U, "100", "350", 200000U, 0, 110, 111);
   vstorage_->LevelFiles(1)[0]->being_compacted = true;
   UpdateVersionStorageInfo();
 
@@ -1518,12 +1681,12 @@ TEST_F(CompactionPickerTest, IntraL0MaxCompactionBytesHit) {
   // max_compaction_bytes limit (the minimum number of files for triggering
   // intra L0 compaction is 4). The one L1 file spans entire L0 key range and
   // is marked as being compacted to avoid L0->L1 compaction.
-  Add(0, 1U, "100", "150", 200000U);
-  Add(0, 2U, "151", "200", 200000U);
-  Add(0, 3U, "201", "250", 200000U);
-  Add(0, 4U, "251", "300", 200000U);
-  Add(0, 5U, "301", "350", 200000U);
-  Add(1, 6U, "100", "350", 200000U);
+  Add(0, 1U, "100", "150", 200000U, 0, 100, 101);
+  Add(0, 2U, "151", "200", 200000U, 0, 102, 103);
+  Add(0, 3U, "201", "250", 200000U, 0, 104, 105);
+  Add(0, 4U, "251", "300", 200000U, 0, 106, 107);
+  Add(0, 5U, "301", "350", 200000U, 0, 108, 109);
+  Add(1, 6U, "100", "350", 200000U, 0, 109, 110);
   vstorage_->LevelFiles(1)[0]->being_compacted = true;
   UpdateVersionStorageInfo();
 
@@ -1537,6 +1700,38 @@ TEST_F(CompactionPickerTest, IntraL0MaxCompactionBytesHit) {
   ASSERT_EQ(0, compaction->output_level());
 }
 
+TEST_F(CompactionPickerTest, IntraL0ForEarliestSeqno) {
+  // Intra L0 compaction triggers only if there are at least
+  // level0_file_num_compaction_trigger + 2 L0 files.
+  mutable_cf_options_.level0_file_num_compaction_trigger = 3;
+  mutable_cf_options_.max_compaction_bytes = 999999u;
+  NewVersionStorage(6, kCompactionStyleLevel);
+
+  // 4 out of 6 L0 files will be picked for intra L0 compaction due to
+  // being_compact limit. And the latest one L0 will be skipped due to earliest
+  // seqno. The one L1 file spans entire L0 key range and is marked as being
+  // compacted to avoid L0->L1 compaction.
+  Add(1, 1U, "100", "350", 200000U, 0, 110, 111);
+  Add(0, 2U, "301", "350", 1U, 0, 108, 109);
+  Add(0, 3U, "251", "300", 1U, 0, 106, 107);
+  Add(0, 4U, "201", "250", 1U, 0, 104, 105);
+  Add(0, 5U, "151", "200", 1U, 0, 102, 103);
+  Add(0, 6U, "100", "150", 1U, 0, 100, 101);
+  Add(0, 7U, "100", "100", 1U, 0, 99, 100);
+  vstorage_->LevelFiles(0)[5]->being_compacted = true;
+  vstorage_->LevelFiles(1)[0]->being_compacted = true;
+  UpdateVersionStorageInfo();
+
+  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_, 107));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(1U, compaction->num_input_levels());
+  ASSERT_EQ(4U, compaction->num_input_files(0));
+  ASSERT_EQ(CompactionReason::kLevelL0FilesNum,
+            compaction->compaction_reason());
+  ASSERT_EQ(0, compaction->output_level());
+}
+
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

db/compaction/compaction_picker_universal.cc

@@ -90,6 +90,19 @@ class UniversalCompactionBuilder {
 
   Compaction* PickDeleteTriggeredCompaction();
 
+  // Form a compaction from the sorted run indicated by start_index to the
+  // oldest sorted run.
+  // The caller is responsible for making sure that those files are not in
+  // compaction.
+  Compaction* PickCompactionToOldest(size_t start_index,
+                                     CompactionReason compaction_reason);
+
+  // Try to pick periodic compaction. The caller should only call it
+  // if there is at least one file marked for periodic compaction.
+  // null will be returned if no such a compaction can be formed
+  // because some files are being compacted.
+  Compaction* PickPeriodicCompaction();
+
   // Used in universal compaction when the enabled_trivial_move
   // option is set. Checks whether there are any overlapping files
   // in the input. Returns true if the input files are non
@@ -234,7 +247,8 @@ bool UniversalCompactionBuilder::IsInputFilesNonOverlapping(Compaction* c) {
 
     next.f = nullptr;
 
-    if (curr.level != 0 && curr.index < c->num_input_files(curr.level) - 1) {
+    if (c->level(curr.level) != 0 &&
+        curr.index < c->num_input_files(curr.level) - 1) {
       next.f = c->input(curr.level, curr.index + 1);
       next.level = curr.level;
       next.index = curr.index + 1;
@@ -253,6 +267,9 @@ bool UniversalCompactionPicker::NeedsCompaction(
   if (vstorage->CompactionScore(kLevel0) >= 1) {
     return true;
   }
+  if (!vstorage->FilesMarkedForPeriodicCompaction().empty()) {
+    return true;
+  }
   if (!vstorage->FilesMarkedForCompaction().empty()) {
     return true;
   }
@@ -261,7 +278,8 @@ bool UniversalCompactionPicker::NeedsCompaction(
 
 Compaction* UniversalCompactionPicker::PickCompaction(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
-    VersionStorageInfo* vstorage, LogBuffer* log_buffer) {
+    VersionStorageInfo* vstorage, LogBuffer* log_buffer,
+    SequenceNumber /* earliest_memtable_seqno */) {
   UniversalCompactionBuilder builder(ioptions_, icmp_, cf_name,
                                      mutable_cf_options, vstorage, this,
                                      log_buffer);
@@ -358,7 +376,8 @@ Compaction* UniversalCompactionBuilder::PickCompaction() {
       CalculateSortedRuns(*vstorage_, ioptions_, mutable_cf_options_);
 
   if (sorted_runs_.size() == 0 ||
-      (vstorage_->FilesMarkedForCompaction().empty() &&
+      (vstorage_->FilesMarkedForPeriodicCompaction().empty() &&
+       vstorage_->FilesMarkedForCompaction().empty() &&
        sorted_runs_.size() < (unsigned int)mutable_cf_options_
                                  .level0_file_num_compaction_trigger)) {
     ROCKS_LOG_BUFFER(log_buffer_, "[%s] Universal: nothing to do\n",
@@ -373,11 +392,19 @@ Compaction* UniversalCompactionBuilder::PickCompaction() {
       "[%s] Universal: sorted runs files(%" ROCKSDB_PRIszt "): %s\n",
       cf_name_.c_str(), sorted_runs_.size(), vstorage_->LevelSummary(&tmp));
 
-  // Check for size amplification first.
   Compaction* c = nullptr;
-  if (sorted_runs_.size() >=
-      static_cast<size_t>(
-          mutable_cf_options_.level0_file_num_compaction_trigger)) {
+  // Periodic compaction has higher priority than other type of compaction
+  // because it's a hard requirement.
+  if (!vstorage_->FilesMarkedForPeriodicCompaction().empty()) {
+    // Always need to do a full compaction for periodic compaction.
+    c = PickPeriodicCompaction();
+  }
+
+  // Check for size amplification.
+  if (c == nullptr &&
+      sorted_runs_.size() >=
+          static_cast<size_t>(
+              mutable_cf_options_.level0_file_num_compaction_trigger)) {
     if ((c = PickCompactionToReduceSizeAmp()) != nullptr) {
       ROCKS_LOG_BUFFER(log_buffer_, "[%s] Universal: compacting for size amp\n",
                        cf_name_.c_str());
@@ -441,7 +468,8 @@ Compaction* UniversalCompactionBuilder::PickCompaction() {
   }
 
   if (mutable_cf_options_.compaction_options_universal.allow_trivial_move ==
-      true) {
+          true &&
+      c->compaction_reason() != CompactionReason::kPeriodicCompaction) {
     c->set_is_trivial_move(IsInputFilesNonOverlapping(c));
   }
 
@@ -815,59 +843,8 @@ Compaction* UniversalCompactionBuilder::PickCompactionToReduceSizeAmp() {
         " earliest-file-size %" PRIu64,
         cf_name_.c_str(), candidate_size, earliest_file_size);
   }
-  assert(start_index < sorted_runs_.size() - 1);
-
-  // Estimate total file size
-  uint64_t estimated_total_size = 0;
-  for (size_t loop = start_index; loop < sorted_runs_.size(); loop++) {
-    estimated_total_size += sorted_runs_[loop].size;
-  }
-  uint32_t path_id =
-      GetPathId(ioptions_, mutable_cf_options_, estimated_total_size);
-  int start_level = sorted_runs_[start_index].level;
-
-  std::vector<CompactionInputFiles> inputs(vstorage_->num_levels());
-  for (size_t i = 0; i < inputs.size(); ++i) {
-    inputs[i].level = start_level + static_cast<int>(i);
-  }
-  // We always compact all the files, so always compress.
-  for (size_t loop = start_index; loop < sorted_runs_.size(); loop++) {
-    auto& picking_sr = sorted_runs_[loop];
-    if (picking_sr.level == 0) {
-      FileMetaData* f = picking_sr.file;
-      inputs[0].files.push_back(f);
-    } else {
-      auto& files = inputs[picking_sr.level - start_level].files;
-      for (auto* f : vstorage_->LevelFiles(picking_sr.level)) {
-        files.push_back(f);
-      }
-    }
-    char file_num_buf[256];
-    picking_sr.DumpSizeInfo(file_num_buf, sizeof(file_num_buf), loop);
-    ROCKS_LOG_BUFFER(log_buffer_, "[%s] Universal: size amp picking %s",
-                     cf_name_.c_str(), file_num_buf);
-  }
-
-  // output files at the bottom most level, unless it's reserved
-  int output_level = vstorage_->num_levels() - 1;
-  // last level is reserved for the files ingested behind
-  if (ioptions_.allow_ingest_behind) {
-    assert(output_level > 1);
-    output_level--;
-  }
-
-  return new Compaction(
-      vstorage_, ioptions_, mutable_cf_options_, std::move(inputs),
-      output_level,
-      MaxFileSizeForLevel(mutable_cf_options_, output_level,
-                          kCompactionStyleUniversal),
-      /* max_grandparent_overlap_bytes */ LLONG_MAX, path_id,
-      GetCompressionType(ioptions_, vstorage_, mutable_cf_options_,
-                         output_level, 1),
-      GetCompressionOptions(ioptions_, vstorage_, output_level),
-      /* max_subcompactions */ 0, /* grandparents */ {}, /* is manual */ false,
-      score_, false /* deletion_compaction */,
-      CompactionReason::kUniversalSizeAmplification);
+  return PickCompactionToOldest(start_index,
+                                CompactionReason::kUniversalSizeAmplification);
 }
 
 // Pick files marked for compaction. Typically, files are marked by
@@ -987,6 +964,142 @@ Compaction* UniversalCompactionBuilder::PickDeleteTriggeredCompaction() {
       score_, false /* deletion_compaction */,
       CompactionReason::kFilesMarkedForCompaction);
 }
+
+Compaction* UniversalCompactionBuilder::PickCompactionToOldest(
+    size_t start_index, CompactionReason compaction_reason) {
+  assert(start_index < sorted_runs_.size());
+
+  // Estimate total file size
+  uint64_t estimated_total_size = 0;
+  for (size_t loop = start_index; loop < sorted_runs_.size(); loop++) {
+    estimated_total_size += sorted_runs_[loop].size;
+  }
+  uint32_t path_id =
+      GetPathId(ioptions_, mutable_cf_options_, estimated_total_size);
+  int start_level = sorted_runs_[start_index].level;
+
+  std::vector<CompactionInputFiles> inputs(vstorage_->num_levels());
+  for (size_t i = 0; i < inputs.size(); ++i) {
+    inputs[i].level = start_level + static_cast<int>(i);
+  }
+  for (size_t loop = start_index; loop < sorted_runs_.size(); loop++) {
+    auto& picking_sr = sorted_runs_[loop];
+    if (picking_sr.level == 0) {
+      FileMetaData* f = picking_sr.file;
+      inputs[0].files.push_back(f);
+    } else {
+      auto& files = inputs[picking_sr.level - start_level].files;
+      for (auto* f : vstorage_->LevelFiles(picking_sr.level)) {
+        files.push_back(f);
+      }
+    }
+    std::string comp_reason_print_string;
+    if (compaction_reason == CompactionReason::kPeriodicCompaction) {
+      comp_reason_print_string = "periodic compaction";
+    } else if (compaction_reason ==
+               CompactionReason::kUniversalSizeAmplification) {
+      comp_reason_print_string = "size amp";
+    } else {
+      assert(false);
+    }
+
+    char file_num_buf[256];
+    picking_sr.DumpSizeInfo(file_num_buf, sizeof(file_num_buf), loop);
+    ROCKS_LOG_BUFFER(log_buffer_, "[%s] Universal: %s picking %s",
+                     cf_name_.c_str(), comp_reason_print_string.c_str(),
+                     file_num_buf);
+  }
+
+  // output files at the bottom most level, unless it's reserved
+  int output_level = vstorage_->num_levels() - 1;
+  // last level is reserved for the files ingested behind
+  if (ioptions_.allow_ingest_behind) {
+    assert(output_level > 1);
+    output_level--;
+  }
+
+  // We never check size for
+  // compaction_options_universal.compression_size_percent,
+  // because we always compact all the files, so always compress.
+  return new Compaction(
+      vstorage_, ioptions_, mutable_cf_options_, std::move(inputs),
+      output_level,
+      MaxFileSizeForLevel(mutable_cf_options_, output_level,
+                          kCompactionStyleUniversal),
+      LLONG_MAX, path_id,
+      GetCompressionType(ioptions_, vstorage_, mutable_cf_options_, start_level,
+                         1, true /* enable_compression */),
+      GetCompressionOptions(ioptions_, vstorage_, start_level,
+                            true /* enable_compression */),
+      /* max_subcompactions */ 0, /* grandparents */ {}, /* is manual */ false,
+      score_, false /* deletion_compaction */, compaction_reason);
+}
+
+Compaction* UniversalCompactionBuilder::PickPeriodicCompaction() {
+  ROCKS_LOG_BUFFER(log_buffer_, "[%s] Universal: Periodic Compaction",
+                   cf_name_.c_str());
+
+  // In universal compaction, sorted runs contain older data are almost always
+  // generated earlier too. To simplify the problem, we just try to trigger
+  // a full compaction. We start from the oldest sorted run and include
+  // all sorted runs, until we hit a sorted already being compacted.
+  // Since usually the largest (which is usually the oldest) sorted run is
+  // included anyway, doing a full compaction won't increase write
+  // amplification much.
+
+  // Get some information from marked files to check whether a file is
+  // included in the compaction.
+
+  size_t start_index = sorted_runs_.size();
+  while (start_index > 0 && !sorted_runs_[start_index - 1].being_compacted) {
+    start_index--;
+  }
+  if (start_index == sorted_runs_.size()) {
+    return nullptr;
+  }
+
+  // There is a rare corner case where we can't pick up all the files
+  // because some files are being compacted and we end up with picking files
+  // but none of them need periodic compaction. Unless we simply recompact
+  // the last sorted run (either the last level or last L0 file), we would just
+  // execute the compaction, in order to simplify  the logic.
+  if (start_index == sorted_runs_.size() - 1) {
+    bool included_file_marked = false;
+    int start_level = sorted_runs_[start_index].level;
+    FileMetaData* start_file = sorted_runs_[start_index].file;
+    for (const std::pair<int, FileMetaData*>& level_file_pair :
+         vstorage_->FilesMarkedForPeriodicCompaction()) {
+      if (start_level != 0) {
+        // Last sorted run is a level
+        if (start_level == level_file_pair.first) {
+          included_file_marked = true;
+          break;
+        }
+      } else {
+        // Last sorted run is a L0 file.
+        if (start_file == level_file_pair.second) {
+          included_file_marked = true;
+          break;
+        }
+      }
+    }
+    if (!included_file_marked) {
+      ROCKS_LOG_BUFFER(log_buffer_,
+                       "[%s] Universal: Cannot form a compaction covering file "
+                       "marked for periodic compaction",
+                       cf_name_.c_str());
+      return nullptr;
+    }
+  }
+
+  Compaction* c = PickCompactionToOldest(start_index,
+                                         CompactionReason::kPeriodicCompaction);
+
+  TEST_SYNC_POINT_CALLBACK(
+      "UniversalCompactionPicker::PickPeriodicCompaction:Return", c);
+
+  return c;
+}
 }  // namespace rocksdb
 
 #endif  // !ROCKSDB_LITE

db/compaction/compaction_picker_universal.h

@@ -18,11 +18,10 @@ class UniversalCompactionPicker : public CompactionPicker {
   UniversalCompactionPicker(const ImmutableCFOptions& ioptions,
                             const InternalKeyComparator* icmp)
       : CompactionPicker(ioptions, icmp) {}
-  virtual Compaction* PickCompaction(const std::string& cf_name,
-                                     const MutableCFOptions& mutable_cf_options,
-                                     VersionStorageInfo* vstorage,
-                                     LogBuffer* log_buffer) override;
-
+  virtual Compaction* PickCompaction(
+      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
+      VersionStorageInfo* vstorage, LogBuffer* log_buffer,
+      SequenceNumber earliest_memtable_seqno = kMaxSequenceNumber) override;
   virtual int MaxOutputLevel() const override { return NumberLevels() - 1; }
 
   virtual bool NeedsCompaction(

db/convenience.cc

@@ -41,14 +41,16 @@ Status VerifySstFileChecksum(const Options& options,
                              const EnvOptions& env_options,
                              const ReadOptions& read_options,
                              const std::string& file_path) {
-  std::unique_ptr<RandomAccessFile> file;
+  std::unique_ptr<FSRandomAccessFile> file;
   uint64_t file_size;
   InternalKeyComparator internal_comparator(options.comparator);
   ImmutableCFOptions ioptions(options);
 
-  Status s = ioptions.env->NewRandomAccessFile(file_path, &file, env_options);
+  Status s = ioptions.fs->NewRandomAccessFile(file_path,
+                                              FileOptions(env_options),
+                                              &file, nullptr);
   if (s.ok()) {
-    s = ioptions.env->GetFileSize(file_path, &file_size);
+    s = ioptions.fs->GetFileSize(file_path, IOOptions(), &file_size, nullptr);
   } else {
     return s;
   }

db/corruption_test.cc

@@ -20,6 +20,7 @@
 #include "db/db_test_util.h"
 #include "db/log_format.h"
 #include "db/version_set.h"
+#include "env/composite_env_wrapper.h"
 #include "file/filename.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/convenience.h"
@@ -189,6 +190,7 @@ class CorruptionTest : public testing::Test {
     ASSERT_TRUE(s.ok()) << s.ToString();
     Options options;
     EnvOptions env_options;
+    options.file_system.reset(new LegacyFileSystemWrapper(options.env));
     ASSERT_NOK(VerifySstFileChecksum(options, env_options, fname));
   }
 
@@ -369,6 +371,13 @@ TEST_F(CorruptionTest, VerifyChecksumReadahead) {
   ASSERT_GE(senv.random_read_counter_.Read(), 213);
   ASSERT_LE(senv.random_read_counter_.Read(), 447);
 
+  // Test readahead shouldn't break mmap mode (where it should be
+  // disabled).
+  options.allow_mmap_reads = true;
+  Reopen(&options);
+  dbi = static_cast<DBImpl*>(db_);
+  ASSERT_OK(dbi->VerifyChecksum(ro));
+
   CloseDb();
 }
 
@@ -532,7 +541,8 @@ TEST_F(CorruptionTest, RangeDeletionCorrupted) {
   std::unique_ptr<RandomAccessFile> file;
   ASSERT_OK(options_.env->NewRandomAccessFile(filename, &file, EnvOptions()));
   std::unique_ptr<RandomAccessFileReader> file_reader(
-      new RandomAccessFileReader(std::move(file), filename));
+      new RandomAccessFileReader(NewLegacyRandomAccessFileWrapper(file),
+                                 filename));
 
   uint64_t file_size;
   ASSERT_OK(options_.env->GetFileSize(filename, &file_size));

db/cuckoo_table_db_test.cc

@@ -285,6 +285,9 @@ TEST_F(CuckooTableDBTest, SameKeyInsertedInTwoDifferentFilesAndCompacted) {
 TEST_F(CuckooTableDBTest, AdaptiveTable) {
   Options options = CurrentOptions();
 
+  // Ensure options compatible with PlainTable
+  options.prefix_extractor.reset(NewCappedPrefixTransform(8));
+
   // Write some keys using cuckoo table.
   options.table_factory.reset(NewCuckooTableFactory());
   Reopen(&options);

db/db_basic_test.cc

@@ -6,11 +6,11 @@
 // 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.
-// #include <iostream>
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
 #include "rocksdb/perf_context.h"
 #include "rocksdb/utilities/debug.h"
+#include "table/block_based/block_based_table_reader.h"
 #include "table/block_based/block_builder.h"
 #include "test_util/fault_injection_test_env.h"
 #if !defined(ROCKSDB_LITE)
@@ -524,6 +524,7 @@ TEST_F(DBBasicTest, Snapshot) {
     ASSERT_EQ(1U, GetNumSnapshots());
     uint64_t time_snap1 = GetTimeOldestSnapshots();
     ASSERT_GT(time_snap1, 0U);
+    ASSERT_EQ(GetSequenceOldestSnapshots(), s1->GetSequenceNumber());
     Put(0, "foo", "0v2");
     Put(1, "foo", "1v2");
 
@@ -532,6 +533,7 @@ TEST_F(DBBasicTest, Snapshot) {
     const Snapshot* s2 = db_->GetSnapshot();
     ASSERT_EQ(2U, GetNumSnapshots());
     ASSERT_EQ(time_snap1, GetTimeOldestSnapshots());
+    ASSERT_EQ(GetSequenceOldestSnapshots(), s1->GetSequenceNumber());
     Put(0, "foo", "0v3");
     Put(1, "foo", "1v3");
 
@@ -539,6 +541,7 @@ TEST_F(DBBasicTest, Snapshot) {
       ManagedSnapshot s3(db_);
       ASSERT_EQ(3U, GetNumSnapshots());
       ASSERT_EQ(time_snap1, GetTimeOldestSnapshots());
+      ASSERT_EQ(GetSequenceOldestSnapshots(), s1->GetSequenceNumber());
 
       Put(0, "foo", "0v4");
       Put(1, "foo", "1v4");
@@ -554,6 +557,7 @@ TEST_F(DBBasicTest, Snapshot) {
 
     ASSERT_EQ(2U, GetNumSnapshots());
     ASSERT_EQ(time_snap1, GetTimeOldestSnapshots());
+    ASSERT_EQ(GetSequenceOldestSnapshots(), s1->GetSequenceNumber());
     ASSERT_EQ("0v1", Get(0, "foo", s1));
     ASSERT_EQ("1v1", Get(1, "foo", s1));
     ASSERT_EQ("0v2", Get(0, "foo", s2));
@@ -568,9 +572,11 @@ TEST_F(DBBasicTest, Snapshot) {
     ASSERT_EQ("1v4", Get(1, "foo"));
     ASSERT_EQ(1U, GetNumSnapshots());
     ASSERT_LT(time_snap1, GetTimeOldestSnapshots());
+    ASSERT_EQ(GetSequenceOldestSnapshots(), s2->GetSequenceNumber());
 
     db_->ReleaseSnapshot(s2);
     ASSERT_EQ(0U, GetNumSnapshots());
+    ASSERT_EQ(GetSequenceOldestSnapshots(), 0);
     ASSERT_EQ("0v4", Get(0, "foo"));
     ASSERT_EQ("1v4", Get(1, "foo"));
   } while (ChangeOptions());
@@ -871,11 +877,12 @@ TEST_F(DBBasicTest, ChecksumTest) {
     ASSERT_OK(Flush());
   }
 
-  // verify data with each type of checksum
-  for (int i = 0; i <= kxxHash64; ++i) {
+  // with each valid checksum type setting...
+  for (int i = 0; i <= max_checksum; ++i) {
     table_options.checksum = static_cast<ChecksumType>(i);
     options.table_factory.reset(NewBlockBasedTableFactory(table_options));
     Reopen(options);
+    // verify every type of checksum (should be regardless of that setting)
     for (int j = 0; j < (max_checksum + 1) * kNumPerFile; ++j) {
       ASSERT_EQ(Key(j), Get(Key(j)));
     }
@@ -913,30 +920,30 @@ TEST_F(DBBasicTest, MmapAndBufferOptions) {
 
 class TestEnv : public EnvWrapper {
   public:
-    explicit TestEnv() : EnvWrapper(Env::Default()),
-                close_count(0) { }
+   explicit TestEnv(Env* base_env) : EnvWrapper(base_env), close_count(0) {}
 
-    class TestLogger : public Logger {
-      public:
-        using Logger::Logv;
-        TestLogger(TestEnv *env_ptr) : Logger() { env = env_ptr; }
-        ~TestLogger() override {
-          if (!closed_) {
-            CloseHelper();
-          }
-        }
-        void Logv(const char* /*format*/, va_list /*ap*/) override{};
+   class TestLogger : public Logger {
+    public:
+     using Logger::Logv;
+     explicit TestLogger(TestEnv* env_ptr) : Logger() { env = env_ptr; }
+     ~TestLogger() override {
+       if (!closed_) {
+         CloseHelper();
+       }
+     }
+     void Logv(const char* /*format*/, va_list /*ap*/) override {}
 
-       protected:
-        Status CloseImpl() override { return CloseHelper(); }
+    protected:
+     Status CloseImpl() override { return CloseHelper(); }
 
-       private:
-        Status CloseHelper() {
-          env->CloseCountInc();;
-          return Status::IOError();
-        }
-        TestEnv *env;
-    };
+    private:
+     Status CloseHelper() {
+       env->CloseCountInc();
+       ;
+       return Status::IOError();
+     }
+     TestEnv* env;
+   };
 
     void CloseCountInc() { close_count++; }
 
@@ -958,7 +965,8 @@ TEST_F(DBBasicTest, DBClose) {
   ASSERT_OK(DestroyDB(dbname, options));
 
   DB* db = nullptr;
-  TestEnv* env = new TestEnv();
+  TestEnv* env = new TestEnv(env_);
+  std::unique_ptr<TestEnv> local_env_guard(env);
   options.create_if_missing = true;
   options.env = env;
   Status s = DB::Open(options, dbname, &db);
@@ -992,13 +1000,11 @@ TEST_F(DBBasicTest, DBClose) {
   ASSERT_EQ(env->GetCloseCount(), 2);
   options.info_log.reset();
   ASSERT_EQ(env->GetCloseCount(), 3);
-
-  delete options.env;
 }
 
 TEST_F(DBBasicTest, DBCloseFlushError) {
   std::unique_ptr<FaultInjectionTestEnv> fault_injection_env(
-      new FaultInjectionTestEnv(Env::Default()));
+      new FaultInjectionTestEnv(env_));
   Options options = GetDefaultOptions();
   options.create_if_missing = true;
   options.manual_wal_flush = true;
@@ -1018,15 +1024,27 @@ TEST_F(DBBasicTest, DBCloseFlushError) {
   Destroy(options);
 }
 
-TEST_F(DBBasicTest, MultiGetMultiCF) {
+class DBMultiGetTestWithParam : public DBBasicTest,
+                                public testing::WithParamInterface<bool> {};
+
+TEST_P(DBMultiGetTestWithParam, MultiGetMultiCF) {
   Options options = CurrentOptions();
   CreateAndReopenWithCF({"pikachu", "ilya", "muromec", "dobrynia", "nikitich",
                          "alyosha", "popovich"},
                         options);
+  // <CF, key, value> tuples
+  std::vector<std::tuple<int, std::string, std::string>> cf_kv_vec;
+  static const int num_keys = 24;
+  cf_kv_vec.reserve(num_keys);
 
-  for (int i = 0; i < 8; ++i) {
-    ASSERT_OK(Put(i, "cf" + std::to_string(i) + "_key",
-                  "cf" + std::to_string(i) + "_val"));
+  for (int i = 0; i < num_keys; ++i) {
+    int cf = i / 3;
+    int cf_key = 1 % 3;
+    cf_kv_vec.emplace_back(std::make_tuple(
+        cf, "cf" + std::to_string(cf) + "_key_" + std::to_string(cf_key),
+        "cf" + std::to_string(cf) + "_val_" + std::to_string(cf_key)));
+    ASSERT_OK(Put(std::get<0>(cf_kv_vec[i]), std::get<1>(cf_kv_vec[i]),
+                  std::get<2>(cf_kv_vec[i])));
   }
 
   int get_sv_count = 0;
@@ -1036,10 +1054,14 @@ TEST_F(DBBasicTest, MultiGetMultiCF) {
         if (++get_sv_count == 2) {
           // After MultiGet refs a couple of CFs, flush all CFs so MultiGet
           // is forced to repeat the process
-          for (int i = 0; i < 8; ++i) {
-            ASSERT_OK(Flush(i));
-            ASSERT_OK(Put(i, "cf" + std::to_string(i) + "_key",
-                          "cf" + std::to_string(i) + "_val2"));
+          for (int i = 0; i < num_keys; ++i) {
+            int cf = i / 3;
+            int cf_key = i % 8;
+            if (cf_key == 0) {
+              ASSERT_OK(Flush(cf));
+            }
+            ASSERT_OK(Put(std::get<0>(cf_kv_vec[i]), std::get<1>(cf_kv_vec[i]),
+                          std::get<2>(cf_kv_vec[i]) + "_2"));
           }
         }
         if (get_sv_count == 11) {
@@ -1057,26 +1079,53 @@ TEST_F(DBBasicTest, MultiGetMultiCF) {
   std::vector<std::string> keys;
   std::vector<std::string> values;
 
-  for (int i = 0; i < 8; ++i) {
-    cfs.push_back(i);
-    keys.push_back("cf" + std::to_string(i) + "_key");
+  for (int i = 0; i < num_keys; ++i) {
+    cfs.push_back(std::get<0>(cf_kv_vec[i]));
+    keys.push_back(std::get<1>(cf_kv_vec[i]));
   }
 
-  values = MultiGet(cfs, keys, nullptr);
-  ASSERT_EQ(values.size(), 8);
+  values = MultiGet(cfs, keys, nullptr, GetParam());
+  ASSERT_EQ(values.size(), num_keys);
   for (unsigned int j = 0; j < values.size(); ++j) {
-    ASSERT_EQ(values[j], "cf" + std::to_string(j) + "_val2");
+    ASSERT_EQ(values[j], std::get<2>(cf_kv_vec[j]) + "_2");
   }
-  for (int i = 0; i < 8; ++i) {
+
+  keys.clear();
+  cfs.clear();
+  cfs.push_back(std::get<0>(cf_kv_vec[0]));
+  keys.push_back(std::get<1>(cf_kv_vec[0]));
+  cfs.push_back(std::get<0>(cf_kv_vec[3]));
+  keys.push_back(std::get<1>(cf_kv_vec[3]));
+  cfs.push_back(std::get<0>(cf_kv_vec[4]));
+  keys.push_back(std::get<1>(cf_kv_vec[4]));
+  values = MultiGet(cfs, keys, nullptr, GetParam());
+  ASSERT_EQ(values[0], std::get<2>(cf_kv_vec[0]) + "_2");
+  ASSERT_EQ(values[1], std::get<2>(cf_kv_vec[3]) + "_2");
+  ASSERT_EQ(values[2], std::get<2>(cf_kv_vec[4]) + "_2");
+
+  keys.clear();
+  cfs.clear();
+  cfs.push_back(std::get<0>(cf_kv_vec[7]));
+  keys.push_back(std::get<1>(cf_kv_vec[7]));
+  cfs.push_back(std::get<0>(cf_kv_vec[6]));
+  keys.push_back(std::get<1>(cf_kv_vec[6]));
+  cfs.push_back(std::get<0>(cf_kv_vec[1]));
+  keys.push_back(std::get<1>(cf_kv_vec[1]));
+  values = MultiGet(cfs, keys, nullptr, GetParam());
+  ASSERT_EQ(values[0], std::get<2>(cf_kv_vec[7]) + "_2");
+  ASSERT_EQ(values[1], std::get<2>(cf_kv_vec[6]) + "_2");
+  ASSERT_EQ(values[2], std::get<2>(cf_kv_vec[1]) + "_2");
+
+  for (int cf = 0; cf < 8; ++cf) {
     auto* cfd = reinterpret_cast<ColumnFamilyHandleImpl*>(
-                    reinterpret_cast<DBImpl*>(db_)->GetColumnFamilyHandle(i))
+                    reinterpret_cast<DBImpl*>(db_)->GetColumnFamilyHandle(cf))
                     ->cfd();
     ASSERT_NE(cfd->TEST_GetLocalSV()->Get(), SuperVersion::kSVInUse);
     ASSERT_NE(cfd->TEST_GetLocalSV()->Get(), SuperVersion::kSVObsolete);
   }
 }
 
-TEST_F(DBBasicTest, MultiGetMultiCFMutex) {
+TEST_P(DBMultiGetTestWithParam, MultiGetMultiCFMutex) {
   Options options = CurrentOptions();
   CreateAndReopenWithCF({"pikachu", "ilya", "muromec", "dobrynia", "nikitich",
                          "alyosha", "popovich"},
@@ -1122,7 +1171,7 @@ TEST_F(DBBasicTest, MultiGetMultiCFMutex) {
     keys.push_back("cf" + std::to_string(i) + "_key");
   }
 
-  values = MultiGet(cfs, keys, nullptr);
+  values = MultiGet(cfs, keys, nullptr, GetParam());
   ASSERT_TRUE(last_try);
   ASSERT_EQ(values.size(), 8);
   for (unsigned int j = 0; j < values.size(); ++j) {
@@ -1137,7 +1186,7 @@ TEST_F(DBBasicTest, MultiGetMultiCFMutex) {
   }
 }
 
-TEST_F(DBBasicTest, MultiGetMultiCFSnapshot) {
+TEST_P(DBMultiGetTestWithParam, MultiGetMultiCFSnapshot) {
   Options options = CurrentOptions();
   CreateAndReopenWithCF({"pikachu", "ilya", "muromec", "dobrynia", "nikitich",
                          "alyosha", "popovich"},
@@ -1182,7 +1231,7 @@ TEST_F(DBBasicTest, MultiGetMultiCFSnapshot) {
   }
 
   const Snapshot* snapshot = db_->GetSnapshot();
-  values = MultiGet(cfs, keys, snapshot);
+  values = MultiGet(cfs, keys, snapshot, GetParam());
   db_->ReleaseSnapshot(snapshot);
   ASSERT_EQ(values.size(), 8);
   for (unsigned int j = 0; j < values.size(); ++j) {
@@ -1196,6 +1245,9 @@ TEST_F(DBBasicTest, MultiGetMultiCFSnapshot) {
   }
 }
 
+INSTANTIATE_TEST_CASE_P(DBMultiGetTestWithParam, DBMultiGetTestWithParam,
+                        testing::Bool());
+
 TEST_F(DBBasicTest, MultiGetBatchedSimpleUnsorted) {
   do {
     CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
@@ -1363,6 +1415,7 @@ class MultiGetPrefixExtractorTest : public DBBasicTest,
 TEST_P(MultiGetPrefixExtractorTest, Batched) {
   Options options = CurrentOptions();
   options.prefix_extractor.reset(NewFixedPrefixTransform(2));
+  options.memtable_prefix_bloom_size_ratio = 10;
   BlockBasedTableOptions bbto;
   if (GetParam()) {
     bbto.index_type = BlockBasedTableOptions::IndexType::kTwoLevelIndexSearch;
@@ -1374,17 +1427,30 @@ TEST_P(MultiGetPrefixExtractorTest, Batched) {
   options.table_factory.reset(NewBlockBasedTableFactory(bbto));
   Reopen(options);
 
+  SetPerfLevel(kEnableCount);
+  get_perf_context()->Reset();
+
   // First key is not in the prefix_extractor domain
   ASSERT_OK(Put("k", "v0"));
   ASSERT_OK(Put("kk1", "v1"));
   ASSERT_OK(Put("kk2", "v2"));
   ASSERT_OK(Put("kk3", "v3"));
   ASSERT_OK(Put("kk4", "v4"));
+  std::vector<std::string> mem_keys(
+      {"k", "kk1", "kk2", "kk3", "kk4", "rofl", "lmho"});
+  std::vector<std::string> inmem_values;
+  inmem_values = MultiGet(mem_keys, nullptr);
+  ASSERT_EQ(inmem_values[0], "v0");
+  ASSERT_EQ(inmem_values[1], "v1");
+  ASSERT_EQ(inmem_values[2], "v2");
+  ASSERT_EQ(inmem_values[3], "v3");
+  ASSERT_EQ(inmem_values[4], "v4");
+  ASSERT_EQ(get_perf_context()->bloom_memtable_miss_count, 2);
+  ASSERT_EQ(get_perf_context()->bloom_memtable_hit_count, 5);
   ASSERT_OK(Flush());
 
   std::vector<std::string> keys({"k", "kk1", "kk2", "kk3", "kk4"});
   std::vector<std::string> values;
-  SetPerfLevel(kEnableCount);
   get_perf_context()->Reset();
   values = MultiGet(keys, nullptr);
   ASSERT_EQ(values[0], "v0");
@@ -1539,6 +1605,45 @@ TEST_F(DBBasicTest, GetAllKeyVersions) {
 }
 #endif  // !ROCKSDB_LITE
 
+TEST_F(DBBasicTest, MultiGetIOBufferOverrun) {
+  Options options = CurrentOptions();
+  Random rnd(301);
+  BlockBasedTableOptions table_options;
+  table_options.pin_l0_filter_and_index_blocks_in_cache = true;
+  table_options.block_size = 16 * 1024;
+  assert(table_options.block_size >
+          BlockBasedTable::kMultiGetReadStackBufSize);
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  Reopen(options);
+
+  std::string zero_str(128, '\0');
+  for (int i = 0; i < 100; ++i) {
+    // Make the value compressible. A purely random string doesn't compress
+    // and the resultant data block will not be compressed
+    std::string value(RandomString(&rnd, 128) + zero_str);
+    assert(Put(Key(i), value) == Status::OK());
+  }
+  Flush();
+
+  std::vector<std::string> key_data(10);
+  std::vector<Slice> keys;
+  // We cannot resize a PinnableSlice vector, so just set initial size to
+  // largest we think we will need
+  std::vector<PinnableSlice> values(10);
+  std::vector<Status> statuses;
+  ReadOptions ro;
+
+  // Warm up the cache first
+  key_data.emplace_back(Key(0));
+  keys.emplace_back(Slice(key_data.back()));
+  key_data.emplace_back(Key(50));
+  keys.emplace_back(Slice(key_data.back()));
+  statuses.resize(keys.size());
+
+  dbfull()->MultiGet(ro, dbfull()->DefaultColumnFamily(), keys.size(),
+                     keys.data(), values.data(), statuses.data(), true);
+}
+
 class DBBasicTestWithParallelIO
     : public DBTestBase,
       public testing::WithParamInterface<std::tuple<bool, bool, bool, bool>> {
@@ -1563,8 +1668,32 @@ class DBBasicTestWithParallelIO
     Options options = CurrentOptions();
     Random rnd(301);
     BlockBasedTableOptions table_options;
-    table_options.pin_l0_filter_and_index_blocks_in_cache = true;
+
+#ifndef ROCKSDB_LITE
+    if (compression_enabled_) {
+      std::vector<CompressionType> compression_types;
+      compression_types = GetSupportedCompressions();
+      // Not every platform may have compression libraries available, so
+      // dynamically pick based on what's available
+      if (compression_types.size() == 0) {
+        compression_enabled_ = false;
+      } else {
+        options.compression = compression_types[0];
+      }
+    }
+#else
+    // GetSupportedCompressions() is not available in LITE build
+    if (!Snappy_Supported()) {
+      compression_enabled_ = false;
+    }
+#endif //ROCKSDB_LITE
+
     table_options.block_cache = uncompressed_cache_;
+    if (table_options.block_cache == nullptr) {
+      table_options.no_block_cache = true;
+    } else {
+      table_options.pin_l0_filter_and_index_blocks_in_cache = true;
+    }
     table_options.block_cache_compressed = compressed_cache_;
     table_options.flush_block_policy_factory.reset(
         new MyFlushBlockPolicyFactory());
@@ -1582,6 +1711,14 @@ class DBBasicTestWithParallelIO
       assert(Put(Key(i), values_[i]) == Status::OK());
     }
     Flush();
+
+    for (int i = 0; i < 100; ++i) {
+      // block cannot gain space by compression
+      uncompressable_values_.emplace_back(RandomString(&rnd, 256) + '\0');
+      std::string tmp_key = "a" + Key(i);
+      assert(Put(tmp_key, uncompressable_values_[i]) == Status::OK());
+    }
+    Flush();
   }
 
   bool CheckValue(int i, const std::string& value) {
@@ -1591,6 +1728,13 @@ class DBBasicTestWithParallelIO
     return false;
   }
 
+  bool CheckUncompressableValue(int i, const std::string& value) {
+    if (uncompressable_values_[i].compare(value) == 0) {
+      return true;
+    }
+    return false;
+  }
+
   int num_lookups() { return uncompressed_cache_->num_lookups(); }
   int num_found() { return uncompressed_cache_->num_found(); }
   int num_inserts() { return uncompressed_cache_->num_inserts(); }
@@ -1600,6 +1744,9 @@ class DBBasicTestWithParallelIO
   int num_inserts_compressed() { return compressed_cache_->num_inserts(); }
 
   bool fill_cache() { return fill_cache_; }
+  bool compression_enabled() { return compression_enabled_; }
+  bool has_compressed_cache() { return compressed_cache_ != nullptr; }
+  bool has_uncompressed_cache() { return uncompressed_cache_ != nullptr; }
 
   static void SetUpTestCase() {}
   static void TearDownTestCase() {}
@@ -1737,6 +1884,7 @@ class DBBasicTestWithParallelIO
   std::shared_ptr<MyBlockCache> uncompressed_cache_;
   bool compression_enabled_;
   std::vector<std::string> values_;
+  std::vector<std::string> uncompressable_values_;
   bool fill_cache_;
 };
 
@@ -1774,7 +1922,16 @@ TEST_P(DBBasicTestWithParallelIO, MultiGet) {
   ASSERT_TRUE(CheckValue(1, values[0].ToString()));
   ASSERT_TRUE(CheckValue(51, values[1].ToString()));
 
-  int expected_reads = random_reads + (fill_cache() ? 0 : 2);
+  bool read_from_cache = false;
+  if (fill_cache()) {
+    if (has_uncompressed_cache()) {
+      read_from_cache = true;
+    } else if (has_compressed_cache() && compression_enabled()) {
+      read_from_cache = true;
+    }
+  }
+
+  int expected_reads = random_reads + (read_from_cache ? 0 : 2);
   ASSERT_EQ(env_->random_read_counter_.Read(), expected_reads);
 
   keys.resize(10);
@@ -1792,8 +1949,66 @@ TEST_P(DBBasicTestWithParallelIO, MultiGet) {
     ASSERT_OK(statuses[i]);
     ASSERT_TRUE(CheckValue(key_ints[i], values[i].ToString()));
   }
-  expected_reads += (fill_cache() ? 2 : 4);
+  if (compression_enabled() && !has_compressed_cache()) {
+    expected_reads += (read_from_cache ? 2 : 3);
+  } else {
+    expected_reads += (read_from_cache ? 2 : 4);
+  }
   ASSERT_EQ(env_->random_read_counter_.Read(), expected_reads);
+
+  keys.resize(10);
+  statuses.resize(10);
+  std::vector<int> key_uncmp{1, 2, 15, 16, 55, 81, 82, 83, 84, 85};
+  for (size_t i = 0; i < key_uncmp.size(); ++i) {
+    key_data[i] = "a" + Key(key_uncmp[i]);
+    keys[i] = Slice(key_data[i]);
+    statuses[i] = Status::OK();
+    values[i].Reset();
+  }
+  dbfull()->MultiGet(ro, dbfull()->DefaultColumnFamily(), keys.size(),
+                     keys.data(), values.data(), statuses.data(), true);
+  for (size_t i = 0; i < key_uncmp.size(); ++i) {
+    ASSERT_OK(statuses[i]);
+    ASSERT_TRUE(CheckUncompressableValue(key_uncmp[i], values[i].ToString()));
+  }
+  if (compression_enabled() && !has_compressed_cache()) {
+    expected_reads += (read_from_cache ? 3 : 3);
+  } else {
+    expected_reads += (read_from_cache ? 4 : 4);
+  }
+  ASSERT_EQ(env_->random_read_counter_.Read(), expected_reads);
+
+  keys.resize(5);
+  statuses.resize(5);
+  std::vector<int> key_tr{1, 2, 15, 16, 55};
+  for (size_t i = 0; i < key_tr.size(); ++i) {
+    key_data[i] = "a" + Key(key_tr[i]);
+    keys[i] = Slice(key_data[i]);
+    statuses[i] = Status::OK();
+    values[i].Reset();
+  }
+  dbfull()->MultiGet(ro, dbfull()->DefaultColumnFamily(), keys.size(),
+                     keys.data(), values.data(), statuses.data(), true);
+  for (size_t i = 0; i < key_tr.size(); ++i) {
+    ASSERT_OK(statuses[i]);
+    ASSERT_TRUE(CheckUncompressableValue(key_tr[i], values[i].ToString()));
+  }
+  if (compression_enabled() && !has_compressed_cache()) {
+    expected_reads += (read_from_cache ? 0 : 2);
+    ASSERT_EQ(env_->random_read_counter_.Read(), expected_reads);
+  } else {
+    if (has_uncompressed_cache()) {
+      expected_reads += (read_from_cache ? 0 : 3);
+      ASSERT_EQ(env_->random_read_counter_.Read(), expected_reads);
+    } else {
+      // A rare case, even we enable the block compression but some of data
+      // blocks are not compressed due to content. If user only enable the
+      // compressed cache, the uncompressed blocks will not tbe cached, and
+      // block reads will be triggered. The number of reads is related to
+      // the compression algorithm.
+      ASSERT_TRUE(env_->random_read_counter_.Read() >= expected_reads);
+    }
+  }
 }
 
 INSTANTIATE_TEST_CASE_P(
@@ -1803,12 +2018,8 @@ INSTANTIATE_TEST_CASE_P(
     // Param 1 - Uncompressed cache enabled
     // Param 2 - Data compression enabled
     // Param 3 - ReadOptions::fill_cache
-    ::testing::Values(std::make_tuple(false, true, true, true),
-                      std::make_tuple(true, true, true, true),
-                      std::make_tuple(false, true, false, true),
-                      std::make_tuple(false, true, true, false),
-                      std::make_tuple(true, true, true, false),
-                      std::make_tuple(false, true, false, false)));
+    ::testing::Combine(::testing::Bool(), ::testing::Bool(),
+                       ::testing::Bool(), ::testing::Bool()));
 
 class DBBasicTestWithTimestampWithParam
     : public DBTestBase,
@@ -1913,48 +2124,81 @@ TEST_P(DBBasicTestWithTimestampWithParam, PutAndGet) {
       10 /*bits_per_key*/, false /*use_block_based_builder*/));
   bbto.whole_key_filtering = true;
   options.table_factory.reset(NewBlockBasedTableFactory(bbto));
-  DestroyAndReopen(options);
-  CreateAndReopenWithCF({"pikachu"}, options);
-  size_t num_cfs = handles_.size();
-  ASSERT_EQ(2, num_cfs);
-  std::vector<std::string> write_ts_strs(kNumTimestamps);
-  std::vector<std::string> read_ts_strs(kNumTimestamps);
-  std::vector<Slice> write_ts_list;
-  std::vector<Slice> read_ts_list;
 
-  for (size_t i = 0; i != kNumTimestamps; ++i) {
-    write_ts_list.emplace_back(EncodeTimestamp(i * 2, 0, &write_ts_strs[i]));
-    read_ts_list.emplace_back(EncodeTimestamp(1 + i * 2, 0, &read_ts_strs[i]));
-    const Slice& write_ts = write_ts_list.back();
-    WriteOptions wopts;
-    wopts.timestamp = &write_ts;
-    for (int cf = 0; cf != static_cast<int>(num_cfs); ++cf) {
-      for (size_t j = 0; j != (kNumKeysPerFile - 1) / kNumTimestamps; ++j) {
-        ASSERT_OK(Put(cf, "key" + std::to_string(j),
-                      "value_" + std::to_string(j) + "_" + std::to_string(i),
-                      wopts));
-      }
-      if (!memtable_only) {
-        ASSERT_OK(Flush(cf));
-      }
-    }
+  std::vector<CompressionType> compression_types;
+  compression_types.push_back(kNoCompression);
+  if (Zlib_Supported()) {
+    compression_types.push_back(kZlibCompression);
   }
-  const auto& verify_db_func = [&]() {
-    for (size_t i = 0; i != kNumTimestamps; ++i) {
-      ReadOptions ropts;
-      ropts.timestamp = &read_ts_list[i];
-      for (int cf = 0; cf != static_cast<int>(num_cfs); ++cf) {
-        ColumnFamilyHandle* cfh = handles_[cf];
-        for (size_t j = 0; j != (kNumKeysPerFile - 1) / kNumTimestamps; ++j) {
-          std::string value;
-          ASSERT_OK(db_->Get(ropts, cfh, "key" + std::to_string(j), &value));
-          ASSERT_EQ("value_" + std::to_string(j) + "_" + std::to_string(i),
-                    value);
+#if LZ4_VERSION_NUMBER >= 10400  // r124+
+  compression_types.push_back(kLZ4Compression);
+  compression_types.push_back(kLZ4HCCompression);
+#endif  // LZ4_VERSION_NUMBER >= 10400
+  if (ZSTD_Supported()) {
+    compression_types.push_back(kZSTD);
+  }
+
+  // Switch compression dictionary on/off to check key extraction
+  // correctness in kBuffered state
+  std::vector<uint32_t> max_dict_bytes_list = {0, 1 << 14};  // 0 or 16KB
+
+  for (auto compression_type : compression_types) {
+    for (uint32_t max_dict_bytes : max_dict_bytes_list) {
+      options.compression = compression_type;
+      options.compression_opts.max_dict_bytes = max_dict_bytes;
+      if (compression_type == kZSTD) {
+        options.compression_opts.zstd_max_train_bytes = max_dict_bytes;
+      }
+      options.target_file_size_base = 1 << 26;  // 64MB
+
+      DestroyAndReopen(options);
+      CreateAndReopenWithCF({"pikachu"}, options);
+      size_t num_cfs = handles_.size();
+      ASSERT_EQ(2, num_cfs);
+      std::vector<std::string> write_ts_strs(kNumTimestamps);
+      std::vector<std::string> read_ts_strs(kNumTimestamps);
+      std::vector<Slice> write_ts_list;
+      std::vector<Slice> read_ts_list;
+
+      for (size_t i = 0; i != kNumTimestamps; ++i) {
+        write_ts_list.emplace_back(
+            EncodeTimestamp(i * 2, 0, &write_ts_strs[i]));
+        read_ts_list.emplace_back(
+            EncodeTimestamp(1 + i * 2, 0, &read_ts_strs[i]));
+        const Slice& write_ts = write_ts_list.back();
+        WriteOptions wopts;
+        wopts.timestamp = &write_ts;
+        for (int cf = 0; cf != static_cast<int>(num_cfs); ++cf) {
+          for (size_t j = 0; j != (kNumKeysPerFile - 1) / kNumTimestamps; ++j) {
+            ASSERT_OK(Put(
+                cf, "key" + std::to_string(j),
+                "value_" + std::to_string(j) + "_" + std::to_string(i), wopts));
+          }
+          if (!memtable_only) {
+            ASSERT_OK(Flush(cf));
+          }
         }
       }
+      const auto& verify_db_func = [&]() {
+        for (size_t i = 0; i != kNumTimestamps; ++i) {
+          ReadOptions ropts;
+          ropts.timestamp = &read_ts_list[i];
+          for (int cf = 0; cf != static_cast<int>(num_cfs); ++cf) {
+            ColumnFamilyHandle* cfh = handles_[cf];
+            for (size_t j = 0; j != (kNumKeysPerFile - 1) / kNumTimestamps;
+                 ++j) {
+              std::string value;
+              ASSERT_OK(
+                  db_->Get(ropts, cfh, "key" + std::to_string(j), &value));
+              ASSERT_EQ("value_" + std::to_string(j) + "_" + std::to_string(i),
+                        value);
+            }
+          }
+        }
+      };
+      verify_db_func();
     }
-  };
-  verify_db_func();
+  }
 }
 
 INSTANTIATE_TEST_CASE_P(Timestamp, DBBasicTestWithTimestampWithParam,

db/db_blob_index_test.cc

@@ -398,6 +398,29 @@ TEST_F(DBBlobIndexTest, Iterate) {
     verify(15, Status::kOk, get_value(16, 0), get_value(14, 0),
            create_blob_iterator, check_is_blob(false));
 
+#ifndef ROCKSDB_LITE
+    // Iterator with blob support and using seek.
+    ASSERT_OK(dbfull()->SetOptions(
+        cfh(), {{"max_sequential_skip_in_iterations", "0"}}));
+    verify(1, Status::kOk, get_value(1, 0), get_value(1, 0),
+           create_blob_iterator, check_is_blob(true));
+    verify(3, Status::kOk, get_value(3, 0), get_value(3, 0),
+           create_blob_iterator, check_is_blob(true));
+    verify(5, Status::kOk, get_value(5, 0), get_value(5, 0),
+           create_blob_iterator, check_is_blob(false));
+    verify(7, Status::kOk, get_value(8, 0), get_value(6, 0),
+           create_blob_iterator, check_is_blob(false));
+    verify(9, Status::kOk, get_value(10, 0), get_value(8, 0),
+           create_blob_iterator, check_is_blob(false));
+    verify(11, Status::kNotSupported, "", "", create_blob_iterator);
+    verify(13, Status::kOk,
+           get_value(13, 2) + "," + get_value(13, 1) + "," + get_value(13, 0),
+           get_value(13, 2) + "," + get_value(13, 1) + "," + get_value(13, 0),
+           create_blob_iterator, check_is_blob(false));
+    verify(15, Status::kOk, get_value(16, 0), get_value(14, 0),
+           create_blob_iterator, check_is_blob(false));
+#endif  // !ROCKSDB_LITE
+
     for (auto* snapshot : snapshots) {
       dbfull()->ReleaseSnapshot(snapshot);
     }

db/db_block_cache_test.cc

@@ -10,6 +10,7 @@
 #include "cache/lru_cache.h"
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
+#include "util/compression.h"
 
 namespace rocksdb {
 
@@ -685,16 +686,18 @@ TEST_F(DBBlockCacheTest, CacheCompressionDict) {
 
   // Try all the available libraries that support dictionary compression
   std::vector<CompressionType> compression_types;
-#ifdef ZLIB
-  compression_types.push_back(kZlibCompression);
-#endif  // ZLIB
-#if LZ4_VERSION_NUMBER >= 10400
-  compression_types.push_back(kLZ4Compression);
-  compression_types.push_back(kLZ4HCCompression);
-#endif  // LZ4_VERSION_NUMBER >= 10400
-#if ZSTD_VERSION_NUMBER >= 500
-  compression_types.push_back(kZSTD);
-#endif  // ZSTD_VERSION_NUMBER >= 500
+  if (Zlib_Supported()) {
+    compression_types.push_back(kZlibCompression);
+  }
+  if (LZ4_Supported()) {
+    compression_types.push_back(kLZ4Compression);
+    compression_types.push_back(kLZ4HCCompression);
+  }
+  if (ZSTD_Supported()) {
+    compression_types.push_back(kZSTD);
+  } else if (ZSTDNotFinal_Supported()) {
+    compression_types.push_back(kZSTDNotFinalCompression);
+  }
   Random rnd(301);
   for (auto compression_type : compression_types) {
     Options options = CurrentOptions();

db/db_bloom_filter_test.cc

@@ -10,9 +10,20 @@
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
 #include "rocksdb/perf_context.h"
+#include "table/block_based/filter_policy_internal.h"
 
 namespace rocksdb {
 
+namespace {
+using BFP = BloomFilterPolicy;
+
+namespace BFP2 {
+// Extends BFP::Mode with option to use Plain table
+using PseudoMode = int;
+static constexpr PseudoMode kPlainTable = -1;
+}  // namespace BFP2
+}  // namespace
+
 // DB tests related to bloom filter.
 
 class DBBloomFilterTest : public DBTestBase {
@@ -20,12 +31,12 @@ class DBBloomFilterTest : public DBTestBase {
   DBBloomFilterTest() : DBTestBase("/db_bloom_filter_test") {}
 };
 
-class DBBloomFilterTestWithParam
-    : public DBTestBase,
-      public testing::WithParamInterface<std::tuple<bool, bool, uint32_t>> {
+class DBBloomFilterTestWithParam : public DBTestBase,
+                                   public testing::WithParamInterface<
+                                       std::tuple<BFP::Mode, bool, uint32_t>> {
   //                             public testing::WithParamInterface<bool> {
  protected:
-  bool use_block_based_filter_;
+  BFP::Mode bfp_impl_;
   bool partition_filters_;
   uint32_t format_version_;
 
@@ -35,7 +46,7 @@ class DBBloomFilterTestWithParam
   ~DBBloomFilterTestWithParam() override {}
 
   void SetUp() override {
-    use_block_based_filter_ = std::get<0>(GetParam());
+    bfp_impl_ = std::get<0>(GetParam());
     partition_filters_ = std::get<1>(GetParam());
     format_version_ = std::get<2>(GetParam());
   }
@@ -71,8 +82,7 @@ TEST_P(DBBloomFilterTestDefFormatVersion, KeyMayExist) {
     ReadOptions ropts;
     std::string value;
     anon::OptionsOverride options_override;
-    options_override.filter_policy.reset(
-        NewBloomFilterPolicy(20, use_block_based_filter_));
+    options_override.filter_policy.reset(new BFP(20, bfp_impl_));
     options_override.partition_filters = partition_filters_;
     options_override.metadata_block_size = 32;
     Options options = CurrentOptions(options_override);
@@ -432,8 +442,7 @@ TEST_P(DBBloomFilterTestWithParam, BloomFilter) {
     // trigger reset of table_factory
     BlockBasedTableOptions table_options;
     table_options.no_block_cache = true;
-    table_options.filter_policy.reset(
-        NewBloomFilterPolicy(10, use_block_based_filter_));
+    table_options.filter_policy.reset(new BFP(10, bfp_impl_));
     table_options.partition_filters = partition_filters_;
     if (partition_filters_) {
       table_options.index_type =
@@ -502,24 +511,27 @@ TEST_P(DBBloomFilterTestWithParam, BloomFilter) {
 #ifndef ROCKSDB_VALGRIND_RUN
 INSTANTIATE_TEST_CASE_P(
     FormatDef, DBBloomFilterTestDefFormatVersion,
-    ::testing::Values(std::make_tuple(true, false, test::kDefaultFormatVersion),
-                      std::make_tuple(false, true, test::kDefaultFormatVersion),
-                      std::make_tuple(false, false,
-                                      test::kDefaultFormatVersion)));
+    ::testing::Values(
+        std::make_tuple(BFP::kDeprecatedBlock, false,
+                        test::kDefaultFormatVersion),
+        std::make_tuple(BFP::kAuto, true, test::kDefaultFormatVersion),
+        std::make_tuple(BFP::kAuto, false, test::kDefaultFormatVersion)));
 
 INSTANTIATE_TEST_CASE_P(
     FormatDef, DBBloomFilterTestWithParam,
-    ::testing::Values(std::make_tuple(true, false, test::kDefaultFormatVersion),
-                      std::make_tuple(false, true, test::kDefaultFormatVersion),
-                      std::make_tuple(false, false,
-                                      test::kDefaultFormatVersion)));
+    ::testing::Values(
+        std::make_tuple(BFP::kDeprecatedBlock, false,
+                        test::kDefaultFormatVersion),
+        std::make_tuple(BFP::kAuto, true, test::kDefaultFormatVersion),
+        std::make_tuple(BFP::kAuto, false, test::kDefaultFormatVersion)));
 
 INSTANTIATE_TEST_CASE_P(
     FormatLatest, DBBloomFilterTestWithParam,
-    ::testing::Values(std::make_tuple(true, false, test::kLatestFormatVersion),
-                      std::make_tuple(false, true, test::kLatestFormatVersion),
-                      std::make_tuple(false, false,
-                                      test::kLatestFormatVersion)));
+    ::testing::Values(
+        std::make_tuple(BFP::kDeprecatedBlock, false,
+                        test::kLatestFormatVersion),
+        std::make_tuple(BFP::kAuto, true, test::kLatestFormatVersion),
+        std::make_tuple(BFP::kAuto, false, test::kLatestFormatVersion)));
 #endif  // ROCKSDB_VALGRIND_RUN
 
 TEST_F(DBBloomFilterTest, BloomFilterRate) {
@@ -636,15 +648,17 @@ TEST_F(DBBloomFilterTest, BloomFilterReverseCompatibility) {
 }
 
 namespace {
-// A wrapped bloom over default FilterPolicy
-class WrappedBloom : public FilterPolicy {
+// A wrapped bloom over block-based FilterPolicy
+class TestingWrappedBlockBasedFilterPolicy : public FilterPolicy {
  public:
-  explicit WrappedBloom(int bits_per_key)
-      : filter_(NewBloomFilterPolicy(bits_per_key)), counter_(0) {}
+  explicit TestingWrappedBlockBasedFilterPolicy(int bits_per_key)
+      : filter_(NewBloomFilterPolicy(bits_per_key, true)), counter_(0) {}
 
-  ~WrappedBloom() override { delete filter_; }
+  ~TestingWrappedBlockBasedFilterPolicy() override { delete filter_; }
 
-  const char* Name() const override { return "WrappedRocksDbFilterPolicy"; }
+  const char* Name() const override {
+    return "TestingWrappedBlockBasedFilterPolicy";
+  }
 
   void CreateFilter(const rocksdb::Slice* keys, int n,
                     std::string* dst) const override {
@@ -671,12 +685,13 @@ class WrappedBloom : public FilterPolicy {
 };
 }  // namespace
 
-TEST_F(DBBloomFilterTest, BloomFilterWrapper) {
+TEST_F(DBBloomFilterTest, WrappedBlockBasedFilterPolicy) {
   Options options = CurrentOptions();
   options.statistics = rocksdb::CreateDBStatistics();
 
   BlockBasedTableOptions table_options;
-  WrappedBloom* policy = new WrappedBloom(10);
+  TestingWrappedBlockBasedFilterPolicy* policy =
+      new TestingWrappedBlockBasedFilterPolicy(10);
   table_options.filter_policy.reset(policy);
   options.table_factory.reset(NewBlockBasedTableFactory(table_options));
 
@@ -706,6 +721,166 @@ TEST_F(DBBloomFilterTest, BloomFilterWrapper) {
   ASSERT_EQ(2U * maxKey, policy->GetCounter());
 }
 
+namespace {
+// NOTE: This class is referenced by HISTORY.md as a model for a wrapper
+// FilterPolicy selecting among configurations based on context.
+class LevelAndStyleCustomFilterPolicy : public FilterPolicy {
+ public:
+  explicit LevelAndStyleCustomFilterPolicy(int bpk_fifo, int bpk_l0_other,
+                                           int bpk_otherwise)
+      : policy_fifo_(NewBloomFilterPolicy(bpk_fifo)),
+        policy_l0_other_(NewBloomFilterPolicy(bpk_l0_other)),
+        policy_otherwise_(NewBloomFilterPolicy(bpk_otherwise)) {}
+
+  // OK to use built-in policy name because we are deferring to a
+  // built-in builder. We aren't changing the serialized format.
+  const char* Name() const override { return policy_fifo_->Name(); }
+
+  FilterBitsBuilder* GetBuilderWithContext(
+      const FilterBuildingContext& context) const override {
+    if (context.compaction_style == kCompactionStyleFIFO) {
+      return policy_fifo_->GetBuilderWithContext(context);
+    } else if (context.level_at_creation == 0) {
+      return policy_l0_other_->GetBuilderWithContext(context);
+    } else {
+      return policy_otherwise_->GetBuilderWithContext(context);
+    }
+  }
+
+  FilterBitsReader* GetFilterBitsReader(const Slice& contents) const override {
+    // OK to defer to any of them; they all can parse built-in filters
+    // from any settings.
+    return policy_fifo_->GetFilterBitsReader(contents);
+  }
+
+  // Defer just in case configuration uses block-based filter
+  void CreateFilter(const Slice* keys, int n, std::string* dst) const override {
+    policy_otherwise_->CreateFilter(keys, n, dst);
+  }
+  bool KeyMayMatch(const Slice& key, const Slice& filter) const override {
+    return policy_otherwise_->KeyMayMatch(key, filter);
+  }
+
+ private:
+  const std::unique_ptr<const FilterPolicy> policy_fifo_;
+  const std::unique_ptr<const FilterPolicy> policy_l0_other_;
+  const std::unique_ptr<const FilterPolicy> policy_otherwise_;
+};
+
+class TestingContextCustomFilterPolicy
+    : public LevelAndStyleCustomFilterPolicy {
+ public:
+  explicit TestingContextCustomFilterPolicy(int bpk_fifo, int bpk_l0_other,
+                                            int bpk_otherwise)
+      : LevelAndStyleCustomFilterPolicy(bpk_fifo, bpk_l0_other, bpk_otherwise) {
+  }
+
+  FilterBitsBuilder* GetBuilderWithContext(
+      const FilterBuildingContext& context) const override {
+    test_report_ += "cf=";
+    test_report_ += context.column_family_name;
+    test_report_ += ",cs=";
+    test_report_ +=
+        OptionsHelper::compaction_style_to_string[context.compaction_style];
+    test_report_ += ",lv=";
+    test_report_ += std::to_string(context.level_at_creation);
+    test_report_ += "\n";
+
+    return LevelAndStyleCustomFilterPolicy::GetBuilderWithContext(context);
+  }
+
+  std::string DumpTestReport() {
+    std::string rv;
+    std::swap(rv, test_report_);
+    return rv;
+  }
+
+ private:
+  mutable std::string test_report_;
+};
+}  // namespace
+
+TEST_F(DBBloomFilterTest, ContextCustomFilterPolicy) {
+  for (bool fifo : {true, false}) {
+    Options options = CurrentOptions();
+    options.statistics = rocksdb::CreateDBStatistics();
+    options.compaction_style =
+        fifo ? kCompactionStyleFIFO : kCompactionStyleLevel;
+
+    BlockBasedTableOptions table_options;
+    auto policy = std::make_shared<TestingContextCustomFilterPolicy>(15, 8, 5);
+    table_options.filter_policy = policy;
+    table_options.format_version = 5;
+    options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+
+    CreateAndReopenWithCF({fifo ? "abe" : "bob"}, options);
+
+    const int maxKey = 10000;
+    for (int i = 0; i < maxKey / 2; i++) {
+      ASSERT_OK(Put(1, Key(i), Key(i)));
+    }
+    // Add a large key to make the file contain wide range
+    ASSERT_OK(Put(1, Key(maxKey + 55555), Key(maxKey + 55555)));
+    Flush(1);
+    EXPECT_EQ(policy->DumpTestReport(),
+              fifo ? "cf=abe,cs=kCompactionStyleFIFO,lv=0\n"
+                   : "cf=bob,cs=kCompactionStyleLevel,lv=0\n");
+
+    for (int i = maxKey / 2; i < maxKey; i++) {
+      ASSERT_OK(Put(1, Key(i), Key(i)));
+    }
+    Flush(1);
+    EXPECT_EQ(policy->DumpTestReport(),
+              fifo ? "cf=abe,cs=kCompactionStyleFIFO,lv=0\n"
+                   : "cf=bob,cs=kCompactionStyleLevel,lv=0\n");
+
+    // Check that they can be found
+    for (int i = 0; i < maxKey; i++) {
+      ASSERT_EQ(Key(i), Get(1, Key(i)));
+    }
+    // Since we have two tables / two filters, we might have Bloom checks on
+    // our queries, but no more than one "useful" per query on a found key.
+    EXPECT_LE(TestGetAndResetTickerCount(options, BLOOM_FILTER_USEFUL), maxKey);
+
+    // Check that we have two filters, each about
+    // fifo: 0.12% FP rate (15 bits per key)
+    // level: 2.3% FP rate (8 bits per key)
+    for (int i = 0; i < maxKey; i++) {
+      ASSERT_EQ("NOT_FOUND", Get(1, Key(i + 33333)));
+    }
+    {
+      auto useful_count =
+          TestGetAndResetTickerCount(options, BLOOM_FILTER_USEFUL);
+      EXPECT_GE(useful_count, maxKey * 2 * (fifo ? 0.9980 : 0.975));
+      EXPECT_LE(useful_count, maxKey * 2 * (fifo ? 0.9995 : 0.98));
+    }
+
+    if (!fifo) {  // FIFO only has L0
+      // Full compaction
+      ASSERT_OK(db_->CompactRange(CompactRangeOptions(), handles_[1], nullptr,
+                                  nullptr));
+      EXPECT_EQ(policy->DumpTestReport(),
+                "cf=bob,cs=kCompactionStyleLevel,lv=1\n");
+
+      // Check that we now have one filter, about 9.2% FP rate (5 bits per key)
+      for (int i = 0; i < maxKey; i++) {
+        ASSERT_EQ("NOT_FOUND", Get(1, Key(i + 33333)));
+      }
+      {
+        auto useful_count =
+            TestGetAndResetTickerCount(options, BLOOM_FILTER_USEFUL);
+        EXPECT_GE(useful_count, maxKey * 0.90);
+        EXPECT_LE(useful_count, maxKey * 0.91);
+      }
+    }
+
+    // Destroy
+    ASSERT_OK(dbfull()->DropColumnFamily(handles_[1]));
+    dbfull()->DestroyColumnFamilyHandle(handles_[1]);
+    handles_[1] = nullptr;
+  }
+}
+
 class SliceTransformLimitedDomain : public SliceTransform {
   const char* Name() const override { return "SliceTransformLimitedDomain"; }
 
@@ -858,33 +1033,32 @@ TEST_F(DBBloomFilterTest, MemtablePrefixBloomOutOfDomain) {
 #ifndef ROCKSDB_LITE
 class BloomStatsTestWithParam
     : public DBBloomFilterTest,
-      public testing::WithParamInterface<std::tuple<bool, bool, bool>> {
+      public testing::WithParamInterface<std::tuple<BFP2::PseudoMode, bool>> {
  public:
   BloomStatsTestWithParam() {
-    use_block_table_ = std::get<0>(GetParam());
-    use_block_based_builder_ = std::get<1>(GetParam());
-    partition_filters_ = std::get<2>(GetParam());
+    bfp_impl_ = std::get<0>(GetParam());
+    partition_filters_ = std::get<1>(GetParam());
 
     options_.create_if_missing = true;
     options_.prefix_extractor.reset(rocksdb::NewFixedPrefixTransform(4));
     options_.memtable_prefix_bloom_size_ratio =
         8.0 * 1024.0 / static_cast<double>(options_.write_buffer_size);
-    if (use_block_table_) {
+    if (bfp_impl_ == BFP2::kPlainTable) {
+      assert(!partition_filters_);  // not supported in plain table
+      PlainTableOptions table_options;
+      options_.table_factory.reset(NewPlainTableFactory(table_options));
+    } else {
       BlockBasedTableOptions table_options;
       table_options.hash_index_allow_collision = false;
       if (partition_filters_) {
-        assert(!use_block_based_builder_);
+        assert(bfp_impl_ != BFP::kDeprecatedBlock);
         table_options.partition_filters = partition_filters_;
         table_options.index_type =
             BlockBasedTableOptions::IndexType::kTwoLevelIndexSearch;
       }
       table_options.filter_policy.reset(
-          NewBloomFilterPolicy(10, use_block_based_builder_));
+          new BFP(10, static_cast<BFP::Mode>(bfp_impl_)));
       options_.table_factory.reset(NewBlockBasedTableFactory(table_options));
-    } else {
-      assert(!partition_filters_);  // not supported in plain table
-      PlainTableOptions table_options;
-      options_.table_factory.reset(NewPlainTableFactory(table_options));
     }
     options_.env = env_;
 
@@ -901,8 +1075,7 @@ class BloomStatsTestWithParam
   static void SetUpTestCase() {}
   static void TearDownTestCase() {}
 
-  bool use_block_table_;
-  bool use_block_based_builder_;
+  BFP2::PseudoMode bfp_impl_;
   bool partition_filters_;
   Options options_;
 };
@@ -1006,7 +1179,7 @@ TEST_P(BloomStatsTestWithParam, BloomStatsTestWithIter) {
   ASSERT_EQ(value3, iter->value().ToString());
   // The seek doesn't check block-based bloom filter because last index key
   // starts with the same prefix we're seeking to.
-  uint64_t expected_hits = use_block_based_builder_ ? 1 : 2;
+  uint64_t expected_hits = bfp_impl_ == BFP::kDeprecatedBlock ? 1 : 2;
   ASSERT_EQ(expected_hits, get_perf_context()->bloom_sst_hit_count);
 
   iter->Seek(key2);
@@ -1016,12 +1189,14 @@ TEST_P(BloomStatsTestWithParam, BloomStatsTestWithIter) {
   ASSERT_EQ(expected_hits, get_perf_context()->bloom_sst_hit_count);
 }
 
-INSTANTIATE_TEST_CASE_P(BloomStatsTestWithParam, BloomStatsTestWithParam,
-                        ::testing::Values(std::make_tuple(true, true, false),
-                                          std::make_tuple(true, false, false),
-                                          std::make_tuple(true, false, true),
-                                          std::make_tuple(false, false,
-                                                          false)));
+INSTANTIATE_TEST_CASE_P(
+    BloomStatsTestWithParam, BloomStatsTestWithParam,
+    ::testing::Values(std::make_tuple(BFP::kDeprecatedBlock, false),
+                      std::make_tuple(BFP::kLegacyBloom, false),
+                      std::make_tuple(BFP::kLegacyBloom, true),
+                      std::make_tuple(BFP::kFastLocalBloom, false),
+                      std::make_tuple(BFP::kFastLocalBloom, true),
+                      std::make_tuple(BFP2::kPlainTable, false)));
 
 namespace {
 void PrefixScanInit(DBBloomFilterTest* dbtest) {
@@ -1147,6 +1322,7 @@ TEST_F(DBBloomFilterTest, OptimizeFiltersForHits) {
   options.table_factory.reset(NewBlockBasedTableFactory(bbto));
   options.optimize_filters_for_hits = true;
   options.statistics = rocksdb::CreateDBStatistics();
+  get_perf_context()->Reset();
   get_perf_context()->EnablePerLevelPerfContext();
   CreateAndReopenWithCF({"mypikachu"}, options);
 
@@ -1328,8 +1504,8 @@ int CountIter(std::unique_ptr<Iterator>& iter, const Slice& key) {
 // into the same string, or 2) the transformed seek key is of the same length
 // as the upper bound and two keys are adjacent according to the comparator.
 TEST_F(DBBloomFilterTest, DynamicBloomFilterUpperBound) {
-  int iteration = 0;
-  for (bool use_block_based_builder : {true, false}) {
+  for (auto bfp_impl : BFP::kAllFixedImpls) {
+    int using_full_builder = bfp_impl != BFP::kDeprecatedBlock;
     Options options;
     options.create_if_missing = true;
     options.prefix_extractor.reset(NewCappedPrefixTransform(4));
@@ -1338,8 +1514,7 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterUpperBound) {
     // Enable prefix bloom for SST files
     BlockBasedTableOptions table_options;
     table_options.cache_index_and_filter_blocks = true;
-    table_options.filter_policy.reset(
-        NewBloomFilterPolicy(10, use_block_based_builder));
+    table_options.filter_policy.reset(new BFP(10, bfp_impl));
     table_options.index_shortening = BlockBasedTableOptions::
         IndexShorteningMode::kShortenSeparatorsAndSuccessor;
     options.table_factory.reset(NewBlockBasedTableFactory(table_options));
@@ -1382,7 +1557,7 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterUpperBound) {
       ASSERT_EQ(CountIter(iter, "abcdxx00"), 4);
       // should check bloom filter since upper bound meets requirement
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-                2 + iteration);
+                2 + using_full_builder);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 0);
     }
     {
@@ -1396,7 +1571,7 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterUpperBound) {
       ASSERT_EQ(CountIter(iter, "abcdxx01"), 4);
       // should skip bloom filter since upper bound is too long
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-                2 + iteration);
+                2 + using_full_builder);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 0);
     }
     {
@@ -1410,7 +1585,7 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterUpperBound) {
       // should check bloom filter since upper bound matches transformed seek
       // key
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-                2 + iteration * 2);
+                2 + using_full_builder * 2);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 0);
     }
     {
@@ -1424,7 +1599,7 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterUpperBound) {
       ASSERT_EQ(CountIter(iter, "aaaaaaaa"), 0);
       // should skip bloom filter since mismatch is found
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-                2 + iteration * 2);
+                2 + using_full_builder * 2);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 0);
     }
     ASSERT_OK(dbfull()->SetOptions({{"prefix_extractor", "fixed:3"}}));
@@ -1438,7 +1613,7 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterUpperBound) {
       std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
       ASSERT_EQ(CountIter(iter, "abc"), 4);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-                2 + iteration * 2);
+                2 + using_full_builder * 2);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 0);
     }
     ASSERT_OK(dbfull()->SetOptions({{"prefix_extractor", "capped:4"}}));
@@ -1451,18 +1626,17 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterUpperBound) {
       std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
       ASSERT_EQ(CountIter(iter, "abc"), 0);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-                3 + iteration * 2);
+                3 + using_full_builder * 2);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 1);
     }
-    iteration++;
   }
 }
 
 // Create multiple SST files each with a different prefix_extractor config,
 // verify iterators can read all SST files using the latest config.
 TEST_F(DBBloomFilterTest, DynamicBloomFilterMultipleSST) {
-  int iteration = 0;
-  for (bool use_block_based_builder : {true, false}) {
+  for (auto bfp_impl : BFP::kAllFixedImpls) {
+    int using_full_builder = bfp_impl != BFP::kDeprecatedBlock;
     Options options;
     options.create_if_missing = true;
     options.prefix_extractor.reset(NewFixedPrefixTransform(1));
@@ -1470,8 +1644,7 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterMultipleSST) {
     options.statistics = CreateDBStatistics();
     // Enable prefix bloom for SST files
     BlockBasedTableOptions table_options;
-    table_options.filter_policy.reset(
-        NewBloomFilterPolicy(10, use_block_based_builder));
+    table_options.filter_policy.reset(new BFP(10, bfp_impl));
     table_options.cache_index_and_filter_blocks = true;
     options.table_factory.reset(NewBlockBasedTableFactory(table_options));
     DestroyAndReopen(options);
@@ -1497,10 +1670,10 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterMultipleSST) {
     std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
     ASSERT_EQ(CountIter(iter, "foo"), 2);
     ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-              1 + iteration);
+              1 + using_full_builder);
     ASSERT_EQ(CountIter(iter, "gpk"), 0);
     ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-              1 + iteration);
+              1 + using_full_builder);
     ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 0);
 
     // second SST with capped:3 BF
@@ -1514,13 +1687,13 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterMultipleSST) {
       std::unique_ptr<Iterator> iter_tmp(db_->NewIterator(read_options));
       ASSERT_EQ(CountIter(iter_tmp, "foo"), 4);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-                2 + iteration * 2);
+                2 + using_full_builder * 2);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 0);
       ASSERT_EQ(CountIter(iter_tmp, "gpk"), 0);
       // both counters are incremented because BF is "not changed" for 1 of the
       // 2 SST files, so filter is checked once and found no match.
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-                3 + iteration * 2);
+                3 + using_full_builder * 2);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 1);
     }
 
@@ -1539,24 +1712,24 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterMultipleSST) {
       ASSERT_EQ(CountIter(iter_tmp, "foo"), 9);
       // the first and last BF are checked
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-                4 + iteration * 3);
+                4 + using_full_builder * 3);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 1);
       ASSERT_EQ(CountIter(iter_tmp, "gpk"), 0);
       // only last BF is checked and not found
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-                5 + iteration * 3);
+                5 + using_full_builder * 3);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 2);
     }
 
     // iter_old can only see the first SST, so checked plus 1
     ASSERT_EQ(CountIter(iter_old, "foo"), 4);
     ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-              6 + iteration * 3);
+              6 + using_full_builder * 3);
     // iter was created after the first setoptions call so only full filter
     // will check the filter
     ASSERT_EQ(CountIter(iter, "foo"), 2);
     ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-              6 + iteration * 4);
+              6 + using_full_builder * 4);
 
     {
       // keys in all three SSTs are visible to iterator
@@ -1565,11 +1738,11 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterMultipleSST) {
       std::unique_ptr<Iterator> iter_all(db_->NewIterator(read_options));
       ASSERT_EQ(CountIter(iter_all, "foo"), 9);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-                7 + iteration * 5);
+                7 + using_full_builder * 5);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 2);
       ASSERT_EQ(CountIter(iter_all, "gpk"), 0);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-                8 + iteration * 5);
+                8 + using_full_builder * 5);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 3);
     }
     ASSERT_OK(dbfull()->SetOptions({{"prefix_extractor", "capped:3"}}));
@@ -1581,15 +1754,14 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterMultipleSST) {
       // all three SST are checked because the current options has the same as
       // the remaining SST (capped:3)
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-                9 + iteration * 7);
+                9 + using_full_builder * 7);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 3);
       ASSERT_EQ(CountIter(iter_all, "gpk"), 0);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED),
-                10 + iteration * 7);
+                10 + using_full_builder * 7);
       ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 4);
     }
     // TODO(Zhongyi): Maybe also need to add Get calls to test point look up?
-    iteration++;
   }
 }
 
@@ -1598,7 +1770,7 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterMultipleSST) {
 // as expected
 TEST_F(DBBloomFilterTest, DynamicBloomFilterNewColumnFamily) {
   int iteration = 0;
-  for (bool use_block_based_builder : {true, false}) {
+  for (auto bfp_impl : BFP::kAllFixedImpls) {
     Options options = CurrentOptions();
     options.create_if_missing = true;
     options.prefix_extractor.reset(NewFixedPrefixTransform(1));
@@ -1607,8 +1779,7 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterNewColumnFamily) {
     // Enable prefix bloom for SST files
     BlockBasedTableOptions table_options;
     table_options.cache_index_and_filter_blocks = true;
-    table_options.filter_policy.reset(
-        NewBloomFilterPolicy(10, use_block_based_builder));
+    table_options.filter_policy.reset(new BFP(10, bfp_impl));
     options.table_factory.reset(NewBlockBasedTableFactory(table_options));
     CreateAndReopenWithCF({"pikachu" + std::to_string(iteration)}, options);
     ReadOptions read_options;
@@ -1657,8 +1828,7 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterNewColumnFamily) {
 // Verify it's possible to change prefix_extractor at runtime and iterators
 // behaves as expected
 TEST_F(DBBloomFilterTest, DynamicBloomFilterOptions) {
-  int iteration = 0;
-  for (bool use_block_based_builder : {true, false}) {
+  for (auto bfp_impl : BFP::kAllFixedImpls) {
     Options options;
     options.create_if_missing = true;
     options.prefix_extractor.reset(NewFixedPrefixTransform(1));
@@ -1667,8 +1837,7 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterOptions) {
     // Enable prefix bloom for SST files
     BlockBasedTableOptions table_options;
     table_options.cache_index_and_filter_blocks = true;
-    table_options.filter_policy.reset(
-        NewBloomFilterPolicy(10, use_block_based_builder));
+    table_options.filter_policy.reset(new BFP(10, bfp_impl));
     options.table_factory.reset(NewBlockBasedTableFactory(table_options));
     DestroyAndReopen(options);
 
@@ -1719,7 +1888,6 @@ TEST_F(DBBloomFilterTest, DynamicBloomFilterOptions) {
     ASSERT_EQ(CountIter(iter_old, "abc"), 0);
     ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_CHECKED), 12);
     ASSERT_EQ(TestGetTickerCount(options, BLOOM_FILTER_PREFIX_USEFUL), 3);
-    iteration++;
   }
 }
 

db/db_compaction_test.cc

@@ -12,6 +12,7 @@
 #include "port/stack_trace.h"
 #include "rocksdb/concurrent_task_limiter.h"
 #include "rocksdb/experimental.h"
+#include "rocksdb/sst_file_writer.h"
 #include "rocksdb/utilities/convenience.h"
 #include "test_util/fault_injection_test_env.h"
 #include "test_util/sync_point.h"
@@ -1502,6 +1503,37 @@ TEST_F(DBCompactionTest, DISABLED_ManualPartialFill) {
   }
 }
 
+TEST_F(DBCompactionTest, ManualCompactionWithUnorderedWrite) {
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      {{"DBImpl::WriteImpl:UnorderedWriteAfterWriteWAL",
+        "DBCompactionTest::ManualCompactionWithUnorderedWrite:WaitWriteWAL"},
+       {"DBImpl::WaitForPendingWrites:BeforeBlock",
+        "DBImpl::WriteImpl:BeforeUnorderedWriteMemtable"}});
+
+  Options options = CurrentOptions();
+  options.unordered_write = true;
+  DestroyAndReopen(options);
+  Put("foo", "v1");
+  ASSERT_OK(Flush());
+
+  Put("bar", "v1");
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+  port::Thread writer([&]() { Put("foo", "v2"); });
+
+  TEST_SYNC_POINT(
+      "DBCompactionTest::ManualCompactionWithUnorderedWrite:WaitWriteWAL");
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+
+  writer.join();
+  ASSERT_EQ(Get("foo"), "v2");
+
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+
+  Reopen(options);
+  ASSERT_EQ(Get("foo"), "v2");
+}
+
 TEST_F(DBCompactionTest, DeleteFileRange) {
   Options options = CurrentOptions();
   options.write_buffer_size = 10 * 1024 * 1024;
@@ -3521,76 +3553,238 @@ TEST_F(DBCompactionTest, LevelCompactExpiredTtlFiles) {
   rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }
 
+TEST_F(DBCompactionTest, LevelTtlCascadingCompactions) {
+  const int kValueSize = 100;
+
+  for (bool if_restart : {false, true}) {
+    for (bool if_open_all_files : {false, true}) {
+      Options options = CurrentOptions();
+      options.compression = kNoCompression;
+      options.ttl = 24 * 60 * 60;  // 24 hours
+      if (if_open_all_files) {
+        options.max_open_files = -1;
+      } else {
+        options.max_open_files = 20;
+      }
+      // RocksDB sanitize max open files to at least 20. Modify it back.
+      rocksdb::SyncPoint::GetInstance()->SetCallBack(
+          "SanitizeOptions::AfterChangeMaxOpenFiles", [&](void* arg) {
+            int* max_open_files = static_cast<int*>(arg);
+            *max_open_files = 2;
+          });
+      // In the case where all files are opened and doing DB restart
+      // forcing the oldest ancester time in manifest file to be 0 to
+      // simulate the case of reading from an old version.
+      rocksdb::SyncPoint::GetInstance()->SetCallBack(
+          "VersionEdit::EncodeTo:VarintOldestAncesterTime", [&](void* arg) {
+            if (if_restart && if_open_all_files) {
+              std::string* encoded_fieled = static_cast<std::string*>(arg);
+              *encoded_fieled = "";
+              PutVarint64(encoded_fieled, 0);
+            }
+          });
+
+      env_->time_elapse_only_sleep_ = false;
+      options.env = env_;
+
+      env_->addon_time_.store(0);
+      DestroyAndReopen(options);
+
+      int ttl_compactions = 0;
+      rocksdb::SyncPoint::GetInstance()->SetCallBack(
+          "LevelCompactionPicker::PickCompaction:Return", [&](void* arg) {
+            Compaction* compaction = reinterpret_cast<Compaction*>(arg);
+            auto compaction_reason = compaction->compaction_reason();
+            if (compaction_reason == CompactionReason::kTtl) {
+              ttl_compactions++;
+            }
+          });
+      rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+      // Add two L6 files with key ranges: [1 .. 100], [101 .. 200].
+      Random rnd(301);
+      for (int i = 1; i <= 100; ++i) {
+        ASSERT_OK(Put(Key(i), RandomString(&rnd, kValueSize)));
+      }
+      Flush();
+      for (int i = 101; i <= 200; ++i) {
+        ASSERT_OK(Put(Key(i), RandomString(&rnd, kValueSize)));
+      }
+      Flush();
+      MoveFilesToLevel(6);
+      ASSERT_EQ("0,0,0,0,0,0,2", FilesPerLevel());
+
+      // Add two L4 files with key ranges: [1 .. 50], [51 .. 150].
+      for (int i = 1; i <= 50; ++i) {
+        ASSERT_OK(Put(Key(i), RandomString(&rnd, kValueSize)));
+      }
+      Flush();
+      for (int i = 51; i <= 150; ++i) {
+        ASSERT_OK(Put(Key(i), RandomString(&rnd, kValueSize)));
+      }
+      Flush();
+      MoveFilesToLevel(4);
+      ASSERT_EQ("0,0,0,0,2,0,2", FilesPerLevel());
+
+      // Add one L1 file with key range: [26, 75].
+      for (int i = 26; i <= 75; ++i) {
+        ASSERT_OK(Put(Key(i), RandomString(&rnd, kValueSize)));
+      }
+      Flush();
+      dbfull()->TEST_WaitForCompact();
+      MoveFilesToLevel(1);
+      ASSERT_EQ("0,1,0,0,2,0,2", FilesPerLevel());
+
+      // LSM tree:
+      // L1:         [26 .. 75]
+      // L4:     [1 .. 50][51 ..... 150]
+      // L6:     [1 ........ 100][101 .... 200]
+      //
+      // On TTL expiry, TTL compaction should be initiated on L1 file, and the
+      // compactions should keep going on until the key range hits bottom level.
+      // In other words: the compaction on this data range "cascasdes" until
+      // reaching the bottom level.
+      //
+      // Order of events on TTL expiry:
+      // 1. L1 file falls to L3 via 2 trivial moves which are initiated by the
+      // ttl
+      //    compaction.
+      // 2. A TTL compaction happens between L3 and L4 files. Output file in L4.
+      // 3. The new output file from L4 falls to L5 via 1 trival move initiated
+      //    by the ttl compaction.
+      // 4. A TTL compaction happens between L5 and L6 files. Ouptut in L6.
+
+      // Add 25 hours and do a write
+      env_->addon_time_.fetch_add(25 * 60 * 60);
+
+      ASSERT_OK(Put(Key(1), "1"));
+      if (if_restart) {
+        Reopen(options);
+      } else {
+        Flush();
+      }
+      dbfull()->TEST_WaitForCompact();
+      ASSERT_EQ("1,0,0,0,0,0,1", FilesPerLevel());
+      ASSERT_EQ(5, ttl_compactions);
+
+      env_->addon_time_.fetch_add(25 * 60 * 60);
+      ASSERT_OK(Put(Key(2), "1"));
+      if (if_restart) {
+        Reopen(options);
+      } else {
+        Flush();
+      }
+      dbfull()->TEST_WaitForCompact();
+      ASSERT_EQ("1,0,0,0,0,0,1", FilesPerLevel());
+      ASSERT_GE(ttl_compactions, 6);
+
+      rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+    }
+  }
+}
+
 TEST_F(DBCompactionTest, LevelPeriodicCompaction) {
   const int kNumKeysPerFile = 32;
   const int kNumLevelFiles = 2;
   const int kValueSize = 100;
 
-  Options options = CurrentOptions();
-  options.periodic_compaction_seconds = 48 * 60 * 60;  // 2 days
-  options.max_open_files = -1;       // needed for ttl compaction
-  env_->time_elapse_only_sleep_ = false;
-  options.env = env_;
+  for (bool if_restart : {false, true}) {
+    for (bool if_open_all_files : {false, true}) {
+      Options options = CurrentOptions();
+      options.periodic_compaction_seconds = 48 * 60 * 60;  // 2 days
+      if (if_open_all_files) {
+        options.max_open_files = -1;  // needed for ttl compaction
+      } else {
+        options.max_open_files = 20;
+      }
+      // RocksDB sanitize max open files to at least 20. Modify it back.
+      rocksdb::SyncPoint::GetInstance()->SetCallBack(
+          "SanitizeOptions::AfterChangeMaxOpenFiles", [&](void* arg) {
+            int* max_open_files = static_cast<int*>(arg);
+            *max_open_files = 0;
+          });
+      // In the case where all files are opened and doing DB restart
+      // forcing the file creation time in manifest file to be 0 to
+      // simulate the case of reading from an old version.
+      rocksdb::SyncPoint::GetInstance()->SetCallBack(
+          "VersionEdit::EncodeTo:VarintFileCreationTime", [&](void* arg) {
+            if (if_restart && if_open_all_files) {
+              std::string* encoded_fieled = static_cast<std::string*>(arg);
+              *encoded_fieled = "";
+              PutVarint64(encoded_fieled, 0);
+            }
+          });
 
-  env_->addon_time_.store(0);
-  DestroyAndReopen(options);
+      env_->time_elapse_only_sleep_ = false;
+      options.env = env_;
 
-  int periodic_compactions = 0;
-  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "LevelCompactionPicker::PickCompaction:Return", [&](void* arg) {
-        Compaction* compaction = reinterpret_cast<Compaction*>(arg);
-        auto compaction_reason = compaction->compaction_reason();
-        if (compaction_reason == CompactionReason::kPeriodicCompaction) {
-          periodic_compactions++;
+      env_->addon_time_.store(0);
+      DestroyAndReopen(options);
+
+      int periodic_compactions = 0;
+      rocksdb::SyncPoint::GetInstance()->SetCallBack(
+          "LevelCompactionPicker::PickCompaction:Return", [&](void* arg) {
+            Compaction* compaction = reinterpret_cast<Compaction*>(arg);
+            auto compaction_reason = compaction->compaction_reason();
+            if (compaction_reason == CompactionReason::kPeriodicCompaction) {
+              periodic_compactions++;
+            }
+          });
+      rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+      Random rnd(301);
+      for (int i = 0; i < kNumLevelFiles; ++i) {
+        for (int j = 0; j < kNumKeysPerFile; ++j) {
+          ASSERT_OK(Put(Key(i * kNumKeysPerFile + j),
+                        RandomString(&rnd, kValueSize)));
         }
-      });
-  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+        Flush();
+      }
+      dbfull()->TEST_WaitForCompact();
 
-  Random rnd(301);
-  for (int i = 0; i < kNumLevelFiles; ++i) {
-    for (int j = 0; j < kNumKeysPerFile; ++j) {
-      ASSERT_OK(
-          Put(Key(i * kNumKeysPerFile + j), RandomString(&rnd, kValueSize)));
+      ASSERT_EQ("2", FilesPerLevel());
+      ASSERT_EQ(0, periodic_compactions);
+
+      // Add 50 hours and do a write
+      env_->addon_time_.fetch_add(50 * 60 * 60);
+      ASSERT_OK(Put("a", "1"));
+      Flush();
+      dbfull()->TEST_WaitForCompact();
+      // Assert that the files stay in the same level
+      ASSERT_EQ("3", FilesPerLevel());
+      // The two old files go through the periodic compaction process
+      ASSERT_EQ(2, periodic_compactions);
+
+      MoveFilesToLevel(1);
+      ASSERT_EQ("0,3", FilesPerLevel());
+
+      // Add another 50 hours and do another write
+      env_->addon_time_.fetch_add(50 * 60 * 60);
+      ASSERT_OK(Put("b", "2"));
+      if (if_restart) {
+        Reopen(options);
+      } else {
+        Flush();
+      }
+      dbfull()->TEST_WaitForCompact();
+      ASSERT_EQ("1,3", FilesPerLevel());
+      // The three old files now go through the periodic compaction process. 2
+      // + 3.
+      ASSERT_EQ(5, periodic_compactions);
+
+      // Add another 50 hours and do another write
+      env_->addon_time_.fetch_add(50 * 60 * 60);
+      ASSERT_OK(Put("c", "3"));
+      Flush();
+      dbfull()->TEST_WaitForCompact();
+      ASSERT_EQ("2,3", FilesPerLevel());
+      // The four old files now go through the periodic compaction process. 5
+      // + 4.
+      ASSERT_EQ(9, periodic_compactions);
+
+      rocksdb::SyncPoint::GetInstance()->DisableProcessing();
     }
-    Flush();
   }
-  dbfull()->TEST_WaitForCompact();
-
-  ASSERT_EQ("2", FilesPerLevel());
-  ASSERT_EQ(0, periodic_compactions);
-
-  // Add 50 hours and do a write
-  env_->addon_time_.fetch_add(50 * 60 * 60);
-  ASSERT_OK(Put("a", "1"));
-  Flush();
-  dbfull()->TEST_WaitForCompact();
-  // Assert that the files stay in the same level
-  ASSERT_EQ("3", FilesPerLevel());
-  // The two old files go through the periodic compaction process
-  ASSERT_EQ(2, periodic_compactions);
-
-  MoveFilesToLevel(1);
-  ASSERT_EQ("0,3", FilesPerLevel());
-
-  // Add another 50 hours and do another write
-  env_->addon_time_.fetch_add(50 * 60 * 60);
-  ASSERT_OK(Put("b", "2"));
-  Flush();
-  dbfull()->TEST_WaitForCompact();
-  ASSERT_EQ("1,3", FilesPerLevel());
-  // The three old files now go through the periodic compaction process. 2 + 3.
-  ASSERT_EQ(5, periodic_compactions);
-
-  // Add another 50 hours and do another write
-  env_->addon_time_.fetch_add(50 * 60 * 60);
-  ASSERT_OK(Put("c", "3"));
-  Flush();
-  dbfull()->TEST_WaitForCompact();
-  ASSERT_EQ("2,3", FilesPerLevel());
-  // The four old files now go through the periodic compaction process. 5 + 4.
-  ASSERT_EQ(9, periodic_compactions);
-
-  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }
 
 TEST_F(DBCompactionTest, LevelPeriodicCompactionWithOldDB) {
@@ -3603,7 +3797,6 @@ TEST_F(DBCompactionTest, LevelPeriodicCompactionWithOldDB) {
   const int kValueSize = 100;
 
   Options options = CurrentOptions();
-  options.max_open_files = -1;  // needed for ttl compaction
   env_->time_elapse_only_sleep_ = false;
   options.env = env_;
 
@@ -3748,6 +3941,91 @@ TEST_F(DBCompactionTest, LevelPeriodicAndTtlCompaction) {
   rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }
 
+TEST_F(DBCompactionTest, LevelPeriodicCompactionWithCompactionFilters) {
+  class TestCompactionFilter : public CompactionFilter {
+    const char* Name() const override { return "TestCompactionFilter"; }
+  };
+  class TestCompactionFilterFactory : public CompactionFilterFactory {
+    const char* Name() const override { return "TestCompactionFilterFactory"; }
+    std::unique_ptr<CompactionFilter> CreateCompactionFilter(
+        const CompactionFilter::Context& /*context*/) override {
+      return std::unique_ptr<CompactionFilter>(new TestCompactionFilter());
+    }
+  };
+
+  const int kNumKeysPerFile = 32;
+  const int kNumLevelFiles = 2;
+  const int kValueSize = 100;
+
+  Random rnd(301);
+
+  Options options = CurrentOptions();
+  TestCompactionFilter test_compaction_filter;
+  env_->time_elapse_only_sleep_ = false;
+  options.env = env_;
+  env_->addon_time_.store(0);
+
+  enum CompactionFilterType {
+    kUseCompactionFilter,
+    kUseCompactionFilterFactory
+  };
+
+  for (CompactionFilterType comp_filter_type :
+       {kUseCompactionFilter, kUseCompactionFilterFactory}) {
+    // Assert that periodic compactions are not enabled.
+    ASSERT_EQ(port::kMaxUint64 - 1, options.periodic_compaction_seconds);
+
+    if (comp_filter_type == kUseCompactionFilter) {
+      options.compaction_filter = &test_compaction_filter;
+      options.compaction_filter_factory.reset();
+    } else if (comp_filter_type == kUseCompactionFilterFactory) {
+      options.compaction_filter = nullptr;
+      options.compaction_filter_factory.reset(
+          new TestCompactionFilterFactory());
+    }
+    DestroyAndReopen(options);
+
+    // periodic_compaction_seconds should be set to the sanitized value when
+    // a compaction filter or a compaction filter factory is used.
+    ASSERT_EQ(30 * 24 * 60 * 60,
+              dbfull()->GetOptions().periodic_compaction_seconds);
+
+    int periodic_compactions = 0;
+    rocksdb::SyncPoint::GetInstance()->SetCallBack(
+        "LevelCompactionPicker::PickCompaction:Return", [&](void* arg) {
+          Compaction* compaction = reinterpret_cast<Compaction*>(arg);
+          auto compaction_reason = compaction->compaction_reason();
+          if (compaction_reason == CompactionReason::kPeriodicCompaction) {
+            periodic_compactions++;
+          }
+        });
+    rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+    for (int i = 0; i < kNumLevelFiles; ++i) {
+      for (int j = 0; j < kNumKeysPerFile; ++j) {
+        ASSERT_OK(
+            Put(Key(i * kNumKeysPerFile + j), RandomString(&rnd, kValueSize)));
+      }
+      Flush();
+    }
+    dbfull()->TEST_WaitForCompact();
+
+    ASSERT_EQ("2", FilesPerLevel());
+    ASSERT_EQ(0, periodic_compactions);
+
+    // Add 31 days and do a write
+    env_->addon_time_.fetch_add(31 * 24 * 60 * 60);
+    ASSERT_OK(Put("a", "1"));
+    Flush();
+    dbfull()->TEST_WaitForCompact();
+    // Assert that the files stay in the same level
+    ASSERT_EQ("3", FilesPerLevel());
+    // The two old files go through the periodic compaction process
+    ASSERT_EQ(2, periodic_compactions);
+
+    rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  }
+}
 
 TEST_F(DBCompactionTest, CompactRangeDelayedByL0FileCount) {
   // Verify that, when `CompactRangeOptions::allow_write_stall == false`, manual
@@ -4658,6 +4936,7 @@ TEST_P(DBCompactionTestWithParam, FixFileIngestionCompactionDeadlock) {
   ASSERT_OK(dbfull()->TEST_WaitForCompact());
   Close();
 }
+
 TEST_F(DBCompactionTest, ConsistencyFailTest) {
   Options options = CurrentOptions();
   DestroyAndReopen(options);
@@ -4683,6 +4962,176 @@ TEST_F(DBCompactionTest, ConsistencyFailTest) {
   ASSERT_NOK(Put("foo", "bar"));
   rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }
+
+void IngestOneKeyValue(DBImpl* db, const std::string& key,
+                       const std::string& value, const Options& options) {
+  ExternalSstFileInfo info;
+  std::string f = test::PerThreadDBPath("sst_file" + key);
+  EnvOptions env;
+  rocksdb::SstFileWriter writer(env, options);
+  auto s = writer.Open(f);
+  ASSERT_OK(s);
+  // ASSERT_OK(writer.Put(Key(), ""));
+  ASSERT_OK(writer.Put(key, value));
+
+  ASSERT_OK(writer.Finish(&info));
+  IngestExternalFileOptions ingest_opt;
+
+  ASSERT_OK(db->IngestExternalFile({info.file_path}, ingest_opt));
+}
+
+TEST_P(DBCompactionTestWithParam,
+       FlushAfterIntraL0CompactionCheckConsistencyFail) {
+  Options options = CurrentOptions();
+  options.force_consistency_checks = true;
+  options.compression = kNoCompression;
+  options.level0_file_num_compaction_trigger = 5;
+  options.max_background_compactions = 2;
+  options.max_subcompactions = max_subcompactions_;
+  DestroyAndReopen(options);
+
+  const size_t kValueSize = 1 << 20;
+  Random rnd(301);
+  std::atomic<int> pick_intra_l0_count(0);
+  std::string value(RandomString(&rnd, kValueSize));
+
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      {{"DBCompactionTestWithParam::FlushAfterIntraL0:1",
+        "CompactionJob::Run():Start"}});
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "FindIntraL0Compaction",
+      [&](void* /*arg*/) { pick_intra_l0_count.fetch_add(1); });
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  // prevents trivial move
+  for (int i = 0; i < 10; ++i) {
+    ASSERT_OK(Put(Key(i), ""));  // prevents trivial move
+  }
+  ASSERT_OK(Flush());
+  Compact("", Key(99));
+  ASSERT_EQ(0, NumTableFilesAtLevel(0));
+
+  // Flush 5 L0 sst.
+  for (int i = 0; i < 5; ++i) {
+    ASSERT_OK(Put(Key(i + 1), value));
+    ASSERT_OK(Flush());
+  }
+  ASSERT_EQ(5, NumTableFilesAtLevel(0));
+
+  // Put one key, to make smallest log sequence number in this memtable is less
+  // than sst which would be ingested in next step.
+  ASSERT_OK(Put(Key(0), "a"));
+
+  ASSERT_EQ(5, NumTableFilesAtLevel(0));
+
+  // Ingest 5 L0 sst. And this files would trigger PickIntraL0Compaction.
+  for (int i = 5; i < 10; i++) {
+    IngestOneKeyValue(dbfull(), Key(i), value, options);
+    ASSERT_EQ(i + 1, NumTableFilesAtLevel(0));
+  }
+
+  TEST_SYNC_POINT("DBCompactionTestWithParam::FlushAfterIntraL0:1");
+  // Put one key, to make biggest log sequence number in this memtable is bigger
+  // than sst which would be ingested in next step.
+  ASSERT_OK(Put(Key(2), "b"));
+  ASSERT_EQ(10, NumTableFilesAtLevel(0));
+  dbfull()->TEST_WaitForCompact();
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  std::vector<std::vector<FileMetaData>> level_to_files;
+  dbfull()->TEST_GetFilesMetaData(dbfull()->DefaultColumnFamily(),
+                                  &level_to_files);
+  ASSERT_GT(level_to_files[0].size(), 0);
+  ASSERT_GT(pick_intra_l0_count.load(), 0);
+
+  ASSERT_OK(Flush());
+}
+
+TEST_P(DBCompactionTestWithParam,
+       IntraL0CompactionAfterFlushCheckConsistencyFail) {
+  Options options = CurrentOptions();
+  options.force_consistency_checks = true;
+  options.compression = kNoCompression;
+  options.level0_file_num_compaction_trigger = 5;
+  options.max_background_compactions = 2;
+  options.max_subcompactions = max_subcompactions_;
+  options.write_buffer_size = 2 << 20;
+  options.max_write_buffer_number = 6;
+  DestroyAndReopen(options);
+
+  const size_t kValueSize = 1 << 20;
+  Random rnd(301);
+  std::string value(RandomString(&rnd, kValueSize));
+  std::string value2(RandomString(&rnd, kValueSize));
+  std::string bigvalue = value + value;
+
+  // prevents trivial move
+  for (int i = 0; i < 10; ++i) {
+    ASSERT_OK(Put(Key(i), ""));  // prevents trivial move
+  }
+  ASSERT_OK(Flush());
+  Compact("", Key(99));
+  ASSERT_EQ(0, NumTableFilesAtLevel(0));
+
+  std::atomic<int> pick_intra_l0_count(0);
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      {{"DBCompactionTestWithParam::IntraL0CompactionAfterFlush:1",
+        "CompactionJob::Run():Start"}});
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "FindIntraL0Compaction",
+      [&](void* /*arg*/) { pick_intra_l0_count.fetch_add(1); });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+  // Make 6 L0 sst.
+  for (int i = 0; i < 6; ++i) {
+    if (i % 2 == 0) {
+      IngestOneKeyValue(dbfull(), Key(i), value, options);
+    } else {
+      ASSERT_OK(Put(Key(i), value));
+      ASSERT_OK(Flush());
+    }
+  }
+
+  ASSERT_EQ(6, NumTableFilesAtLevel(0));
+
+  // Stop run flush job
+  env_->SetBackgroundThreads(1, Env::HIGH);
+  test::SleepingBackgroundTask sleeping_tasks;
+  env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_tasks,
+                 Env::Priority::HIGH);
+  sleeping_tasks.WaitUntilSleeping();
+
+  // Put many keys to make memtable request to flush
+  for (int i = 0; i < 6; ++i) {
+    ASSERT_OK(Put(Key(i), bigvalue));
+  }
+
+  ASSERT_EQ(6, NumTableFilesAtLevel(0));
+  // ingest file to trigger IntraL0Compaction
+  for (int i = 6; i < 10; ++i) {
+    ASSERT_EQ(i, NumTableFilesAtLevel(0));
+    IngestOneKeyValue(dbfull(), Key(i), value2, options);
+  }
+  ASSERT_EQ(10, NumTableFilesAtLevel(0));
+
+  // Wake up flush job
+  sleeping_tasks.WakeUp();
+  sleeping_tasks.WaitUntilDone();
+  TEST_SYNC_POINT("DBCompactionTestWithParam::IntraL0CompactionAfterFlush:1");
+  dbfull()->TEST_WaitForCompact();
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+
+  uint64_t error_count = 0;
+  db_->GetIntProperty("rocksdb.background-errors", &error_count);
+  ASSERT_EQ(error_count, 0);
+  ASSERT_GT(pick_intra_l0_count.load(), 0);
+  for (int i = 0; i < 6; ++i) {
+    ASSERT_EQ(bigvalue, Get(Key(i)));
+  }
+  for (int i = 6; i < 10; ++i) {
+    ASSERT_EQ(value2, Get(Key(i)));
+  }
+}
+
 #endif // !defined(ROCKSDB_LITE)
 }  // namespace rocksdb
 

db/db_filesnapshot.cc

@@ -102,7 +102,7 @@ Status DBImpl::GetLiveFiles(std::vector<std::string>& ret,
         TEST_SYNC_POINT("DBImpl::GetLiveFiles:1");
         TEST_SYNC_POINT("DBImpl::GetLiveFiles:2");
         mutex_.Lock();
-        cfd->Unref();
+        cfd->UnrefAndTryDelete();
         if (!status.ok()) {
           break;
         }

db/db_flush_test.cc

@@ -7,10 +7,16 @@
 // 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 <atomic>
+
+#include "db/db_impl/db_impl.h"
 #include "db/db_test_util.h"
+#include "port/port.h"
 #include "port/stack_trace.h"
 #include "test_util/fault_injection_test_env.h"
 #include "test_util/sync_point.h"
+#include "util/cast_util.h"
+#include "util/mutexlock.h"
 
 namespace rocksdb {
 
@@ -206,6 +212,30 @@ TEST_F(DBFlushTest, ManualFlushWithMinWriteBufferNumberToMerge) {
   t.join();
 }
 
+TEST_F(DBFlushTest, ScheduleOnlyOneBgThread) {
+  Options options = CurrentOptions();
+  Reopen(options);
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+  int called = 0;
+  SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::MaybeScheduleFlushOrCompaction:AfterSchedule:0", [&](void* arg) {
+        ASSERT_NE(nullptr, arg);
+        auto unscheduled_flushes = *reinterpret_cast<int*>(arg);
+        ASSERT_EQ(0, unscheduled_flushes);
+        ++called;
+      });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  ASSERT_OK(Put("a", "foo"));
+  FlushOptions flush_opts;
+  ASSERT_OK(dbfull()->Flush(flush_opts));
+  ASSERT_EQ(1, called);
+
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+}
+
 TEST_P(DBFlushDirectIOTest, DirectIO) {
   Options options;
   options.create_if_missing = true;
@@ -323,6 +353,96 @@ TEST_F(DBFlushTest, CFDropRaceWithWaitForFlushMemTables) {
   SyncPoint::GetInstance()->DisableProcessing();
 }
 
+#ifndef ROCKSDB_LITE
+TEST_F(DBFlushTest, FireOnFlushCompletedAfterCommittedResult) {
+  class TestListener : public EventListener {
+   public:
+    void OnFlushCompleted(DB* db, const FlushJobInfo& info) override {
+      // There's only one key in each flush.
+      ASSERT_EQ(info.smallest_seqno, info.largest_seqno);
+      ASSERT_NE(0, info.smallest_seqno);
+      if (info.smallest_seqno == seq1) {
+        // First flush completed
+        ASSERT_FALSE(completed1);
+        completed1 = true;
+        CheckFlushResultCommitted(db, seq1);
+      } else {
+        // Second flush completed
+        ASSERT_FALSE(completed2);
+        completed2 = true;
+        ASSERT_EQ(info.smallest_seqno, seq2);
+        CheckFlushResultCommitted(db, seq2);
+      }
+    }
+
+    void CheckFlushResultCommitted(DB* db, SequenceNumber seq) {
+      DBImpl* db_impl = static_cast_with_check<DBImpl>(db);
+      InstrumentedMutex* mutex = db_impl->mutex();
+      mutex->Lock();
+      auto* cfd =
+          reinterpret_cast<ColumnFamilyHandleImpl*>(db->DefaultColumnFamily())
+              ->cfd();
+      ASSERT_LT(seq, cfd->imm()->current()->GetEarliestSequenceNumber());
+      mutex->Unlock();
+    }
+
+    std::atomic<SequenceNumber> seq1{0};
+    std::atomic<SequenceNumber> seq2{0};
+    std::atomic<bool> completed1{false};
+    std::atomic<bool> completed2{false};
+  };
+  std::shared_ptr<TestListener> listener = std::make_shared<TestListener>();
+
+  SyncPoint::GetInstance()->LoadDependency(
+      {{"DBImpl::BackgroundCallFlush:start",
+        "DBFlushTest::FireOnFlushCompletedAfterCommittedResult:WaitFirst"},
+       {"DBImpl::FlushMemTableToOutputFile:Finish",
+        "DBFlushTest::FireOnFlushCompletedAfterCommittedResult:WaitSecond"}});
+  SyncPoint::GetInstance()->SetCallBack(
+      "FlushJob::WriteLevel0Table", [&listener](void* arg) {
+        // Wait for the second flush finished, out of mutex.
+        auto* mems = reinterpret_cast<autovector<MemTable*>*>(arg);
+        if (mems->front()->GetEarliestSequenceNumber() == listener->seq1 - 1) {
+          TEST_SYNC_POINT(
+              "DBFlushTest::FireOnFlushCompletedAfterCommittedResult:"
+              "WaitSecond");
+        }
+      });
+
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.listeners.push_back(listener);
+  // Setting max_flush_jobs = max_background_jobs / 4 = 2.
+  options.max_background_jobs = 8;
+  // Allow 2 immutable memtables.
+  options.max_write_buffer_number = 3;
+  Reopen(options);
+  SyncPoint::GetInstance()->EnableProcessing();
+  ASSERT_OK(Put("foo", "v"));
+  listener->seq1 = db_->GetLatestSequenceNumber();
+  // t1 will wait for the second flush complete before committing flush result.
+  auto t1 = port::Thread([&]() {
+    // flush_opts.wait = true
+    ASSERT_OK(db_->Flush(FlushOptions()));
+  });
+  // Wait for first flush started.
+  TEST_SYNC_POINT(
+      "DBFlushTest::FireOnFlushCompletedAfterCommittedResult:WaitFirst");
+  // The second flush will exit early without commit its result. The work
+  // is delegated to the first flush.
+  ASSERT_OK(Put("bar", "v"));
+  listener->seq2 = db_->GetLatestSequenceNumber();
+  FlushOptions flush_opts;
+  flush_opts.wait = false;
+  ASSERT_OK(db_->Flush(flush_opts));
+  t1.join();
+  ASSERT_TRUE(listener->completed1);
+  ASSERT_TRUE(listener->completed2);
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+}
+#endif  // !ROCKSDB_LITE
+
 TEST_P(DBAtomicFlushTest, ManualAtomicFlush) {
   Options options = CurrentOptions();
   options.create_if_missing = true;

db/db_impl/db_impl.h

@@ -199,11 +199,15 @@ class DBImpl : public DB {
                         PinnableSlice* values, Status* statuses,
                         const bool sorted_input = false) override;
 
-  void MultiGetImpl(
+  virtual void MultiGet(const ReadOptions& options, const size_t num_keys,
+                        ColumnFamilyHandle** column_families, const Slice* keys,
+                        PinnableSlice* values, Status* statuses,
+                        const bool sorted_input = false) override;
+
+  virtual void MultiGetWithCallback(
       const ReadOptions& options, ColumnFamilyHandle* column_family,
-      autovector<KeyContext, MultiGetContext::MAX_BATCH_SIZE>& key_context,
-      bool sorted_input, ReadCallback* callback = nullptr,
-      bool* is_blob_index = nullptr);
+      ReadCallback* callback,
+      autovector<KeyContext*, MultiGetContext::MAX_BATCH_SIZE>* sorted_keys);
 
   virtual Status CreateColumnFamily(const ColumnFamilyOptions& cf_options,
                                     const std::string& column_family,
@@ -302,6 +306,7 @@ class DBImpl : public DB {
       ColumnFamilyHandle* column_family) override;
   virtual const std::string& GetName() const override;
   virtual Env* GetEnv() const override;
+  virtual FileSystem* GetFileSystem() const override;
   using DB::GetOptions;
   virtual Options GetOptions(ColumnFamilyHandle* column_family) const override;
   using DB::GetDBOptions;
@@ -349,6 +354,8 @@ class DBImpl : public DB {
   virtual Status GetSortedWalFiles(VectorLogPtr& files) override;
   virtual Status GetCurrentWalFile(
       std::unique_ptr<LogFile>* current_log_file) override;
+  virtual Status GetCreationTimeOfOldestFile(
+      uint64_t* creation_time) override;
 
   virtual Status GetUpdatesSince(
       SequenceNumber seq_number, std::unique_ptr<TransactionLogIterator>* iter,
@@ -792,6 +799,10 @@ class DBImpl : public DB {
     logs_to_free_queue_.push_back(log_writer);
   }
 
+  void AddSuperVersionsToFreeQueue(SuperVersion* sv) {
+    superversions_to_free_queue_.push_back(sv);
+  }
+
   void SetSnapshotChecker(SnapshotChecker* snapshot_checker);
 
   // Fill JobContext with snapshot information needed by flush and compaction.
@@ -935,13 +946,14 @@ class DBImpl : public DB {
 #endif  // NDEBUG
 
  protected:
-  Env* const env_;
   const std::string dbname_;
   std::string db_id_;
   std::unique_ptr<VersionSet> versions_;
   // Flag to check whether we allocated and own the info log file
   bool own_info_log_;
   const DBOptions initial_db_options_;
+  Env* const env_;
+  std::shared_ptr<FileSystem> fs_;
   const ImmutableDBOptions immutable_db_options_;
   MutableDBOptions mutable_db_options_;
   Statistics* stats_;
@@ -969,10 +981,10 @@ class DBImpl : public DB {
   bool single_column_family_mode_;
 
   // The options to access storage files
-  const EnvOptions env_options_;
+  const FileOptions file_options_;
 
   // Additonal options for compaction and flush
-  EnvOptions env_options_for_compaction_;
+  FileOptions file_options_for_compaction_;
 
   std::unique_ptr<ColumnFamilyMemTablesImpl> column_family_memtables_;
 
@@ -1005,11 +1017,11 @@ class DBImpl : public DB {
 
   void NotifyOnFlushBegin(ColumnFamilyData* cfd, FileMetaData* file_meta,
                           const MutableCFOptions& mutable_cf_options,
-                          int job_id, TableProperties prop);
+                          int job_id);
 
-  void NotifyOnFlushCompleted(ColumnFamilyData* cfd, FileMetaData* file_meta,
-                              const MutableCFOptions& mutable_cf_options,
-                              int job_id, TableProperties prop);
+  void NotifyOnFlushCompleted(
+      ColumnFamilyData* cfd, const MutableCFOptions& mutable_cf_options,
+      std::list<std::unique_ptr<FlushJobInfo>>* flush_jobs_info);
 
   void NotifyOnCompactionBegin(ColumnFamilyData* cfd, Compaction* c,
                                const Status& st,
@@ -1103,6 +1115,8 @@ class DBImpl : public DB {
       bool read_only = false, bool error_if_log_file_exist = false,
       bool error_if_data_exists_in_logs = false);
 
+  virtual bool OwnTablesAndLogs() const { return true; }
+
  private:
   friend class DB;
   friend class ErrorHandler;
@@ -1187,7 +1201,7 @@ class DBImpl : public DB {
   };
 
   // PurgeFileInfo is a structure to hold information of files to be deleted in
-  // purge_queue_
+  // purge_files_
   struct PurgeFileInfo {
     std::string fname;
     std::string dir_to_sync;
@@ -1313,7 +1327,8 @@ class DBImpl : public DB {
   // created between the calls CaptureCurrentFileNumberInPendingOutputs() and
   // ReleaseFileNumberFromPendingOutputs() can now be deleted (if it's not live
   // and blocked by any other pending_outputs_ calls)
-  void ReleaseFileNumberFromPendingOutputs(std::list<uint64_t>::iterator v);
+  void ReleaseFileNumberFromPendingOutputs(
+      std::unique_ptr<std::list<uint64_t>::iterator>& v);
 
   Status SyncClosedLogs(JobContext* job_context);
 
@@ -1406,10 +1421,23 @@ class DBImpl : public DB {
       bool resuming_from_bg_err);
 
   inline void WaitForPendingWrites() {
+    mutex_.AssertHeld();
+    TEST_SYNC_POINT("DBImpl::WaitForPendingWrites:BeforeBlock");
+    // In case of pipelined write is enabled, wait for all pending memtable
+    // writers.
+    if (immutable_db_options_.enable_pipelined_write) {
+      // Memtable writers may call DB::Get in case max_successive_merges > 0,
+      // which may lock mutex. Unlocking mutex here to avoid deadlock.
+      mutex_.Unlock();
+      write_thread_.WaitForMemTableWriters();
+      mutex_.Lock();
+    }
+
     if (!immutable_db_options_.unordered_write) {
       // Then the writes are finished before the next write group starts
       return;
     }
+
     // Wait for the ones who already wrote to the WAL to finish their
     // memtable write.
     if (pending_memtable_writes_.load() != 0) {
@@ -1605,7 +1633,7 @@ class DBImpl : public DB {
   // Write a version edit to the MANIFEST.
   Status ReserveFileNumbersBeforeIngestion(
       ColumnFamilyData* cfd, uint64_t num,
-      std::list<uint64_t>::iterator* pending_output_elem,
+      std::unique_ptr<std::list<uint64_t>::iterator>& pending_output_elem,
       uint64_t* next_file_number);
 #endif  //! ROCKSDB_LITE
 
@@ -1625,6 +1653,81 @@ class DBImpl : public DB {
       const DBOptions& db_options,
       const std::vector<ColumnFamilyDescriptor>& column_families);
 
+  // Utility function to do some debug validation and sort the given vector
+  // of MultiGet keys
+  void PrepareMultiGetKeys(
+      const size_t num_keys, bool sorted,
+      autovector<KeyContext*, MultiGetContext::MAX_BATCH_SIZE>* key_ptrs);
+
+  // A structure to hold the information required to process MultiGet of keys
+  // belonging to one column family. For a multi column family MultiGet, there
+  // will be a container of these objects.
+  struct MultiGetColumnFamilyData {
+    ColumnFamilyHandle* cf;
+    ColumnFamilyData* cfd;
+
+    // For the batched MultiGet which relies on sorted keys, start specifies
+    // the index of first key belonging to this column family in the sorted
+    // list.
+    size_t start;
+
+    // For the batched MultiGet case, num_keys specifies the number of keys
+    // belonging to this column family in the sorted list
+    size_t num_keys;
+
+    // SuperVersion for the column family obtained in a manner that ensures a
+    // consistent view across all column families in the DB
+    SuperVersion* super_version;
+    MultiGetColumnFamilyData(ColumnFamilyHandle* column_family,
+                             SuperVersion* sv)
+        : cf(column_family),
+          cfd(static_cast<ColumnFamilyHandleImpl*>(cf)->cfd()),
+          start(0),
+          num_keys(0),
+          super_version(sv) {}
+
+    MultiGetColumnFamilyData(ColumnFamilyHandle* column_family, size_t first,
+                             size_t count, SuperVersion* sv)
+        : cf(column_family),
+          cfd(static_cast<ColumnFamilyHandleImpl*>(cf)->cfd()),
+          start(first),
+          num_keys(count),
+          super_version(sv) {}
+
+    MultiGetColumnFamilyData() = default;
+  };
+
+  // A common function to obtain a consistent snapshot, which can be implicit
+  // if the user doesn't specify a snapshot in read_options, across
+  // multiple column families for MultiGet. It will attempt to get an implicit
+  // snapshot without acquiring the db_mutes, but will give up after a few
+  // tries and acquire the mutex if a memtable flush happens. The template
+  // allows both the batched and non-batched MultiGet to call this with
+  // either an std::unordered_map or autovector of column families.
+  //
+  // If callback is non-null, the callback is refreshed with the snapshot
+  // sequence number
+  //
+  // A return value of true indicates that the SuperVersions were obtained
+  // from the ColumnFamilyData, whereas false indicates they are thread
+  // local
+  template <class T>
+  bool MultiCFSnapshot(
+      const ReadOptions& read_options, ReadCallback* callback,
+      std::function<MultiGetColumnFamilyData*(typename T::iterator&)>&
+          iter_deref_func,
+      T* cf_list, SequenceNumber* snapshot);
+
+  // The actual implementation of the batching MultiGet. The caller is expected
+  // to have acquired the SuperVersion and pass in a snapshot sequence number
+  // in order to construct the LookupKeys. The start_key and num_keys specify
+  // the range of keys in the sorted_keys vector for a single column family.
+  void MultiGetImpl(
+      const ReadOptions& read_options, size_t start_key, size_t num_keys,
+      autovector<KeyContext*, MultiGetContext::MAX_BATCH_SIZE>* sorted_keys,
+      SuperVersion* sv, SequenceNumber snap_seqnum, ReadCallback* callback,
+      bool* is_blob_index);
+
   // table_cache_ provides its own synchronization
   std::shared_ptr<Cache> table_cache_;
 
@@ -1735,8 +1838,6 @@ class DBImpl : public DB {
 
   WriteController write_controller_;
 
-  std::unique_ptr<RateLimiter> low_pri_write_rate_limiter_;
-
   // Size of the last batch group. In slowdown mode, next write needs to
   // sleep if it uses up the quota.
   // Note: This is to protect memtable and compaction. If the batch only writes
@@ -1794,6 +1895,7 @@ class DBImpl : public DB {
 
   // A queue to store log writers to close
   std::deque<log::Writer*> logs_to_free_queue_;
+  std::deque<SuperVersion*> superversions_to_free_queue_;
   int unscheduled_flushes_;
   int unscheduled_compactions_;
 

db/db_impl/db_impl_compaction_flush.cc

@@ -148,7 +148,7 @@ Status DBImpl::FlushMemTableToOutputFile(
 
   FlushJob flush_job(
       dbname_, cfd, immutable_db_options_, mutable_cf_options,
-      nullptr /* memtable_id */, env_options_for_compaction_, versions_.get(),
+      nullptr /* memtable_id */, file_options_for_compaction_, versions_.get(),
       &mutex_, &shutting_down_, snapshot_seqs, earliest_write_conflict_snapshot,
       snapshot_checker, job_context, log_buffer, directories_.GetDbDir(),
       GetDataDir(cfd, 0U),
@@ -164,8 +164,7 @@ Status DBImpl::FlushMemTableToOutputFile(
 
 #ifndef ROCKSDB_LITE
   // may temporarily unlock and lock the mutex.
-  NotifyOnFlushBegin(cfd, &file_meta, mutable_cf_options, job_context->job_id,
-                     flush_job.GetTableProperties());
+  NotifyOnFlushBegin(cfd, &file_meta, mutable_cf_options, job_context->job_id);
 #endif  // ROCKSDB_LITE
 
   Status s;
@@ -213,8 +212,8 @@ Status DBImpl::FlushMemTableToOutputFile(
   if (s.ok()) {
 #ifndef ROCKSDB_LITE
     // may temporarily unlock and lock the mutex.
-    NotifyOnFlushCompleted(cfd, &file_meta, mutable_cf_options,
-                           job_context->job_id, flush_job.GetTableProperties());
+    NotifyOnFlushCompleted(cfd, mutable_cf_options,
+                           flush_job.GetCommittedFlushJobsInfo());
     auto sfm = static_cast<SstFileManagerImpl*>(
         immutable_db_options_.sst_file_manager.get());
     if (sfm) {
@@ -233,6 +232,7 @@ Status DBImpl::FlushMemTableToOutputFile(
     }
 #endif  // ROCKSDB_LITE
   }
+  TEST_SYNC_POINT("DBImpl::FlushMemTableToOutputFile:Finish");
   return s;
 }
 
@@ -303,7 +303,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
 
   autovector<Directory*> distinct_output_dirs;
   autovector<std::string> distinct_output_dir_paths;
-  std::vector<FlushJob> jobs;
+  std::vector<std::unique_ptr<FlushJob>> jobs;
   std::vector<MutableCFOptions> all_mutable_cf_options;
   int num_cfs = static_cast<int>(cfds.size());
   all_mutable_cf_options.reserve(num_cfs);
@@ -330,16 +330,16 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
     all_mutable_cf_options.emplace_back(*cfd->GetLatestMutableCFOptions());
     const MutableCFOptions& mutable_cf_options = all_mutable_cf_options.back();
     const uint64_t* max_memtable_id = &(bg_flush_args[i].max_memtable_id_);
-    jobs.emplace_back(
+    jobs.emplace_back(new FlushJob(
         dbname_, cfd, immutable_db_options_, mutable_cf_options,
-        max_memtable_id, env_options_for_compaction_, versions_.get(), &mutex_,
+        max_memtable_id, file_options_for_compaction_, versions_.get(), &mutex_,
         &shutting_down_, snapshot_seqs, earliest_write_conflict_snapshot,
         snapshot_checker, job_context, log_buffer, directories_.GetDbDir(),
         data_dir, 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);
-    jobs.back().PickMemTable();
+        thread_pri));
+    jobs.back()->PickMemTable();
   }
 
   std::vector<FileMetaData> file_meta(num_cfs);
@@ -351,7 +351,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
     const MutableCFOptions& mutable_cf_options = all_mutable_cf_options.at(i);
     // may temporarily unlock and lock the mutex.
     NotifyOnFlushBegin(cfds[i], &file_meta[i], mutable_cf_options,
-                       job_context->job_id, jobs[i].GetTableProperties());
+                       job_context->job_id);
   }
 #endif /* !ROCKSDB_LITE */
 
@@ -373,7 +373,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
     // TODO (yanqin): parallelize jobs with threads.
     for (int i = 1; i != num_cfs; ++i) {
       exec_status[i].second =
-          jobs[i].Run(&logs_with_prep_tracker_, &file_meta[i]);
+          jobs[i]->Run(&logs_with_prep_tracker_, &file_meta[i]);
       exec_status[i].first = true;
     }
     if (num_cfs > 1) {
@@ -382,8 +382,10 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
       TEST_SYNC_POINT(
           "DBImpl::AtomicFlushMemTablesToOutputFiles:SomeFlushJobsComplete:2");
     }
+    assert(exec_status.size() > 0);
+    assert(!file_meta.empty());
     exec_status[0].second =
-        jobs[0].Run(&logs_with_prep_tracker_, &file_meta[0]);
+        jobs[0]->Run(&logs_with_prep_tracker_, &file_meta[0]);
     exec_status[0].first = true;
 
     Status error_status;
@@ -424,7 +426,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
     auto wait_to_install_func = [&]() {
       bool ready = true;
       for (size_t i = 0; i != cfds.size(); ++i) {
-        const auto& mems = jobs[i].GetMemTables();
+        const auto& mems = jobs[i]->GetMemTables();
         if (cfds[i]->IsDropped()) {
           // If the column family is dropped, then do not wait.
           continue;
@@ -465,7 +467,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
     autovector<const MutableCFOptions*> mutable_cf_options_list;
     autovector<FileMetaData*> tmp_file_meta;
     for (int i = 0; i != num_cfs; ++i) {
-      const auto& mems = jobs[i].GetMemTables();
+      const auto& mems = jobs[i]->GetMemTables();
       if (!cfds[i]->IsDropped() && !mems.empty()) {
         tmp_cfds.emplace_back(cfds[i]);
         mems_list.emplace_back(&mems);
@@ -501,12 +503,13 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
 #ifndef ROCKSDB_LITE
     auto sfm = static_cast<SstFileManagerImpl*>(
         immutable_db_options_.sst_file_manager.get());
+    assert(all_mutable_cf_options.size() == static_cast<size_t>(num_cfs));
     for (int i = 0; i != num_cfs; ++i) {
       if (cfds[i]->IsDropped()) {
         continue;
       }
-      NotifyOnFlushCompleted(cfds[i], &file_meta[i], all_mutable_cf_options[i],
-                             job_context->job_id, jobs[i].GetTableProperties());
+      NotifyOnFlushCompleted(cfds[i], all_mutable_cf_options[i],
+                             jobs[i]->GetCommittedFlushJobsInfo());
       if (sfm) {
         std::string file_path = MakeTableFileName(
             cfds[i]->ioptions()->cf_paths[0].path, file_meta[i].fd.GetNumber());
@@ -530,12 +533,12 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
     // unref the versions.
     for (int i = 0; i != num_cfs; ++i) {
       if (!exec_status[i].first) {
-        jobs[i].Cancel();
+        jobs[i]->Cancel();
       }
     }
     for (int i = 0; i != num_cfs; ++i) {
       if (exec_status[i].first && exec_status[i].second.ok()) {
-        auto& mems = jobs[i].GetMemTables();
+        auto& mems = jobs[i]->GetMemTables();
         cfds[i]->imm()->RollbackMemtableFlush(mems,
                                               file_meta[i].fd.GetNumber());
       }
@@ -549,7 +552,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
 
 void DBImpl::NotifyOnFlushBegin(ColumnFamilyData* cfd, FileMetaData* file_meta,
                                 const MutableCFOptions& mutable_cf_options,
-                                int job_id, TableProperties prop) {
+                                int job_id) {
 #ifndef ROCKSDB_LITE
   if (immutable_db_options_.listeners.size() == 0U) {
     return;
@@ -567,20 +570,21 @@ void DBImpl::NotifyOnFlushBegin(ColumnFamilyData* cfd, FileMetaData* file_meta,
   // release lock while notifying events
   mutex_.Unlock();
   {
-    FlushJobInfo info;
+    FlushJobInfo info{};
     info.cf_id = cfd->GetID();
     info.cf_name = cfd->GetName();
     // TODO(yhchiang): make db_paths dynamic in case flush does not
     //                 go to L0 in the future.
-    info.file_path = MakeTableFileName(cfd->ioptions()->cf_paths[0].path,
-                                       file_meta->fd.GetNumber());
+    const uint64_t file_number = file_meta->fd.GetNumber();
+    info.file_path =
+        MakeTableFileName(cfd->ioptions()->cf_paths[0].path, file_number);
+    info.file_number = file_number;
     info.thread_id = env_->GetThreadID();
     info.job_id = job_id;
     info.triggered_writes_slowdown = triggered_writes_slowdown;
     info.triggered_writes_stop = triggered_writes_stop;
     info.smallest_seqno = file_meta->fd.smallest_seqno;
     info.largest_seqno = file_meta->fd.largest_seqno;
-    info.table_properties = prop;
     info.flush_reason = cfd->GetFlushReason();
     for (auto listener : immutable_db_options_.listeners) {
       listener->OnFlushBegin(this, info);
@@ -594,15 +598,14 @@ void DBImpl::NotifyOnFlushBegin(ColumnFamilyData* cfd, FileMetaData* file_meta,
   (void)file_meta;
   (void)mutable_cf_options;
   (void)job_id;
-  (void)prop;
 #endif  // ROCKSDB_LITE
 }
 
-void DBImpl::NotifyOnFlushCompleted(ColumnFamilyData* cfd,
-                                    FileMetaData* file_meta,
-                                    const MutableCFOptions& mutable_cf_options,
-                                    int job_id, TableProperties prop) {
+void DBImpl::NotifyOnFlushCompleted(
+    ColumnFamilyData* cfd, const MutableCFOptions& mutable_cf_options,
+    std::list<std::unique_ptr<FlushJobInfo>>* flush_jobs_info) {
 #ifndef ROCKSDB_LITE
+  assert(flush_jobs_info != nullptr);
   if (immutable_db_options_.listeners.size() == 0U) {
     return;
   }
@@ -619,34 +622,22 @@ void DBImpl::NotifyOnFlushCompleted(ColumnFamilyData* cfd,
   // release lock while notifying events
   mutex_.Unlock();
   {
-    FlushJobInfo info;
-    info.cf_id = cfd->GetID();
-    info.cf_name = cfd->GetName();
-    // TODO(yhchiang): make db_paths dynamic in case flush does not
-    //                 go to L0 in the future.
-    info.file_path = MakeTableFileName(cfd->ioptions()->cf_paths[0].path,
-                                       file_meta->fd.GetNumber());
-    info.thread_id = env_->GetThreadID();
-    info.job_id = job_id;
-    info.triggered_writes_slowdown = triggered_writes_slowdown;
-    info.triggered_writes_stop = triggered_writes_stop;
-    info.smallest_seqno = file_meta->fd.smallest_seqno;
-    info.largest_seqno = file_meta->fd.largest_seqno;
-    info.table_properties = prop;
-    info.flush_reason = cfd->GetFlushReason();
-    for (auto listener : immutable_db_options_.listeners) {
-      listener->OnFlushCompleted(this, info);
+    for (auto& info : *flush_jobs_info) {
+      info->triggered_writes_slowdown = triggered_writes_slowdown;
+      info->triggered_writes_stop = triggered_writes_stop;
+      for (auto listener : immutable_db_options_.listeners) {
+        listener->OnFlushCompleted(this, *info);
+      }
     }
+    flush_jobs_info->clear();
   }
   mutex_.Lock();
   // no need to signal bg_cv_ as it will be signaled at the end of the
   // flush process.
 #else
   (void)cfd;
-  (void)file_meta;
   (void)mutable_cf_options;
-  (void)job_id;
-  (void)prop;
+  (void)flush_jobs_info;
 #endif  // ROCKSDB_LITE
 }
 
@@ -668,7 +659,7 @@ Status DBImpl::CompactRange(const CompactRangeOptions& options,
     // one/both sides of the interval are unbounded. But it requires more
     // changes to RangesOverlapWithMemtables.
     Range range(*begin, *end);
-    SuperVersion* super_version = cfd->GetReferencedSuperVersion(&mutex_);
+    SuperVersion* super_version = cfd->GetReferencedSuperVersion(this);
     cfd->RangesOverlapWithMemtables({range}, super_version, &flush_needed);
     CleanupSuperVersion(super_version);
   }
@@ -969,14 +960,15 @@ Status DBImpl::CompactFilesImpl(
   GetSnapshotContext(job_context, &snapshot_seqs,
                      &earliest_write_conflict_snapshot, &snapshot_checker);
 
-  auto pending_outputs_inserted_elem =
-      CaptureCurrentFileNumberInPendingOutputs();
+  std::unique_ptr<std::list<uint64_t>::iterator> pending_outputs_inserted_elem(
+      new std::list<uint64_t>::iterator(
+          CaptureCurrentFileNumberInPendingOutputs()));
 
   assert(is_snapshot_supported_ || snapshots_.empty());
   CompactionJobStats compaction_job_stats;
   CompactionJob compaction_job(
       job_context->job_id, c.get(), immutable_db_options_,
-      env_options_for_compaction_, versions_.get(), &shutting_down_,
+      file_options_for_compaction_, versions_.get(), &shutting_down_,
       preserve_deletes_seqnum_.load(), log_buffer, directories_.GetDbDir(),
       GetDataDir(c->column_family_data(), c->output_path_id()), stats_, &mutex_,
       &error_handler_, snapshot_seqs, earliest_write_conflict_snapshot,
@@ -1115,7 +1107,7 @@ void DBImpl::NotifyOnCompactionBegin(ColumnFamilyData* cfd, Compaction* c,
   mutex_.Unlock();
   TEST_SYNC_POINT("DBImpl::NotifyOnCompactionBegin::UnlockMutex");
   {
-    CompactionJobInfo info;
+    CompactionJobInfo info{};
     info.cf_name = cfd->GetName();
     info.status = st;
     info.thread_id = env_->GetThreadID();
@@ -1128,9 +1120,13 @@ void DBImpl::NotifyOnCompactionBegin(ColumnFamilyData* cfd, Compaction* c,
     info.compression = c->output_compression();
     for (size_t i = 0; i < c->num_input_levels(); ++i) {
       for (const auto fmd : *c->inputs(i)) {
+        const FileDescriptor& desc = fmd->fd;
+        const uint64_t file_number = desc.GetNumber();
         auto fn = TableFileName(c->immutable_cf_options()->cf_paths,
-                                fmd->fd.GetNumber(), fmd->fd.GetPathId());
+                                file_number, desc.GetPathId());
         info.input_files.push_back(fn);
+        info.input_file_infos.push_back(CompactionFileInfo{
+            static_cast<int>(i), file_number, fmd->oldest_blob_file_number});
         if (info.table_properties.count(fn) == 0) {
           std::shared_ptr<const TableProperties> tp;
           auto s = current->GetTableProperties(&tp, fmd, &fn);
@@ -1141,9 +1137,13 @@ void DBImpl::NotifyOnCompactionBegin(ColumnFamilyData* cfd, Compaction* c,
       }
     }
     for (const auto newf : c->edit()->GetNewFiles()) {
+      const FileMetaData& meta = newf.second;
+      const FileDescriptor& desc = meta.fd;
+      const uint64_t file_number = desc.GetNumber();
       info.output_files.push_back(TableFileName(
-          c->immutable_cf_options()->cf_paths, newf.second.fd.GetNumber(),
-          newf.second.fd.GetPathId()));
+          c->immutable_cf_options()->cf_paths, file_number, desc.GetPathId()));
+      info.output_file_infos.push_back(CompactionFileInfo{
+          newf.first, file_number, meta.oldest_blob_file_number});
     }
     for (auto listener : immutable_db_options_.listeners) {
       listener->OnCompactionBegin(this, info);
@@ -1181,7 +1181,7 @@ void DBImpl::NotifyOnCompactionCompleted(
   mutex_.Unlock();
   TEST_SYNC_POINT("DBImpl::NotifyOnCompactionCompleted::UnlockMutex");
   {
-    CompactionJobInfo info;
+    CompactionJobInfo info{};
     BuildCompactionJobInfo(cfd, c, st, compaction_job_stats, job_id, current,
                            &info);
     for (auto listener : immutable_db_options_.listeners) {
@@ -1256,7 +1256,8 @@ Status DBImpl::ReFitLevel(ColumnFamilyData* cfd, int level, int target_level) {
       edit.AddFile(to_level, f->fd.GetNumber(), f->fd.GetPathId(),
                    f->fd.GetFileSize(), f->smallest, f->largest,
                    f->fd.smallest_seqno, f->fd.largest_seqno,
-                   f->marked_for_compaction);
+                   f->marked_for_compaction, f->oldest_blob_file_number,
+                   f->oldest_ancester_time, f->file_creation_time);
     }
     ROCKS_LOG_DEBUG(immutable_db_options_.info_log,
                     "[%s] Apply version edit:\n%s", cfd->GetName().c_str(),
@@ -1532,9 +1533,14 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
     InstrumentedMutexLock guard_lock(&mutex_);
 
     WriteThread::Writer w;
+    WriteThread::Writer nonmem_w;
     if (!writes_stopped) {
       write_thread_.EnterUnbatched(&w, &mutex_);
+      if (two_write_queues_) {
+        nonmem_write_thread_.EnterUnbatched(&nonmem_w, &mutex_);
+      }
     }
+    WaitForPendingWrites();
 
     if (!cfd->mem()->IsEmpty() || !cached_recoverable_state_empty_.load()) {
       s = SwitchMemtable(cfd, &context);
@@ -1596,6 +1602,9 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
 
     if (!writes_stopped) {
       write_thread_.ExitUnbatched(&w);
+      if (two_write_queues_) {
+        nonmem_write_thread_.ExitUnbatched(&nonmem_w);
+      }
     }
   }
   TEST_SYNC_POINT("DBImpl::FlushMemTable:AfterScheduleFlush");
@@ -1609,19 +1618,16 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
     }
     s = WaitForFlushMemTables(cfds, flush_memtable_ids,
                               (flush_reason == FlushReason::kErrorRecovery));
+    InstrumentedMutexLock lock_guard(&mutex_);
     for (auto* tmp_cfd : cfds) {
-      if (tmp_cfd->Unref()) {
-        // Only one thread can reach here.
-        InstrumentedMutexLock lock_guard(&mutex_);
-        delete tmp_cfd;
-      }
+      tmp_cfd->UnrefAndTryDelete();
     }
   }
-  TEST_SYNC_POINT("FlushMemTableFinished");
+  TEST_SYNC_POINT("DBImpl::FlushMemTable:FlushMemTableFinished");
   return s;
 }
 
-// Flush all elments in 'column_family_datas'
+// Flush all elements in 'column_family_datas'
 // and atomically record the result to the MANIFEST.
 Status DBImpl::AtomicFlushMemTables(
     const autovector<ColumnFamilyData*>& column_family_datas,
@@ -1650,9 +1656,14 @@ Status DBImpl::AtomicFlushMemTables(
     InstrumentedMutexLock guard_lock(&mutex_);
 
     WriteThread::Writer w;
+    WriteThread::Writer nonmem_w;
     if (!writes_stopped) {
       write_thread_.EnterUnbatched(&w, &mutex_);
+      if (two_write_queues_) {
+        nonmem_write_thread_.EnterUnbatched(&nonmem_w, &mutex_);
+      }
     }
+    WaitForPendingWrites();
 
     for (auto cfd : column_family_datas) {
       if (cfd->IsDropped()) {
@@ -1669,7 +1680,7 @@ Status DBImpl::AtomicFlushMemTables(
       }
       cfd->Ref();
       s = SwitchMemtable(cfd, &context);
-      cfd->Unref();
+      cfd->UnrefAndTryDelete();
       if (!s.ok()) {
         break;
       }
@@ -1695,6 +1706,9 @@ Status DBImpl::AtomicFlushMemTables(
 
     if (!writes_stopped) {
       write_thread_.ExitUnbatched(&w);
+      if (two_write_queues_) {
+        nonmem_write_thread_.ExitUnbatched(&nonmem_w);
+      }
     }
   }
   TEST_SYNC_POINT("DBImpl::AtomicFlushMemTables:AfterScheduleFlush");
@@ -1706,12 +1720,9 @@ Status DBImpl::AtomicFlushMemTables(
     }
     s = WaitForFlushMemTables(cfds, flush_memtable_ids,
                               (flush_reason == FlushReason::kErrorRecovery));
+    InstrumentedMutexLock lock_guard(&mutex_);
     for (auto* cfd : cfds) {
-      if (cfd->Unref()) {
-        // Only one thread can reach here.
-        InstrumentedMutexLock lock_guard(&mutex_);
-        delete cfd;
-      }
+      cfd->UnrefAndTryDelete();
     }
   }
   return s;
@@ -1903,6 +1914,10 @@ void DBImpl::MaybeScheduleFlushOrCompaction() {
     fta->thread_pri_ = Env::Priority::HIGH;
     env_->Schedule(&DBImpl::BGWorkFlush, fta, Env::Priority::HIGH, this,
                    &DBImpl::UnscheduleFlushCallback);
+    --unscheduled_flushes_;
+    TEST_SYNC_POINT_CALLBACK(
+        "DBImpl::MaybeScheduleFlushOrCompaction:AfterSchedule:0",
+        &unscheduled_flushes_);
   }
 
   // special case -- if high-pri (flush) thread pool is empty, then schedule
@@ -1917,6 +1932,7 @@ void DBImpl::MaybeScheduleFlushOrCompaction() {
       fta->thread_pri_ = Env::Priority::LOW;
       env_->Schedule(&DBImpl::BGWorkFlush, fta, Env::Priority::LOW, this,
                      &DBImpl::UnscheduleFlushCallback);
+      --unscheduled_flushes_;
     }
   }
 
@@ -2000,8 +2016,6 @@ ColumnFamilyData* DBImpl::PopFirstFromCompactionQueue() {
 DBImpl::FlushRequest DBImpl::PopFirstFromFlushQueue() {
   assert(!flush_queue_.empty());
   FlushRequest flush_req = flush_queue_.front();
-  assert(unscheduled_flushes_ >= static_cast<int>(flush_req.size()));
-  unscheduled_flushes_ -= static_cast<int>(flush_req.size());
   flush_queue_.pop_front();
   // TODO: need to unset flush reason?
   return flush_req;
@@ -2043,7 +2057,7 @@ void DBImpl::SchedulePendingFlush(const FlushRequest& flush_req,
     cfd->Ref();
     cfd->SetFlushReason(flush_reason);
   }
-  unscheduled_flushes_ += static_cast<int>(flush_req.size());
+  ++unscheduled_flushes_;
   flush_queue_.push_back(flush_req);
 }
 
@@ -2189,16 +2203,13 @@ Status DBImpl::BackgroundFlush(bool* made_progress, JobContext* job_context,
     *reason = bg_flush_args[0].cfd_->GetFlushReason();
     for (auto& arg : bg_flush_args) {
       ColumnFamilyData* cfd = arg.cfd_;
-      if (cfd->Unref()) {
-        delete cfd;
+      if (cfd->UnrefAndTryDelete()) {
         arg.cfd_ = nullptr;
       }
     }
   }
   for (auto cfd : column_families_not_to_flush) {
-    if (cfd->Unref()) {
-      delete cfd;
-    }
+    cfd->UnrefAndTryDelete();
   }
   return status;
 }
@@ -2216,8 +2227,9 @@ void DBImpl::BackgroundCallFlush(Env::Priority thread_pri) {
     assert(bg_flush_scheduled_);
     num_running_flushes_++;
 
-    auto pending_outputs_inserted_elem =
-        CaptureCurrentFileNumberInPendingOutputs();
+    std::unique_ptr<std::list<uint64_t>::iterator>
+        pending_outputs_inserted_elem(new std::list<uint64_t>::iterator(
+            CaptureCurrentFileNumberInPendingOutputs()));
     FlushReason reason;
 
     Status s = BackgroundFlush(&made_progress, &job_context, &log_buffer,
@@ -2298,8 +2310,9 @@ void DBImpl::BackgroundCallCompaction(PrepickedCompaction* prepicked_compaction,
 
     num_running_compactions_++;
 
-    auto pending_outputs_inserted_elem =
-        CaptureCurrentFileNumberInPendingOutputs();
+    std::unique_ptr<std::list<uint64_t>::iterator>
+        pending_outputs_inserted_elem(new std::list<uint64_t>::iterator(
+            CaptureCurrentFileNumberInPendingOutputs()));
 
     assert((bg_thread_pri == Env::Priority::BOTTOM &&
             bg_bottom_compaction_scheduled_) ||
@@ -2525,10 +2538,9 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
     // reference).
     // This will all happen under a mutex so we don't have to be afraid of
     // somebody else deleting it.
-    if (cfd->Unref()) {
+    if (cfd->UnrefAndTryDelete()) {
       // This was the last reference of the column family, so no need to
       // compact.
-      delete cfd;
       return Status::OK();
     }
 
@@ -2654,7 +2666,9 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
         c->edit()->AddFile(c->output_level(), f->fd.GetNumber(),
                            f->fd.GetPathId(), f->fd.GetFileSize(), f->smallest,
                            f->largest, f->fd.smallest_seqno,
-                           f->fd.largest_seqno, f->marked_for_compaction);
+                           f->fd.largest_seqno, f->marked_for_compaction,
+                           f->oldest_blob_file_number, f->oldest_ancester_time,
+                           f->file_creation_time);
 
         ROCKS_LOG_BUFFER(
             log_buffer,
@@ -2733,7 +2747,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
     assert(is_snapshot_supported_ || snapshots_.empty());
     CompactionJob compaction_job(
         job_context->job_id, c.get(), immutable_db_options_,
-        env_options_for_compaction_, versions_.get(), &shutting_down_,
+        file_options_for_compaction_, versions_.get(), &shutting_down_,
         preserve_deletes_seqnum_.load(), log_buffer, directories_.GetDbDir(),
         GetDataDir(c->column_family_data(), c->output_path_id()), stats_,
         &mutex_, &error_handler_, snapshot_seqs,
@@ -2963,9 +2977,13 @@ void DBImpl::BuildCompactionJobInfo(
   compaction_job_info->compression = c->output_compression();
   for (size_t i = 0; i < c->num_input_levels(); ++i) {
     for (const auto fmd : *c->inputs(i)) {
-      auto fn = TableFileName(c->immutable_cf_options()->cf_paths,
-                              fmd->fd.GetNumber(), fmd->fd.GetPathId());
+      const FileDescriptor& desc = fmd->fd;
+      const uint64_t file_number = desc.GetNumber();
+      auto fn = TableFileName(c->immutable_cf_options()->cf_paths, file_number,
+                              desc.GetPathId());
       compaction_job_info->input_files.push_back(fn);
+      compaction_job_info->input_file_infos.push_back(CompactionFileInfo{
+          static_cast<int>(i), file_number, fmd->oldest_blob_file_number});
       if (compaction_job_info->table_properties.count(fn) == 0) {
         std::shared_ptr<const TableProperties> tp;
         auto s = current->GetTableProperties(&tp, fmd, &fn);
@@ -2976,9 +2994,13 @@ void DBImpl::BuildCompactionJobInfo(
     }
   }
   for (const auto& newf : c->edit()->GetNewFiles()) {
-    compaction_job_info->output_files.push_back(
-        TableFileName(c->immutable_cf_options()->cf_paths,
-                      newf.second.fd.GetNumber(), newf.second.fd.GetPathId()));
+    const FileMetaData& meta = newf.second;
+    const FileDescriptor& desc = meta.fd;
+    const uint64_t file_number = desc.GetNumber();
+    compaction_job_info->output_files.push_back(TableFileName(
+        c->immutable_cf_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});
   }
 }
 #endif

db/db_impl/db_impl_debug.cc

@@ -105,7 +105,16 @@ Status DBImpl::TEST_SwitchMemtable(ColumnFamilyData* cfd) {
   if (cfd == nullptr) {
     cfd = default_cf_handle_->cfd();
   }
-  return SwitchMemtable(cfd, &write_context);
+
+  if (two_write_queues_) {
+    WriteThread::Writer nonmem_w;
+    nonmem_write_thread_.EnterUnbatched(&nonmem_w, &mutex_);
+    Status s = SwitchMemtable(cfd, &write_context);
+    nonmem_write_thread_.ExitUnbatched(&nonmem_w);
+    return s;
+  } else {
+    return SwitchMemtable(cfd, &write_context);
+  }
 }
 
 Status DBImpl::TEST_FlushMemTable(bool wait, bool allow_write_stall,

db/db_impl/db_impl_experimental.cc

@@ -128,7 +128,8 @@ Status DBImpl::PromoteL0(ColumnFamilyHandle* column_family, int target_level) {
       edit.AddFile(target_level, f->fd.GetNumber(), f->fd.GetPathId(),
                    f->fd.GetFileSize(), f->smallest, f->largest,
                    f->fd.smallest_seqno, f->fd.largest_seqno,
-                   f->marked_for_compaction);
+                   f->marked_for_compaction, f->oldest_blob_file_number,
+                   f->oldest_ancester_time, f->file_creation_time);
     }
 
     status = versions_->LogAndApply(cfd, *cfd->GetLatestMutableCFOptions(),

db/db_impl/db_impl_files.cc

@@ -385,6 +385,7 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
     w->Close();
   }
 
+  bool own_files = OwnTablesAndLogs();
   std::unordered_set<uint64_t> files_to_del;
   for (const auto& candidate_file : candidate_files) {
     const std::string& to_delete = candidate_file.file_name;
@@ -484,6 +485,12 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
     }
 #endif  // !ROCKSDB_LITE
 
+    // If I do not own these files, e.g. secondary instance with max_open_files
+    // = -1, then no need to delete or schedule delete these files since they
+    // will be removed by their owner, e.g. the primary instance.
+    if (!own_files) {
+      continue;
+    }
     Status file_deletion_status;
     if (schedule_only) {
       InstrumentedMutexLock guard_lock(&mutex_);
@@ -495,7 +502,6 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
 
   {
     // After purging obsolete files, remove them from files_grabbed_for_purge_.
-    // Use a temporary vector to perform bulk deletion via swap.
     InstrumentedMutexLock guard_lock(&mutex_);
     autovector<uint64_t> to_be_removed;
     for (auto fn : files_grabbed_for_purge_) {

db/db_impl/db_impl_open.cc

@@ -12,6 +12,7 @@
 
 #include "db/builder.h"
 #include "db/error_handler.h"
+#include "env/composite_env_wrapper.h"
 #include "file/read_write_util.h"
 #include "file/sst_file_manager_impl.h"
 #include "file/writable_file_writer.h"
@@ -34,6 +35,18 @@ Options SanitizeOptions(const std::string& dbname, const Options& src) {
 DBOptions SanitizeOptions(const std::string& dbname, const DBOptions& src) {
   DBOptions result(src);
 
+  if (result.file_system == nullptr) {
+    if (result.env == Env::Default()) {
+      result.file_system = FileSystem::Default();
+    } else {
+      result.file_system.reset(new LegacyFileSystemWrapper(result.env));
+    }
+  } else {
+    if (result.env == nullptr) {
+      result.env = Env::Default();
+    }
+  }
+
   // result.max_open_files means an "infinite" open files.
   if (result.max_open_files != -1) {
     int max_max_open_files = port::GetMaxOpenFiles();
@@ -254,16 +267,16 @@ Status DBImpl::NewDB() {
   ROCKS_LOG_INFO(immutable_db_options_.info_log, "Creating manifest 1 \n");
   const std::string manifest = DescriptorFileName(dbname_, 1);
   {
-    std::unique_ptr<WritableFile> file;
-    EnvOptions env_options = env_->OptimizeForManifestWrite(env_options_);
-    s = NewWritableFile(env_, manifest, &file, env_options);
+    std::unique_ptr<FSWritableFile> file;
+    FileOptions file_options = fs_->OptimizeForManifestWrite(file_options_);
+    s = NewWritableFile(fs_.get(), manifest, &file, file_options);
     if (!s.ok()) {
       return s;
     }
     file->SetPreallocationBlockSize(
         immutable_db_options_.manifest_preallocation_size);
     std::unique_ptr<WritableFileWriter> file_writer(new WritableFileWriter(
-        std::move(file), manifest, env_options, env_, nullptr /* stats */,
+        std::move(file), manifest, file_options, env_, nullptr /* stats */,
         immutable_db_options_.listeners));
     log::Writer log(std::move(file_writer), 0, false);
     std::string record;
@@ -277,7 +290,7 @@ Status DBImpl::NewDB() {
     // Make "CURRENT" file that points to the new manifest file.
     s = SetCurrentFile(env_, dbname_, 1, directories_.GetDbDir());
   } else {
-    env_->DeleteFile(manifest);
+    fs_->DeleteFile(manifest, IOOptions(), nullptr);
   }
   return s;
 }
@@ -350,7 +363,8 @@ Status DBImpl::Recover(
       return s;
     }
 
-    s = env_->FileExists(CurrentFileName(dbname_));
+    std::string current_fname = CurrentFileName(dbname_);
+    s = env_->FileExists(current_fname);
     if (s.IsNotFound()) {
       if (immutable_db_options_.create_if_missing) {
         // Has to be called only after Identity File creation is successful
@@ -363,7 +377,7 @@ Status DBImpl::Recover(
         }
       } else {
         return Status::InvalidArgument(
-            dbname_, "does not exist (create_if_missing is false)");
+            current_fname, "does not exist (create_if_missing is false)");
       }
     } else if (s.ok()) {
       if (immutable_db_options_.error_if_exists) {
@@ -375,20 +389,20 @@ Status DBImpl::Recover(
       assert(s.IsIOError());
       return s;
     }
-    // Verify compatibility of env_options_ and filesystem
+    // Verify compatibility of file_options_ and filesystem
     {
-      std::unique_ptr<RandomAccessFile> idfile;
-      EnvOptions customized_env(env_options_);
-      customized_env.use_direct_reads |=
+      std::unique_ptr<FSRandomAccessFile> idfile;
+      FileOptions customized_fs(file_options_);
+      customized_fs.use_direct_reads |=
           immutable_db_options_.use_direct_io_for_flush_and_compaction;
-      s = env_->NewRandomAccessFile(CurrentFileName(dbname_), &idfile,
-                                    customized_env);
+      s = fs_->NewRandomAccessFile(current_fname, customized_fs, &idfile,
+                                   nullptr);
       if (!s.ok()) {
         std::string error_str = s.ToString();
         // Check if unsupported Direct I/O is the root cause
-        customized_env.use_direct_reads = false;
-        s = env_->NewRandomAccessFile(CurrentFileName(dbname_), &idfile,
-                                      customized_env);
+        customized_fs.use_direct_reads = false;
+        s = fs_->NewRandomAccessFile(current_fname, customized_fs, &idfile,
+                                     nullptr);
         if (s.ok()) {
           return Status::InvalidArgument(
               "Direct I/O is not supported by the specified DB.");
@@ -408,7 +422,7 @@ Status DBImpl::Recover(
   // the very first time.
   if (db_id_.empty()) {
     // Check for the IDENTITY file and create it if not there.
-    s = env_->FileExists(IdentityFileName(dbname_));
+    s = fs_->FileExists(IdentityFileName(dbname_), IOOptions(), nullptr);
     // Typically Identity file is created in NewDB() and for some reason if
     // it is no longer available then at this point DB ID is not in Identity
     // file or Manifest.
@@ -748,9 +762,10 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
 
     std::unique_ptr<SequentialFileReader> file_reader;
     {
-      std::unique_ptr<SequentialFile> file;
-      status = env_->NewSequentialFile(fname, &file,
-                                       env_->OptimizeForLogRead(env_options_));
+      std::unique_ptr<FSSequentialFile> file;
+      status = fs_->NewSequentialFile(fname,
+                                      fs_->OptimizeForLogRead(file_options_),
+                                      &file, nullptr);
       if (!status.ok()) {
         MaybeIgnoreError(&status);
         if (!status.ok()) {
@@ -920,7 +935,7 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
         ColumnFamilyData* cfd;
 
         while ((cfd = flush_scheduler_.TakeNextColumnFamily()) != nullptr) {
-          cfd->Unref();
+          cfd->UnrefAndTryDelete();
           // If this asserts, it means that InsertInto failed in
           // filtering updates to already-flushed column families
           assert(cfd->GetLogNumber() <= log_number);
@@ -1023,6 +1038,9 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
         continue;
       }
 
+      TEST_SYNC_POINT_CALLBACK(
+          "DBImpl::RecoverLogFiles:BeforeFlushFinalMemtable", /*arg=*/nullptr);
+
       // flush the final memtable (if non-empty)
       if (cfd->mem()->GetFirstSequenceNumber() != 0) {
         // If flush happened in the middle of recovery (e.g. due to memtable
@@ -1042,7 +1060,9 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
         data_seen = true;
       }
 
-      // write MANIFEST with update
+      // Update the log number info in the version edit corresponding to this
+      // column family. Note that the version edits will be written to MANIFEST
+      // together later.
       // writing log_number in the manifest means that any log file
       // with number strongly less than (log_number + 1) is already
       // recovered and should be ignored on next reincarnation.
@@ -1051,19 +1071,28 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
       if (flushed || cfd->mem()->GetFirstSequenceNumber() == 0) {
         edit->SetLogNumber(max_log_number + 1);
       }
+    }
+    if (status.ok()) {
       // we must mark the next log number as used, even though it's
       // not actually used. that is because VersionSet assumes
       // VersionSet::next_file_number_ always to be strictly greater than any
       // log number
       versions_->MarkFileNumberUsed(max_log_number + 1);
-      status = versions_->LogAndApply(cfd, *cfd->GetLatestMutableCFOptions(),
-                                      edit, &mutex_);
-      TEST_SYNC_POINT_CALLBACK("DBImpl::RecoverLogFiles:AfterLogAndApply",
-                               nullptr);
-      if (!status.ok()) {
-        // Recovery failed
-        break;
+
+      autovector<ColumnFamilyData*> cfds;
+      autovector<const MutableCFOptions*> cf_opts;
+      autovector<autovector<VersionEdit*>> edit_lists;
+      for (auto* cfd : *versions_->GetColumnFamilySet()) {
+        cfds.push_back(cfd);
+        cf_opts.push_back(cfd->GetLatestMutableCFOptions());
+        auto iter = version_edits.find(cfd->GetID());
+        assert(iter != version_edits.end());
+        edit_lists.push_back({&iter->second});
       }
+      // write MANIFEST with update
+      status = versions_->LogAndApply(cfds, cf_opts, edit_lists, &mutex_,
+                                      directories_.GetDbDir(),
+                                      /*new_descriptor_log=*/true);
     }
   }
 
@@ -1105,17 +1134,18 @@ Status DBImpl::RestoreAliveLogFiles(const std::vector<uint64_t>& log_numbers) {
     // preallocated space are not needed anymore. It is likely only the last
     // log has such preallocated space, so we only truncate for the last log.
     if (log_number == log_numbers.back()) {
-      std::unique_ptr<WritableFile> last_log;
-      Status truncate_status = env_->ReopenWritableFile(
-          fname, &last_log,
-          env_->OptimizeForLogWrite(
-              env_options_,
-              BuildDBOptions(immutable_db_options_, mutable_db_options_)));
+      std::unique_ptr<FSWritableFile> last_log;
+      Status truncate_status = fs_->ReopenWritableFile(
+          fname,
+          fs_->OptimizeForLogWrite(
+              file_options_,
+              BuildDBOptions(immutable_db_options_, mutable_db_options_)),
+          &last_log, nullptr);
       if (truncate_status.ok()) {
-        truncate_status = last_log->Truncate(log.size);
+        truncate_status = last_log->Truncate(log.size, IOOptions(), nullptr);
       }
       if (truncate_status.ok()) {
-        truncate_status = last_log->Close();
+        truncate_status = last_log->Close(IOOptions(), nullptr);
       }
       // Not a critical error if fail to truncate.
       if (!truncate_status.ok()) {
@@ -1136,8 +1166,9 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
   mutex_.AssertHeld();
   const uint64_t start_micros = env_->NowMicros();
   FileMetaData meta;
-  auto pending_outputs_inserted_elem =
-      CaptureCurrentFileNumberInPendingOutputs();
+  std::unique_ptr<std::list<uint64_t>::iterator> pending_outputs_inserted_elem(
+      new std::list<uint64_t>::iterator(
+          CaptureCurrentFileNumberInPendingOutputs()));
   meta.fd = FileDescriptor(versions_->NewFileNumber(), 0, 0);
   ReadOptions ro;
   ro.total_order_seek = true;
@@ -1160,6 +1191,7 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
     int64_t _current_time = 0;
     env_->GetCurrentTime(&_current_time);  // ignore error
     const uint64_t current_time = static_cast<uint64_t>(_current_time);
+    meta.oldest_ancester_time = current_time;
 
     {
       auto write_hint = cfd->CalculateSSTWriteHint(0);
@@ -1180,8 +1212,8 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
         range_del_iters.emplace_back(range_del_iter);
       }
       s = BuildTable(
-          dbname_, env_, *cfd->ioptions(), mutable_cf_options,
-          env_options_for_compaction_, cfd->table_cache(), iter.get(),
+          dbname_, env_, fs_.get(), *cfd->ioptions(), mutable_cf_options,
+          file_options_for_compaction_, cfd->table_cache(), iter.get(),
           std::move(range_del_iters), &meta, cfd->internal_comparator(),
           cfd->int_tbl_prop_collector_factories(), cfd->GetID(), cfd->GetName(),
           snapshot_seqs, earliest_write_conflict_snapshot, snapshot_checker,
@@ -1209,7 +1241,8 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
     edit->AddFile(level, meta.fd.GetNumber(), meta.fd.GetPathId(),
                   meta.fd.GetFileSize(), meta.smallest, meta.largest,
                   meta.fd.smallest_seqno, meta.fd.largest_seqno,
-                  meta.marked_for_compaction);
+                  meta.marked_for_compaction, meta.oldest_blob_file_number,
+                  meta.oldest_ancester_time, meta.file_creation_time);
   }
 
   InternalStats::CompactionStats stats(CompactionReason::kFlush, 1);
@@ -1263,12 +1296,12 @@ Status DB::Open(const DBOptions& db_options, const std::string& dbname,
 Status DBImpl::CreateWAL(uint64_t log_file_num, uint64_t recycle_log_number,
                          size_t preallocate_block_size, log::Writer** new_log) {
   Status s;
-  std::unique_ptr<WritableFile> lfile;
+  std::unique_ptr<FSWritableFile> lfile;
 
   DBOptions db_options =
       BuildDBOptions(immutable_db_options_, mutable_db_options_);
-  EnvOptions opt_env_options =
-      env_->OptimizeForLogWrite(env_options_, db_options);
+  FileOptions opt_file_options =
+      fs_->OptimizeForLogWrite(file_options_, db_options);
   std::string log_fname =
       LogFileName(immutable_db_options_.wal_dir, log_file_num);
 
@@ -1278,10 +1311,10 @@ Status DBImpl::CreateWAL(uint64_t log_file_num, uint64_t recycle_log_number,
                    recycle_log_number);
     std::string old_log_fname =
         LogFileName(immutable_db_options_.wal_dir, recycle_log_number);
-    s = env_->ReuseWritableFile(log_fname, old_log_fname, &lfile,
-                                opt_env_options);
+    s = fs_->ReuseWritableFile(log_fname, old_log_fname, opt_file_options,
+                               &lfile, /*dbg=*/nullptr);
   } else {
-    s = NewWritableFile(env_, log_fname, &lfile, opt_env_options);
+    s = NewWritableFile(fs_.get(), log_fname, &lfile, opt_file_options);
   }
 
   if (s.ok()) {
@@ -1290,7 +1323,7 @@ Status DBImpl::CreateWAL(uint64_t log_file_num, uint64_t recycle_log_number,
 
     const auto& listeners = immutable_db_options_.listeners;
     std::unique_ptr<WritableFileWriter> file_writer(
-        new WritableFileWriter(std::move(lfile), log_fname, opt_env_options,
+        new WritableFileWriter(std::move(lfile), log_fname, opt_file_options,
                                env_, nullptr /* stats */, listeners));
     *new_log = new log::Writer(std::move(file_writer), log_file_num,
                                immutable_db_options_.recycle_log_file_num > 0,

db/db_impl/db_impl_secondary.cc

@@ -11,6 +11,7 @@
 #include "db/merge_context.h"
 #include "logging/auto_roll_logger.h"
 #include "monitoring/perf_context_imp.h"
+#include "util/cast_util.h"
 
 namespace rocksdb {
 
@@ -143,9 +144,10 @@ Status DBImplSecondary::MaybeInitLogReader(
 
     std::unique_ptr<SequentialFileReader> file_reader;
     {
-      std::unique_ptr<SequentialFile> file;
-      Status status = env_->NewSequentialFile(
-          fname, &file, env_->OptimizeForLogRead(env_options_));
+      std::unique_ptr<FSSequentialFile> file;
+      Status status = fs_->NewSequentialFile(
+          fname, fs_->OptimizeForLogRead(file_options_), &file,
+          nullptr);
       if (!status.ok()) {
         *log_reader = nullptr;
         return status;
@@ -405,7 +407,7 @@ ArenaWrappedDBIter* DBImplSecondary::NewIteratorImpl(
     const ReadOptions& read_options, ColumnFamilyData* cfd,
     SequenceNumber snapshot, ReadCallback* read_callback) {
   assert(nullptr != cfd);
-  SuperVersion* super_version = cfd->GetReferencedSuperVersion(&mutex_);
+  SuperVersion* super_version = cfd->GetReferencedSuperVersion(this);
   auto db_iter = NewArenaWrappedDbIterator(
       env_, read_options, *cfd->ioptions(), super_version->mutable_cf_options,
       snapshot,
@@ -497,45 +499,61 @@ Status DBImplSecondary::TryCatchUpWithPrimary() {
   // read the manifest and apply new changes to the secondary instance
   std::unordered_set<ColumnFamilyData*> cfds_changed;
   JobContext job_context(0, true /*create_superversion*/);
-  InstrumentedMutexLock lock_guard(&mutex_);
-  s = static_cast<ReactiveVersionSet*>(versions_.get())
-          ->ReadAndApply(&mutex_, &manifest_reader_, &cfds_changed);
+  {
+    InstrumentedMutexLock lock_guard(&mutex_);
+    s = static_cast_with_check<ReactiveVersionSet>(versions_.get())
+            ->ReadAndApply(&mutex_, &manifest_reader_, &cfds_changed);
 
-  ROCKS_LOG_INFO(immutable_db_options_.info_log, "Last sequence is %" PRIu64,
-                 static_cast<uint64_t>(versions_->LastSequence()));
-  for (ColumnFamilyData* cfd : cfds_changed) {
-    if (cfd->IsDropped()) {
-      ROCKS_LOG_DEBUG(immutable_db_options_.info_log, "[%s] is dropped\n",
-                      cfd->GetName().c_str());
-      continue;
+    ROCKS_LOG_INFO(immutable_db_options_.info_log, "Last sequence is %" PRIu64,
+                   static_cast<uint64_t>(versions_->LastSequence()));
+    for (ColumnFamilyData* cfd : cfds_changed) {
+      if (cfd->IsDropped()) {
+        ROCKS_LOG_DEBUG(immutable_db_options_.info_log, "[%s] is dropped\n",
+                        cfd->GetName().c_str());
+        continue;
+      }
+      VersionStorageInfo::LevelSummaryStorage tmp;
+      ROCKS_LOG_DEBUG(immutable_db_options_.info_log,
+                      "[%s] Level summary: %s\n", cfd->GetName().c_str(),
+                      cfd->current()->storage_info()->LevelSummary(&tmp));
     }
-    VersionStorageInfo::LevelSummaryStorage tmp;
-    ROCKS_LOG_DEBUG(immutable_db_options_.info_log, "[%s] Level summary: %s\n",
-                    cfd->GetName().c_str(),
-                    cfd->current()->storage_info()->LevelSummary(&tmp));
-  }
 
-  // list wal_dir to discover new WALs and apply new changes to the secondary
-  // instance
-  if (s.ok()) {
-    s = FindAndRecoverLogFiles(&cfds_changed, &job_context);
-  }
-  if (s.IsPathNotFound()) {
-    ROCKS_LOG_INFO(immutable_db_options_.info_log,
-                   "Secondary tries to read WAL, but WAL file(s) have already "
-                   "been purged by primary.");
-    s = Status::OK();
-  }
-  if (s.ok()) {
-    for (auto cfd : cfds_changed) {
-      cfd->imm()->RemoveOldMemTables(cfd->GetLogNumber(),
-                                     &job_context.memtables_to_free);
-      auto& sv_context = job_context.superversion_contexts.back();
-      cfd->InstallSuperVersion(&sv_context, &mutex_);
-      sv_context.NewSuperVersion();
+    // list wal_dir to discover new WALs and apply new changes to the secondary
+    // instance
+    if (s.ok()) {
+      s = FindAndRecoverLogFiles(&cfds_changed, &job_context);
+    }
+    if (s.IsPathNotFound()) {
+      ROCKS_LOG_INFO(
+          immutable_db_options_.info_log,
+          "Secondary tries to read WAL, but WAL file(s) have already "
+          "been purged by primary.");
+      s = Status::OK();
+    }
+    if (s.ok()) {
+      for (auto cfd : cfds_changed) {
+        cfd->imm()->RemoveOldMemTables(cfd->GetLogNumber(),
+                                       &job_context.memtables_to_free);
+        auto& sv_context = job_context.superversion_contexts.back();
+        cfd->InstallSuperVersion(&sv_context, &mutex_);
+        sv_context.NewSuperVersion();
+      }
     }
-    job_context.Clean();
   }
+  job_context.Clean();
+
+  // Cleanup unused, obsolete files.
+  JobContext purge_files_job_context(0);
+  {
+    InstrumentedMutexLock lock_guard(&mutex_);
+    // Currently, secondary instance does not own the database files, thus it
+    // is unnecessary for the secondary to force full scan.
+    FindObsoleteFiles(&purge_files_job_context, /*force=*/false);
+  }
+  if (purge_files_job_context.HaveSomethingToDelete()) {
+    PurgeObsoleteFiles(purge_files_job_context);
+  }
+  purge_files_job_context.Clean();
   return s;
 }
 
@@ -583,7 +601,7 @@ Status DB::OpenAsSecondary(
   handles->clear();
   DBImplSecondary* impl = new DBImplSecondary(tmp_opts, dbname);
   impl->versions_.reset(new ReactiveVersionSet(
-      dbname, &impl->immutable_db_options_, impl->env_options_,
+      dbname, &impl->immutable_db_options_, impl->file_options_,
       impl->table_cache_.get(), impl->write_buffer_manager_,
       &impl->write_controller_));
   impl->column_family_memtables_.reset(

db/db_impl/db_impl_secondary.h

@@ -172,6 +172,24 @@ class DBImplSecondary : public DBImpl {
     return Status::NotSupported("Not supported operation in secondary mode.");
   }
 
+  using DBImpl::SetDBOptions;
+  Status SetDBOptions(const std::unordered_map<std::string, std::string>&
+                      /*options_map*/) override {
+    // Currently not supported because changing certain options may cause
+    // flush/compaction.
+    return Status::NotSupported("Not supported operation in secondary mode.");
+  }
+
+  using DBImpl::SetOptions;
+  Status SetOptions(
+      ColumnFamilyHandle* /*cfd*/,
+      const std::unordered_map<std::string, std::string>& /*options_map*/)
+      override {
+    // Currently not supported because changing certain options may cause
+    // flush/compaction and/or write to MANIFEST.
+    return Status::NotSupported("Not supported operation in secondary mode.");
+  }
+
   using DBImpl::SyncWAL;
   Status SyncWAL() override {
     return Status::NotSupported("Not supported operation in secondary mode.");
@@ -269,6 +287,14 @@ class DBImplSecondary : public DBImpl {
     return s;
   }
 
+  bool OwnTablesAndLogs() const override {
+    // Currently, the secondary instance does not own the database files. It
+    // simply opens the files of the primary instance and tracks their file
+    // descriptors until they become obsolete. In the future, the secondary may
+    // create links to database files. OwnTablesAndLogs will return true then.
+    return false;
+  }
+
  private:
   friend class DB;
 

db/db_impl/db_impl_write.cc

@@ -131,6 +131,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
                               log_used, log_ref, &seq, sub_batch_cnt,
                               pre_release_callback, kDoAssignOrder,
                               kDoPublishLastSeq, disable_memtable);
+    TEST_SYNC_POINT("DBImpl::WriteImpl:UnorderedWriteAfterWriteWAL");
     if (!status.ok()) {
       return status;
     }
@@ -138,6 +139,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
       *seq_used = seq;
     }
     if (!disable_memtable) {
+      TEST_SYNC_POINT("DBImpl::WriteImpl:BeforeUnorderedWriteMemtable");
       status = UnorderedWriteMemtable(write_options, my_batch, callback,
                                       log_ref, seq, sub_batch_cnt);
     }
@@ -1246,14 +1248,23 @@ Status DBImpl::SwitchWAL(WriteContext* write_context) {
     }
     MaybeFlushStatsCF(&cfds);
   }
+  WriteThread::Writer nonmem_w;
+  if (two_write_queues_) {
+    nonmem_write_thread_.EnterUnbatched(&nonmem_w, &mutex_);
+  }
+
   for (const auto cfd : cfds) {
     cfd->Ref();
     status = SwitchMemtable(cfd, write_context);
-    cfd->Unref();
+    cfd->UnrefAndTryDelete();
     if (!status.ok()) {
       break;
     }
   }
+  if (two_write_queues_) {
+    nonmem_write_thread_.ExitUnbatched(&nonmem_w);
+  }
+
   if (status.ok()) {
     if (immutable_db_options_.atomic_flush) {
       AssignAtomicFlushSeq(cfds);
@@ -1314,17 +1325,25 @@ Status DBImpl::HandleWriteBufferFull(WriteContext* write_context) {
     MaybeFlushStatsCF(&cfds);
   }
 
+  WriteThread::Writer nonmem_w;
+  if (two_write_queues_) {
+    nonmem_write_thread_.EnterUnbatched(&nonmem_w, &mutex_);
+  }
   for (const auto cfd : cfds) {
     if (cfd->mem()->IsEmpty()) {
       continue;
     }
     cfd->Ref();
     status = SwitchMemtable(cfd, write_context);
-    cfd->Unref();
+    cfd->UnrefAndTryDelete();
     if (!status.ok()) {
       break;
     }
   }
+  if (two_write_queues_) {
+    nonmem_write_thread_.ExitUnbatched(&nonmem_w);
+  }
+
   if (status.ok()) {
     if (immutable_db_options_.atomic_flush) {
       AssignAtomicFlushSeq(cfds);
@@ -1499,15 +1518,16 @@ Status DBImpl::TrimMemtableHistory(WriteContext* context) {
   for (auto& cfd : cfds) {
     autovector<MemTable*> to_delete;
     cfd->imm()->TrimHistory(&to_delete, cfd->mem()->ApproximateMemoryUsage());
-    for (auto m : to_delete) {
-      delete m;
+    if (!to_delete.empty()) {
+      for (auto m : to_delete) {
+        delete m;
+      }
+      context->superversion_context.NewSuperVersion();
+      assert(context->superversion_context.new_superversion.get() != nullptr);
+      cfd->InstallSuperVersion(&context->superversion_context, &mutex_);
     }
-    context->superversion_context.NewSuperVersion();
-    assert(context->superversion_context.new_superversion.get() != nullptr);
-    cfd->InstallSuperVersion(&context->superversion_context, &mutex_);
 
-    if (cfd->Unref()) {
-      delete cfd;
+    if (cfd->UnrefAndTryDelete()) {
       cfd = nullptr;
     }
   }
@@ -1530,18 +1550,27 @@ Status DBImpl::ScheduleFlushes(WriteContext* context) {
     MaybeFlushStatsCF(&cfds);
   }
   Status status;
+  WriteThread::Writer nonmem_w;
+  if (two_write_queues_) {
+    nonmem_write_thread_.EnterUnbatched(&nonmem_w, &mutex_);
+  }
+
   for (auto& cfd : cfds) {
     if (!cfd->mem()->IsEmpty()) {
       status = SwitchMemtable(cfd, context);
     }
-    if (cfd->Unref()) {
-      delete cfd;
+    if (cfd->UnrefAndTryDelete()) {
       cfd = nullptr;
     }
     if (!status.ok()) {
       break;
     }
   }
+
+  if (two_write_queues_) {
+    nonmem_write_thread_.ExitUnbatched(&nonmem_w);
+  }
+
   if (status.ok()) {
     if (immutable_db_options_.atomic_flush) {
       AssignAtomicFlushSeq(cfds);
@@ -1572,15 +1601,11 @@ void DBImpl::NotifyOnMemTableSealed(ColumnFamilyData* /*cfd*/,
 
 // 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) {
   mutex_.AssertHeld();
   WriteThread::Writer nonmem_w;
-  if (two_write_queues_) {
-    // SwitchMemtable is a rare event. To simply the reasoning, we make sure
-    // that there is no concurrent thread writing to WAL.
-    nonmem_write_thread_.EnterUnbatched(&nonmem_w, &mutex_);
-  }
-
   std::unique_ptr<WritableFile> lfile;
   log::Writer* new_log = nullptr;
   MemTable* new_mem = nullptr;
@@ -1592,16 +1617,6 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
     return s;
   }
 
-  // In case of pipelined write is enabled, wait for all pending memtable
-  // writers.
-  if (immutable_db_options_.enable_pipelined_write) {
-    // Memtable writers may call DB::Get in case max_successive_merges > 0,
-    // which may lock mutex. Unlocking mutex here to avoid deadlock.
-    mutex_.Unlock();
-    write_thread_.WaitForMemTableWriters();
-    mutex_.Lock();
-  }
-
   // Attempt to switch to a new memtable and trigger flush of old.
   // Do this without holding the dbmutex lock.
   assert(versions_->prev_log_number() == 0);
@@ -1697,10 +1712,6 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
     error_handler_.SetBGError(s, BackgroundErrorReason::kMemTable);
     // Read back bg_error in order to get the right severity
     s = error_handler_.GetBGError();
-
-    if (two_write_queues_) {
-      nonmem_write_thread_.ExitUnbatched(&nonmem_w);
-    }
     return s;
   }
 
@@ -1731,9 +1742,6 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
   NotifyOnMemTableSealed(cfd, memtable_info);
   mutex_.Lock();
 #endif  // ROCKSDB_LITE
-  if (two_write_queues_) {
-    nonmem_write_thread_.ExitUnbatched(&nonmem_w);
-  }
   return s;
 }
 

db/db_impl/db_secondary_test.cc

@@ -195,6 +195,90 @@ TEST_F(DBSecondaryTest, OpenAsSecondary) {
   verify_db_func("new_foo_value", "new_bar_value");
 }
 
+namespace {
+class TraceFileEnv : public EnvWrapper {
+ public:
+  explicit TraceFileEnv(Env* _target) : EnvWrapper(_target) {}
+  Status NewRandomAccessFile(const std::string& f,
+                             std::unique_ptr<RandomAccessFile>* r,
+                             const EnvOptions& env_options) override {
+    class TracedRandomAccessFile : public RandomAccessFile {
+     public:
+      TracedRandomAccessFile(std::unique_ptr<RandomAccessFile>&& target,
+                             std::atomic<int>& counter)
+          : target_(std::move(target)), files_closed_(counter) {}
+      ~TracedRandomAccessFile() override {
+        files_closed_.fetch_add(1, std::memory_order_relaxed);
+      }
+      Status Read(uint64_t offset, size_t n, Slice* result,
+                  char* scratch) const override {
+        return target_->Read(offset, n, result, scratch);
+      }
+
+     private:
+      std::unique_ptr<RandomAccessFile> target_;
+      std::atomic<int>& files_closed_;
+    };
+    Status s = target()->NewRandomAccessFile(f, r, env_options);
+    if (s.ok()) {
+      r->reset(new TracedRandomAccessFile(std::move(*r), files_closed_));
+    }
+    return s;
+  }
+
+  int files_closed() const {
+    return files_closed_.load(std::memory_order_relaxed);
+  }
+
+ private:
+  std::atomic<int> files_closed_{0};
+};
+}  // namespace
+
+TEST_F(DBSecondaryTest, SecondaryCloseFiles) {
+  Options options;
+  options.env = env_;
+  options.max_open_files = 1;
+  options.disable_auto_compactions = true;
+  Reopen(options);
+  Options options1;
+  std::unique_ptr<Env> traced_env(new TraceFileEnv(env_));
+  options1.env = traced_env.get();
+  OpenSecondary(options1);
+
+  static const auto verify_db = [&]() {
+    std::unique_ptr<Iterator> iter1(dbfull()->NewIterator(ReadOptions()));
+    std::unique_ptr<Iterator> iter2(db_secondary_->NewIterator(ReadOptions()));
+    for (iter1->SeekToFirst(), iter2->SeekToFirst();
+         iter1->Valid() && iter2->Valid(); iter1->Next(), iter2->Next()) {
+      ASSERT_EQ(iter1->key(), iter2->key());
+      ASSERT_EQ(iter1->value(), iter2->value());
+    }
+    ASSERT_FALSE(iter1->Valid());
+    ASSERT_FALSE(iter2->Valid());
+  };
+
+  ASSERT_OK(Put("a", "value"));
+  ASSERT_OK(Put("c", "value"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
+  verify_db();
+
+  ASSERT_OK(Put("b", "value"));
+  ASSERT_OK(Put("d", "value"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
+  verify_db();
+
+  ASSERT_OK(dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
+  ASSERT_EQ(2, static_cast<TraceFileEnv*>(traced_env.get())->files_closed());
+
+  Status s = db_secondary_->SetDBOptions({{"max_open_files", "-1"}});
+  ASSERT_TRUE(s.IsNotSupported());
+  CloseSecondary();
+}
+
 TEST_F(DBSecondaryTest, OpenAsSecondaryWALTailing) {
   Options options;
   options.env = env_;

db/db_iter.cc

@@ -856,6 +856,7 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
   // In case read_callback presents, the value we seek to may not be visible.
   // Find the next value that's visible.
   ParsedInternalKey ikey;
+  is_blob_ = false;
   while (true) {
     if (!iter_.Valid()) {
       valid_ = false;
@@ -897,6 +898,7 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
   if (ikey.type == kTypeValue || ikey.type == kTypeBlobIndex) {
     assert(iter_.iter()->IsValuePinned());
     pinned_value_ = iter_.value();
+    is_blob_ = (ikey.type == kTypeBlobIndex);
     valid_ = true;
     return true;
   }

db/db_iterator_test.cc

@@ -804,6 +804,50 @@ TEST_P(DBIteratorTest, IteratorPinsRef) {
   } while (ChangeCompactOptions());
 }
 
+TEST_P(DBIteratorTest, IteratorDeleteAfterCfDelete) {
+  CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
+
+  Put(1, "foo", "delete-cf-then-delete-iter");
+  Put(1, "hello", "value2");
+
+  ColumnFamilyHandle* cf = handles_[1];
+  ReadOptions ro;
+
+  auto* iter = db_->NewIterator(ro, cf);
+  iter->SeekToFirst();
+  ASSERT_EQ(IterStatus(iter), "foo->delete-cf-then-delete-iter");
+
+  // delete CF handle
+  db_->DestroyColumnFamilyHandle(cf);
+  handles_.erase(std::begin(handles_) + 1);
+
+  // delete Iterator after CF handle is deleted
+  iter->Next();
+  ASSERT_EQ(IterStatus(iter), "hello->value2");
+  delete iter;
+}
+
+TEST_P(DBIteratorTest, IteratorDeleteAfterCfDrop) {
+  CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
+
+  Put(1, "foo", "drop-cf-then-delete-iter");
+
+  ReadOptions ro;
+  ColumnFamilyHandle* cf = handles_[1];
+
+  auto* iter = db_->NewIterator(ro, cf);
+  iter->SeekToFirst();
+  ASSERT_EQ(IterStatus(iter), "foo->drop-cf-then-delete-iter");
+
+  // drop and delete CF
+  db_->DropColumnFamily(cf);
+  db_->DestroyColumnFamilyHandle(cf);
+  handles_.erase(std::begin(handles_) + 1);
+
+  // delete Iterator after CF handle is dropped
+  delete iter;
+}
+
 // SetOptions not defined in ROCKSDB LITE
 #ifndef ROCKSDB_LITE
 TEST_P(DBIteratorTest, DBIteratorBoundTest) {
@@ -1297,19 +1341,19 @@ class DBIteratorTestForPinnedData : public DBIteratorTest {
 
       // Insert data to true_data map and to DB
       true_data[k] = v;
-      if (rnd.OneIn(static_cast<int>(100.0 / merge_percentage))) {
+      if (rnd.PercentTrue(merge_percentage)) {
         ASSERT_OK(db_->Merge(WriteOptions(), k, v));
       } else {
         ASSERT_OK(Put(k, v));
       }
 
       // Pick random keys to be used to test Seek()
-      if (rnd.OneIn(static_cast<int>(100.0 / seeks_percentage))) {
+      if (rnd.PercentTrue(seeks_percentage)) {
         random_keys.push_back(k);
       }
 
       // Delete some random keys
-      if (rnd.OneIn(static_cast<int>(100.0 / delete_percentage))) {
+      if (rnd.PercentTrue(delete_percentage)) {
         deleted_keys.push_back(k);
         true_data.erase(k);
         ASSERT_OK(Delete(k));
@@ -1867,7 +1911,7 @@ TEST_P(DBIteratorTest, IterPrevKeyCrossingBlocksRandomized) {
   // make sure that we dont merge them while flushing but actually
   // merge them in the read path
   for (int i = 0; i < kNumKeys; i++) {
-    if (rnd.OneIn(static_cast<int>(100.0 / kNoMergeOpPercentage))) {
+    if (rnd.PercentTrue(kNoMergeOpPercentage)) {
       // Dont give merge operations for some keys
       continue;
     }
@@ -1883,7 +1927,7 @@ TEST_P(DBIteratorTest, IterPrevKeyCrossingBlocksRandomized) {
   ASSERT_OK(Flush());
 
   for (int i = 0; i < kNumKeys; i++) {
-    if (rnd.OneIn(static_cast<int>(100.0 / kDeletePercentage))) {
+    if (rnd.PercentTrue(kDeletePercentage)) {
       gen_key = Key(i);
 
       ASSERT_OK(Delete(gen_key));

db/db_merge_operator_test.cc

@@ -150,6 +150,8 @@ TEST_F(DBMergeOperatorTest, MergeErrorOnWrite) {
   options.create_if_missing = true;
   options.merge_operator.reset(new TestPutOperator());
   options.max_successive_merges = 3;
+  // allow_concurrent_memtable_write is incompatible with max_successive_merges
+  options.allow_concurrent_memtable_write = false;
   options.env = env_;
   Reopen(options);
   ASSERT_OK(Merge("k1", "v1"));

db/db_options_test.cc

@@ -606,6 +606,76 @@ TEST_F(DBOptionsTest, SanitizeDelayedWriteRate) {
   ASSERT_EQ(31 * 1024 * 1024, dbfull()->GetDBOptions().delayed_write_rate);
 }
 
+TEST_F(DBOptionsTest, SanitizeUniversalTTLCompaction) {
+  Options options;
+  options.compaction_style = kCompactionStyleUniversal;
+
+  options.ttl = 0;
+  options.periodic_compaction_seconds = 0;
+  Reopen(options);
+  ASSERT_EQ(0, dbfull()->GetOptions().ttl);
+  ASSERT_EQ(0, dbfull()->GetOptions().periodic_compaction_seconds);
+
+  options.ttl = 0;
+  options.periodic_compaction_seconds = 100;
+  Reopen(options);
+  ASSERT_EQ(0, dbfull()->GetOptions().ttl);
+  ASSERT_EQ(100, dbfull()->GetOptions().periodic_compaction_seconds);
+
+  options.ttl = 100;
+  options.periodic_compaction_seconds = 0;
+  Reopen(options);
+  ASSERT_EQ(100, dbfull()->GetOptions().ttl);
+  ASSERT_EQ(100, dbfull()->GetOptions().periodic_compaction_seconds);
+
+  options.ttl = 100;
+  options.periodic_compaction_seconds = 500;
+  Reopen(options);
+  ASSERT_EQ(100, dbfull()->GetOptions().ttl);
+  ASSERT_EQ(100, dbfull()->GetOptions().periodic_compaction_seconds);
+}
+
+TEST_F(DBOptionsTest, SanitizeTtlDefault) {
+  Options options;
+  Reopen(options);
+  ASSERT_EQ(30 * 24 * 60 * 60, dbfull()->GetOptions().ttl);
+
+  options.compaction_style = kCompactionStyleLevel;
+  options.ttl = 0;
+  Reopen(options);
+  ASSERT_EQ(0, dbfull()->GetOptions().ttl);
+
+  options.ttl = 100;
+  Reopen(options);
+  ASSERT_EQ(100, dbfull()->GetOptions().ttl);
+}
+
+TEST_F(DBOptionsTest, SanitizeFIFOPeriodicCompaction) {
+  Options options;
+  options.compaction_style = kCompactionStyleFIFO;
+  options.ttl = 0;
+  Reopen(options);
+  ASSERT_EQ(30 * 24 * 60 * 60, dbfull()->GetOptions().ttl);
+
+  options.ttl = 100;
+  Reopen(options);
+  ASSERT_EQ(100, dbfull()->GetOptions().ttl);
+
+  options.ttl = 100 * 24 * 60 * 60;
+  Reopen(options);
+  ASSERT_EQ(100 * 24 * 60 * 60, dbfull()->GetOptions().ttl);
+
+  options.ttl = 200;
+  options.periodic_compaction_seconds = 300;
+  Reopen(options);
+  ASSERT_EQ(200, dbfull()->GetOptions().ttl);
+
+  options.ttl = 500;
+  options.periodic_compaction_seconds = 300;
+  Reopen(options);
+  ASSERT_EQ(300, dbfull()->GetOptions().ttl);
+}
+
 TEST_F(DBOptionsTest, SetFIFOCompactionOptions) {
   Options options;
   options.compaction_style = kCompactionStyleFIFO;

db/db_range_del_test.cc

@@ -1427,6 +1427,47 @@ TEST_F(DBRangeDelTest, SnapshotPreventsDroppedKeys) {
   db_->ReleaseSnapshot(snapshot);
 }
 
+TEST_F(DBRangeDelTest, SnapshotPreventsDroppedKeysInImmMemTables) {
+  const int kFileBytes = 1 << 20;
+
+  Options options = CurrentOptions();
+  options.compression = kNoCompression;
+  options.disable_auto_compactions = true;
+  options.target_file_size_base = kFileBytes;
+  Reopen(options);
+
+  // block flush thread -> pin immtables in memory
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->LoadDependency({
+      {"SnapshotPreventsDroppedKeysInImmMemTables:AfterNewIterator",
+       "DBImpl::BGWorkFlush"},
+  });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  ASSERT_OK(Put(Key(0), "a"));
+  std::unique_ptr<const Snapshot, std::function<void(const Snapshot*)>>
+      snapshot(db_->GetSnapshot(),
+               [this](const Snapshot* s) { db_->ReleaseSnapshot(s); });
+
+  ASSERT_OK(db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), Key(0),
+                             Key(10)));
+
+  ASSERT_OK(dbfull()->TEST_SwitchMemtable());
+
+  ReadOptions read_opts;
+  read_opts.snapshot = snapshot.get();
+  std::unique_ptr<Iterator> iter(db_->NewIterator(read_opts));
+
+  TEST_SYNC_POINT("SnapshotPreventsDroppedKeysInImmMemTables:AfterNewIterator");
+
+  iter->SeekToFirst();
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ(Key(0), iter->key());
+
+  iter->Next();
+  ASSERT_FALSE(iter->Valid());
+}
+
 TEST_F(DBRangeDelTest, RangeTombstoneWrittenToMinimalSsts) {
   // Adapted from
   // https://github.com/cockroachdb/cockroach/blob/de8b3ea603dd1592d9dc26443c2cc92c356fbc2f/pkg/storage/engine/rocksdb_test.go#L1267-L1398.
@@ -1521,6 +1562,84 @@ TEST_F(DBRangeDelTest, RangeTombstoneWrittenToMinimalSsts) {
   ASSERT_EQ(1, num_range_deletions);
 }
 
+TEST_F(DBRangeDelTest, OverlappedTombstones) {
+  const int kNumPerFile = 4, kNumFiles = 2;
+  Options options = CurrentOptions();
+  options.disable_auto_compactions = true;
+  options.max_compaction_bytes = 9 * 1024;
+  DestroyAndReopen(options);
+  Random rnd(301);
+  for (int i = 0; i < kNumFiles; ++i) {
+    std::vector<std::string> values;
+    // Write 12K (4 values, each 3K)
+    for (int j = 0; j < kNumPerFile; j++) {
+      values.push_back(RandomString(&rnd, 3 << 10));
+      ASSERT_OK(Put(Key(i * kNumPerFile + j), values[j]));
+    }
+  }
+  ASSERT_OK(db_->Flush(FlushOptions()));
+  ASSERT_EQ(1, NumTableFilesAtLevel(0));
+  MoveFilesToLevel(2);
+  ASSERT_EQ(2, NumTableFilesAtLevel(2));
+
+  ASSERT_OK(db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), Key(1),
+                             Key((kNumFiles)*kNumPerFile + 1)));
+  ASSERT_OK(db_->Flush(FlushOptions()));
+
+  ASSERT_EQ(1, NumTableFilesAtLevel(0));
+
+  dbfull()->TEST_CompactRange(0, nullptr, nullptr, nullptr,
+                              true /* disallow_trivial_move */);
+
+  // The tombstone range is not broken up into multiple SSTs which may incur a
+  // large compaction with L2.
+  ASSERT_EQ(1, NumTableFilesAtLevel(1));
+  std::vector<std::vector<FileMetaData>> files;
+  dbfull()->TEST_CompactRange(1, nullptr, nullptr, nullptr,
+                              true /* disallow_trivial_move */);
+  ASSERT_EQ(1, NumTableFilesAtLevel(2));
+  ASSERT_EQ(0, NumTableFilesAtLevel(1));
+}
+
+TEST_F(DBRangeDelTest, OverlappedKeys) {
+  const int kNumPerFile = 4, kNumFiles = 2;
+  Options options = CurrentOptions();
+  options.disable_auto_compactions = true;
+  options.max_compaction_bytes = 9 * 1024;
+  DestroyAndReopen(options);
+  Random rnd(301);
+  for (int i = 0; i < kNumFiles; ++i) {
+    std::vector<std::string> values;
+    // Write 12K (4 values, each 3K)
+    for (int j = 0; j < kNumPerFile; j++) {
+      values.push_back(RandomString(&rnd, 3 << 10));
+      ASSERT_OK(Put(Key(i * kNumPerFile + j), values[j]));
+    }
+  }
+  ASSERT_OK(db_->Flush(FlushOptions()));
+  ASSERT_EQ(1, NumTableFilesAtLevel(0));
+  MoveFilesToLevel(2);
+  ASSERT_EQ(2, NumTableFilesAtLevel(2));
+
+  for (int i = 1; i < kNumFiles * kNumPerFile + 1; i++) {
+    ASSERT_OK(Put(Key(i), "0x123"));
+  }
+  ASSERT_OK(db_->Flush(FlushOptions()));
+  ASSERT_EQ(1, NumTableFilesAtLevel(0));
+
+  // The key range is broken up into three SSTs to avoid a future big compaction
+  // with the grandparent
+  dbfull()->TEST_CompactRange(0, nullptr, nullptr, nullptr,
+                              true /* disallow_trivial_move */);
+  ASSERT_EQ(3, NumTableFilesAtLevel(1));
+
+  std::vector<std::vector<FileMetaData>> files;
+  dbfull()->TEST_CompactRange(1, nullptr, nullptr, nullptr,
+                              true /* disallow_trivial_move */);
+  ASSERT_EQ(1, NumTableFilesAtLevel(2));
+  ASSERT_EQ(0, NumTableFilesAtLevel(1));
+}
+
 #endif  // ROCKSDB_LITE
 
 }  // namespace rocksdb

db/db_test.cc

@@ -24,6 +24,7 @@
 #endif
 
 #include "cache/lru_cache.h"
+#include "db/blob_index.h"
 #include "db/db_impl/db_impl.h"
 #include "db/db_test_util.h"
 #include "db/dbformat.h"
@@ -1019,39 +1020,149 @@ TEST_F(DBTest, FailMoreDbPaths) {
   ASSERT_TRUE(TryReopen(options).IsNotSupported());
 }
 
-void CheckColumnFamilyMeta(const ColumnFamilyMetaData& cf_meta) {
+void CheckColumnFamilyMeta(
+    const ColumnFamilyMetaData& cf_meta,
+    const std::vector<std::vector<FileMetaData>>& files_by_level,
+    uint64_t start_time, uint64_t end_time) {
+  ASSERT_EQ(cf_meta.name, kDefaultColumnFamilyName);
+  ASSERT_EQ(cf_meta.levels.size(), files_by_level.size());
+
   uint64_t cf_size = 0;
-  uint64_t cf_csize = 0;
   size_t file_count = 0;
-  for (auto level_meta : cf_meta.levels) {
+
+  for (size_t i = 0; i < cf_meta.levels.size(); ++i) {
+    const auto& level_meta_from_cf = cf_meta.levels[i];
+    const auto& level_meta_from_files = files_by_level[i];
+
+    ASSERT_EQ(level_meta_from_cf.level, i);
+    ASSERT_EQ(level_meta_from_cf.files.size(), level_meta_from_files.size());
+
+    file_count += level_meta_from_cf.files.size();
+
     uint64_t level_size = 0;
-    uint64_t level_csize = 0;
-    file_count += level_meta.files.size();
-    for (auto file_meta : level_meta.files) {
-      level_size += file_meta.size;
+    for (size_t j = 0; j < level_meta_from_cf.files.size(); ++j) {
+      const auto& file_meta_from_cf = level_meta_from_cf.files[j];
+      const auto& file_meta_from_files = level_meta_from_files[j];
+
+      level_size += file_meta_from_cf.size;
+
+      ASSERT_EQ(file_meta_from_cf.file_number,
+                file_meta_from_files.fd.GetNumber());
+      ASSERT_EQ(file_meta_from_cf.file_number,
+                TableFileNameToNumber(file_meta_from_cf.name));
+      ASSERT_EQ(file_meta_from_cf.size, file_meta_from_files.fd.file_size);
+      ASSERT_EQ(file_meta_from_cf.smallest_seqno,
+                file_meta_from_files.fd.smallest_seqno);
+      ASSERT_EQ(file_meta_from_cf.largest_seqno,
+                file_meta_from_files.fd.largest_seqno);
+      ASSERT_EQ(file_meta_from_cf.smallestkey,
+                file_meta_from_files.smallest.user_key().ToString());
+      ASSERT_EQ(file_meta_from_cf.largestkey,
+                file_meta_from_files.largest.user_key().ToString());
+      ASSERT_EQ(file_meta_from_cf.oldest_blob_file_number,
+                file_meta_from_files.oldest_blob_file_number);
+      ASSERT_EQ(file_meta_from_cf.oldest_ancester_time,
+                file_meta_from_files.oldest_ancester_time);
+      ASSERT_EQ(file_meta_from_cf.file_creation_time,
+                file_meta_from_files.file_creation_time);
+      ASSERT_GE(file_meta_from_cf.file_creation_time, start_time);
+      ASSERT_LE(file_meta_from_cf.file_creation_time, end_time);
+      ASSERT_GE(file_meta_from_cf.oldest_ancester_time, start_time);
+      ASSERT_LE(file_meta_from_cf.oldest_ancester_time, end_time);
     }
-    ASSERT_EQ(level_meta.size, level_size);
+
+    ASSERT_EQ(level_meta_from_cf.size, level_size);
     cf_size += level_size;
-    cf_csize += level_csize;
   }
+
   ASSERT_EQ(cf_meta.file_count, file_count);
   ASSERT_EQ(cf_meta.size, cf_size);
 }
 
+void CheckLiveFilesMeta(
+    const std::vector<LiveFileMetaData>& live_file_meta,
+    const std::vector<std::vector<FileMetaData>>& files_by_level) {
+  size_t total_file_count = 0;
+  for (const auto& f : files_by_level) {
+    total_file_count += f.size();
+  }
+
+  ASSERT_EQ(live_file_meta.size(), total_file_count);
+
+  int level = 0;
+  int i = 0;
+
+  for (const auto& meta : live_file_meta) {
+    if (level != meta.level) {
+      level = meta.level;
+      i = 0;
+    }
+
+    ASSERT_LT(i, files_by_level[level].size());
+
+    const auto& expected_meta = files_by_level[level][i];
+
+    ASSERT_EQ(meta.column_family_name, kDefaultColumnFamilyName);
+    ASSERT_EQ(meta.file_number, expected_meta.fd.GetNumber());
+    ASSERT_EQ(meta.file_number, TableFileNameToNumber(meta.name));
+    ASSERT_EQ(meta.size, expected_meta.fd.file_size);
+    ASSERT_EQ(meta.smallest_seqno, expected_meta.fd.smallest_seqno);
+    ASSERT_EQ(meta.largest_seqno, expected_meta.fd.largest_seqno);
+    ASSERT_EQ(meta.smallestkey, expected_meta.smallest.user_key().ToString());
+    ASSERT_EQ(meta.largestkey, expected_meta.largest.user_key().ToString());
+    ASSERT_EQ(meta.oldest_blob_file_number,
+              expected_meta.oldest_blob_file_number);
+
+    ++i;
+  }
+}
+
 #ifndef ROCKSDB_LITE
-TEST_F(DBTest, ColumnFamilyMetaDataTest) {
+TEST_F(DBTest, MetaDataTest) {
   Options options = CurrentOptions();
   options.create_if_missing = true;
+  options.disable_auto_compactions = true;
+
+  int64_t temp_time = 0;
+  options.env->GetCurrentTime(&temp_time);
+  uint64_t start_time = static_cast<uint64_t>(temp_time);
+
   DestroyAndReopen(options);
 
   Random rnd(301);
   int key_index = 0;
-  ColumnFamilyMetaData cf_meta;
   for (int i = 0; i < 100; ++i) {
-    GenerateNewFile(&rnd, &key_index);
-    db_->GetColumnFamilyMetaData(&cf_meta);
-    CheckColumnFamilyMeta(cf_meta);
+    // Add a single blob reference to each file
+    std::string blob_index;
+    BlobIndex::EncodeBlob(&blob_index, /* blob_file_number */ i + 1000,
+                          /* offset */ 1234, /* size */ 5678, kNoCompression);
+
+    WriteBatch batch;
+    ASSERT_OK(WriteBatchInternal::PutBlobIndex(&batch, 0, Key(key_index),
+                                               blob_index));
+    ASSERT_OK(dbfull()->Write(WriteOptions(), &batch));
+
+    ++key_index;
+
+    // Fill up the rest of the file with random values.
+    GenerateNewFile(&rnd, &key_index, /* nowait */ true);
+
+    Flush();
   }
+
+  std::vector<std::vector<FileMetaData>> files_by_level;
+  dbfull()->TEST_GetFilesMetaData(db_->DefaultColumnFamily(), &files_by_level);
+
+  options.env->GetCurrentTime(&temp_time);
+  uint64_t end_time = static_cast<uint64_t>(temp_time);
+
+  ColumnFamilyMetaData cf_meta;
+  db_->GetColumnFamilyMetaData(&cf_meta);
+  CheckColumnFamilyMeta(cf_meta, files_by_level, start_time, end_time);
+
+  std::vector<LiveFileMetaData> live_file_meta;
+  db_->GetLiveFilesMetaData(&live_file_meta);
+  CheckLiveFilesMeta(live_file_meta, files_by_level);
 }
 
 namespace {
@@ -1594,6 +1705,7 @@ TEST_F(DBTest, Snapshot) {
     ASSERT_EQ(1U, GetNumSnapshots());
     uint64_t time_snap1 = GetTimeOldestSnapshots();
     ASSERT_GT(time_snap1, 0U);
+    ASSERT_EQ(GetSequenceOldestSnapshots(), s1->GetSequenceNumber());
     Put(0, "foo", "0v2");
     Put(1, "foo", "1v2");
 
@@ -1602,6 +1714,7 @@ TEST_F(DBTest, Snapshot) {
     const Snapshot* s2 = db_->GetSnapshot();
     ASSERT_EQ(2U, GetNumSnapshots());
     ASSERT_EQ(time_snap1, GetTimeOldestSnapshots());
+    ASSERT_EQ(GetSequenceOldestSnapshots(), s1->GetSequenceNumber());
     Put(0, "foo", "0v3");
     Put(1, "foo", "1v3");
 
@@ -1609,6 +1722,7 @@ TEST_F(DBTest, Snapshot) {
       ManagedSnapshot s3(db_);
       ASSERT_EQ(3U, GetNumSnapshots());
       ASSERT_EQ(time_snap1, GetTimeOldestSnapshots());
+      ASSERT_EQ(GetSequenceOldestSnapshots(), s1->GetSequenceNumber());
 
       Put(0, "foo", "0v4");
       Put(1, "foo", "1v4");
@@ -1624,6 +1738,7 @@ TEST_F(DBTest, Snapshot) {
 
     ASSERT_EQ(2U, GetNumSnapshots());
     ASSERT_EQ(time_snap1, GetTimeOldestSnapshots());
+    ASSERT_EQ(GetSequenceOldestSnapshots(), s1->GetSequenceNumber());
     ASSERT_EQ("0v1", Get(0, "foo", s1));
     ASSERT_EQ("1v1", Get(1, "foo", s1));
     ASSERT_EQ("0v2", Get(0, "foo", s2));
@@ -1638,9 +1753,11 @@ TEST_F(DBTest, Snapshot) {
     ASSERT_EQ("1v4", Get(1, "foo"));
     ASSERT_EQ(1U, GetNumSnapshots());
     ASSERT_LT(time_snap1, GetTimeOldestSnapshots());
+    ASSERT_EQ(GetSequenceOldestSnapshots(), s2->GetSequenceNumber());
 
     db_->ReleaseSnapshot(s2);
     ASSERT_EQ(0U, GetNumSnapshots());
+    ASSERT_EQ(GetSequenceOldestSnapshots(), 0);
     ASSERT_EQ("0v4", Get(0, "foo"));
     ASSERT_EQ("1v4", Get(1, "foo"));
   } while (ChangeOptions());
@@ -2798,6 +2915,11 @@ class ModelDB : public DB {
     return Status::OK();
   }
 
+  virtual Status GetCreationTimeOfOldestFile(
+      uint64_t* /*creation_time*/) override {
+    return Status::NotSupported();
+  }
+
   Status DeleteFile(std::string /*name*/) override { return Status::OK(); }
 
   Status GetUpdatesSince(
@@ -3219,10 +3341,10 @@ TEST_F(DBTest, FIFOCompactionWithTTLAndMaxOpenFilesTest) {
   options.create_if_missing = true;
   options.ttl = 600;  // seconds
 
-  // Check that it is not supported with max_open_files != -1.
+  // TTL is now supported with max_open_files != -1.
   options.max_open_files = 100;
   options = CurrentOptions(options);
-  ASSERT_TRUE(TryReopen(options).IsNotSupported());
+  ASSERT_OK(TryReopen(options));
 
   options.max_open_files = -1;
   ASSERT_OK(TryReopen(options));
@@ -4729,6 +4851,7 @@ TEST_F(DBTest, DynamicCompactionOptions) {
 // Even more FIFOCompactionTests are at DBTest.FIFOCompaction* .
 TEST_F(DBTest, DynamicFIFOCompactionOptions) {
   Options options;
+  options.ttl = 0;
   options.create_if_missing = true;
   DestroyAndReopen(options);
 
@@ -6088,10 +6211,10 @@ TEST_F(DBTest, CreateColumnFamilyShouldFailOnIncompatibleOptions) {
   Reopen(options);
 
   ColumnFamilyOptions cf_options(options);
-  // ttl is only supported when max_open_files is -1.
+  // ttl is now supported when max_open_files is -1.
   cf_options.ttl = 3600;
   ColumnFamilyHandle* handle;
-  ASSERT_NOK(db_->CreateColumnFamily(cf_options, "pikachu", &handle));
+  ASSERT_OK(db_->CreateColumnFamily(cf_options, "pikachu", &handle));
   delete handle;
 }
 
@@ -6271,6 +6394,116 @@ TEST_F(DBTest, LargeBlockSizeTest) {
   ASSERT_NOK(TryReopenWithColumnFamilies({"default", "pikachu"}, options));
 }
 
+#ifndef ROCKSDB_LITE
+
+TEST_F(DBTest, CreationTimeOfOldestFile) {
+  const int kNumKeysPerFile = 32;
+  const int kNumLevelFiles = 2;
+  const int kValueSize = 100;
+
+  Options options = CurrentOptions();
+  options.max_open_files = -1;
+  env_->time_elapse_only_sleep_ = false;
+  options.env = env_;
+
+  env_->addon_time_.store(0);
+  DestroyAndReopen(options);
+
+  bool set_file_creation_time_to_zero = true;
+  int idx = 0;
+
+  int64_t time_1 = 0;
+  env_->GetCurrentTime(&time_1);
+  const uint64_t uint_time_1 = static_cast<uint64_t>(time_1);
+
+  // Add 50 hours
+  env_->addon_time_.fetch_add(50 * 60 * 60);
+
+  int64_t time_2 = 0;
+  env_->GetCurrentTime(&time_2);
+  const uint64_t uint_time_2 = static_cast<uint64_t>(time_2);
+
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "PropertyBlockBuilder::AddTableProperty:Start", [&](void* arg) {
+        TableProperties* props = reinterpret_cast<TableProperties*>(arg);
+        if (set_file_creation_time_to_zero) {
+          if (idx == 0) {
+            props->file_creation_time = 0;
+            idx++;
+          } else if (idx == 1) {
+            props->file_creation_time = uint_time_1;
+            idx = 0;
+          }
+        } else {
+          if (idx == 0) {
+            props->file_creation_time = uint_time_1;
+            idx++;
+          } else if (idx == 1) {
+            props->file_creation_time = uint_time_2;
+          }
+        }
+      });
+  // Set file creation time in manifest all to 0.
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "FileMetaData::FileMetaData", [&](void* arg) {
+        FileMetaData* meta = static_cast<FileMetaData*>(arg);
+        meta->file_creation_time = 0;
+      });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  Random rnd(301);
+  for (int i = 0; i < kNumLevelFiles; ++i) {
+    for (int j = 0; j < kNumKeysPerFile; ++j) {
+      ASSERT_OK(
+          Put(Key(i * kNumKeysPerFile + j), RandomString(&rnd, kValueSize)));
+    }
+    Flush();
+  }
+
+  // At this point there should be 2 files, one with file_creation_time = 0 and
+  // the other non-zero. GetCreationTimeOfOldestFile API should return 0.
+  uint64_t creation_time;
+  Status s1 = dbfull()->GetCreationTimeOfOldestFile(&creation_time);
+  ASSERT_EQ(0, creation_time);
+  ASSERT_EQ(s1, Status::OK());
+
+  // Testing with non-zero file creation time.
+  set_file_creation_time_to_zero = false;
+  options = CurrentOptions();
+  options.max_open_files = -1;
+  env_->time_elapse_only_sleep_ = false;
+  options.env = env_;
+
+  env_->addon_time_.store(0);
+  DestroyAndReopen(options);
+
+  for (int i = 0; i < kNumLevelFiles; ++i) {
+    for (int j = 0; j < kNumKeysPerFile; ++j) {
+      ASSERT_OK(
+          Put(Key(i * kNumKeysPerFile + j), RandomString(&rnd, kValueSize)));
+    }
+    Flush();
+  }
+
+  // At this point there should be 2 files with non-zero file creation time.
+  // GetCreationTimeOfOldestFile API should return non-zero value.
+  uint64_t ctime;
+  Status s2 = dbfull()->GetCreationTimeOfOldestFile(&ctime);
+  ASSERT_EQ(uint_time_1, ctime);
+  ASSERT_EQ(s2, Status::OK());
+
+  // Testing with max_open_files != -1
+  options = CurrentOptions();
+  options.max_open_files = 10;
+  DestroyAndReopen(options);
+  Status s3 = dbfull()->GetCreationTimeOfOldestFile(&ctime);
+  ASSERT_EQ(s3, Status::NotSupported());
+
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+}
+
+#endif
+
 }  // namespace rocksdb
 
 #ifdef ROCKSDB_UNITTESTS_WITH_CUSTOM_OBJECTS_FROM_STATIC_LIBS

db/db_test2.cc

@@ -153,6 +153,8 @@ TEST_F(DBTest2, MaxSuccessiveMergesChangeWithDBRecovery) {
   options.create_if_missing = true;
   options.statistics = rocksdb::CreateDBStatistics();
   options.max_successive_merges = 3;
+  // allow_concurrent_memtable_write is incompatible with max_successive_merges
+  options.allow_concurrent_memtable_write = false;
   options.merge_operator = MergeOperators::CreatePutOperator();
   options.disable_auto_compactions = true;
   DestroyAndReopen(options);
@@ -1942,7 +1944,10 @@ TEST_F(DBTest2, TestPerfContextIterCpuTime) {
 }
 #endif  // OS_LINUX
 
-#ifndef OS_SOLARIS // GetUniqueIdFromFile is not implemented
+// GetUniqueIdFromFile is not implemented on these platforms. Persistent cache
+// breaks when that function is not implemented and no regular block cache is
+// provided.
+#if !defined(OS_SOLARIS) && !defined(OS_WIN)
 TEST_F(DBTest2, PersistentCache) {
   int num_iter = 80;
 
@@ -2006,7 +2011,7 @@ TEST_F(DBTest2, PersistentCache) {
     }
   }
 }
-#endif // !OS_SOLARIS
+#endif  // !defined(OS_SOLARIS) && !defined(OS_WIN)
 
 namespace {
 void CountSyncPoint() {
@@ -4026,6 +4031,40 @@ TEST_F(DBTest2, PrefixBloomReseek) {
   delete iter;
 }
 
+TEST_F(DBTest2, PrefixBloomFilteredOut) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.prefix_extractor.reset(NewCappedPrefixTransform(3));
+  BlockBasedTableOptions bbto;
+  bbto.filter_policy.reset(NewBloomFilterPolicy(10, false));
+  bbto.whole_key_filtering = false;
+  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
+  DestroyAndReopen(options);
+
+  // Construct two L1 files with keys:
+  // f1:[aaa1 ccc1] f2:[ddd0]
+  ASSERT_OK(Put("aaa1", ""));
+  ASSERT_OK(Put("ccc1", ""));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("ddd0", ""));
+  ASSERT_OK(Flush());
+  CompactRangeOptions cro;
+  cro.bottommost_level_compaction = BottommostLevelCompaction::kSkip;
+  ASSERT_OK(db_->CompactRange(cro, nullptr, nullptr));
+
+  Iterator* iter = db_->NewIterator(ReadOptions());
+
+  // Bloom filter is filterd out by f1.
+  // This is just one of several valid position following the contract.
+  // Postioning to ccc1 or ddd0 is also valid. This is just to validate
+  // the behavior of the current implementation. If underlying implementation
+  // changes, the test might fail here.
+  iter->Seek("bbb1");
+  ASSERT_FALSE(iter->Valid());
+
+  delete iter;
+}
+
 #ifndef ROCKSDB_LITE
 TEST_F(DBTest2, RowCacheSnapshot) {
   Options options = CurrentOptions();
@@ -4072,63 +4111,162 @@ TEST_F(DBTest2, RowCacheSnapshot) {
 }
 #endif  // ROCKSDB_LITE
 
-// Disabled but the test is failing.
 // When DB is reopened with multiple column families, the manifest file
 // is written after the first CF is flushed, and it is written again
 // after each flush. If DB crashes between the flushes, the flushed CF
 // flushed will pass the latest log file, and now we require it not
 // to be corrupted, and triggering a corruption report.
 // We need to fix the bug and enable the test.
-TEST_F(DBTest2, DISABLED_CrashInRecoveryMultipleCF) {
-  Options options = CurrentOptions();
-  options.create_if_missing = true;
-  options.wal_recovery_mode = WALRecoveryMode::kPointInTimeRecovery;
-  CreateAndReopenWithCF({"pikachu"}, options);
-  ASSERT_OK(Put("foo", "bar"));
-  ASSERT_OK(Flush());
-  ASSERT_OK(Put(1, "foo", "bar"));
-  ASSERT_OK(Flush(1));
-  ASSERT_OK(Put("foo", "bar"));
-  ASSERT_OK(Put(1, "foo", "bar"));
-  // The value is large enough to be divided to two blocks.
-  std::string large_value(400, ' ');
-  ASSERT_OK(Put("foo1", large_value));
-  ASSERT_OK(Put("foo2", large_value));
-  Close();
+TEST_F(DBTest2, CrashInRecoveryMultipleCF) {
+  const std::vector<std::string> sync_points = {
+      "DBImpl::RecoverLogFiles:BeforeFlushFinalMemtable",
+      "VersionSet::ProcessManifestWrites:BeforeWriteLastVersionEdit:0"};
+  for (const auto& test_sync_point : sync_points) {
+    Options options = CurrentOptions();
+    // First destroy original db to ensure a clean start.
+    DestroyAndReopen(options);
+    options.create_if_missing = true;
+    options.wal_recovery_mode = WALRecoveryMode::kPointInTimeRecovery;
+    CreateAndReopenWithCF({"pikachu"}, options);
+    ASSERT_OK(Put("foo", "bar"));
+    ASSERT_OK(Flush());
+    ASSERT_OK(Put(1, "foo", "bar"));
+    ASSERT_OK(Flush(1));
+    ASSERT_OK(Put("foo", "bar"));
+    ASSERT_OK(Put(1, "foo", "bar"));
+    // The value is large enough to be divided to two blocks.
+    std::string large_value(400, ' ');
+    ASSERT_OK(Put("foo1", large_value));
+    ASSERT_OK(Put("foo2", large_value));
+    Close();
 
-  // Corrupt the log file in the middle, so that it is not corrupted
-  // in the tail.
-  std::vector<std::string> filenames;
-  ASSERT_OK(env_->GetChildren(dbname_, &filenames));
-  for (const auto& f : filenames) {
-    uint64_t number;
-    FileType type;
-    if (ParseFileName(f, &number, &type) && type == FileType::kLogFile) {
-      std::string fname = dbname_ + "/" + f;
-      std::string file_content;
-      ASSERT_OK(ReadFileToString(env_, fname, &file_content));
-      file_content[400] = 'h';
-      file_content[401] = 'a';
-      ASSERT_OK(WriteStringToFile(env_, file_content, fname));
-      break;
+    // Corrupt the log file in the middle, so that it is not corrupted
+    // in the tail.
+    std::vector<std::string> filenames;
+    ASSERT_OK(env_->GetChildren(dbname_, &filenames));
+    for (const auto& f : filenames) {
+      uint64_t number;
+      FileType type;
+      if (ParseFileName(f, &number, &type) && type == FileType::kLogFile) {
+        std::string fname = dbname_ + "/" + f;
+        std::string file_content;
+        ASSERT_OK(ReadFileToString(env_, fname, &file_content));
+        file_content[400] = 'h';
+        file_content[401] = 'a';
+        ASSERT_OK(WriteStringToFile(env_, file_content, fname));
+        break;
+      }
     }
+
+    // Reopen and freeze the file system after the first manifest write.
+    FaultInjectionTestEnv fit_env(options.env);
+    options.env = &fit_env;
+    rocksdb::SyncPoint::GetInstance()->ClearAllCallBacks();
+    rocksdb::SyncPoint::GetInstance()->SetCallBack(
+        test_sync_point,
+        [&](void* /*arg*/) { fit_env.SetFilesystemActive(false); });
+    rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+    ASSERT_NOK(TryReopenWithColumnFamilies(
+        {kDefaultColumnFamilyName, "pikachu"}, options));
+    rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+
+    fit_env.SetFilesystemActive(true);
+    // If we continue using failure ingestion Env, it will conplain something
+    // when renaming current file, which is not expected. Need to investigate
+    // why.
+    options.env = env_;
+    ASSERT_OK(TryReopenWithColumnFamilies({kDefaultColumnFamilyName, "pikachu"},
+                                          options));
+  }
+}
+
+TEST_F(DBTest2, SeekFileRangeDeleteTail) {
+  Options options = CurrentOptions();
+  options.prefix_extractor.reset(NewCappedPrefixTransform(1));
+  options.num_levels = 3;
+  DestroyAndReopen(options);
+
+  ASSERT_OK(Put("a", "a"));
+  const Snapshot* s1 = db_->GetSnapshot();
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "f"));
+  ASSERT_OK(Put("b", "a"));
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(Put("x", "a"));
+  ASSERT_OK(Put("z", "a"));
+  ASSERT_OK(Flush());
+
+  CompactRangeOptions cro;
+  cro.change_level = true;
+  cro.target_level = 2;
+  ASSERT_OK(db_->CompactRange(cro, nullptr, nullptr));
+
+  {
+    ReadOptions ro;
+    ro.total_order_seek = true;
+    std::unique_ptr<Iterator> iter(db_->NewIterator(ro));
+    iter->Seek("e");
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ("x", iter->key().ToString());
+  }
+  db_->ReleaseSnapshot(s1);
+}
+
+TEST_F(DBTest2, BackgroundPurgeTest) {
+  Options options = CurrentOptions();
+  options.write_buffer_manager = std::make_shared<rocksdb::WriteBufferManager>(1 << 20);
+  options.avoid_unnecessary_blocking_io = true;
+  DestroyAndReopen(options);
+  size_t base_value = options.write_buffer_manager->memory_usage();
+
+  ASSERT_OK(Put("a", "a"));
+  Iterator* iter = db_->NewIterator(ReadOptions());
+  ASSERT_OK(Flush());
+  size_t value = options.write_buffer_manager->memory_usage();
+  ASSERT_GT(value, base_value);
+
+  db_->GetEnv()->SetBackgroundThreads(1, Env::Priority::HIGH);
+  test::SleepingBackgroundTask sleeping_task_after;
+  db_->GetEnv()->Schedule(&test::SleepingBackgroundTask::DoSleepTask,
+                          &sleeping_task_after, Env::Priority::HIGH);
+  delete iter;
+
+  Env::Default()->SleepForMicroseconds(100000);
+  value = options.write_buffer_manager->memory_usage();
+  ASSERT_GT(value, base_value);
+
+  sleeping_task_after.WakeUp();
+  sleeping_task_after.WaitUntilDone();
+
+  test::SleepingBackgroundTask sleeping_task_after2;
+  db_->GetEnv()->Schedule(&test::SleepingBackgroundTask::DoSleepTask,
+                          &sleeping_task_after2, Env::Priority::HIGH);
+  sleeping_task_after2.WakeUp();
+  sleeping_task_after2.WaitUntilDone();
+
+  value = options.write_buffer_manager->memory_usage();
+  ASSERT_EQ(base_value, value);
+}
+
+TEST_F(DBTest2, SwitchMemtableRaceWithNewManifest) {
+  Options options = CurrentOptions();
+  DestroyAndReopen(options);
+  options.max_manifest_file_size = 10;
+  options.create_if_missing = true;
+  CreateAndReopenWithCF({"pikachu"}, options);
+  ASSERT_EQ(2, handles_.size());
+
+  ASSERT_OK(Put("foo", "value"));
+  const int kL0Files = options.level0_file_num_compaction_trigger;
+  for (int i = 0; i < kL0Files; ++i) {
+    ASSERT_OK(Put(/*cf=*/1, "a", std::to_string(i)));
+    ASSERT_OK(Flush(/*cf=*/1));
   }
 
-  // Reopen and freeze the file system after the first manifest write.
-  FaultInjectionTestEnv fit_env(options.env);
-  options.env = &fit_env;
-  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "DBImpl::RecoverLogFiles:AfterLogAndApply",
-      [&](void* /*arg*/) { fit_env.SetFilesystemActive(false); });
-  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-  ASSERT_NOK(TryReopenWithColumnFamilies({"default", "pikachu"}, options));
-  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
-
-  fit_env.SetFilesystemActive(true);
-  // If we continue using failure ingestion Env, it will conplain something
-  // when renaming current file, which is not expected. Need to investigate why.
-  options.env = env_;
-  ASSERT_OK(TryReopenWithColumnFamilies({"default", "pikachu"}, options));
+  port::Thread thread([&]() { ASSERT_OK(Flush()); });
+  ASSERT_OK(dbfull()->TEST_WaitForCompact());
+  thread.join();
 }
 }  // namespace rocksdb
 

db/db_test_util.cc

@@ -53,9 +53,9 @@ DBTestBase::DBTestBase(const std::string path)
 #ifndef ROCKSDB_LITE
   const char* test_env_uri = getenv("TEST_ENV_URI");
   if (test_env_uri) {
-    Status s = ObjectRegistry::NewInstance()->NewSharedObject(test_env_uri,
-                                                              &env_guard_);
-    base_env = env_guard_.get();
+    Env* test_env = nullptr;
+    Status s = Env::LoadEnv(test_env_uri, &test_env, &env_guard_);
+    base_env = test_env;
     EXPECT_OK(s);
     EXPECT_NE(Env::Default(), base_env);
   }
@@ -777,7 +777,8 @@ std::string DBTestBase::Get(int cf, const std::string& k,
 
 std::vector<std::string> DBTestBase::MultiGet(std::vector<int> cfs,
                                               const std::vector<std::string>& k,
-                                              const Snapshot* snapshot) {
+                                              const Snapshot* snapshot,
+                                              const bool batched) {
   ReadOptions options;
   options.verify_checksums = true;
   options.snapshot = snapshot;
@@ -789,12 +790,30 @@ std::vector<std::string> DBTestBase::MultiGet(std::vector<int> cfs,
     handles.push_back(handles_[cfs[i]]);
     keys.push_back(k[i]);
   }
-  std::vector<Status> s = db_->MultiGet(options, handles, keys, &result);
-  for (unsigned int i = 0; i < s.size(); ++i) {
-    if (s[i].IsNotFound()) {
-      result[i] = "NOT_FOUND";
-    } else if (!s[i].ok()) {
-      result[i] = s[i].ToString();
+  std::vector<Status> s;
+  if (!batched) {
+    s = db_->MultiGet(options, handles, keys, &result);
+    for (unsigned int i = 0; i < s.size(); ++i) {
+      if (s[i].IsNotFound()) {
+        result[i] = "NOT_FOUND";
+      } else if (!s[i].ok()) {
+        result[i] = s[i].ToString();
+      }
+    }
+  } else {
+    std::vector<PinnableSlice> pin_values(cfs.size());
+    result.resize(cfs.size());
+    s.resize(cfs.size());
+    db_->MultiGet(options, cfs.size(), handles.data(), keys.data(),
+                  pin_values.data(), s.data());
+    for (unsigned int i = 0; i < s.size(); ++i) {
+      if (s[i].IsNotFound()) {
+        result[i] = "NOT_FOUND";
+      } else if (!s[i].ok()) {
+        result[i] = s[i].ToString();
+      } else {
+        result[i].assign(pin_values[i].data(), pin_values[i].size());
+      }
     }
   }
   return result;
@@ -848,6 +867,13 @@ uint64_t DBTestBase::GetTimeOldestSnapshots() {
   return int_num;
 }
 
+uint64_t DBTestBase::GetSequenceOldestSnapshots() {
+  uint64_t int_num;
+  EXPECT_TRUE(
+      dbfull()->GetIntProperty("rocksdb.oldest-snapshot-sequence", &int_num));
+  return int_num;
+}
+
 // Return a string that contains all key,value pairs in order,
 // formatted like "(k1->v1)(k2->v2)".
 std::string DBTestBase::Contents(int cf) {

db/db_test_util.h

@@ -850,7 +850,8 @@ class DBTestBase : public testing::Test {
 
   std::vector<std::string> MultiGet(std::vector<int> cfs,
                                     const std::vector<std::string>& k,
-                                    const Snapshot* snapshot = nullptr);
+                                    const Snapshot* snapshot,
+                                    const bool batched);
 
   std::vector<std::string> MultiGet(const std::vector<std::string>& k,
                                     const Snapshot* snapshot = nullptr);
@@ -859,6 +860,8 @@ class DBTestBase : public testing::Test {
 
   uint64_t GetTimeOldestSnapshots();
 
+  uint64_t GetSequenceOldestSnapshots();
+
   // Return a string that contains all key,value pairs in order,
   // formatted like "(k1->v1)(k2->v2)".
   std::string Contents(int cf = 0);
@@ -987,6 +990,11 @@ class DBTestBase : public testing::Test {
   uint64_t TestGetTickerCount(const Options& options, Tickers ticker_type) {
     return options.statistics->getTickerCount(ticker_type);
   }
+
+  uint64_t TestGetAndResetTickerCount(const Options& options,
+                                      Tickers ticker_type) {
+    return options.statistics->getAndResetTickerCount(ticker_type);
+  }
 };
 
 }  // namespace rocksdb

db/db_universal_compaction_test.cc

@@ -41,10 +41,9 @@ class DBTestUniversalCompaction : public DBTestUniversalCompactionBase {
       DBTestUniversalCompactionBase("/db_universal_compaction_test") {}
 };
 
-class DBTestUniversalDeleteTrigCompaction : public DBTestBase {
+class DBTestUniversalCompaction2 : public DBTestBase {
  public:
-  DBTestUniversalDeleteTrigCompaction()
-      : DBTestBase("/db_universal_compaction_test") {}
+  DBTestUniversalCompaction2() : DBTestBase("/db_universal_compaction_test2") {}
 };
 
 namespace {
@@ -1754,13 +1753,14 @@ TEST_P(DBTestUniversalCompaction, RecalculateScoreAfterPicking) {
   // scheduling more; otherwise, other CFs/DBs may be delayed unnecessarily.
   const int kNumFilesTrigger = 8;
   Options options = CurrentOptions();
+  options.memtable_factory.reset(
+      new SpecialSkipListFactory(KNumKeysByGenerateNewFile - 1));
   options.compaction_options_universal.max_merge_width = kNumFilesTrigger / 2;
   options.compaction_options_universal.max_size_amplification_percent =
       static_cast<unsigned int>(-1);
   options.compaction_style = kCompactionStyleUniversal;
   options.level0_file_num_compaction_trigger = kNumFilesTrigger;
   options.num_levels = num_levels_;
-  options.write_buffer_size = 100 << 10;  // 100KB
   Reopen(options);
 
   std::atomic<int> num_compactions_attempted(0);
@@ -1915,7 +1915,7 @@ INSTANTIATE_TEST_CASE_P(DBTestUniversalManualCompactionOutputPathId,
                         ::testing::Combine(::testing::Values(1, 8),
                                            ::testing::Bool()));
 
-TEST_F(DBTestUniversalDeleteTrigCompaction, BasicL0toL1) {
+TEST_F(DBTestUniversalCompaction2, BasicL0toL1) {
   const int kNumKeys = 3000;
   const int kWindowSize = 100;
   const int kNumDelsTrigger = 90;
@@ -1956,7 +1956,7 @@ TEST_F(DBTestUniversalDeleteTrigCompaction, BasicL0toL1) {
   ASSERT_GT(NumTableFilesAtLevel(6), 0);
 }
 
-TEST_F(DBTestUniversalDeleteTrigCompaction, SingleLevel) {
+TEST_F(DBTestUniversalCompaction2, SingleLevel) {
   const int kNumKeys = 3000;
   const int kWindowSize = 100;
   const int kNumDelsTrigger = 90;
@@ -1995,7 +1995,7 @@ TEST_F(DBTestUniversalDeleteTrigCompaction, SingleLevel) {
   ASSERT_EQ(1, NumTableFilesAtLevel(0));
 }
 
-TEST_F(DBTestUniversalDeleteTrigCompaction, MultipleLevels) {
+TEST_F(DBTestUniversalCompaction2, MultipleLevels) {
   const int kWindowSize = 100;
   const int kNumDelsTrigger = 90;
 
@@ -2067,7 +2067,7 @@ TEST_F(DBTestUniversalDeleteTrigCompaction, MultipleLevels) {
   ASSERT_GT(NumTableFilesAtLevel(6), 0);
 }
 
-TEST_F(DBTestUniversalDeleteTrigCompaction, OverlappingL0) {
+TEST_F(DBTestUniversalCompaction2, OverlappingL0) {
   const int kWindowSize = 100;
   const int kNumDelsTrigger = 90;
 
@@ -2107,7 +2107,7 @@ TEST_F(DBTestUniversalDeleteTrigCompaction, OverlappingL0) {
   ASSERT_GT(NumTableFilesAtLevel(6), 0);
 }
 
-TEST_F(DBTestUniversalDeleteTrigCompaction, IngestBehind) {
+TEST_F(DBTestUniversalCompaction2, IngestBehind) {
   const int kNumKeys = 3000;
   const int kWindowSize = 100;
   const int kNumDelsTrigger = 90;
@@ -2150,6 +2150,96 @@ TEST_F(DBTestUniversalDeleteTrigCompaction, IngestBehind) {
   ASSERT_GT(NumTableFilesAtLevel(5), 0);
 }
 
+TEST_F(DBTestUniversalCompaction2, PeriodicCompactionDefault) {
+  Options options;
+  options.compaction_style = kCompactionStyleUniversal;
+
+  KeepFilterFactory* filter = new KeepFilterFactory(true);
+  options.compaction_filter_factory.reset(filter);
+  Reopen(options);
+  ASSERT_EQ(30 * 24 * 60 * 60,
+            dbfull()->GetOptions().periodic_compaction_seconds);
+
+  KeepFilter df;
+  options.compaction_filter_factory.reset();
+  options.compaction_filter = &df;
+  Reopen(options);
+  ASSERT_EQ(30 * 24 * 60 * 60,
+            dbfull()->GetOptions().periodic_compaction_seconds);
+
+  options.ttl = 60 * 24 * 60 * 60;
+  options.compaction_filter = nullptr;
+  Reopen(options);
+  ASSERT_EQ(60 * 24 * 60 * 60,
+            dbfull()->GetOptions().periodic_compaction_seconds);
+}
+
+TEST_F(DBTestUniversalCompaction2, PeriodicCompaction) {
+  Options opts = CurrentOptions();
+  opts.env = env_;
+  opts.compaction_style = kCompactionStyleUniversal;
+  opts.level0_file_num_compaction_trigger = 10;
+  opts.max_open_files = -1;
+  opts.compaction_options_universal.size_ratio = 10;
+  opts.compaction_options_universal.min_merge_width = 2;
+  opts.compaction_options_universal.max_size_amplification_percent = 200;
+  opts.periodic_compaction_seconds = 48 * 60 * 60;  // 2 days
+  opts.num_levels = 5;
+  env_->addon_time_.store(0);
+  Reopen(opts);
+
+  int periodic_compactions = 0;
+  int start_level = -1;
+  int output_level = -1;
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "UniversalCompactionPicker::PickPeriodicCompaction:Return",
+      [&](void* arg) {
+        Compaction* compaction = reinterpret_cast<Compaction*>(arg);
+        ASSERT_TRUE(arg != nullptr);
+        ASSERT_TRUE(compaction->compaction_reason() ==
+                    CompactionReason::kPeriodicCompaction);
+        start_level = compaction->start_level();
+        output_level = compaction->output_level();
+        periodic_compactions++;
+      });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  // Case 1: Oldest flushed file excceeds periodic compaction threshold.
+  ASSERT_OK(Put("foo", "bar"));
+  Flush();
+  ASSERT_EQ(0, periodic_compactions);
+  // Move clock forward so that the flushed file would qualify periodic
+  // compaction.
+  env_->addon_time_.store(48 * 60 * 60 + 100);
+
+  // Another flush would trigger compaction the oldest file.
+  ASSERT_OK(Put("foo", "bar2"));
+  Flush();
+  dbfull()->TEST_WaitForCompact();
+
+  ASSERT_EQ(1, periodic_compactions);
+  ASSERT_EQ(0, start_level);
+  ASSERT_EQ(4, output_level);
+
+  // Case 2: Oldest compacted file excceeds periodic compaction threshold
+  periodic_compactions = 0;
+  // A flush doesn't trigger a periodic compaction when threshold not hit
+  ASSERT_OK(Put("foo", "bar2"));
+  Flush();
+  dbfull()->TEST_WaitForCompact();
+  ASSERT_EQ(0, periodic_compactions);
+
+  // After periodic compaction threshold hits, a flush will trigger
+  // a compaction
+  ASSERT_OK(Put("foo", "bar2"));
+  env_->addon_time_.fetch_add(48 * 60 * 60 + 100);
+  Flush();
+  dbfull()->TEST_WaitForCompact();
+  ASSERT_EQ(1, periodic_compactions);
+  ASSERT_EQ(0, start_level);
+  ASSERT_EQ(4, output_level);
+}
+
 }  // namespace rocksdb
 
 #endif  // !defined(ROCKSDB_LITE)

db/db_wal_test.cc

@@ -8,6 +8,7 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #include "db/db_test_util.h"
+#include "env/composite_env_wrapper.h"
 #include "options/options_helper.h"
 #include "port/port.h"
 #include "port/stack_trace.h"
@@ -902,8 +903,8 @@ class RecoveryTestHelper {
       std::string fname = LogFileName(test->dbname_, current_log_number);
       std::unique_ptr<WritableFile> file;
       ASSERT_OK(db_options.env->NewWritableFile(fname, &file, env_options));
-      std::unique_ptr<WritableFileWriter> file_writer(
-          new WritableFileWriter(std::move(file), fname, env_options));
+      std::unique_ptr<WritableFileWriter> file_writer(new WritableFileWriter(
+          NewLegacyWritableFileWrapper(std::move(file)), fname, env_options));
       current_log_writer.reset(
           new log::Writer(std::move(file_writer), current_log_number,
                           db_options.recycle_log_file_num > 0));

db/error_handler_test.cc

@@ -183,7 +183,8 @@ TEST_F(DBErrorHandlingTest, CompactionWriteError) {
       );
   listener->EnableAutoRecovery(false);
   rocksdb::SyncPoint::GetInstance()->LoadDependency(
-      {{"FlushMemTableFinished", "BackgroundCallCompaction:0"}});
+      {{"DBImpl::FlushMemTable:FlushMemTableFinished",
+        "BackgroundCallCompaction:0"}});
   rocksdb::SyncPoint::GetInstance()->SetCallBack(
       "BackgroundCallCompaction:0", [&](void *) {
       fault_env->SetFilesystemActive(false, Status::NoSpace("Out of space"));
@@ -219,7 +220,8 @@ TEST_F(DBErrorHandlingTest, CorruptionError) {
   ASSERT_EQ(s, Status::OK());
 
   rocksdb::SyncPoint::GetInstance()->LoadDependency(
-      {{"FlushMemTableFinished", "BackgroundCallCompaction:0"}});
+      {{"DBImpl::FlushMemTable:FlushMemTableFinished",
+        "BackgroundCallCompaction:0"}});
   rocksdb::SyncPoint::GetInstance()->SetCallBack(
       "BackgroundCallCompaction:0", [&](void *) {
       fault_env->SetFilesystemActive(false, Status::Corruption("Corruption"));

db/event_helpers.cc

@@ -70,8 +70,8 @@ void EventHelpers::LogAndNotifyTableFileCreationFinished(
     const std::vector<std::shared_ptr<EventListener>>& listeners,
     const std::string& db_name, const std::string& cf_name,
     const std::string& file_path, int job_id, const FileDescriptor& fd,
-    const TableProperties& table_properties, TableFileCreationReason reason,
-    const Status& s) {
+    uint64_t oldest_blob_file_number, const TableProperties& table_properties,
+    TableFileCreationReason reason, const Status& s) {
   if (s.ok() && event_logger) {
     JSONWriter jwriter;
     AppendCurrentTime(&jwriter);
@@ -129,6 +129,11 @@ void EventHelpers::LogAndNotifyTableFileCreationFinished(
       }
       jwriter.EndObject();
     }
+
+    if (oldest_blob_file_number != kInvalidBlobFileNumber) {
+      jwriter << "oldest_blob_file_number" << oldest_blob_file_number;
+    }
+
     jwriter.EndObject();
 
     event_logger->Log(jwriter);

db/event_helpers.h

@@ -34,8 +34,8 @@ class EventHelpers {
       const std::vector<std::shared_ptr<EventListener>>& listeners,
       const std::string& db_name, const std::string& cf_name,
       const std::string& file_path, int job_id, const FileDescriptor& fd,
-      const TableProperties& table_properties, TableFileCreationReason reason,
-      const Status& s);
+      uint64_t oldest_blob_file_number, const TableProperties& table_properties,
+      TableFileCreationReason reason, const Status& s);
   static void LogAndNotifyTableFileDeletion(
       EventLogger* event_logger, int job_id,
       uint64_t file_number, const std::string& file_path,

db/external_sst_file_ingestion_job.cc

@@ -101,14 +101,15 @@ Status ExternalSstFileIngestionJob::Prepare(
         TableFileName(cfd_->ioptions()->cf_paths, f.fd.GetNumber(),
                       f.fd.GetPathId());
     if (ingestion_options_.move_files) {
-      status = env_->LinkFile(path_outside_db, path_inside_db);
+      status =
+          fs_->LinkFile(path_outside_db, path_inside_db, IOOptions(), nullptr);
       if (status.ok()) {
         // It is unsafe to assume application had sync the file and file
         // directory before ingest the file. For integrity of RocksDB we need
         // to sync the file.
-        std::unique_ptr<WritableFile> file_to_sync;
-        status = env_->ReopenWritableFile(path_inside_db, &file_to_sync,
-                                          env_options_);
+        std::unique_ptr<FSWritableFile> file_to_sync;
+        status = fs_->ReopenWritableFile(path_inside_db, env_options_,
+                                         &file_to_sync, nullptr);
         if (status.ok()) {
           TEST_SYNC_POINT(
               "ExternalSstFileIngestionJob::BeforeSyncIngestedFile");
@@ -133,7 +134,7 @@ Status ExternalSstFileIngestionJob::Prepare(
       TEST_SYNC_POINT_CALLBACK("ExternalSstFileIngestionJob::Prepare:CopyFile",
                                nullptr);
       // CopyFile also sync the new file.
-      status = CopyFile(env_, path_outside_db, path_inside_db, 0,
+      status = CopyFile(fs_, path_outside_db, path_inside_db, 0,
                         db_options_.use_fsync);
     }
     TEST_SYNC_POINT("ExternalSstFileIngestionJob::Prepare:FileAdded");
@@ -243,10 +244,22 @@ Status ExternalSstFileIngestionJob::Run() {
     if (!status.ok()) {
       return status;
     }
+
+    // We use the import time as the ancester time. This is the time the data
+    // is written to the database.
+    int64_t temp_current_time = 0;
+    uint64_t current_time = kUnknownFileCreationTime;
+    uint64_t oldest_ancester_time = kUnknownOldestAncesterTime;
+    if (env_->GetCurrentTime(&temp_current_time).ok()) {
+      current_time = oldest_ancester_time =
+          static_cast<uint64_t>(temp_current_time);
+    }
+
     edit_.AddFile(f.picked_level, f.fd.GetNumber(), f.fd.GetPathId(),
                   f.fd.GetFileSize(), f.smallest_internal_key,
                   f.largest_internal_key, f.assigned_seqno, f.assigned_seqno,
-                  false);
+                  false, kInvalidBlobFileNumber, oldest_ancester_time,
+                  current_time);
   }
   return status;
 }
@@ -348,17 +361,19 @@ Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
   file_to_ingest->external_file_path = external_file;
 
   // Get external file size
-  Status status = env_->GetFileSize(external_file, &file_to_ingest->file_size);
+  Status status = fs_->GetFileSize(external_file, IOOptions(),
+                                   &file_to_ingest->file_size, nullptr);
   if (!status.ok()) {
     return status;
   }
 
   // Create TableReader for external file
   std::unique_ptr<TableReader> table_reader;
-  std::unique_ptr<RandomAccessFile> sst_file;
+  std::unique_ptr<FSRandomAccessFile> sst_file;
   std::unique_ptr<RandomAccessFileReader> sst_file_reader;
 
-  status = env_->NewRandomAccessFile(external_file, &sst_file, env_options_);
+  status = fs_->NewRandomAccessFile(external_file, env_options_,
+                                    &sst_file, nullptr);
   if (!status.ok()) {
     return status;
   }
@@ -639,13 +654,15 @@ Status ExternalSstFileIngestionJob::AssignGlobalSeqnoForIngestedFile(
     // Determine if we can write global_seqno to a given offset of file.
     // If the file system does not support random write, then we should not.
     // Otherwise we should.
-    std::unique_ptr<RandomRWFile> rwfile;
-    Status status = env_->NewRandomRWFile(file_to_ingest->internal_file_path,
-                                          &rwfile, env_options_);
+    std::unique_ptr<FSRandomRWFile> rwfile;
+    Status status =
+        fs_->NewRandomRWFile(file_to_ingest->internal_file_path, env_options_,
+                             &rwfile, nullptr);
     if (status.ok()) {
       std::string seqno_val;
       PutFixed64(&seqno_val, seqno);
-      status = rwfile->Write(file_to_ingest->global_seqno_offset, seqno_val);
+      status = rwfile->Write(file_to_ingest->global_seqno_offset, seqno_val,
+                             IOOptions(), nullptr);
       if (status.ok()) {
         TEST_SYNC_POINT("ExternalSstFileIngestionJob::BeforeSyncGlobalSeqno");
         status = SyncIngestedFile(rwfile.get());
@@ -703,9 +720,9 @@ template <typename TWritableFile>
 Status ExternalSstFileIngestionJob::SyncIngestedFile(TWritableFile* file) {
   assert(file != nullptr);
   if (db_options_.use_fsync) {
-    return file->Fsync();
+    return file->Fsync(IOOptions(), nullptr);
   } else {
-    return file->Sync();
+    return file->Sync(IOOptions(), nullptr);
   }
 }
 

db/external_sst_file_ingestion_job.h

@@ -74,6 +74,7 @@ class ExternalSstFileIngestionJob {
       const IngestExternalFileOptions& ingestion_options,
       Directories* directories, EventLogger* event_logger)
       : env_(env),
+        fs_(db_options.fs.get()),
         versions_(versions),
         cfd_(cfd),
         db_options_(db_options),
@@ -158,6 +159,7 @@ class ExternalSstFileIngestionJob {
   Status SyncIngestedFile(TWritableFile* file);
 
   Env* env_;
+  FileSystem* fs_;
   VersionSet* versions_;
   ColumnFamilyData* cfd_;
   const ImmutableDBOptions& db_options_;

db/external_sst_file_test.cc

@@ -1700,6 +1700,35 @@ TEST_F(ExternalSSTFileTest, SstFileWriterNonSharedKeys) {
   ASSERT_OK(DeprecatedAddFile({file_path}));
 }
 
+TEST_F(ExternalSSTFileTest, WithUnorderedWrite) {
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->LoadDependency(
+      {{"DBImpl::WriteImpl:UnorderedWriteAfterWriteWAL",
+        "ExternalSSTFileTest::WithUnorderedWrite:WaitWriteWAL"},
+       {"DBImpl::WaitForPendingWrites:BeforeBlock",
+        "DBImpl::WriteImpl:BeforeUnorderedWriteMemtable"}});
+  SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::IngestExternalFile:NeedFlush", [&](void* need_flush) {
+        ASSERT_TRUE(*reinterpret_cast<bool*>(need_flush));
+      });
+
+  Options options = CurrentOptions();
+  options.unordered_write = true;
+  DestroyAndReopen(options);
+  Put("foo", "v1");
+  SyncPoint::GetInstance()->EnableProcessing();
+  port::Thread writer([&]() { Put("bar", "v2"); });
+
+  TEST_SYNC_POINT("ExternalSSTFileTest::WithUnorderedWrite:WaitWriteWAL");
+  ASSERT_OK(GenerateAndAddExternalFile(options, {{"bar", "v3"}}, -1,
+                                       true /* allow_global_seqno */));
+  ASSERT_EQ(Get("bar"), "v3");
+
+  writer.join();
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+}
+
 TEST_P(ExternalSSTFileTest, IngestFileWithGlobalSeqnoRandomized) {
   Options options = CurrentOptions();
   options.IncreaseParallelism(20);
@@ -2750,6 +2779,26 @@ TEST_P(ExternalSSTFileTest,
   Destroy(options, true /* delete_cf_paths */);
 }
 
+TEST_P(ExternalSSTFileTest, IngestFilesTriggerFlushingWithTwoWriteQueue) {
+  Options options = CurrentOptions();
+  // Use large buffer to avoid memtable flush
+  options.write_buffer_size = 1024 * 1024;
+  options.two_write_queues = true;
+  DestroyAndReopen(options);
+
+  ASSERT_OK(dbfull()->Put(WriteOptions(), "1000", "v1"));
+  ASSERT_OK(dbfull()->Put(WriteOptions(), "1001", "v1"));
+  ASSERT_OK(dbfull()->Put(WriteOptions(), "9999", "v1"));
+
+  // Put one key which is overlap with keys in memtable.
+  // It will trigger flushing memtable and require this thread is
+  // currently at the front of the 2nd writer queue. We must make
+  // sure that it won't enter the 2nd writer queue for the second time.
+  std::vector<std::pair<std::string, std::string>> data;
+  data.push_back(std::make_pair("1001", "v2"));
+  GenerateAndAddExternalFile(options, data);
+}
+
 INSTANTIATE_TEST_CASE_P(ExternalSSTFileTest, ExternalSSTFileTest,
                         testing::Values(std::make_tuple(false, false),
                                         std::make_tuple(false, true),

db/flush_job.cc

@@ -86,7 +86,7 @@ FlushJob::FlushJob(const std::string& dbname, ColumnFamilyData* cfd,
                    const ImmutableDBOptions& db_options,
                    const MutableCFOptions& mutable_cf_options,
                    const uint64_t* max_memtable_id,
-                   const EnvOptions& env_options, VersionSet* versions,
+                   const FileOptions& file_options, VersionSet* versions,
                    InstrumentedMutex* db_mutex,
                    std::atomic<bool>* shutting_down,
                    std::vector<SequenceNumber> existing_snapshots,
@@ -103,7 +103,7 @@ FlushJob::FlushJob(const std::string& dbname, ColumnFamilyData* cfd,
       db_options_(db_options),
       mutable_cf_options_(mutable_cf_options),
       max_memtable_id_(max_memtable_id),
-      env_options_(env_options),
+      file_options_(file_options),
       versions_(versions),
       db_mutex_(db_mutex),
       shutting_down_(shutting_down),
@@ -241,7 +241,7 @@ Status FlushJob::Run(LogsWithPrepTracker* prep_tracker,
     s = cfd_->imm()->TryInstallMemtableFlushResults(
         cfd_, mutable_cf_options_, mems_, prep_tracker, versions_, db_mutex_,
         meta_.fd.GetNumber(), &job_context_->memtables_to_free, db_directory_,
-        log_buffer_);
+        log_buffer_, &committed_flush_jobs_info_);
   }
 
   if (s.ok() && file_meta != nullptr) {
@@ -249,7 +249,8 @@ Status FlushJob::Run(LogsWithPrepTracker* prep_tracker,
   }
   RecordFlushIOStats();
 
-  auto stream = event_logger_->LogToBuffer(log_buffer_);
+  // When measure_io_stats_ is true, the default 512 bytes is not enough.
+  auto stream = event_logger_->LogToBuffer(log_buffer_, 1024);
   stream << "job" << job_context_->job_id << "event"
          << "flush_finished";
   stream << "output_compression"
@@ -365,9 +366,14 @@ Status FlushJob::WriteLevel0Table() {
       uint64_t oldest_key_time =
           mems_.front()->ApproximateOldestKeyTime();
 
+      // It's not clear whether oldest_key_time is always available. In case
+      // it is not available, use current_time.
+      meta_.oldest_ancester_time = std::min(current_time, oldest_key_time);
+      meta_.file_creation_time = current_time;
+
       s = BuildTable(
-          dbname_, db_options_.env, *cfd_->ioptions(), mutable_cf_options_,
-          env_options_, cfd_->table_cache(), iter.get(),
+          dbname_, db_options_.env, db_options_.fs.get(), *cfd_->ioptions(),
+          mutable_cf_options_, file_options_, cfd_->table_cache(), iter.get(),
           std::move(range_del_iters), &meta_, cfd_->internal_comparator(),
           cfd_->int_tbl_prop_collector_factories(), cfd_->GetID(),
           cfd_->GetName(), existing_snapshots_,
@@ -392,7 +398,7 @@ Status FlushJob::WriteLevel0Table() {
     if (s.ok() && output_file_directory_ != nullptr && sync_output_directory_) {
       s = output_file_directory_->Fsync();
     }
-    TEST_SYNC_POINT("FlushJob::WriteLevel0Table");
+    TEST_SYNC_POINT_CALLBACK("FlushJob::WriteLevel0Table", &mems_);
     db_mutex_->Lock();
   }
   base_->Unref();
@@ -408,8 +414,13 @@ Status FlushJob::WriteLevel0Table() {
     edit_->AddFile(0 /* level */, meta_.fd.GetNumber(), meta_.fd.GetPathId(),
                    meta_.fd.GetFileSize(), meta_.smallest, meta_.largest,
                    meta_.fd.smallest_seqno, meta_.fd.largest_seqno,
-                   meta_.marked_for_compaction);
+                   meta_.marked_for_compaction, meta_.oldest_blob_file_number,
+                   meta_.oldest_ancester_time, meta_.file_creation_time);
   }
+#ifndef ROCKSDB_LITE
+  // Piggyback FlushJobInfo on the first first flushed memtable.
+  mems_[0]->SetFlushJobInfo(GetFlushJobInfo());
+#endif  // !ROCKSDB_LITE
 
   // Note that here we treat flush as level 0 compaction in internal stats
   InternalStats::CompactionStats stats(CompactionReason::kFlush, 1);
@@ -424,4 +435,26 @@ Status FlushJob::WriteLevel0Table() {
   return s;
 }
 
+#ifndef ROCKSDB_LITE
+std::unique_ptr<FlushJobInfo> FlushJob::GetFlushJobInfo() const {
+  db_mutex_->AssertHeld();
+  std::unique_ptr<FlushJobInfo> info(new FlushJobInfo{});
+  info->cf_id = cfd_->GetID();
+  info->cf_name = cfd_->GetName();
+
+  const uint64_t file_number = meta_.fd.GetNumber();
+  info->file_path =
+      MakeTableFileName(cfd_->ioptions()->cf_paths[0].path, file_number);
+  info->file_number = file_number;
+  info->oldest_blob_file_number = meta_.oldest_blob_file_number;
+  info->thread_id = db_options_.env->GetThreadID();
+  info->job_id = job_context_->job_id;
+  info->smallest_seqno = meta_.fd.smallest_seqno;
+  info->largest_seqno = meta_.fd.largest_seqno;
+  info->table_properties = table_properties_;
+  info->flush_reason = cfd_->GetFlushReason();
+  return info;
+}
+#endif  // !ROCKSDB_LITE
+
 }  // namespace rocksdb

db/flush_job.h

@@ -11,10 +11,11 @@
 #include <atomic>
 #include <deque>
 #include <limits>
+#include <list>
 #include <set>
+#include <string>
 #include <utility>
 #include <vector>
-#include <string>
 
 #include "db/column_family.h"
 #include "db/dbformat.h"
@@ -34,6 +35,7 @@
 #include "port/port.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
+#include "rocksdb/listener.h"
 #include "rocksdb/memtablerep.h"
 #include "rocksdb/transaction_log.h"
 #include "table/scoped_arena_iterator.h"
@@ -59,7 +61,7 @@ class FlushJob {
   FlushJob(const std::string& dbname, ColumnFamilyData* cfd,
            const ImmutableDBOptions& db_options,
            const MutableCFOptions& mutable_cf_options,
-           const uint64_t* max_memtable_id, const EnvOptions& env_options,
+           const uint64_t* max_memtable_id, const FileOptions& file_options,
            VersionSet* versions, InstrumentedMutex* db_mutex,
            std::atomic<bool>* shutting_down,
            std::vector<SequenceNumber> existing_snapshots,
@@ -79,14 +81,22 @@ class FlushJob {
   Status Run(LogsWithPrepTracker* prep_tracker = nullptr,
              FileMetaData* file_meta = nullptr);
   void Cancel();
-  TableProperties GetTableProperties() const { return table_properties_; }
   const autovector<MemTable*>& GetMemTables() const { return mems_; }
 
+#ifndef ROCKSDB_LITE
+  std::list<std::unique_ptr<FlushJobInfo>>* GetCommittedFlushJobsInfo() {
+    return &committed_flush_jobs_info_;
+  }
+#endif  // !ROCKSDB_LITE
+
  private:
   void ReportStartedFlush();
   void ReportFlushInputSize(const autovector<MemTable*>& mems);
   void RecordFlushIOStats();
   Status WriteLevel0Table();
+#ifndef ROCKSDB_LITE
+  std::unique_ptr<FlushJobInfo> GetFlushJobInfo() const;
+#endif  // !ROCKSDB_LITE
 
   const std::string& dbname_;
   ColumnFamilyData* cfd_;
@@ -98,7 +108,7 @@ class FlushJob {
   // equal to *max_memtable_id_ will be selected for flush. If null, then all
   // memtables in the column family will be selected.
   const uint64_t* max_memtable_id_;
-  const EnvOptions env_options_;
+  const FileOptions file_options_;
   VersionSet* versions_;
   InstrumentedMutex* db_mutex_;
   std::atomic<bool>* shutting_down_;
@@ -131,6 +141,10 @@ class FlushJob {
   // In this case, only after all flush jobs succeed in flush can RocksDB
   // commit to the MANIFEST.
   const bool write_manifest_;
+  // The current flush job can commit flush result of a concurrent flush job.
+  // We collect FlushJobInfo of all jobs committed by current job and fire
+  // OnFlushCompleted for them.
+  std::list<std::unique_ptr<FlushJobInfo>> committed_flush_jobs_info_;
 
   // Variables below are set by PickMemTable():
   FileMetaData meta_;

db/flush_job_test.cc

@@ -4,9 +4,11 @@
 //  (found in the LICENSE.Apache file in the root directory).
 
 #include <algorithm>
+#include <array>
 #include <map>
 #include <string>
 
+#include "db/blob_index.h"
 #include "db/column_family.h"
 #include "db/db_impl/db_impl.h"
 #include "db/flush_job.h"
@@ -28,16 +30,13 @@ class FlushJobTest : public testing::Test {
  public:
   FlushJobTest()
       : env_(Env::Default()),
+        fs_(std::make_shared<LegacyFileSystemWrapper>(env_)),
         dbname_(test::PerThreadDBPath("flush_job_test")),
         options_(),
         db_options_(options_),
         column_family_names_({kDefaultColumnFamilyName, "foo", "bar"}),
         table_cache_(NewLRUCache(50000, 16)),
         write_buffer_manager_(db_options_.db_write_buffer_size),
-        versions_(new VersionSet(dbname_, &db_options_, env_options_,
-                                 table_cache_.get(), &write_buffer_manager_,
-                                 &write_controller_,
-                                 /*block_cache_tracer=*/nullptr)),
         shutting_down_(false),
         mock_table_factory_(new mock::MockTableFactory()) {
     EXPECT_OK(env_->CreateDirIfMissing(dbname_));
@@ -52,6 +51,12 @@ class FlushJobTest : public testing::Test {
       column_families.emplace_back(cf_name, cf_options_);
     }
 
+    db_options_.env = env_;
+    db_options_.fs = fs_;
+    versions_.reset(new VersionSet(dbname_, &db_options_, env_options_,
+                                   table_cache_.get(), &write_buffer_manager_,
+                                   &write_controller_,
+                                   /*block_cache_tracer=*/nullptr));
     EXPECT_OK(versions_->Recover(column_families, false));
   }
 
@@ -86,8 +91,8 @@ class FlushJobTest : public testing::Test {
     Status s = env_->NewWritableFile(
         manifest, &file, env_->OptimizeForManifestWrite(env_options_));
     ASSERT_OK(s);
-    std::unique_ptr<WritableFileWriter> file_writer(
-        new WritableFileWriter(std::move(file), manifest, EnvOptions()));
+    std::unique_ptr<WritableFileWriter> file_writer(new WritableFileWriter(
+        NewLegacyWritableFileWrapper(std::move(file)), manifest, EnvOptions()));
     {
       log::Writer log(std::move(file_writer), 0, false);
       std::string record;
@@ -107,6 +112,7 @@ class FlushJobTest : public testing::Test {
   }
 
   Env* env_;
+  std::shared_ptr<FileSystem> fs_;
   std::string dbname_;
   EnvOptions env_options_;
   Options options_;
@@ -154,6 +160,7 @@ TEST_F(FlushJobTest, NonEmpty) {
   //   seqno [    1,    2 ... 8998, 8999, 9000, 9001, 9002 ... 9999 ]
   //   key   [ 1001, 1002 ... 9998, 9999,    0,    1,    2 ...  999 ]
   //   range-delete "9995" -> "9999" at seqno 10000
+  //   blob references with seqnos 10001..10006
   for (int i = 1; i < 10000; ++i) {
     std::string key(ToString((i + 1000) % 10000));
     std::string value("value" + key);
@@ -163,9 +170,43 @@ TEST_F(FlushJobTest, NonEmpty) {
       inserted_keys.insert({internal_key.Encode().ToString(), value});
     }
   }
-  new_mem->Add(SequenceNumber(10000), kTypeRangeDeletion, "9995", "9999a");
-  InternalKey internal_key("9995", SequenceNumber(10000), kTypeRangeDeletion);
-  inserted_keys.insert({internal_key.Encode().ToString(), "9999a"});
+
+  {
+    new_mem->Add(SequenceNumber(10000), kTypeRangeDeletion, "9995", "9999a");
+    InternalKey internal_key("9995", SequenceNumber(10000), kTypeRangeDeletion);
+    inserted_keys.insert({internal_key.Encode().ToString(), "9999a"});
+  }
+
+#ifndef ROCKSDB_LITE
+  // Note: the first two blob references will not be considered when resolving
+  // the oldest blob file referenced (the first one is inlined TTL, while the
+  // second one is TTL and thus points to a TTL blob file).
+  constexpr std::array<uint64_t, 6> blob_file_numbers{
+      kInvalidBlobFileNumber, 5, 103, 17, 102, 101};
+  for (size_t i = 0; i < blob_file_numbers.size(); ++i) {
+    std::string key(ToString(i + 10001));
+    std::string blob_index;
+    if (i == 0) {
+      BlobIndex::EncodeInlinedTTL(&blob_index, /* expiration */ 1234567890ULL,
+                                  "foo");
+    } else if (i == 1) {
+      BlobIndex::EncodeBlobTTL(&blob_index, /* expiration */ 1234567890ULL,
+                               blob_file_numbers[i], /* offset */ i << 10,
+                               /* size */ i << 20, kNoCompression);
+    } else {
+      BlobIndex::EncodeBlob(&blob_index, blob_file_numbers[i],
+                            /* offset */ i << 10, /* size */ i << 20,
+                            kNoCompression);
+    }
+
+    const SequenceNumber seq(i + 10001);
+    new_mem->Add(seq, kTypeBlobIndex, key, blob_index);
+
+    InternalKey internal_key(key, seq, kTypeBlobIndex);
+    inserted_keys.emplace_hint(inserted_keys.end(),
+                               internal_key.Encode().ToString(), blob_index);
+  }
+#endif
 
   autovector<MemTable*> to_delete;
   cfd->imm()->Add(new_mem, &to_delete);
@@ -194,11 +235,14 @@ TEST_F(FlushJobTest, NonEmpty) {
   ASSERT_GT(hist.average, 0.0);
 
   ASSERT_EQ(ToString(0), file_meta.smallest.user_key().ToString());
-  ASSERT_EQ(
-      "9999a",
-      file_meta.largest.user_key().ToString());  // range tombstone end key
+  ASSERT_EQ("9999a", file_meta.largest.user_key().ToString());
   ASSERT_EQ(1, file_meta.fd.smallest_seqno);
-  ASSERT_EQ(10000, file_meta.fd.largest_seqno);  // range tombstone seqnum 10000
+#ifndef ROCKSDB_LITE
+  ASSERT_EQ(10006, file_meta.fd.largest_seqno);
+  ASSERT_EQ(17, file_meta.oldest_blob_file_number);
+#else
+  ASSERT_EQ(10000, file_meta.fd.largest_seqno);
+#endif
   mock_table_factory_->AssertSingleFile(inserted_keys);
   job_context.Clean();
 }
@@ -261,6 +305,7 @@ TEST_F(FlushJobTest, FlushMemTablesSingleColumnFamily) {
   ASSERT_EQ(0, file_meta.fd.smallest_seqno);
   ASSERT_EQ(SequenceNumber(num_mems_to_flush * num_keys_per_table - 1),
             file_meta.fd.largest_seqno);
+  ASSERT_EQ(kInvalidBlobFileNumber, file_meta.oldest_blob_file_number);
 
   for (auto m : to_delete) {
     delete m;
@@ -306,18 +351,18 @@ TEST_F(FlushJobTest, FlushMemtablesMultipleColumnFamilies) {
 
   EventLogger event_logger(db_options_.info_log.get());
   SnapshotChecker* snapshot_checker = nullptr;  // not relevant
-  std::vector<FlushJob> flush_jobs;
+  std::vector<std::unique_ptr<FlushJob>> flush_jobs;
   k = 0;
   for (auto cfd : all_cfds) {
     std::vector<SequenceNumber> snapshot_seqs;
-    flush_jobs.emplace_back(
+    flush_jobs.emplace_back(new FlushJob(
         dbname_, cfd, db_options_, *cfd->GetLatestMutableCFOptions(),
         &memtable_ids[k], env_options_, versions_.get(), &mutex_,
         &shutting_down_, snapshot_seqs, kMaxSequenceNumber, snapshot_checker,
         &job_context, nullptr, nullptr, nullptr, kNoCompression,
         db_options_.statistics.get(), &event_logger, true,
         false /* sync_output_directory */, false /* write_manifest */,
-        Env::Priority::USER);
+        Env::Priority::USER));
     k++;
   }
   HistogramData hist;
@@ -326,12 +371,12 @@ TEST_F(FlushJobTest, FlushMemtablesMultipleColumnFamilies) {
   file_metas.reserve(flush_jobs.size());
   mutex_.Lock();
   for (auto& job : flush_jobs) {
-    job.PickMemTable();
+    job->PickMemTable();
   }
   for (auto& job : flush_jobs) {
     FileMetaData meta;
     // Run will release and re-acquire  mutex
-    ASSERT_OK(job.Run(nullptr /**/, &meta));
+    ASSERT_OK(job->Run(nullptr /**/, &meta));
     file_metas.emplace_back(meta);
   }
   autovector<FileMetaData*> file_meta_ptrs;
@@ -340,7 +385,7 @@ TEST_F(FlushJobTest, FlushMemtablesMultipleColumnFamilies) {
   }
   autovector<const autovector<MemTable*>*> mems_list;
   for (size_t i = 0; i != all_cfds.size(); ++i) {
-    const auto& mems = flush_jobs[i].GetMemTables();
+    const auto& mems = flush_jobs[i]->GetMemTables();
     mems_list.push_back(&mems);
   }
   autovector<const MutableCFOptions*> mutable_cf_options_list;

db/flush_scheduler.cc

@@ -60,9 +60,7 @@ ColumnFamilyData* FlushScheduler::TakeNextColumnFamily() {
     }
 
     // no longer relevant, retry
-    if (cfd->Unref()) {
-      delete cfd;
-    }
+    cfd->UnrefAndTryDelete();
   }
 }
 
@@ -80,9 +78,7 @@ bool FlushScheduler::Empty() {
 void FlushScheduler::Clear() {
   ColumnFamilyData* cfd;
   while ((cfd = TakeNextColumnFamily()) != nullptr) {
-    if (cfd->Unref()) {
-      delete cfd;
-    }
+    cfd->UnrefAndTryDelete();
   }
   assert(head_.load(std::memory_order_relaxed) == nullptr);
 }

db/forward_iterator.cc

@@ -239,9 +239,13 @@ void ForwardIterator::SVCleanup(DBImpl* db, SuperVersion* sv,
     db->FindObsoleteFiles(&job_context, false, true);
     if (background_purge_on_iterator_cleanup) {
       db->ScheduleBgLogWriterClose(&job_context);
+      db->AddSuperVersionsToFreeQueue(sv);
+      db->SchedulePurge();
     }
     db->mutex_.Unlock();
-    delete sv;
+    if (!background_purge_on_iterator_cleanup) {
+      delete sv;
+    }
     if (job_context.HaveSomethingToDelete()) {
       db->PurgeObsoleteFiles(job_context, background_purge_on_iterator_cleanup);
     }
@@ -614,7 +618,7 @@ void ForwardIterator::RebuildIterators(bool refresh_sv) {
   Cleanup(refresh_sv);
   if (refresh_sv) {
     // New
-    sv_ = cfd_->GetReferencedSuperVersion(&(db_->mutex_));
+    sv_ = cfd_->GetReferencedSuperVersion(db_);
   }
   ReadRangeDelAggregator range_del_agg(&cfd_->internal_comparator(),
                                        kMaxSequenceNumber /* upper_bound */);
@@ -668,7 +672,7 @@ void ForwardIterator::RebuildIterators(bool refresh_sv) {
 void ForwardIterator::RenewIterators() {
   SuperVersion* svnew;
   assert(sv_);
-  svnew = cfd_->GetReferencedSuperVersion(&(db_->mutex_));
+  svnew = cfd_->GetReferencedSuperVersion(db_);
 
   if (mutable_iter_ != nullptr) {
     DeleteIterator(mutable_iter_, true /* is_arena */);

db/import_column_family_job.cc

@@ -92,14 +92,15 @@ Status ImportColumnFamilyJob::Prepare(uint64_t next_file_number,
         cfd_->ioptions()->cf_paths, f.fd.GetNumber(), f.fd.GetPathId());
 
     if (hardlink_files) {
-      status = env_->LinkFile(path_outside_db, path_inside_db);
+      status =
+          fs_->LinkFile(path_outside_db, path_inside_db, IOOptions(), nullptr);
       if (status.IsNotSupported()) {
         // Original file is on a different FS, use copy instead of hard linking
         hardlink_files = false;
       }
     }
     if (!hardlink_files) {
-      status = CopyFile(env_, path_outside_db, path_inside_db, 0,
+      status = CopyFile(fs_, path_outside_db, path_inside_db, 0,
                         db_options_.use_fsync);
     }
     if (!status.ok()) {
@@ -115,7 +116,8 @@ Status ImportColumnFamilyJob::Prepare(uint64_t next_file_number,
       if (f.internal_file_path.empty()) {
         break;
       }
-      const auto s = env_->DeleteFile(f.internal_file_path);
+      const auto s =
+          fs_->DeleteFile(f.internal_file_path, IOOptions(), nullptr);
       if (!s.ok()) {
         ROCKS_LOG_WARN(db_options_.info_log,
                        "AddFile() clean up for file %s failed : %s",
@@ -133,13 +135,25 @@ Status ImportColumnFamilyJob::Run() {
   Status status;
   edit_.SetColumnFamily(cfd_->GetID());
 
+  // We use the import time as the ancester time. This is the time the data
+  // is written to the database.
+  int64_t temp_current_time = 0;
+  uint64_t oldest_ancester_time = kUnknownOldestAncesterTime;
+  uint64_t current_time = kUnknownOldestAncesterTime;
+  if (env_->GetCurrentTime(&temp_current_time).ok()) {
+    current_time = oldest_ancester_time =
+        static_cast<uint64_t>(temp_current_time);
+  }
+
   for (size_t i = 0; i < files_to_import_.size(); ++i) {
     const auto& f = files_to_import_[i];
     const auto& file_metadata = metadata_[i];
+
     edit_.AddFile(file_metadata.level, f.fd.GetNumber(), f.fd.GetPathId(),
                   f.fd.GetFileSize(), f.smallest_internal_key,
                   f.largest_internal_key, file_metadata.smallest_seqno,
-                  file_metadata.largest_seqno, false);
+                  file_metadata.largest_seqno, false, kInvalidBlobFileNumber,
+                  oldest_ancester_time, current_time);
 
     // If incoming sequence number is higher, update local sequence number.
     if (file_metadata.largest_seqno > versions_->LastSequence()) {
@@ -156,7 +170,8 @@ void ImportColumnFamilyJob::Cleanup(const Status& status) {
   if (!status.ok()) {
     // We failed to add files to the database remove all the files we copied.
     for (const auto& f : files_to_import_) {
-      const auto s = env_->DeleteFile(f.internal_file_path);
+      const auto s =
+          fs_->DeleteFile(f.internal_file_path, IOOptions(), nullptr);
       if (!s.ok()) {
         ROCKS_LOG_WARN(db_options_.info_log,
                        "AddFile() clean up for file %s failed : %s",
@@ -166,7 +181,8 @@ void ImportColumnFamilyJob::Cleanup(const Status& status) {
   } else if (status.ok() && import_options_.move_files) {
     // The files were moved and added successfully, remove original file links
     for (IngestedFileInfo& f : files_to_import_) {
-      const auto s = env_->DeleteFile(f.external_file_path);
+      const auto s =
+          fs_->DeleteFile(f.external_file_path, IOOptions(), nullptr);
       if (!s.ok()) {
         ROCKS_LOG_WARN(
             db_options_.info_log,
@@ -184,17 +200,19 @@ Status ImportColumnFamilyJob::GetIngestedFileInfo(
   file_to_import->external_file_path = external_file;
 
   // Get external file size
-  auto status = env_->GetFileSize(external_file, &file_to_import->file_size);
+  Status status = fs_->GetFileSize(external_file, IOOptions(),
+                                   &file_to_import->file_size, nullptr);
   if (!status.ok()) {
     return status;
   }
 
   // Create TableReader for external file
   std::unique_ptr<TableReader> table_reader;
-  std::unique_ptr<RandomAccessFile> sst_file;
+  std::unique_ptr<FSRandomAccessFile> sst_file;
   std::unique_ptr<RandomAccessFileReader> sst_file_reader;
 
-  status = env_->NewRandomAccessFile(external_file, &sst_file, env_options_);
+  status = fs_->NewRandomAccessFile(external_file, env_options_,
+                                    &sst_file, nullptr);
   if (!status.ok()) {
     return status;
   }

db/import_column_family_job.h

@@ -29,6 +29,7 @@ class ImportColumnFamilyJob {
         versions_(versions),
         cfd_(cfd),
         db_options_(db_options),
+        fs_(db_options_.fs.get()),
         env_options_(env_options),
         import_options_(import_options),
         metadata_(metadata) {}
@@ -60,6 +61,7 @@ class ImportColumnFamilyJob {
   VersionSet* versions_;
   ColumnFamilyData* cfd_;
   const ImmutableDBOptions& db_options_;
+  FileSystem* fs_;
   const EnvOptions& env_options_;
   autovector<IngestedFileInfo> files_to_import_;
   VersionEdit edit_;

db/internal_stats.cc

@@ -231,6 +231,7 @@ static const std::string is_file_deletions_enabled =
     "is-file-deletions-enabled";
 static const std::string num_snapshots = "num-snapshots";
 static const std::string oldest_snapshot_time = "oldest-snapshot-time";
+static const std::string oldest_snapshot_sequence = "oldest-snapshot-sequence";
 static const std::string num_live_versions = "num-live-versions";
 static const std::string current_version_number =
     "current-super-version-number";
@@ -309,6 +310,8 @@ const std::string DB::Properties::kNumSnapshots =
     rocksdb_prefix + num_snapshots;
 const std::string DB::Properties::kOldestSnapshotTime =
     rocksdb_prefix + oldest_snapshot_time;
+const std::string DB::Properties::kOldestSnapshotSequence =
+    rocksdb_prefix + oldest_snapshot_sequence;
 const std::string DB::Properties::kNumLiveVersions =
     rocksdb_prefix + num_live_versions;
 const std::string DB::Properties::kCurrentSuperVersionNumber =
@@ -427,6 +430,9 @@ const std::unordered_map<std::string, DBPropertyInfo>
         {DB::Properties::kOldestSnapshotTime,
          {false, nullptr, &InternalStats::HandleOldestSnapshotTime, nullptr,
           nullptr}},
+        {DB::Properties::kOldestSnapshotSequence,
+         {false, nullptr, &InternalStats::HandleOldestSnapshotSequence, nullptr,
+          nullptr}},
         {DB::Properties::kNumLiveVersions,
          {false, nullptr, &InternalStats::HandleNumLiveVersions, nullptr,
           nullptr}},
@@ -772,6 +778,12 @@ bool InternalStats::HandleOldestSnapshotTime(uint64_t* value, DBImpl* db,
   return true;
 }
 
+bool InternalStats::HandleOldestSnapshotSequence(uint64_t* value, DBImpl* db,
+                                                 Version* /*version*/) {
+  *value = static_cast<uint64_t>(db->snapshots().GetOldestSnapshotSequence());
+  return true;
+}
+
 bool InternalStats::HandleNumLiveVersions(uint64_t* value, DBImpl* /*db*/,
                                           Version* /*version*/) {
   *value = cfd_->GetNumLiveVersions();

db/internal_stats.h

@@ -553,6 +553,8 @@ class InternalStats {
   bool HandleEstimateNumKeys(uint64_t* value, DBImpl* db, Version* version);
   bool HandleNumSnapshots(uint64_t* value, DBImpl* db, Version* version);
   bool HandleOldestSnapshotTime(uint64_t* value, DBImpl* db, Version* version);
+  bool HandleOldestSnapshotSequence(uint64_t* value, DBImpl* db,
+                                    Version* version);
   bool HandleNumLiveVersions(uint64_t* value, DBImpl* db, Version* version);
   bool HandleCurrentSuperVersionNumber(uint64_t* value, DBImpl* db,
                                        Version* version);

db/listener_test.cc

@@ -3,6 +3,7 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).
 
+#include "db/blob_index.h"
 #include "db/db_impl/db_impl.h"
 #include "db/db_test_util.h"
 #include "db/dbformat.h"
@@ -42,6 +43,14 @@ class EventListenerTest : public DBTestBase {
  public:
   EventListenerTest() : DBTestBase("/listener_test") {}
 
+  static std::string BlobStr(uint64_t blob_file_number, uint64_t offset,
+                             uint64_t size) {
+    std::string blob_index;
+    BlobIndex::EncodeBlob(&blob_index, blob_file_number, offset, size,
+                          kNoCompression);
+    return blob_index;
+  }
+
   const size_t k110KB = 110 << 10;
 };
 
@@ -79,11 +88,47 @@ class TestPropertiesCollectorFactory : public TablePropertiesCollectorFactory {
 
 class TestCompactionListener : public EventListener {
  public:
+  explicit TestCompactionListener(EventListenerTest* test) : test_(test) {}
+
   void OnCompactionCompleted(DB *db, const CompactionJobInfo& ci) override {
     std::lock_guard<std::mutex> lock(mutex_);
     compacted_dbs_.push_back(db);
     ASSERT_GT(ci.input_files.size(), 0U);
+    ASSERT_EQ(ci.input_files.size(), ci.input_file_infos.size());
+
+    for (size_t i = 0; i < ci.input_file_infos.size(); ++i) {
+      ASSERT_EQ(ci.input_file_infos[i].level, ci.base_input_level);
+      ASSERT_EQ(ci.input_file_infos[i].file_number,
+                TableFileNameToNumber(ci.input_files[i]));
+    }
+
     ASSERT_GT(ci.output_files.size(), 0U);
+    ASSERT_EQ(ci.output_files.size(), ci.output_file_infos.size());
+
+    ASSERT_TRUE(test_);
+    ASSERT_EQ(test_->db_, db);
+
+    std::vector<std::vector<FileMetaData>> files_by_level;
+    test_->dbfull()->TEST_GetFilesMetaData(test_->handles_[ci.cf_id],
+                                           &files_by_level);
+    ASSERT_GT(files_by_level.size(), ci.output_level);
+
+    for (size_t i = 0; i < ci.output_file_infos.size(); ++i) {
+      ASSERT_EQ(ci.output_file_infos[i].level, ci.output_level);
+      ASSERT_EQ(ci.output_file_infos[i].file_number,
+                TableFileNameToNumber(ci.output_files[i]));
+
+      auto it = std::find_if(
+          files_by_level[ci.output_level].begin(),
+          files_by_level[ci.output_level].end(), [&](const FileMetaData& meta) {
+            return meta.fd.GetNumber() == ci.output_file_infos[i].file_number;
+          });
+      ASSERT_NE(it, files_by_level[ci.output_level].end());
+
+      ASSERT_EQ(ci.output_file_infos[i].oldest_blob_file_number,
+                it->oldest_blob_file_number);
+    }
+
     ASSERT_EQ(db->GetEnv()->GetThreadID(), ci.thread_id);
     ASSERT_GT(ci.thread_id, 0U);
 
@@ -98,6 +143,7 @@ class TestCompactionListener : public EventListener {
     }
   }
 
+  EventListenerTest* test_;
   std::vector<DB*> compacted_dbs_;
   std::mutex mutex_;
 };
@@ -125,13 +171,19 @@ TEST_F(EventListenerTest, OnSingleDBCompactionTest) {
   options.table_properties_collector_factories.push_back(
       std::make_shared<TestPropertiesCollectorFactory>());
 
-  TestCompactionListener* listener = new TestCompactionListener();
+  TestCompactionListener* listener = new TestCompactionListener(this);
   options.listeners.emplace_back(listener);
   std::vector<std::string> cf_names = {
       "pikachu", "ilya", "muromec", "dobrynia",
       "nikitich", "alyosha", "popovich"};
   CreateAndReopenWithCF(cf_names, options);
   ASSERT_OK(Put(1, "pikachu", std::string(90000, 'p')));
+
+  WriteBatch batch;
+  ASSERT_OK(WriteBatchInternal::PutBlobIndex(&batch, 1, "ditto",
+                                             BlobStr(123, 0, 1 << 10)));
+  ASSERT_OK(dbfull()->Write(WriteOptions(), &batch));
+
   ASSERT_OK(Put(2, "ilya", std::string(90000, 'i')));
   ASSERT_OK(Put(3, "muromec", std::string(90000, 'm')));
   ASSERT_OK(Put(4, "dobrynia", std::string(90000, 'd')));
@@ -140,11 +192,9 @@ TEST_F(EventListenerTest, OnSingleDBCompactionTest) {
   ASSERT_OK(Put(7, "popovich", std::string(90000, 'p')));
   for (int i = 1; i < 8; ++i) {
     ASSERT_OK(Flush(i));
-    const Slice kRangeStart = "a";
-    const Slice kRangeEnd = "z";
-    ASSERT_OK(dbfull()->CompactRange(CompactRangeOptions(), handles_[i],
-                                     &kRangeStart, &kRangeEnd));
     dbfull()->TEST_WaitForFlushMemTable();
+    ASSERT_OK(dbfull()->CompactRange(CompactRangeOptions(), handles_[i],
+                                     nullptr, nullptr));
     dbfull()->TEST_WaitForCompact();
   }
 
@@ -157,8 +207,8 @@ TEST_F(EventListenerTest, OnSingleDBCompactionTest) {
 // This simple Listener can only handle one flush at a time.
 class TestFlushListener : public EventListener {
  public:
-  explicit TestFlushListener(Env* env)
-      : slowdown_count(0), stop_count(0), db_closed(), env_(env) {
+  TestFlushListener(Env* env, EventListenerTest* test)
+      : slowdown_count(0), stop_count(0), db_closed(), env_(env), test_(test) {
     db_closed = false;
   }
   void OnTableFileCreated(
@@ -210,6 +260,27 @@ class TestFlushListener : public EventListener {
     ASSERT_EQ(prev_fc_info_.cf_name, info.cf_name);
     ASSERT_EQ(prev_fc_info_.job_id, info.job_id);
     ASSERT_EQ(prev_fc_info_.file_path, info.file_path);
+    ASSERT_EQ(TableFileNameToNumber(info.file_path), info.file_number);
+
+    // Note: the following chunk relies on the notification pertaining to the
+    // database pointed to by DBTestBase::db_, and is thus bypassed when
+    // that assumption does not hold (see the test case MultiDBMultiListeners
+    // below).
+    ASSERT_TRUE(test_);
+    if (db == test_->db_) {
+      std::vector<std::vector<FileMetaData>> files_by_level;
+      test_->dbfull()->TEST_GetFilesMetaData(test_->handles_[info.cf_id],
+                                             &files_by_level);
+
+      ASSERT_FALSE(files_by_level.empty());
+      auto it = std::find_if(files_by_level[0].begin(), files_by_level[0].end(),
+                             [&](const FileMetaData& meta) {
+                               return meta.fd.GetNumber() == info.file_number;
+                             });
+      ASSERT_NE(it, files_by_level[0].end());
+      ASSERT_EQ(info.oldest_blob_file_number, it->oldest_blob_file_number);
+    }
+
     ASSERT_EQ(db->GetEnv()->GetThreadID(), info.thread_id);
     ASSERT_GT(info.thread_id, 0U);
     ASSERT_EQ(info.table_properties.user_collected_properties.find("0")->second,
@@ -226,6 +297,7 @@ class TestFlushListener : public EventListener {
 
  protected:
   Env* env_;
+  EventListenerTest* test_;
 };
 
 TEST_F(EventListenerTest, OnSingleDBFlushTest) {
@@ -235,7 +307,7 @@ TEST_F(EventListenerTest, OnSingleDBFlushTest) {
 #ifdef ROCKSDB_USING_THREAD_STATUS
   options.enable_thread_tracking = true;
 #endif  // ROCKSDB_USING_THREAD_STATUS
-  TestFlushListener* listener = new TestFlushListener(options.env);
+  TestFlushListener* listener = new TestFlushListener(options.env, this);
   options.listeners.emplace_back(listener);
   std::vector<std::string> cf_names = {
       "pikachu", "ilya", "muromec", "dobrynia",
@@ -245,6 +317,12 @@ TEST_F(EventListenerTest, OnSingleDBFlushTest) {
   CreateAndReopenWithCF(cf_names, options);
 
   ASSERT_OK(Put(1, "pikachu", std::string(90000, 'p')));
+
+  WriteBatch batch;
+  ASSERT_OK(WriteBatchInternal::PutBlobIndex(&batch, 1, "ditto",
+                                             BlobStr(456, 0, 1 << 10)));
+  ASSERT_OK(dbfull()->Write(WriteOptions(), &batch));
+
   ASSERT_OK(Put(2, "ilya", std::string(90000, 'i')));
   ASSERT_OK(Put(3, "muromec", std::string(90000, 'm')));
   ASSERT_OK(Put(4, "dobrynia", std::string(90000, 'd')));
@@ -272,7 +350,7 @@ TEST_F(EventListenerTest, MultiCF) {
 #ifdef ROCKSDB_USING_THREAD_STATUS
   options.enable_thread_tracking = true;
 #endif  // ROCKSDB_USING_THREAD_STATUS
-  TestFlushListener* listener = new TestFlushListener(options.env);
+  TestFlushListener* listener = new TestFlushListener(options.env, this);
   options.listeners.emplace_back(listener);
   options.table_properties_collector_factories.push_back(
       std::make_shared<TestPropertiesCollectorFactory>());
@@ -313,7 +391,7 @@ TEST_F(EventListenerTest, MultiDBMultiListeners) {
   const int kNumDBs = 5;
   const int kNumListeners = 10;
   for (int i = 0; i < kNumListeners; ++i) {
-    listeners.emplace_back(new TestFlushListener(options.env));
+    listeners.emplace_back(new TestFlushListener(options.env, this));
   }
 
   std::vector<std::string> cf_names = {
@@ -390,7 +468,7 @@ TEST_F(EventListenerTest, DisableBGCompaction) {
 #ifdef ROCKSDB_USING_THREAD_STATUS
   options.enable_thread_tracking = true;
 #endif  // ROCKSDB_USING_THREAD_STATUS
-  TestFlushListener* listener = new TestFlushListener(options.env);
+  TestFlushListener* listener = new TestFlushListener(options.env, this);
   const int kCompactionTrigger = 1;
   const int kSlowdownTrigger = 5;
   const int kStopTrigger = 100;