Bump up version to 6.5.3

Update HISTORY.md with the recent memtable trimming fixes
Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/6194 Differential Revision: D19125292 Pulled By: ltamasi fbshipit-source-id: d41aca2755ec4bec07feedd6b561e8d18606a931
2026-11-06 11:02:36 -05:00 · 2020-01-10 09:58:14 -08:00 · 2020-01-10 09:57:50 -08:00 · 2020-01-10 09:54:56 -08:00 · 2020-01-10 09:53:31 -08:00 · 2020-01-10 09:53:31 -08:00
662 changed files with 54812 additions and 16548 deletions
@@ -24,6 +24,7 @@ make_config.mk
 *.vcxproj.filters
 *.sln
 *.cmake
+.watchmanconfig
 CMakeCache.txt
 CMakeFiles/
 build/
@@ -32,6 +33,8 @@ ldb
 manifest_dump
 sst_dump
 blob_dump
+block_cache_trace_analyzer
+tools/block_cache_analyzer/*.pyc
 column_aware_encoding_exp
 util/build_version.cc
 build_tools/VALGRIND_LOGS/
@@ -47,6 +50,8 @@ rocksdb_undump
 db_test2
 trace_analyzer
 trace_analyzer_test
+block_cache_trace_analyzer
+.DS_Store

 java/out
 java/target
@@ -1,5 +1,4 @@
-sudo: false
-dist: trusty
+dist: xenial
 language: cpp
 os:
  - linux
@@ -9,7 +8,7 @@ compiler:
  - gcc
 osx_image: xcode8.3
 jdk:
-  - oraclejdk7
+  - openjdk7
 cache:
  - ccache
  - apt
@@ -71,7 +70,10 @@ install:
      CC=gcc-8 && CXX=g++-8;
    fi
  - if [[ "${JOB_NAME}" == cmake* ]] && [ "${TRAVIS_OS_NAME}" == linux ]; then
-      mkdir cmake-dist && curl -sfSL https://cmake.org/files/v3.8/cmake-3.8.1-Linux-x86_64.tar.gz | tar --strip-components=1 -C cmake-dist -xz && export PATH=$PWD/cmake-dist/bin:$PATH;
+      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
+  - if [[ "${JOB_NAME}" == java_test ]]; then
+      java -version && echo "JAVA_HOME=${JAVA_HOME}";
    fi

 before_script:
@@ -101,7 +103,7 @@ script:
    esac
  - case $JOB_NAME in
    java_test)
-      OPT=-DTRAVIS V=1 make clean jclean && make rocksdbjava jtest
+      OPT=-DTRAVIS V=1 make rocksdbjava jtest
      ;;
    lite_build)
      OPT='-DTRAVIS -DROCKSDB_LITE' V=1 make -j4 static_lib tools
@@ -110,6 +112,7 @@ script:
      OPT=-DTRAVIS V=1 make -j4 static_lib && cd examples && make -j4
      ;;
    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
      ;;
    cmake*)
@@ -0,0 +1,6 @@
+{
+  "content_hash_warming": true,
+  "content_hash_max_items": 333333,
+  "hint_num_files_per_dir": 8,
+  "fsevents_latency": 0.05
+}
@@ -32,17 +32,24 @@
 # 3. cmake ..
 # 4. make -j

-cmake_minimum_required(VERSION 2.8.12)
-project(rocksdb)
-enable_language(CXX)
-enable_language(C)
-enable_language(ASM)
+cmake_minimum_required(VERSION 3.5.1)
+
+list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_LIST_DIR}/cmake/modules/")
+include(ReadVersion)
+get_rocksdb_version(rocksdb_VERSION)
+project(rocksdb
+  VERSION ${rocksdb_VERSION}
+  LANGUAGES CXX C ASM)

 if(POLICY CMP0042)
  cmake_policy(SET CMP0042 NEW)
 endif()

-list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_LIST_DIR}/cmake/modules/")
+find_program(CCACHE_FOUND ccache)
+if(CCACHE_FOUND)
+  set_property(GLOBAL PROPERTY RULE_LAUNCH_COMPILE ccache)
+  set_property(GLOBAL PROPERTY RULE_LAUNCH_LINK ccache)
+endif(CCACHE_FOUND)

 option(WITH_JEMALLOC "build with JeMalloc" OFF)
 option(WITH_SNAPPY "build with SNAPPY" OFF)
@@ -53,6 +60,13 @@ option(WITH_WINDOWS_UTF8_FILENAMES "use UTF8 as characterset for opening files,
 if (WITH_WINDOWS_UTF8_FILENAMES)
  add_definitions(-DROCKSDB_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)
+else()
+  option(WITH_FOLLY_DISTRIBUTED_MUTEX "build with folly::DistributedMutex" OFF)
+endif()
 if(MSVC)
  # Defaults currently different for GFLAGS.
  #  We will address find_package work a little later
@@ -68,8 +82,7 @@ else()
    if(WITH_JEMALLOC)
      find_package(JeMalloc REQUIRED)
      add_definitions(-DROCKSDB_JEMALLOC -DJEMALLOC_NO_DEMANGLE)
-      include_directories(${JEMALLOC_INCLUDE_DIR})
-      list(APPEND THIRDPARTY_LIBS ${JEMALLOC_LIBRARIES})
+      list(APPEND THIRDPARTY_LIBS JeMalloc::JeMalloc)
    endif()
  endif()

@@ -87,47 +100,43 @@ else()
  if(WITH_SNAPPY)
    find_package(snappy REQUIRED)
    add_definitions(-DSNAPPY)
-    include_directories(${SNAPPY_INCLUDE_DIR})
-    list(APPEND THIRDPARTY_LIBS ${SNAPPY_LIBRARIES})
+    list(APPEND THIRDPARTY_LIBS snappy::snappy)
  endif()

  if(WITH_ZLIB)
    find_package(ZLIB REQUIRED)
    add_definitions(-DZLIB)
-    if(ZLIB_INCLUDE_DIRS)
-      # CMake 3
-      include_directories(${ZLIB_INCLUDE_DIRS})
-    else()
-      # CMake 2
-      include_directories(${ZLIB_INCLUDE_DIR})
-    endif()
-    list(APPEND THIRDPARTY_LIBS ${ZLIB_LIBRARIES})
+    list(APPEND THIRDPARTY_LIBS ZLIB::ZLIB)
  endif()

  option(WITH_BZ2 "build with bzip2" OFF)
  if(WITH_BZ2)
-    find_package(bzip2 REQUIRED)
+    find_package(BZip2 REQUIRED)
    add_definitions(-DBZIP2)
-    include_directories(${BZIP2_INCLUDE_DIR})
+    if(BZIP2_INCLUDE_DIRS)
+      include_directories(${BZIP2_INCLUDE_DIRS})
+    else()
+      include_directories(${BZIP2_INCLUDE_DIR})
+    endif()
    list(APPEND THIRDPARTY_LIBS ${BZIP2_LIBRARIES})
  endif()

  if(WITH_LZ4)
    find_package(lz4 REQUIRED)
    add_definitions(-DLZ4)
-    include_directories(${LZ4_INCLUDE_DIR})
-    list(APPEND THIRDPARTY_LIBS ${LZ4_LIBRARIES})
+    list(APPEND THIRDPARTY_LIBS lz4::lz4)
  endif()

  if(WITH_ZSTD)
    find_package(zstd REQUIRED)
    add_definitions(-DZSTD)
    include_directories(${ZSTD_INCLUDE_DIR})
-    list(APPEND THIRDPARTY_LIBS ${ZSTD_LIBRARIES})
+    list(APPEND THIRDPARTY_LIBS zstd::zstd)
  endif()
 endif()

-string(TIMESTAMP GIT_DATE_TIME "%Y/%m/%d %H:%M:%S" UTC)
+string(TIMESTAMP TS "%Y/%m/%d %H:%M:%S" UTC)
+set(GIT_DATE_TIME "${TS}" CACHE STRING "the time we first built rocksdb")

 find_package(Git)

@@ -144,17 +153,6 @@ endif()
 string(REGEX REPLACE "[^0-9a-f]+" "" GIT_SHA "${GIT_SHA}")


-# Read rocksdb version from version.h header file.
-file(READ include/rocksdb/version.h version_header_file)
-string(REGEX MATCH "#define ROCKSDB_MAJOR ([0-9]+)" _ ${version_header_file})
-set(ROCKSDB_VERSION_MAJOR ${CMAKE_MATCH_1})
-string(REGEX MATCH "#define ROCKSDB_MINOR ([0-9]+)" _ ${version_header_file})
-set(ROCKSDB_VERSION_MINOR ${CMAKE_MATCH_1})
-string(REGEX MATCH "#define ROCKSDB_PATCH ([0-9]+)" _ ${version_header_file})
-set(ROCKSDB_VERSION_PATCH ${CMAKE_MATCH_1})
-set(ROCKSDB_VERSION ${ROCKSDB_VERSION_MAJOR}.${ROCKSDB_VERSION_MINOR}.${ROCKSDB_VERSION_PATCH})
-
-
 option(WITH_MD_LIBRARY "build with MD" ON)
 if(WIN32 AND MSVC)
  if(WITH_MD_LIBRARY)
@@ -177,6 +175,7 @@ else()
  set(CMAKE_CXX_FLAGS "${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 "${CMAKE_CXX_FLAGS} -Wno-format")
+    add_definitions(-D_POSIX_C_SOURCE=1)
  endif()
  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
  if(NOT CMAKE_BUILD_TYPE STREQUAL "Debug")
@@ -201,6 +200,15 @@ if(CMAKE_SYSTEM_PROCESSOR MATCHES "ppc64le")
  endif(HAS_ALTIVEC)
 endif(CMAKE_SYSTEM_PROCESSOR MATCHES "ppc64le")

+if(CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64|AARCH64")
+        CHECK_C_COMPILER_FLAG("-march=armv8-a+crc+crypto" HAS_ARMV8_CRC)
+  if(HAS_ARMV8_CRC)
+    message(STATUS " HAS_ARMV8_CRC yes")
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=armv8-a+crc+crypto -Wno-unused-function")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=armv8-a+crc+crypto -Wno-unused-function")
+  endif(HAS_ARMV8_CRC)
+endif(CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64|AARCH64")
+
 option(PORTABLE "build a portable binary" OFF)
 option(FORCE_SSE42 "force building with SSE4.2, even when PORTABLE=ON" OFF)
 if(PORTABLE)
@@ -213,7 +221,7 @@ else()
  if(MSVC)
    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /arch:AVX2")
  else()
-    if(NOT HAVE_POWER8)
+    if(NOT HAVE_POWER8 AND NOT HAS_ARMV8_CRC)
      set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=native")
    endif()
  endif()
@@ -300,15 +308,14 @@ if(WITH_NUMA)
  find_package(NUMA REQUIRED)
  add_definitions(-DNUMA)
  include_directories(${NUMA_INCLUDE_DIR})
-  list(APPEND THIRDPARTY_LIBS ${NUMA_LIBRARIES})
+  list(APPEND THIRDPARTY_LIBS NUMA::NUMA)
 endif()

 option(WITH_TBB "build with Threading Building Blocks (TBB)" OFF)
 if(WITH_TBB)
  find_package(TBB REQUIRED)
  add_definitions(-DTBB)
-  include_directories(${TBB_INCLUDE_DIR})
-  list(APPEND THIRDPARTY_LIBS ${TBB_LIBRARIES})
+  list(APPEND THIRDPARTY_LIBS TBB::TBB)
 endif()

 # Stall notifications eat some performance from inserts
@@ -317,6 +324,10 @@ if(DISABLE_STALL_NOTIF)
  add_definitions(-DROCKSDB_DISABLE_STALL_NOTIFICATION)
 endif()

+option(WITH_DYNAMIC_EXTENSION "build with dynamic extension support" OFF)
+if(NOT WITH_DYNAMIC_EXTENSION)
+  add_definitions(-DROCKSDB_NO_DYNAMIC_EXTENSION)
+endif()

 if(DEFINED USE_RTTI)
  if(USE_RTTI)
@@ -458,7 +469,10 @@ endif()

 include_directories(${PROJECT_SOURCE_DIR})
 include_directories(${PROJECT_SOURCE_DIR}/include)
-include_directories(SYSTEM ${PROJECT_SOURCE_DIR}/third-party/gtest-1.7.0/fused-src)
+include_directories(SYSTEM ${PROJECT_SOURCE_DIR}/third-party/gtest-1.8.1/fused-src)
+if(WITH_FOLLY_DISTRIBUTED_MUTEX)
+  include_directories(${PROJECT_SOURCE_DIR}/third-party/folly)
+endif()
 find_package(Threads REQUIRED)

 # Main library source code
@@ -471,23 +485,24 @@ set(SOURCES
        db/c.cc
        db/column_family.cc
        db/compacted_db_impl.cc
-        db/compaction.cc
-        db/compaction_iterator.cc
-        db/compaction_job.cc
-        db/compaction_picker.cc
-        db/compaction_picker_fifo.cc
-        db/compaction_picker_universal.cc
+        db/compaction/compaction.cc
+        db/compaction/compaction_iterator.cc
+        db/compaction/compaction_picker.cc
+        db/compaction/compaction_job.cc
+        db/compaction/compaction_picker_fifo.cc
+        db/compaction/compaction_picker_level.cc
+        db/compaction/compaction_picker_universal.cc
        db/convenience.cc
        db/db_filesnapshot.cc
-        db/db_impl.cc
-        db/db_impl_write.cc
-        db/db_impl_compaction_flush.cc
-        db/db_impl_files.cc
-        db/db_impl_open.cc
-        db/db_impl_debug.cc
-        db/db_impl_experimental.cc
-        db/db_impl_readonly.cc
-        db/db_impl_secondary.cc
+        db/db_impl/db_impl.cc
+        db/db_impl/db_impl_write.cc
+        db/db_impl/db_impl_compaction_flush.cc
+        db/db_impl/db_impl_files.cc
+        db/db_impl/db_impl_open.cc
+        db/db_impl/db_impl_debug.cc
+        db/db_impl/db_impl_experimental.cc
+        db/db_impl/db_impl_readonly.cc
+        db/db_impl/db_impl_secondary.cc
        db/db_info_dumper.cc
        db/db_iter.cc
        db/dbformat.cc
@@ -499,8 +514,8 @@ set(SOURCES
        db/flush_job.cc
        db/flush_scheduler.cc
        db/forward_iterator.cc
+        db/import_column_family_job.cc
        db/internal_stats.cc
-        db/in_memory_stats_history.cc
        db/logs_with_prep_tracker.cc
        db/log_reader.cc
        db/log_writer.cc
@@ -516,6 +531,7 @@ set(SOURCES
        db/table_cache.cc
        db/table_properties_collector.cc
        db/transaction_log_impl.cc
+        db/trim_history_scheduler.cc
        db/version_builder.cc
        db/version_edit.cc
        db/version_set.cc
@@ -529,6 +545,16 @@ set(SOURCES
        env/env_encryption.cc
        env/env_hdfs.cc
        env/mock_env.cc
+        file/delete_scheduler.cc
+        file/file_util.cc
+        file/filename.cc
+        file/sst_file_manager_impl.cc
+        logging/auto_roll_logger.cc
+        logging/event_logger.cc
+        logging/log_buffer.cc
+        memory/arena.cc
+        memory/concurrent_arena.cc
+        memory/jemalloc_nodump_allocator.cc
        memtable/alloc_tracker.cc
        memtable/hash_linklist_rep.cc
        memtable/hash_skiplist_rep.cc
@@ -537,10 +563,12 @@ set(SOURCES
        memtable/write_buffer_manager.cc
        monitoring/histogram.cc
        monitoring/histogram_windowing.cc
+        monitoring/in_memory_stats_history.cc
        monitoring/instrumented_mutex.cc
        monitoring/iostats_context.cc
        monitoring/perf_context.cc
        monitoring/perf_level.cc
+        monitoring/persistent_stats_history.cc
        monitoring/statistics.cc
        monitoring/thread_status_impl.cc
        monitoring/thread_status_updater.cc
@@ -553,80 +581,74 @@ set(SOURCES
        options/options_parser.cc
        options/options_sanity_check.cc
        port/stack_trace.cc
-        table/adaptive_table_factory.cc
-        table/block.cc
-        table/block_based_filter_block.cc
-        table/block_based_table_builder.cc
-        table/block_based_table_factory.cc
-        table/block_based_table_reader.cc
-        table/block_builder.cc
+        table/adaptive/adaptive_table_factory.cc
+        table/block_based/block.cc
+        table/block_based/block_based_filter_block.cc
+        table/block_based/block_based_table_builder.cc
+        table/block_based/block_based_table_factory.cc
+        table/block_based/block_based_table_reader.cc
+        table/block_based/block_builder.cc
+        table/block_based/block_prefix_index.cc
+        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/flush_block_policy.cc
+        table/block_based/full_filter_block.cc
+        table/block_based/index_builder.cc
+        table/block_based/partitioned_filter_block.cc
+        table/block_based/uncompression_dict_reader.cc
        table/block_fetcher.cc
-        table/block_prefix_index.cc
-        table/bloom_block.cc
-        table/cuckoo_table_builder.cc
-        table/cuckoo_table_factory.cc
-        table/cuckoo_table_reader.cc
-        table/data_block_hash_index.cc
-        table/data_block_footer.cc
-        table/flush_block_policy.cc
+        table/cuckoo/cuckoo_table_builder.cc
+        table/cuckoo/cuckoo_table_factory.cc
+        table/cuckoo/cuckoo_table_reader.cc
        table/format.cc
-        table/full_filter_block.cc
        table/get_context.cc
-        table/index_builder.cc
        table/iterator.cc
        table/merging_iterator.cc
        table/meta_blocks.cc
-        table/partitioned_filter_block.cc
        table/persistent_cache_helper.cc
-        table/plain_table_builder.cc
-        table/plain_table_factory.cc
-        table/plain_table_index.cc
-        table/plain_table_key_coding.cc
-        table/plain_table_reader.cc
+        table/plain/plain_table_bloom.cc
+        table/plain/plain_table_builder.cc
+        table/plain/plain_table_factory.cc
+        table/plain/plain_table_index.cc
+        table/plain/plain_table_key_coding.cc
+        table/plain/plain_table_reader.cc
        table/sst_file_reader.cc
        table/sst_file_writer.cc
        table/table_properties.cc
        table/two_level_iterator.cc
+        test_util/sync_point.cc
+        test_util/sync_point_impl.cc
+        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
        tools/sst_dump_tool.cc
        tools/trace_analyzer_tool.cc
-        util/arena.cc
-        util/auto_roll_logger.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
        util/compression_context_cache.cc
-        util/concurrent_arena.cc
        util/concurrent_task_limiter_impl.cc
        util/crc32c.cc
-        util/delete_scheduler.cc
        util/dynamic_bloom.cc
-        util/event_logger.cc
        util/file_reader_writer.cc
-        util/file_util.cc
-        util/filename.cc
        util/filter_policy.cc
        util/hash.cc
-        util/jemalloc_nodump_allocator.cc
-        util/log_buffer.cc
        util/murmurhash.cc
        util/random.cc
        util/rate_limiter.cc
        util/slice.cc
-        util/sst_file_manager_impl.cc
        util/status.cc
        util/string_util.cc
-        util/sync_point.cc
-        util/sync_point_impl.cc
-        util/testutil.cc
        util/thread_local.cc
        util/threadpool_imp.cc
-        util/trace_replay.cc
-        util/transaction_test_util.cc
        util/xxhash.cc
        utilities/backupable/backupable_db.cc
        utilities/blob_db/blob_compaction_filter.cc
@@ -651,9 +673,11 @@ set(SOURCES
        utilities/merge_operators/bytesxor.cc
        utilities/merge_operators/max.cc
        utilities/merge_operators/put.cc
+        utilities/merge_operators/sortlist.cc
        utilities/merge_operators/string_append/stringappend.cc
        utilities/merge_operators/string_append/stringappend2.cc
        utilities/merge_operators/uint64add.cc
+        utilities/object_registry.cc
        utilities/option_change_migration/option_change_migration.cc
        utilities/options/options_util.cc
        utilities/persistent_cache/block_cache_tier.cc
@@ -661,6 +685,7 @@ set(SOURCES
        utilities/persistent_cache/block_cache_tier_metadata.cc
        utilities/persistent_cache/persistent_cache_tier.cc
        utilities/persistent_cache/volatile_tier_impl.cc
+        utilities/simulator_cache/cache_simulator.cc
        utilities/simulator_cache/sim_cache.cc
        utilities/table_properties_collectors/compact_on_deletion_collector.cc
        utilities/trace/file_trace_reader_writer.cc
@@ -694,6 +719,11 @@ if(HAVE_POWER8)
    util/crc32c_ppc_asm.S)
 endif(HAVE_POWER8)

+if(HAS_ARMV8_CRC)
+  list(APPEND SOURCES
+    util/crc32c_arm64.cc)
+endif(HAS_ARMV8_CRC)
+
 if(WIN32)
  list(APPEND SOURCES
    port/win/io_win.cc
@@ -720,6 +750,15 @@ else()
    env/io_posix.cc)
 endif()

+if(WITH_FOLLY_DISTRIBUTED_MUTEX)
+  list(APPEND SOURCES
+    third-party/folly/folly/detail/Futex.cpp
+    third-party/folly/folly/synchronization/AtomicNotification.cpp
+    third-party/folly/folly/synchronization/DistributedMutex.cpp
+    third-party/folly/folly/synchronization/ParkingLot.cpp
+    third-party/folly/folly/synchronization/WaitOptions.cpp)
+endif()
+
 set(ROCKSDB_STATIC_LIB rocksdb${ARTIFACT_SUFFIX})
 set(ROCKSDB_SHARED_LIB rocksdb-shared${ARTIFACT_SUFFIX})
 set(ROCKSDB_IMPORT_LIB ${ROCKSDB_SHARED_LIB})
@@ -743,8 +782,8 @@ else()
    ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})
  set_target_properties(${ROCKSDB_SHARED_LIB} PROPERTIES
                        LINKER_LANGUAGE CXX
-                        VERSION ${ROCKSDB_VERSION}
-                        SOVERSION ${ROCKSDB_VERSION_MAJOR}
+                        VERSION ${rocksdb_VERSION}
+                        SOVERSION ${rocksdb_VERSION_MAJOR}
                        CXX_STANDARD 11
                        OUTPUT_NAME "rocksdb")
 endif()
@@ -799,7 +838,7 @@ if(NOT WIN32 OR ROCKSDB_INSTALL_ON_WINDOWS)

  write_basic_package_version_file(
    RocksDBConfigVersion.cmake
-    VERSION ${ROCKSDB_VERSION}
+    VERSION ${rocksdb_VERSION}
    COMPATIBILITY SameMajorVersion
  )

@@ -841,16 +880,16 @@ endif()

 option(WITH_TESTS "build with tests" ON)
 if(WITH_TESTS)
-  add_subdirectory(third-party/gtest-1.7.0/fused-src/gtest)
+  add_subdirectory(third-party/gtest-1.8.1/fused-src/gtest)
  set(TESTS
        cache/cache_test.cc
        cache/lru_cache_test.cc
        db/column_family_test.cc
        db/compact_files_test.cc
-        db/compaction_iterator_test.cc
-        db/compaction_job_stats_test.cc
-        db/compaction_job_test.cc
-        db/compaction_picker_test.cc
+        db/compaction/compaction_job_stats_test.cc
+        db/compaction/compaction_job_test.cc
+        db/compaction/compaction_iterator_test.cc
+        db/compaction/compaction_picker_test.cc
        db/comparator_db_test.cc
        db/corruption_test.cc
        db/cuckoo_table_db_test.cc
@@ -870,10 +909,11 @@ if(WITH_TESTS)
        db/db_log_iter_test.cc
        db/db_memtable_test.cc
        db/db_merge_operator_test.cc
+        db/db_merge_operand_test.cc
        db/db_options_test.cc
        db/db_properties_test.cc
        db/db_range_del_test.cc
-        db/db_secondary_test.cc
+        db/db_impl/db_secondary_test.cc
        db/db_sst_test.cc
        db/db_statistics_test.cc
        db/db_table_properties_test.cc
@@ -917,37 +957,41 @@ if(WITH_TESTS)
        env/env_basic_test.cc
        env/env_test.cc
        env/mock_env_test.cc
+        file/delete_scheduler_test.cc
+        logging/auto_roll_logger_test.cc
+        logging/env_logger_test.cc
+        logging/event_logger_test.cc
+        memory/arena_test.cc
        memtable/inlineskiplist_test.cc
        memtable/skiplist_test.cc
        memtable/write_buffer_manager_test.cc
        monitoring/histogram_test.cc
        monitoring/iostats_context_test.cc
        monitoring/statistics_test.cc
+        monitoring/stats_history_test.cc
        options/options_settable_test.cc
        options/options_test.cc
-        table/block_based_filter_block_test.cc
-        table/block_test.cc
+        table/block_based/block_based_filter_block_test.cc
+        table/block_based/block_test.cc
+        table/block_based/data_block_hash_index_test.cc
+        table/block_based/full_filter_block_test.cc
+        table/block_based/partitioned_filter_block_test.cc
        table/cleanable_test.cc
-        table/cuckoo_table_builder_test.cc
-        table/cuckoo_table_reader_test.cc
-        table/data_block_hash_index_test.cc
-        table/full_filter_block_test.cc
+        table/cuckoo/cuckoo_table_builder_test.cc
+        table/cuckoo/cuckoo_table_reader_test.cc
        table/merger_test.cc
        table/sst_file_reader_test.cc
        table/table_test.cc
+        tools/block_cache_analyzer/block_cache_trace_analyzer_test.cc
        tools/ldb_cmd_test.cc
        tools/reduce_levels_test.cc
        tools/sst_dump_test.cc
        tools/trace_analyzer_test.cc
-        util/arena_test.cc
-        util/auto_roll_logger_test.cc
        util/autovector_test.cc
        util/bloom_test.cc
        util/coding_test.cc
        util/crc32c_test.cc
-        util/delete_scheduler_test.cc
        util/dynamic_bloom_test.cc
-        util/event_logger_test.cc
        util/file_reader_writer_test.cc
        util/filelock_test.cc
        util/hash_test.cc
@@ -972,6 +1016,7 @@ if(WITH_TESTS)
        utilities/options/options_util_test.cc
        utilities/persistent_cache/hash_table_test.cc
        utilities/persistent_cache/persistent_cache_test.cc
+        utilities/simulator_cache/cache_simulator_test.cc
        utilities/simulator_cache/sim_cache_test.cc
        utilities/table_properties_collectors/compact_on_deletion_collector_test.cc
        utilities/transactions/optimistic_transaction_test.cc
@@ -985,6 +1030,10 @@ if(WITH_TESTS)
    list(APPEND TESTS utilities/env_librados_test.cc)
  endif()

+  if(WITH_FOLLY_DISTRIBUTED_MUTEX)
+    list(APPEND TESTS third-party/folly/folly/synchronization/test/DistributedMutexTest.cpp)
+  endif()
+
  set(BENCHMARKS
    cache/cache_bench.cc
    memtable/memtablerep_bench.cc
@@ -992,7 +1041,7 @@ if(WITH_TESTS)
    tools/db_bench.cc
    table/table_reader_bench.cc
    utilities/persistent_cache/hash_table_bench.cc)
-  add_library(testharness OBJECT util/testharness.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}
@@ -1006,7 +1055,7 @@ if(WITH_TESTS)
      db/db_test_util.cc
      monitoring/thread_status_updater_debug.cc
      table/mock_table.cc
-      util/fault_injection_test_env.cc
+      test_util/fault_injection_test_env.cc
      utilities/cassandra/test_utils.cc
  )
  # test utilities are only build in debug
@@ -1,3 +1,77 @@
 # Code of Conduct

-Facebook has adopted a Code of Conduct that we expect project participants to adhere to. Please [read the full text](https://code.facebook.com/codeofconduct) so that you can understand what actions will and will not be tolerated.
+## Our Pledge
+
+In the interest of fostering an open and welcoming environment, we as
+contributors and maintainers pledge to make participation in our project and
+our community a harassment-free experience for everyone, regardless of age, body
+size, disability, ethnicity, sex characteristics, gender identity and expression,
+level of experience, education, socio-economic status, nationality, personal
+appearance, race, religion, or sexual identity and orientation.
+
+## Our Standards
+
+Examples of behavior that contributes to creating a positive environment
+include:
+
+* Using welcoming and inclusive language
+* Being respectful of differing viewpoints and experiences
+* Gracefully accepting constructive criticism
+* Focusing on what is best for the community
+* Showing empathy towards other community members
+
+Examples of unacceptable behavior by participants include:
+
+* The use of sexualized language or imagery and unwelcome sexual attention or
+  advances
+* Trolling, insulting/derogatory comments, and personal or political attacks
+* Public or private harassment
+* Publishing others' private information, such as a physical or electronic
+  address, without explicit permission
+* Other conduct which could reasonably be considered inappropriate in a
+  professional setting
+
+## Our Responsibilities
+
+Project maintainers are responsible for clarifying the standards of acceptable
+behavior and are expected to take appropriate and fair corrective action in
+response to any instances of unacceptable behavior.
+
+Project maintainers have the right and responsibility to remove, edit, or
+reject comments, commits, code, wiki edits, issues, and other contributions
+that are not aligned to this Code of Conduct, or to ban temporarily or
+permanently any contributor for other behaviors that they deem inappropriate,
+threatening, offensive, or harmful.
+
+## Scope
+
+This Code of Conduct applies within all project spaces, and it also applies when
+an individual is representing the project or its community in public spaces.
+Examples of representing a project or community include using an official
+project e-mail address, posting via an official social media account, or acting
+as an appointed representative at an online or offline event. Representation of
+a project may be further defined and clarified by project maintainers.
+
+## Enforcement
+
+Instances of abusive, harassing, or otherwise unacceptable behavior may be
+reported by contacting the project team at <opensource-conduct@fb.com>. All
+complaints will be reviewed and investigated and will result in a response that
+is deemed necessary and appropriate to the circumstances. The project team is
+obligated to maintain confidentiality with regard to the reporter of an incident.
+Further details of specific enforcement policies may be posted separately.
+
+Project maintainers who do not follow or enforce the Code of Conduct in good
+faith may face temporary or permanent repercussions as determined by other
+members of the project's leadership.
+
+## Attribution
+
+This Code of Conduct is adapted from the [Contributor Covenant][homepage], version 1.4,
+available at https://www.contributor-covenant.org/version/1/4/code-of-conduct.html
+
+[homepage]: https://www.contributor-covenant.org
+
+For answers to common questions about this code of conduct, see
+https://www.contributor-covenant.org/faq
+
@@ -1,13 +1,132 @@
 # Rocksdb Change Log
-## Unreleased
+## 6.5.3 (1/10/2020)
+### Bug Fixes
+* 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.
+
+## 6.5.2 (11/15/2019)
+### Bug Fixes
+* Fix a assertion failure in MultiGe4t() 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.
+### 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.
+### 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.
+### 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.
+* Change BlockBasedTableOptions.cache_index_and_filter_blocks_with_high_priority's default value from false to true.
+
+### Public API Change
+* Filter and compression dictionary blocks are now handled similarly to data blocks with regards to the block cache: instead of storing objects in the cache, only the blocks themselves are cached. In addition, filter and compression dictionary blocks (as well as filter partitions) no longer get evicted from the cache when a table is closed.
+* Due to the above refactoring, block cache eviction statistics for filter and compression dictionary blocks are temporarily broken. We plan to reintroduce them in a later phase.
+* The semantics of the per-block-type block read counts in the performance context now match those of the generic block_read_count.
+* Errors related to the retrieval of the compression dictionary are now propagated to the user.
+* db_bench adds a "benchmark" stats_history, which prints out the whole stats history.
+* Overload GetAllKeyVersions() to support non-default column family.
+* Added new APIs ExportColumnFamily() and CreateColumnFamilyWithImport() to support export and import of a Column Family. https://github.com/facebook/rocksdb/issues/3469
+* ldb sometimes uses a string-append merge operator if no merge operator is passed in. This is to allow users to print keys from a DB with a merge operator.
+* Replaces old Registra with ObjectRegistry to allow user to create custom object from string, also add LoadEnv() to Env.
+* Added new overload of GetApproximateSizes which gets SizeApproximationOptions object and returns a Status. The older overloads are redirecting their calls to this new method and no longer assert if the include_flags doesn't have either of INCLUDE_MEMTABLES or INCLUDE_FILES bits set. It's recommended to use the new method only, as it is more type safe and returns a meaningful status in case of errors.
+* LDBCommandRunner::RunCommand() to return the status code as an integer, rather than call exit() using the code.
+
+### New Features
+* Add argument `--secondary_path` to ldb to open the database as the secondary instance. This would keep the original DB intact.
+* Compression dictionary blocks are now prefetched and pinned in the cache (based on the customer's settings) the same way as index and filter blocks.
+* Added DBOptions::log_readahead_size which specifies the number of bytes to prefetch when reading the log. This is mostly useful for reading a remotely located log, as it can save the number of round-trips. If 0 (default), then the prefetching is disabled.
+* Added new option in SizeApproximationOptions used with DB::GetApproximateSizes. When approximating the files total size that is used to store a keys range, allow approximation with an error margin of up to total_files_size * files_size_error_margin. This allows to take some shortcuts in files size approximation, resulting in better performance, while guaranteeing the resulting error is within a reasonable margin.
+* 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.
+* 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.
+
+### Bug Fixes
+* Fix ingested file and directory not being fsync.
+* Return TryAgain status in place of Corruption when new tail is not visible to TransactionLogIterator.
+* Fixed a regression where the fill_cache read option also affected index blocks.
+* Fixed an issue where using cache_index_and_filter_blocks==false affected partitions of partitioned indexes/filters as well.
+
+## 6.3.2 (8/15/2019)
+### Public API Change
+* The semantics of the per-block-type block read counts in the performance context now match those of the generic block_read_count.
+
+### Bug Fixes
+* Fixed a regression where the fill_cache read option also affected index blocks.
+* Fixed an issue where using cache_index_and_filter_blocks==false affected partitions of partitioned indexes as well.
+
+## 6.3.1 (7/24/2019)
+### Bug Fixes
+* Fix auto rolling bug introduced in 6.3.0, which causes segfault if log file creation fails.
+
+## 6.3.0 (6/18/2019)
+### Public API Change
+* Now DB::Close() will return Aborted() error when there is unreleased snapshot. Users can retry after all snapshots are released.
+* Index blocks are now handled similarly to data blocks with regards to the block cache: instead of storing objects in the cache, only the blocks themselves are cached. In addition, index blocks no longer get evicted from the cache when a table is closed, can now use the compressed block cache (if any), and can be shared among multiple table readers.
+* Partitions of partitioned indexes no longer affect the read amplification statistics.
+* Due to the above refactoring, block cache eviction statistics for indexes are temporarily broken. We plan to reintroduce them in a later phase.
+* options.keep_log_file_num will be enforced strictly all the time. File names of all log files will be tracked, which may take significantly amount of memory if options.keep_log_file_num is large and either of options.max_log_file_size or options.log_file_time_to_roll is set.
+* Add initial support for Get/Put with user timestamps. Users can specify timestamps via ReadOptions and WriteOptions when calling DB::Get and DB::Put.
+* Accessing a partition of a partitioned filter or index through a pinned reference is no longer considered a cache hit.
+* Add C bindings for secondary instance, i.e. DBImplSecondary.
+* Rate limited deletion of WALs is only enabled if DBOptions::wal_dir is not set, or explicitly set to db_name passed to DB::Open and DBOptions::db_paths is empty, or same as db_paths[0].path
+
+### New Features
+* Add an option `snap_refresh_nanos` (default to 0) to periodically refresh the snapshot list in compaction jobs. Assign to 0 to disable the feature.
+* Add an option `unordered_write` which trades snapshot guarantees with higher write throughput. When used with WRITE_PREPARED transactions with two_write_queues=true, it offers higher throughput with however no compromise on guarantees.
+* Allow DBImplSecondary to remove memtables with obsolete data after replaying MANIFEST and WAL.
+* Add an option `failed_move_fall_back_to_copy` (default is true) for external SST ingestion. When `move_files` is true and hard link fails, ingestion falls back to copy if `failed_move_fall_back_to_copy` is true. Otherwise, ingestion reports an error.
+* Add command `list_file_range_deletes` in ldb, which prints out tombstones in SST files.
+
+### Performance Improvements
+* Reduce binary search when iterator reseek into the same data block.
+* DBIter::Next() can skip user key checking if previous entry's seqnum is 0.
+* Merging iterator to avoid child iterator reseek for some cases
+* Log Writer will flush after finishing the whole record, rather than a fragment.
+* Lower MultiGet batching API latency by reading data blocks from disk in parallel
+
+### General Improvements
+* Added new status code kColumnFamilyDropped to distinguish between Column Family Dropped and DB Shutdown in progress.
+* Improve ColumnFamilyOptions validation when creating a new column family.
+
+### Bug Fixes
+* Fix a bug in WAL replay of secondary instance by skipping write batches with older sequence numbers than the current last sequence number.
+* Fix flush's/compaction's merge processing logic which allowed `Put`s covered by range tombstones to reappear. Note `Put`s may exist even if the user only ever called `Merge()` due to an internal conversion during compaction to the bottommost level.
+* Fix/improve memtable earliest sequence assignment and WAL replay so that WAL entries of unflushed column families will not be skipped after replaying the MANIFEST and increasing db sequence due to another flushed/compacted column family.
+* Fix a bug caused by secondary not skipping the beginning of new MANIFEST.
+* On DB open, delete WAL trash files left behind in wal_dir
+
+## 6.2.0 (4/30/2019)
 ### New Features
 * Add an option `strict_bytes_per_sync` that causes a file-writing thread to block rather than exceed the limit on bytes pending writeback specified by `bytes_per_sync` or `wal_bytes_per_sync`.
 * Improve range scan performance by avoiding per-key upper bound check in BlockBasedTableIterator.
 * Introduce Periodic Compaction for Level style compaction. Files are re-compacted periodically and put in the same level.
 * Block-based table index now contains exact highest key in the file, rather than an upper bound. This may improve Get() and iterator Seek() performance in some situations, especially when direct IO is enabled and block cache is disabled. A setting BlockBasedTableOptions::index_shortening is introduced to control this behavior. Set it to kShortenSeparatorsAndSuccessor to get the old behavior.
 * When reading from option file/string/map, customized envs can be filled according to object registry.
-* Add an option `snap_refresh_nanos` (default to 0.5s) to periodically refresh the snapshot list in compaction jobs. Assign to 0 to disable the feature.
 * Improve range scan performance when using explicit user readahead by not creating new table readers for every iterator.
+* Add index type BlockBasedTableOptions::IndexType::kBinarySearchWithFirstKey. It significantly reduces read amplification in some setups, especially for iterator seeks. It's not fully implemented yet: IO errors are not handled right.

 ### Public API Change
 * Change the behavior of OptimizeForPointLookup(): move away from hash-based block-based-table index, and use whole key memtable filtering.
@@ -19,6 +138,7 @@
 * Fix a race condition between WritePrepared::Get and ::Put with duplicate keys.
 * Fix crash when memtable prefix bloom is enabled and read/write a key out of domain of prefix extractor.
 * Close a WAL file before another thread deletes it.
+* Fix an assertion failure `IsFlushPending() == true` caused by one bg thread releasing the db mutex in ~ColumnFamilyData and another thread clearing `flush_requested_` flag.

 ## 6.1.1 (4/9/2019)
 ### New Features
@@ -82,17 +82,23 @@ ifeq ($(MAKECMDGOALS),rocksdbjavastatic)
 endif

 ifeq ($(MAKECMDGOALS),rocksdbjavastaticrelease)
-	DEBUG_LEVEL=0
+	ifneq ($(DEBUG_LEVEL),2)
+		DEBUG_LEVEL=0
+	endif
 endif

 ifeq ($(MAKECMDGOALS),rocksdbjavastaticreleasedocker)
-        DEBUG_LEVEL=0
+	ifneq ($(DEBUG_LEVEL),2)
+		DEBUG_LEVEL=0
+	endif
 endif

 ifeq ($(MAKECMDGOALS),rocksdbjavastaticpublish)
 	DEBUG_LEVEL=0
 endif

+$(info $$DEBUG_LEVEL is ${DEBUG_LEVEL})
+
 # Lite build flag.
 LITE ?= 0
 ifeq ($(LITE), 0)
@@ -137,9 +143,9 @@ CFLAGS +=  -DHAVE_POWER8
 HAVE_POWER8=1
 endif

-ifeq (,$(shell $(CXX) -fsyntax-only -march=armv8-a+crc -xc /dev/null 2>&1))
-CXXFLAGS += -march=armv8-a+crc
-CFLAGS += -march=armv8-a+crc
+ifeq (,$(shell $(CXX) -fsyntax-only -march=armv8-a+crc+crypto -xc /dev/null 2>&1))
+CXXFLAGS += -march=armv8-a+crc+crypto
+CFLAGS += -march=armv8-a+crc+crypto
 ARMCRC_SOURCE=1
 endif

@@ -250,8 +256,8 @@ endif
 ifdef COMPILE_WITH_TSAN
 	DISABLE_JEMALLOC=1
 	EXEC_LDFLAGS += -fsanitize=thread
-	PLATFORM_CCFLAGS += -fsanitize=thread -fPIC
-	PLATFORM_CXXFLAGS += -fsanitize=thread -fPIC
+	PLATFORM_CCFLAGS += -fsanitize=thread -fPIC -DFOLLY_SANITIZE_THREAD
+	PLATFORM_CXXFLAGS += -fsanitize=thread -fPIC -DFOLLY_SANITIZE_THREAD
        # Turn off -pg when enabling TSAN testing, because that induces
        # a link failure.  TODO: find the root cause
 	PROFILING_FLAGS =
@@ -298,9 +304,13 @@ ifndef DISABLE_JEMALLOC
 	PLATFORM_CCFLAGS += $(JEMALLOC_INCLUDE)
 endif

+ifndef USE_FOLLY_DISTRIBUTED_MUTEX
+	USE_FOLLY_DISTRIBUTED_MUTEX=0
+endif
+
 export GTEST_THROW_ON_FAILURE=1
 export GTEST_HAS_EXCEPTIONS=1
-GTEST_DIR = ./third-party/gtest-1.7.0/fused-src
+GTEST_DIR = ./third-party/gtest-1.8.1/fused-src
 # AIX: pre-defined system headers are surrounded by an extern "C" block
 ifeq ($(PLATFORM), OS_AIX)
 	PLATFORM_CCFLAGS += -I$(GTEST_DIR)
@@ -310,6 +320,23 @@ else
 	PLATFORM_CXXFLAGS += -isystem $(GTEST_DIR)
 endif

+ifeq ($(USE_FOLLY_DISTRIBUTED_MUTEX),1)
+	FOLLY_DIR = ./third-party/folly
+	# AIX: pre-defined system headers are surrounded by an extern "C" block
+	ifeq ($(PLATFORM), OS_AIX)
+		PLATFORM_CCFLAGS += -I$(FOLLY_DIR)
+		PLATFORM_CXXFLAGS += -I$(FOLLY_DIR)
+	else
+		PLATFORM_CCFLAGS += -isystem $(FOLLY_DIR)
+		PLATFORM_CXXFLAGS += -isystem $(FOLLY_DIR)
+	endif
+endif
+
+ifdef TEST_CACHE_LINE_SIZE
+  PLATFORM_CCFLAGS += -DTEST_CACHE_LINE_SIZE=$(TEST_CACHE_LINE_SIZE)
+  PLATFORM_CXXFLAGS += -DTEST_CACHE_LINE_SIZE=$(TEST_CACHE_LINE_SIZE)
+endif
+
 # This (the first rule) must depend on "all".
 default: all

@@ -396,10 +423,13 @@ endif

 LIBOBJECTS += $(TOOL_LIB_SOURCES:.cc=.o)
 MOCKOBJECTS = $(MOCK_LIB_SOURCES:.cc=.o)
+ifeq ($(USE_FOLLY_DISTRIBUTED_MUTEX),1)
+  FOLLYOBJECTS = $(FOLLY_SOURCES:.cpp=.o)
+endif

 GTEST = $(GTEST_DIR)/gtest/gtest-all.o
-TESTUTIL = ./util/testutil.o
-TESTHARNESS = ./util/testharness.o $(TESTUTIL) $(MOCKOBJECTS) $(GTEST)
+TESTUTIL = ./test_util/testutil.o
+TESTHARNESS = ./test_util/testharness.o $(TESTUTIL) $(MOCKOBJECTS) $(GTEST)
 VALGRIND_ERROR = 2
 VALGRIND_VER := $(join $(VALGRIND_VER),valgrind)

@@ -426,6 +456,7 @@ TESTS = \
 	inlineskiplist_test \
 	env_basic_test \
 	env_test \
+	env_logger_test \
 	hash_test \
 	thread_local_test \
 	rate_limiter_test \
@@ -435,10 +466,10 @@ TESTS = \
 	db_block_cache_test \
 	db_test \
 	db_blob_index_test \
-	db_bloom_filter_test \
 	db_iter_test \
 	db_iter_stress_test \
 	db_log_iter_test \
+	db_bloom_filter_test \
 	db_compaction_filter_test \
 	db_compaction_test \
 	db_dynamic_level_test \
@@ -447,6 +478,7 @@ TESTS = \
 	db_iterator_test \
 	db_memtable_test \
 	db_merge_operator_test \
+	db_merge_operand_test \
 	db_options_test \
 	db_range_del_test \
 	db_secondary_test \
@@ -493,6 +525,7 @@ TESTS = \
 	plain_table_db_test \
 	comparator_db_test \
 	external_sst_file_test \
+	import_column_family_test \
 	prefix_test \
 	skiplist_test \
 	write_buffer_manager_test \
@@ -503,6 +536,7 @@ TESTS = \
 	cassandra_serialize_test \
 	ttl_test \
 	backupable_db_test \
+	cache_simulator_test \
 	sim_cache_test \
 	version_edit_test \
 	version_set_test \
@@ -542,6 +576,7 @@ TESTS = \
 	ldb_cmd_test \
 	persistent_cache_test \
 	statistics_test \
+	stats_history_test \
 	lru_cache_test \
 	object_registry_test \
 	repair_test \
@@ -555,9 +590,16 @@ TESTS = \
 	range_del_aggregator_test \
 	sst_file_reader_test \
 	db_secondary_test \
+	block_cache_tracer_test \
+	block_cache_trace_analyzer_test \
+
+ifeq ($(USE_FOLLY_DISTRIBUTED_MUTEX),1)
+	TESTS += folly_synchronization_distributed_mutex_test
+endif

 PARALLEL_TEST = \
 	backupable_db_test \
+	db_bloom_filter_test \
 	db_compaction_filter_test \
 	db_compaction_test \
 	db_merge_operator_test \
@@ -566,7 +608,9 @@ PARALLEL_TEST = \
 	db_universal_compaction_test \
 	db_wal_test \
 	external_sst_file_test \
+	import_column_family_test \
 	fault_injection_test \
+	file_reader_writer_test \
 	inlineskiplist_test \
 	manual_compaction_test \
 	persistent_cache_test \
@@ -599,6 +643,7 @@ TOOLS = \
 	rocksdb_undump \
 	blob_dump \
 	trace_analyzer \
+	block_cache_trace_analyzer \

 TEST_LIBS = \
 	librocksdb_env_basic_test.a
@@ -910,7 +955,7 @@ blackbox_crash_test: db_stress
 	python -u tools/db_crashtest.py blackbox $(CRASH_TEST_EXT_ARGS)

 blackbox_crash_test_with_atomic_flush: db_stress
-	python -u tools/db_crashtest.py --enable_atomic_flush blackbox $(CRASH_TEST_EXT_ARGS)
+	python -u tools/db_crashtest.py --cf_consistency blackbox $(CRASH_TEST_EXT_ARGS)

 ifeq ($(CRASH_TEST_KILL_ODD),)
  CRASH_TEST_KILL_ODD=888887
@@ -923,7 +968,7 @@ whitebox_crash_test: db_stress
      $(CRASH_TEST_KILL_ODD) $(CRASH_TEST_EXT_ARGS)

 whitebox_crash_test_with_atomic_flush: db_stress
-	python -u tools/db_crashtest.py --enable_atomic_flush whitebox  --random_kill_odd \
+	python -u tools/db_crashtest.py --cf_consistency whitebox  --random_kill_odd \
      $(CRASH_TEST_KILL_ODD) $(CRASH_TEST_EXT_ARGS)

 asan_check:
@@ -1054,7 +1099,7 @@ 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:
-	rm -f $(BENCHMARKS) $(TOOLS) $(TESTS) $(LIBRARY) $(SHARED)
+	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 {} \;
@@ -1100,6 +1145,14 @@ db_bench: tools/db_bench.o $(BENCHTOOLOBJECTS)
 trace_analyzer: tools/trace_analyzer.o $(ANALYZETOOLOBJECTS) $(LIBOBJECTS)
 	$(AM_LINK)

+block_cache_trace_analyzer: tools/block_cache_analyzer/block_cache_trace_analyzer_tool.o $(ANALYZETOOLOBJECTS) $(LIBOBJECTS)
+	$(AM_LINK)
+
+ifeq ($(USE_FOLLY_DISTRIBUTED_MUTEX),1)
+folly_synchronization_distributed_mutex_test: $(LIBOBJECTS) $(TESTHARNESS) $(FOLLYOBJECTS) third-party/folly/folly/synchronization/test/DistributedMutexTest.o
+	$(AM_LINK)
+endif
+
 cache_bench: cache/cache_bench.o $(LIBOBJECTS) $(TESTUTIL)
 	$(AM_LINK)

@@ -1121,7 +1174,7 @@ db_sanity_test: tools/db_sanity_test.o $(LIBOBJECTS) $(TESTUTIL)
 db_repl_stress: tools/db_repl_stress.o $(LIBOBJECTS) $(TESTUTIL)
 	$(AM_LINK)

-arena_test: util/arena_test.o $(LIBOBJECTS) $(TESTHARNESS)
+arena_test: memory/arena_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 autovector_test: util/autovector_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -1178,7 +1231,7 @@ histogram_test: monitoring/histogram_test.o $(LIBOBJECTS) $(TESTHARNESS)
 thread_local_test: util/thread_local_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-corruption_test: db/corruption_test.o $(LIBOBJECTS) $(TESTHARNESS)
+corruption_test: db/corruption_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 crc32c_test: util/crc32c_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -1235,6 +1288,9 @@ db_memtable_test: db/db_memtable_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHA
 db_merge_operator_test: db/db_merge_operator_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+db_merge_operand_test: db/db_merge_operand_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 db_options_test: db/db_options_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

@@ -1259,6 +1315,9 @@ external_sst_file_basic_test: db/external_sst_file_basic_test.o db/db_test_util.
 external_sst_file_test: db/external_sst_file_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+import_column_family_test: db/import_column_family_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 db_tailing_iter_test: db/db_tailing_iter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

@@ -1307,6 +1366,9 @@ backupable_db_test: utilities/backupable/backupable_db_test.o $(LIBOBJECTS) $(TE
 checkpoint_test: utilities/checkpoint/checkpoint_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+cache_simulator_test: utilities/simulator_cache/cache_simulator_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 sim_cache_test: utilities/simulator_cache/sim_cache_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

@@ -1333,13 +1395,13 @@ write_batch_with_index_test: utilities/write_batch_with_index/write_batch_with_i
 flush_job_test: db/flush_job_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-compaction_iterator_test: db/compaction_iterator_test.o $(LIBOBJECTS) $(TESTHARNESS)
+compaction_iterator_test: db/compaction/compaction_iterator_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-compaction_job_test: db/compaction_job_test.o $(LIBOBJECTS) $(TESTHARNESS)
+compaction_job_test: db/compaction/compaction_job_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-compaction_job_stats_test: db/compaction_job_stats_test.o $(LIBOBJECTS) $(TESTHARNESS)
+compaction_job_stats_test: db/compaction/compaction_job_stats_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 compact_on_deletion_collector_test: utilities/table_properties_collectors/compact_on_deletion_collector_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -1363,7 +1425,7 @@ fault_injection_test: db/fault_injection_test.o $(LIBOBJECTS) $(TESTHARNESS)
 rate_limiter_test: util/rate_limiter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-delete_scheduler_test: util/delete_scheduler_test.o $(LIBOBJECTS) $(TESTHARNESS)
+delete_scheduler_test: file/delete_scheduler_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 filename_test: db/filename_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -1372,13 +1434,13 @@ filename_test: db/filename_test.o $(LIBOBJECTS) $(TESTHARNESS)
 file_reader_writer_test: util/file_reader_writer_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-block_based_filter_block_test: table/block_based_filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS)
+block_based_filter_block_test: table/block_based/block_based_filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-full_filter_block_test: table/full_filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS)
+full_filter_block_test: table/block_based/full_filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-partitioned_filter_block_test: table/partitioned_filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS)
+partitioned_filter_block_test: table/block_based/partitioned_filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 log_test: db/log_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -1390,10 +1452,10 @@ cleanable_test: table/cleanable_test.o $(LIBOBJECTS) $(TESTHARNESS)
 table_test: table/table_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-block_test: table/block_test.o $(LIBOBJECTS) $(TESTHARNESS)
+block_test: table/block_based/block_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-data_block_hash_index_test: table/data_block_hash_index_test.o $(LIBOBJECTS) $(TESTHARNESS)
+data_block_hash_index_test: table/block_based/data_block_hash_index_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 inlineskiplist_test: memtable/inlineskiplist_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -1411,7 +1473,7 @@ version_edit_test: db/version_edit_test.o $(LIBOBJECTS) $(TESTHARNESS)
 version_set_test: db/version_set_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-compaction_picker_test: db/compaction_picker_test.o $(LIBOBJECTS) $(TESTHARNESS)
+compaction_picker_test: db/compaction/compaction_picker_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 version_builder_test: db/version_builder_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -1459,10 +1521,10 @@ rocksdb_dump: tools/dump/rocksdb_dump.o $(LIBOBJECTS)
 rocksdb_undump: tools/dump/rocksdb_undump.o $(LIBOBJECTS)
 	$(AM_LINK)

-cuckoo_table_builder_test: table/cuckoo_table_builder_test.o $(LIBOBJECTS) $(TESTHARNESS)
+cuckoo_table_builder_test: table/cuckoo/cuckoo_table_builder_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-cuckoo_table_reader_test: table/cuckoo_table_reader_test.o $(LIBOBJECTS) $(TESTHARNESS)
+cuckoo_table_reader_test: table/cuckoo/cuckoo_table_reader_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 cuckoo_table_db_test: db/cuckoo_table_db_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -1492,7 +1554,7 @@ db_bench_tool_test: tools/db_bench_tool_test.o $(BENCHTOOLOBJECTS) $(TESTHARNESS
 trace_analyzer_test: tools/trace_analyzer_test.o $(LIBOBJECTS) $(ANALYZETOOLOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-event_logger_test: util/event_logger_test.o $(LIBOBJECTS) $(TESTHARNESS)
+event_logger_test: logging/event_logger_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 timer_queue_test: util/timer_queue_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -1513,7 +1575,10 @@ manual_compaction_test: db/manual_compaction_test.o $(LIBOBJECTS) $(TESTHARNESS)
 filelock_test: util/filelock_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-auto_roll_logger_test: util/auto_roll_logger_test.o $(LIBOBJECTS) $(TESTHARNESS)
+auto_roll_logger_test: logging/auto_roll_logger_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
+env_logger_test: logging/env_logger_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 memtable_list_test: db/memtable_list_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -1558,6 +1623,9 @@ persistent_cache_test: utilities/persistent_cache/persistent_cache_test.o  db/db
 statistics_test: monitoring/statistics_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+stats_history_test: monitoring/stats_history_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 lru_cache_test: cache/lru_cache_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

@@ -1579,7 +1647,13 @@ range_tombstone_fragmenter_test: db/range_tombstone_fragmenter_test.o db/db_test
 sst_file_reader_test: table/sst_file_reader_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-db_secondary_test: db/db_secondary_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+db_secondary_test: db/db_impl/db_secondary_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
+block_cache_tracer_test: trace_replay/block_cache_tracer_test.o trace_replay/block_cache_tracer.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
+block_cache_trace_analyzer_test: tools/block_cache_analyzer/block_cache_trace_analyzer_test.o tools/block_cache_analyzer/block_cache_trace_analyzer.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 #-------------------------------------------------
@@ -1627,7 +1701,7 @@ 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 $(MACHINE), amd64 arm64 sparc64))
+	ifneq (,$(filter $(MACHINE), amd64 arm64 aarch64 sparc64))
 		ARCH := 64
 	else
 		ARCH := 32
@@ -1636,7 +1710,7 @@ else
 	ARCH := $(shell getconf LONG_BIT)
 endif

-ifeq (,$(findstring ppc,$(MACHINE)))
+ifeq (,$(filter $(MACHINE), ppc arm64 aarch64 sparc64))
        ROCKSDBJNILIB = librocksdbjni-linux$(ARCH).so
 else
        ROCKSDBJNILIB = librocksdbjni-linux-$(MACHINE).so
@@ -1659,8 +1733,8 @@ SNAPPY_DOWNLOAD_BASE ?= https://github.com/google/snappy/archive
 LZ4_VER ?= 1.8.3
 LZ4_SHA256 ?= 33af5936ac06536805f9745e0b6d61da606a1f8b4cc5c04dd3cbaca3b9b4fc43
 LZ4_DOWNLOAD_BASE ?= https://github.com/lz4/lz4/archive
-ZSTD_VER ?= 1.3.7
-ZSTD_SHA256 ?= 5dd1e90eb16c25425880c8a91327f63de22891ffed082fcc17e5ae84fce0d5fb
+ZSTD_VER ?= 1.4.0
+ZSTD_SHA256 ?= 63be339137d2b683c6d19a9e34f4fb684790e864fee13c7dd40e197a64c705c1
 ZSTD_DOWNLOAD_BASE ?= https://github.com/facebook/zstd/archive
 CURL_SSL_OPTS ?= --tlsv1

@@ -1827,27 +1901,23 @@ rocksdbjavastaticreleasedocker: rocksdbjavastatic rocksdbjavastaticdockerx86 roc

 rocksdbjavastaticdockerx86:
 	mkdir -p java/target
-	DOCKER_LINUX_X86_CONTAINER=`docker ps -aqf name=rocksdb_linux_x86-be`; \
-	if [ -z "$$DOCKER_LINUX_X86_CONTAINER" ]; then \
-		docker container create --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host --name rocksdb_linux_x86-be evolvedbinary/rocksjava:centos6_x86-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh; \
-	fi
-	docker start -a rocksdb_linux_x86-be
+	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

 rocksdbjavastaticdockerx86_64:
 	mkdir -p java/target
-	DOCKER_LINUX_X64_CONTAINER=`docker ps -aqf name=rocksdb_linux_x64-be`; \
-	if [ -z "$$DOCKER_LINUX_X64_CONTAINER" ]; then \
-		docker container create --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host --name rocksdb_linux_x64-be evolvedbinary/rocksjava:centos6_x64-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh; \
-	fi
-	docker start -a rocksdb_linux_x64-be
+	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

 rocksdbjavastaticdockerppc64le:
 	mkdir -p java/target
-	DOCKER_LINUX_PPC64LE_CONTAINER=`docker ps -aqf name=rocksdb_linux_ppc64le-be`; \
-	if [ -z "$$DOCKER_LINUX_PPC64LE_CONTAINER" ]; then \
-		docker container create --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host --name rocksdb_linux_ppc64le-be evolvedbinary/rocksjava:centos7_ppc64le-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh; \
+	docker run --rm --name rocksdb_linux_ppc64le-be --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host --env DEBUG_LEVEL=$(DEBUG_LEVEL) evolvedbinary/rocksjava:centos7_ppc64le-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh
+
+rocksdbjavastaticdockerarm64v8:
+	mkdir -p java/target
+	DOCKER_LINUX_ARM64V8_CONTAINER=`docker ps -aqf name=rocksdb_linux_arm64v8-be`; \
+	if [ -z "$$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_ppc64le-be
+	docker start -a rocksdb_linux_arm64v8-be

 rocksdbjavastaticpublish: rocksdbjavastaticrelease rocksdbjavastaticpublishcentral

@@ -50,6 +50,10 @@ Check out their blog post: http://blog.cloudera.com/blog/2015/08/inside-santande
 ## Airbnb
 Airbnb is using RocksDB as a storage engine for their personalized search service. You can learn more about it here: https://www.youtube.com/watch?v=ASQ6XMtogMs

+## Alluxio
+[Alluxio](https://www.alluxio.io) uses RocksDB to serve and scale file system metadata to beyond 1 Billion files. The detailed design and implementation is described in this engineering blog:
+https://www.alluxio.io/blog/scalable-metadata-service-in-alluxio-storing-billions-of-files/
+
 ## Pinterest
 Pinterest's Object Retrieval System uses RocksDB for storage: https://www.youtube.com/watch?v=MtFEVEs_2Vo

@@ -91,4 +95,7 @@ LzLabs is using RocksDB as a storage engine in their multi-database distributed
 [ProfaneDB](https://profanedb.gitlab.io/) is a database for Protocol Buffers, and uses RocksDB for storage. It is accessible via gRPC, and the schema is defined using directly `.proto` files.

 ## IOTA Foundation
- [IOTA Foundation](https://www.iota.org/) is using RocksDB in the [IOTA Reference Implementation (IRI)](https://github.com/iotaledger/iri) to store the local state of the Tangle. The Tangle is the first open-source distributed ledger powering the future of the Internet of Things.
+ [IOTA Foundation](https://www.iota.org/) is using RocksDB in the [IOTA Reference Implementation (IRI)](https://github.com/iotaledger/iri) to store the local state of the Tangle. The Tangle is the first open-source distributed ledger powering the future of the Internet of Things.
+ 
+## Crux
+[Crux](https://github.com/juxt/crux) is a document database that uses RocksDB for local [EAV](https://en.wikipedia.org/wiki/Entity%E2%80%93attribute%E2%80%93value_model) index storage to enable point-in-time bitemporal Datalog queries. The "unbundled" architecture uses Kafka to provide horizontal scalability.
@@ -1,15 +1,67 @@
 version: 1.0.{build}
+
 image: Visual Studio 2017
+
+environment:
+  JAVA_HOME: C:\Program Files\Java\jdk1.8.0
+  THIRDPARTY_HOME: $(APPVEYOR_BUILD_FOLDER)\thirdparty
+  SNAPPY_HOME: $(THIRDPARTY_HOME)\snappy-1.1.7
+  SNAPPY_INCLUDE: $(SNAPPY_HOME);$(SNAPPY_HOME)\build
+  SNAPPY_LIB_DEBUG: $(SNAPPY_HOME)\build\Debug\snappy.lib
+  SNAPPY_LIB_RELEASE: $(SNAPPY_HOME)\build\Release\snappy.lib
+  LZ4_HOME: $(THIRDPARTY_HOME)\lz4-1.8.3
+  LZ4_INCLUDE: $(LZ4_HOME)\lib
+  LZ4_LIB_DEBUG: $(LZ4_HOME)\visual\VS2010\bin\x64_Debug\liblz4_static.lib
+  LZ4_LIB_RELEASE: $(LZ4_HOME)\visual\VS2010\bin\x64_Release\liblz4_static.lib
+  ZSTD_HOME: $(THIRDPARTY_HOME)\zstd-1.4.0
+  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
+
+install:
+  - md %THIRDPARTY_HOME%
+  - echo "Building Snappy dependency..."
+  - cd %THIRDPARTY_HOME%
+  - curl -fsSL -o snappy-1.1.7.zip https://github.com/google/snappy/archive/1.1.7.zip
+  - unzip snappy-1.1.7.zip
+  - cd snappy-1.1.7
+  - mkdir build
+  - cd build
+  - cmake -DCMAKE_GENERATOR_PLATFORM=x64 ..
+  - msbuild Snappy.sln /p:Configuration=Debug /p:Platform=x64
+  - msbuild Snappy.sln /p:Configuration=Release /p:Platform=x64
+  - echo "Building LZ4 dependency..."
+  - cd %THIRDPARTY_HOME%
+  - 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
+  - msbuild lz4.sln /p:Configuration=Debug /p:Platform=x64
+  - msbuild lz4.sln /p:Configuration=Release /p:Platform=x64
+  - echo "Building ZStd dependency..."
+  - cd %THIRDPARTY_HOME%
+  - 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
+  - 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 -DWITH_XPRESS=1 -DPORTABLE=1 -DJNI=1 ..
- cd ..
+  - 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 ..
+  - cd ..
 build:
  project: build\rocksdb.sln
  parallel: true
  verbosity: normal
-test:
-test_script:
- ps: build_tools\run_ci_db_test.ps1 -SuiteRun db_basic_test,db_test2,db_test,env_basic_test,env_test -Concurrency 8
+
+test:
+
+test_script:
+  - ps: build_tools\run_ci_db_test.ps1 -SuiteRun db_basic_test,db_test2,db_test,env_basic_test,env_test,db_merge_operand_test -Concurrency 8
+
+on_failure:
+  - cmd: 7z a build-failed.zip %APPVEYOR_BUILD_FOLDER%\build\ && appveyor PushArtifact build-failed.zip

@@ -4,12 +4,31 @@ from __future__ import division
 from __future__ import print_function
 from __future__ import unicode_literals
 from targets_builder import TARGETSBuilder
+import json
 import os
 import fnmatch
 import sys

 from util import ColorString

+# This script generates TARGETS file for Buck.
+# Buck is a build tool specifying dependencies among different build targets.
+# User can pass extra dependencies as a JSON object via command line, and this
+# script can include these dependencies in the generate TARGETS file.
+# Usage:
+# $python buckifier/buckify_rocksdb.py
+# (This generates a TARGET file without user-specified dependency for unit
+# tests.)
+# $python buckifier/buckify_rocksdb.py \
+#        '{"fake": { \
+#                      "extra_deps": [":test_dep", "//fakes/module:mock1"],  \
+#                      "extra_compiler_flags": ["-DROCKSDB_LITE", "-Os"], \
+#                  } \
+#         }'
+# (Generated TARGETS file has test_dep and mock1 as dependencies for RocksDB
+# unit tests, and will use the extra_compiler_flags to compile the unit test
+# source.)
+
 # tests to export as libraries for inclusion in other projects
 _EXPORTED_TEST_LIBS = ["env_basic_test"]

@@ -86,8 +105,38 @@ def get_tests(repo_path):
    return tests


+# 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'): []
+        }
+    }
+    if len(sys.argv) < 2:
+        return deps_map
+
+    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):
+                v = encode_dict(v)
+            rv[k] = v
+        return rv
+    extra_deps = json.loads(sys.argv[1], object_hook=encode_dict)
+    for target_alias, deps in extra_deps.items():
+        deps_map[target_alias] = deps
+    return deps_map
+
+
 # Prepare TARGETS file for buck
-def generate_targets(repo_path):
+def generate_targets(repo_path, deps_map):
    print(ColorString.info("Generating TARGETS"))
    # parsed src.mk file
    src_mk = parse_src_mk(repo_path)
@@ -118,27 +167,36 @@ def generate_targets(repo_path):
        "rocksdb_tools_lib",
        src_mk.get("BENCH_LIB_SOURCES", []) +
        src_mk.get("ANALYZER_LIB_SOURCES", []) +
-        ["util/testutil.cc"],
+        ["test_util/testutil.cc"],
        [":rocksdb_lib"])

+    print("Extra dependencies:\n{0}".format(str(deps_map)))
    # test for every test we found in the Makefile
-    for test in sorted(tests):
-        match_src = [src for src in cc_files if ("/%s.c" % test) in src]
-        if len(match_src) == 0:
-            print(ColorString.warning("Cannot find .cc file for %s" % test))
-            continue
-        elif len(match_src) > 1:
-            print(ColorString.warning("Found more than one .cc for %s" % test))
-            print(match_src)
-            continue
+    for target_alias, deps in deps_map.items():
+        for test in sorted(tests):
+            match_src = [src for src in cc_files if ("/%s.c" % test) in src]
+            if len(match_src) == 0:
+                print(ColorString.warning("Cannot find .cc file for %s" % test))
+                continue
+            elif len(match_src) > 1:
+                print(ColorString.warning("Found more than one .cc for %s" % test))
+                print(match_src)
+                continue

-        assert(len(match_src) == 1)
-        is_parallel = tests[test]
-        TARGETS.register_test(test, match_src[0], is_parallel)
+            assert(len(match_src) == 1)
+            is_parallel = tests[test]
+            test_target_name = \
+                test if not target_alias else test + "_" + target_alias
+            TARGETS.register_test(
+                test_target_name,
+                match_src[0],
+                is_parallel,
+                deps['extra_deps'],
+                deps['extra_compiler_flags'])

-        if test in _EXPORTED_TEST_LIBS:
-            test_library = "%s_lib" % test
-            TARGETS.add_library(test_library, match_src, [":rocksdb_test_lib"])
+            if test in _EXPORTED_TEST_LIBS:
+                test_library = "%s_lib" % test_target_name
+                TARGETS.add_library(test_library, match_src, [":rocksdb_test_lib"])
    TARGETS.flush_tests()

    print(ColorString.info("Generated TARGETS Summary:"))
@@ -163,8 +221,9 @@ def exit_with_error(msg):


 def main():
+    deps_map = get_dependencies()
    # Generate TARGETS file for buck
-    ok = generate_targets(get_rocksdb_path())
+    ok = generate_targets(get_rocksdb_path(), deps_map)
    if not ok:
        exit_with_error("Failed to generate TARGETS files")

@@ -51,14 +51,21 @@ class TARGETSBuilder:
            pretty_list(deps)))
        self.total_bin = self.total_bin + 1

-    def register_test(self, test_name, src, is_parallel):
+    def register_test(self,
+                      test_name,
+                      src,
+                      is_parallel,
+                      extra_deps,
+                      extra_compiler_flags):
        exec_mode = "serial"
        if is_parallel:
            exec_mode = "parallel"
        self.tests_cfg += targets_cfg.test_cfg_template % (
            test_name,
            str(src),
-            str(exec_mode))
+            str(exec_mode),
+            extra_deps,
+            extra_compiler_flags)

        self.total_test = self.total_test + 1

@@ -11,30 +11,11 @@ REPO_PATH = package_name() + "/"

 ROCKSDB_COMPILER_FLAGS = [
    "-fno-builtin-memcmp",
-    "-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",
-    "-DOS_LINUX",
-    # Flags to enable libs we include
-    "-DSNAPPY",
-    "-DZLIB",
-    "-DBZIP2",
-    "-DLZ4",
-    "-DZSTD",
-    "-DZSTD_STATIC_LINKING_ONLY",
-    "-DGFLAGS=gflags",
-    "-DNUMA",
-    "-DTBB",
    # Needed to compile in fbcode
    "-Wno-expansion-to-defined",
    # Added missing flags from output of build_detect_platform
-    "-DROCKSDB_PTHREAD_ADAPTIVE_MUTEX",
-    "-DROCKSDB_BACKTRACE",
    "-Wnarrowing",
+    "-DROCKSDB_NO_DYNAMIC_EXTENSION",
 ]

 ROCKSDB_EXTERNAL_DEPS = [
@@ -45,11 +26,54 @@ ROCKSDB_EXTERNAL_DEPS = [
    ("lz4", None, "lz4"),
    ("zstd", None),
    ("tbb", None),
-    ("numa", None, "numa"),
    ("googletest", None, "gtest"),
 ]

+ROCKSDB_OS_DEPS = [
+    (
+        "linux",
+        ["third-party//numa:numa"],
+    ),
+]
+
+ROCKSDB_OS_PREPROCESSOR_FLAGS = [
+    (
+        "linux",
+        [
+            "-DOS_LINUX",
+            "-DROCKSDB_FALLOCATE_PRESENT",
+            "-DROCKSDB_MALLOC_USABLE_SIZE",
+            "-DROCKSDB_PTHREAD_ADAPTIVE_MUTEX",
+            "-DROCKSDB_RANGESYNC_PRESENT",
+            "-DROCKSDB_SCHED_GETCPU_PRESENT",
+            "-DHAVE_SSE42",
+            "-DNUMA",
+        ],
+    ),
+    (
+        "macos",
+        ["-DOS_MACOSX"],
+    ),
+]
+
 ROCKSDB_PREPROCESSOR_FLAGS = [
+    "-DROCKSDB_PLATFORM_POSIX",
+    "-DROCKSDB_LIB_IO_POSIX",
+    "-DROCKSDB_SUPPORT_THREAD_LOCAL",
+
+    # Flags to enable libs we include
+    "-DSNAPPY",
+    "-DZLIB",
+    "-DBZIP2",
+    "-DLZ4",
+    "-DZSTD",
+    "-DZSTD_STATIC_LINKING_ONLY",
+    "-DGFLAGS=gflags",
+    "-DTBB",
+
+    # Added missing flags from output of build_detect_platform
+    "-DROCKSDB_BACKTRACE",
+
    # Directories with files for #include
    "-I" + REPO_PATH + "include/",
    "-I" + REPO_PATH,
@@ -57,7 +81,6 @@ ROCKSDB_PREPROCESSOR_FLAGS = [

 ROCKSDB_ARCH_PREPROCESSOR_FLAGS = {
    "x86_64": [
-        "-DHAVE_SSE42",
        "-DHAVE_PCLMUL",
    ],
 }
@@ -74,9 +97,15 @@ sanitizer = read_config("fbcode", "sanitizer")

 # Do not enable jemalloc if sanitizer presents. RocksDB will further detect
 # whether the binary is linked with jemalloc at runtime.
-ROCKSDB_COMPILER_FLAGS += (["-DROCKSDB_JEMALLOC"] if sanitizer == "" else [])
+ROCKSDB_OS_PREPROCESSOR_FLAGS += ([(
+    "linux",
+    ["-DROCKSDB_JEMALLOC"],
+)] if sanitizer == "" else [])

-ROCKSDB_EXTERNAL_DEPS += ([("jemalloc", None, "headers")] if sanitizer == "" else [])
+ROCKSDB_OS_DEPS += ([(
+    "linux",
+    ["third-party//jemalloc:headers"],
+)] if sanitizer == "" else [])
 """


@@ -87,6 +116,8 @@ cpp_library(
    {headers_attr_prefix}headers = {headers},
    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 = [{deps}],
    external_deps = ROCKSDB_EXTERNAL_DEPS,
@@ -109,11 +140,13 @@ test_cfg_template = """    [
        "%s",
        "%s",
        "%s",
+        %s,
+        %s,
    ],
 """

 unittests_template = """
-# [test_name, test_src, test_type]
+# [test_name, test_src, test_type, extra_deps, extra_compiler_flags]
 ROCKS_TESTS = [
 %s]

@@ -122,15 +155,19 @@ ROCKS_TESTS = [
 # will not be included.
 [
    test_binary(
+        extra_compiler_flags = extra_compiler_flags,
+        extra_deps = extra_deps,
        parallelism = parallelism,
        rocksdb_arch_preprocessor_flags = ROCKSDB_ARCH_PREPROCESSOR_FLAGS,
        rocksdb_compiler_flags = ROCKSDB_COMPILER_FLAGS,
        rocksdb_external_deps = ROCKSDB_EXTERNAL_DEPS,
+        rocksdb_os_deps = ROCKSDB_OS_DEPS,
+        rocksdb_os_preprocessor_flags = ROCKSDB_OS_PREPROCESSOR_FLAGS,
        rocksdb_preprocessor_flags = ROCKSDB_PREPROCESSOR_FLAGS,
        test_cc = test_cc,
        test_name = test_name,
    )
-    for test_name, test_cc, parallelism in ROCKS_TESTS
+    for test_name, test_cc, parallelism, extra_deps, extra_compiler_flags in ROCKS_TESTS
    if not is_opt_mode
 ]
 """
@@ -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 +150,9 @@ case "$TARGET_OS" in
            PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -latomic"
        fi
        PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -lpthread -lrt"
+        if test -z "$USE_FOLLY_DISTRIBUTED_MUTEX"; then
+          USE_FOLLY_DISTRIBUTED_MUTEX=1
+        fi
        # PORT_FILES=port/linux/linux_specific.cc
        ;;
    SunOS)
@@ -397,6 +400,7 @@ EOF
          #include <malloc.h>
          int main() {
            size_t res = malloc_usable_size(0);
+            (void)res;
            return 0;
          }
 EOF
@@ -411,6 +415,7 @@ EOF
          #include <pthread.h>
          int main() {
            int x = PTHREAD_MUTEX_ADAPTIVE_NP;
+            (void)x;
            return 0;
          }
 EOF
@@ -422,7 +427,7 @@ EOF
    if ! test $ROCKSDB_DISABLE_BACKTRACE; then
        # Test whether backtrace is available
        $CXX $CFLAGS -x c++ - -o /dev/null 2>/dev/null  <<EOF
-          #include <execinfo.h>>
+          #include <execinfo.h>
          int main() {
            void* frames[1];
            backtrace_symbols(frames, backtrace(frames, 1));
@@ -480,6 +485,7 @@ EOF
          #include <sched.h>
          int main() {
            int cpuid = sched_getcpu();
+            (void)cpuid;
          }
 EOF
        if [ "$?" = 0 ]; then
@@ -515,7 +521,7 @@ fi

 if test "$USE_HDFS"; then
  if test -z "$JAVA_HOME"; then
-    echo "JAVA_HOME has to be set for HDFS usage."
+    echo "JAVA_HOME has to be set for HDFS usage." >&2
    exit 1
  fi
  HDFS_CCFLAGS="$HDFS_CCFLAGS -I$JAVA_HOME/include -I$JAVA_HOME/include/linux -DUSE_HDFS -I$HADOOP_HOME/include"
@@ -527,42 +533,62 @@ if test "$USE_HDFS"; then
  JAVA_LDFLAGS="$JAVA_LDFLAGS $HDFS_LDFLAGS"
 fi

-if test -z "$PORTABLE"; then
+if test "0$PORTABLE" -eq 0; then
  if test -n "`echo $TARGET_ARCHITECTURE | grep ^ppc64`"; then
    # Tune for this POWER processor, treating '+' models as base models
    POWER=`LD_SHOW_AUXV=1 /bin/true | grep AT_PLATFORM | grep -E -o power[0-9]+`
    COMMON_FLAGS="$COMMON_FLAGS -mcpu=$POWER -mtune=$POWER "
  elif test -n "`echo $TARGET_ARCHITECTURE | grep ^s390x`"; then
    COMMON_FLAGS="$COMMON_FLAGS -march=z10 "
-  elif test -n "`echo $TARGET_ARCHITECTURE | grep ^arm`"; then
+  elif test -n "`echo $TARGET_ARCHITECTURE | grep -e^arm -e^aarch64`"; then
    # TODO: Handle this with approprite options.
    COMMON_FLAGS="$COMMON_FLAGS"
+  elif test -n "`echo $TARGET_ARCHITECTURE | grep ^aarch64`"; then
+    COMMON_FLAGS="$COMMON_FLAGS"
  elif [ "$TARGET_OS" == "IOS" ]; then
    COMMON_FLAGS="$COMMON_FLAGS"
-  elif [ "$TARGET_OS" != "AIX" ] && [ "$TARGET_OS" != "SunOS" ]; then
+  elif [ "$TARGET_OS" == "AIX" ] || [ "$TARGET_OS" == "SunOS" ]; then
+    # TODO: Not sure why we don't use -march=native on these OSes
+    if test "$USE_SSE"; then
+      TRY_SSE_ETC="1"
+    fi
+  else
    COMMON_FLAGS="$COMMON_FLAGS -march=native "
-  elif test "$USE_SSE"; then
-    COMMON_FLAGS="$COMMON_FLAGS -msse4.2 -mpclmul"
  fi
-elif test "$USE_SSE"; then
-  COMMON_FLAGS="$COMMON_FLAGS -msse4.2 -mpclmul"
+else
+  # PORTABLE=1
+  if test "$USE_SSE"; then
+    TRY_SSE_ETC="1"
+  fi
 fi

-$CXX $PLATFORM_CXXFLAGS $COMMON_FLAGS -x c++ - -o /dev/null 2>/dev/null <<EOF
+if test "$TRY_SSE_ETC"; then
+  # The USE_SSE flag now means "attempt to compile with widely-available
+  # Intel architecture extensions utilized by specific optimizations in the
+  # source code." It's a qualifier on PORTABLE=1 that means "mostly portable."
+  # It doesn't even really check that your current CPU is compatible.
+  #
+  # SSE4.2 available since nehalem, ca. 2008-2010
+  TRY_SSE42="-msse4.2"
+  # PCLMUL available since westmere, ca. 2010-2011
+  TRY_PCLMUL="-mpclmul"
+fi
+
+$CXX $PLATFORM_CXXFLAGS $COMMON_FLAGS $TRY_SSE42 -x c++ - -o /dev/null 2>/dev/null <<EOF
  #include <cstdint>
  #include <nmmintrin.h>
  int main() {
    volatile uint32_t x = _mm_crc32_u32(0, 0);
+    (void)x;
  }
 EOF
 if [ "$?" = 0 ]; then
-  COMMON_FLAGS="$COMMON_FLAGS -DHAVE_SSE42"
+  COMMON_FLAGS="$COMMON_FLAGS $TRY_SSE42 -DHAVE_SSE42"
 elif test "$USE_SSE"; then
-  echo "warning: USE_SSE specified but compiler could not use SSE intrinsics, disabling"
-  exit 1
+  echo "warning: USE_SSE specified but compiler could not use SSE intrinsics, disabling" >&2
 fi

-$CXX $PLATFORM_CXXFLAGS $COMMON_FLAGS -x c++ - -o /dev/null 2>/dev/null <<EOF
+$CXX $PLATFORM_CXXFLAGS $COMMON_FLAGS $TRY_PCLMUL -x c++ - -o /dev/null 2>/dev/null <<EOF
  #include <cstdint>
  #include <wmmintrin.h>
  int main() {
@@ -570,13 +596,13 @@ $CXX $PLATFORM_CXXFLAGS $COMMON_FLAGS -x c++ - -o /dev/null 2>/dev/null <<EOF
    const auto b = _mm_set_epi64x(0, 0);
    const auto c = _mm_clmulepi64_si128(a, b, 0x00);
    auto d = _mm_cvtsi128_si64(c);
+    (void)d;
  }
 EOF
 if [ "$?" = 0 ]; then
-  COMMON_FLAGS="$COMMON_FLAGS -DHAVE_PCLMUL"
+  COMMON_FLAGS="$COMMON_FLAGS $TRY_PCLMUL -DHAVE_PCLMUL"
 elif test "$USE_SSE"; then
-  echo "warning: USE_SSE specified but compiler could not use PCLMUL intrinsics, disabling"
-  exit 1
+  echo "warning: USE_SSE specified but compiler could not use PCLMUL intrinsics, disabling" >&2
 fi

 # iOS doesn't support thread-local storage, but this check would erroneously
@@ -589,6 +615,7 @@ if [ "$PLATFORM" != IOS ]; then
  #endif
  int main() {
    static __thread int tls;
+    (void)tls;
  }
 EOF
  if [ "$?" = 0 ]; then
@@ -596,6 +623,19 @@ EOF
  fi
 fi

+if [ "$FBCODE_BUILD" != "true" -a "$PLATFORM" = OS_LINUX ]; then
+  $CXX $COMMON_FLAGS $PLATFORM_SHARED_CFLAGS -x c++ -c - -o test_dl.o 2>/dev/null <<EOF
+  void dummy_func() {}
+EOF
+  if [ "$?" = 0 ]; then
+    $CXX $COMMON_FLAGS $PLATFORM_SHARED_LDFLAGS test_dl.o -o /dev/null 2>/dev/null
+    if [ "$?" = 0 ]; then
+      EXEC_LDFLAGS+="-ldl"
+      rm -f test_dl.o
+    fi
+  fi
+fi
+
 PLATFORM_CCFLAGS="$PLATFORM_CCFLAGS $COMMON_FLAGS"
 PLATFORM_CXXFLAGS="$PLATFORM_CXXFLAGS $COMMON_FLAGS"

@@ -640,3 +680,6 @@ if test -n "$WITH_JEMALLOC_FLAG"; then
  echo "WITH_JEMALLOC_FLAG=$WITH_JEMALLOC_FLAG" >> "$OUTPUT"
 fi
 echo "LUA_PATH=$LUA_PATH" >> "$OUTPUT"
+if test -n "$USE_FOLLY_DISTRIBUTED_MUTEX"; then
+  echo "USE_FOLLY_DISTRIBUTED_MUTEX=$USE_FOLLY_DISTRIBUTED_MUTEX" >> "$OUTPUT"
+fi
@@ -86,9 +86,10 @@ else
 fi
 CFLAGS+=" -DTBB"

-# use Intel SSE support for checksum calculations
-export USE_SSE=1
-export PORTABLE=1
+test "$USE_SSE" || USE_SSE=1
+export USE_SSE
+test "$PORTABLE" || PORTABLE=1
+export PORTABLE

 BINUTILS="$BINUTILS_BASE/bin"
 AR="$BINUTILS/ar"
@@ -159,4 +160,6 @@ else
  LUA_LIB=" $LUA_PATH/lib/liblua_pic.a"
 fi

+USE_FOLLY_DISTRIBUTED_MUTEX=1
+
 export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD LUA_PATH LUA_LIB
@@ -53,9 +53,10 @@ LIBUNWIND="$LIBUNWIND_BASE/lib/libunwind.a"
 TBB_INCLUDE=" -isystem $TBB_BASE/include/"
 TBB_LIBS="$TBB_BASE/lib/libtbb.a"

-# use Intel SSE support for checksum calculations
-export USE_SSE=1
-export PORTABLE=1
+test "$USE_SSE" || USE_SSE=1
+export USE_SSE
+test "$PORTABLE" || PORTABLE=1
+export PORTABLE

 BINUTILS="$BINUTILS_BASE/bin"
 AR="$BINUTILS/ar"
@@ -86,9 +86,10 @@ else
 fi
 CFLAGS+=" -DTBB"

-# use Intel SSE support for checksum calculations
-export USE_SSE=1
-export PORTABLE=1
+test "$USE_SSE" || USE_SSE=1
+export USE_SSE
+test "$PORTABLE" || PORTABLE=1
+export PORTABLE

 BINUTILS="$BINUTILS_BASE/bin"
 AR="$BINUTILS/ar"
@@ -155,4 +156,6 @@ VALGRIND_VER="$VALGRIND_BASE/bin/"
 LUA_PATH=
 LUA_LIB=

+USE_FOLLY_DISTRIBUTED_MUTEX=1
+
 export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD LUA_PATH LUA_LIB
@@ -63,6 +63,21 @@ CLEANUP_ENV="
    'user':'root'
 }"

+UPLOAD_DB_DIR="
+{
+    'name':'Upload database directory',
+    'shell':'tar -cvzf rocksdb_db.tar.gz /dev/shm/rocksdb/',
+    'user':'root',
+    'cleanup':true,
+    'provide_artifacts': [
+    {
+      'name':'rocksdb_db_dir',
+      'paths': ['rocksdb_db.tar.gz'],
+      'bundle': false,
+    },
+    ],
+}"
+
 # We will eventually set the RATIO to 1, but we want do this
 # in steps. RATIO=$(nproc) will make it work as J=1
 if [ -z $RATIO ]; then
@@ -109,13 +124,6 @@ else
  TASK_CREATION_TOOL="false"
 fi

-ARTIFACTS=" 'artifacts': [
-    {
-        'name':'database',
-        'paths':[ '/dev/shm/rocksdb' ],
-    }
-]"
-
 #
 # A mechanism to disable tests temporarily
 #
@@ -140,6 +148,7 @@ UNIT_TEST_COMMANDS="[
    {
        'name':'Rocksdb Unit Test',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -160,6 +169,7 @@ UNIT_TEST_NON_SHM_COMMANDS="[
    {
        'name':'Rocksdb Unit Test',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'timeout': 86400,
        'steps': [
            $CLEANUP_ENV,
@@ -182,6 +192,7 @@ RELEASE_BUILD_COMMANDS="[
    {
        'name':'Rocksdb Release Build',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -202,6 +213,7 @@ UNIT_TEST_COMMANDS_481="[
    {
        'name':'Rocksdb Unit Test on GCC 4.8.1',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -222,6 +234,7 @@ RELEASE_BUILD_COMMANDS_481="[
    {
        'name':'Rocksdb Release on GCC 4.8.1',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -242,6 +255,7 @@ CLANG_UNIT_TEST_COMMANDS="[
    {
        'name':'Rocksdb Unit Test',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -262,6 +276,7 @@ CLANG_RELEASE_BUILD_COMMANDS="[
    {
        'name':'Rocksdb CLANG Release Build',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -282,6 +297,7 @@ CLANG_ANALYZE_COMMANDS="[
    {
        'name':'Rocksdb analyze',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -302,6 +318,7 @@ CODE_COV_COMMANDS="[
    {
        'name':'Rocksdb Unit Test Code Coverage',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -322,6 +339,7 @@ UNITY_COMMANDS="[
    {
        'name':'Rocksdb Unity',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -342,6 +360,7 @@ LITE_BUILD_COMMANDS="[
    {
        'name':'Rocksdb Lite build',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -361,6 +380,7 @@ REPORT_LITE_BINARY_SIZE_COMMANDS="[
    {
        'name':'Rocksdb Lite Binary Size',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -376,8 +396,9 @@ REPORT_LITE_BINARY_SIZE_COMMANDS="[
 #
 STRESS_CRASH_TEST_COMMANDS="[
    {
-        'name':'Rocksdb Stress/Crash Test',
+        'name':'Rocksdb Stress and Crash Test',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'timeout': 86400,
        'steps': [
            $CLEANUP_ENV,
@@ -395,7 +416,6 @@ STRESS_CRASH_TEST_COMMANDS="[
                $PARSER
            }
        ],
-        $ARTIFACTS,
        $REPORT
    }
 ]"
@@ -405,8 +425,9 @@ STRESS_CRASH_TEST_COMMANDS="[
 #
 STRESS_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS="[
    {
-        'name':'Rocksdb Stress/Crash Test (atomic flush)',
+        'name':'Rocksdb Stress and Crash Test with atomic flush',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'timeout': 86400,
        'steps': [
            $CLEANUP_ENV,
@@ -422,9 +443,9 @@ STRESS_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS="[
                'shell':'$SHM $DEBUG $NON_TSAN_CRASH make J=1 crash_test_with_atomic_flush || $CONTRUN_NAME=crash_test_with_atomic_flush $TASK_CREATION_TOOL',
                'user':'root',
                $PARSER
-            }
+            },
+            $UPLOAD_DB_DIR,
        ],
-        $ARTIFACTS,
        $REPORT
    }
 ]"
@@ -436,6 +457,7 @@ WRITE_STRESS_COMMANDS="[
    {
        'name':'Rocksdb Write Stress Test',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -458,6 +480,7 @@ ASAN_TEST_COMMANDS="[
    {
        'name':'Rocksdb Unit Test under ASAN',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -478,6 +501,7 @@ ASAN_CRASH_TEST_COMMANDS="[
    {
        'name':'Rocksdb crash test under ASAN',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'timeout': 86400,
        'steps': [
            $CLEANUP_ENV,
@@ -498,8 +522,9 @@ ASAN_CRASH_TEST_COMMANDS="[
 #
 ASAN_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS="[
    {
-        'name':'Rocksdb crash test (atomic flush) under ASAN',
+        'name':'Rocksdb crash test with atomic flush under ASAN',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'timeout': 86400,
        'steps': [
            $CLEANUP_ENV,
@@ -510,6 +535,7 @@ ASAN_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS="[
                'user':'root',
                $PARSER
            },
+            $UPLOAD_DB_DIR,
        ],
        $REPORT
    }
@@ -522,6 +548,7 @@ UBSAN_TEST_COMMANDS="[
    {
        'name':'Rocksdb Unit Test under UBSAN',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -542,6 +569,7 @@ UBSAN_CRASH_TEST_COMMANDS="[
    {
        'name':'Rocksdb crash test under UBSAN',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'timeout': 86400,
        'steps': [
            $CLEANUP_ENV,
@@ -562,8 +590,9 @@ UBSAN_CRASH_TEST_COMMANDS="[
 #
 UBSAN_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS="[
    {
-        'name':'Rocksdb crash test (atomic flush) under UBSAN',
+        'name':'Rocksdb crash test with atomic flush under UBSAN',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'timeout': 86400,
        'steps': [
            $CLEANUP_ENV,
@@ -574,6 +603,7 @@ UBSAN_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS="[
                'user':'root',
                $PARSER
            },
+            $UPLOAD_DB_DIR,
        ],
        $REPORT
    }
@@ -586,6 +616,7 @@ VALGRIND_TEST_COMMANDS="[
    {
        'name':'Rocksdb Unit Test under valgrind',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'timeout': 86400,
        'steps': [
            $CLEANUP_ENV,
@@ -608,6 +639,7 @@ TSAN_UNIT_TEST_COMMANDS="[
    {
        'name':'Rocksdb Unit Test under TSAN',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'timeout': 86400,
        'steps': [
            $CLEANUP_ENV,
@@ -630,6 +662,7 @@ TSAN_CRASH_TEST_COMMANDS="[
    {
        'name':'Rocksdb Crash Test under TSAN',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'timeout': 86400,
        'steps': [
            $CLEANUP_ENV,
@@ -652,6 +685,7 @@ TSAN_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS="[
    {
        'name':'Rocksdb Crash Test with atomic flush under TSAN',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'timeout': 86400,
        'steps': [
            $CLEANUP_ENV,
@@ -662,6 +696,7 @@ TSAN_CRASH_TEST_WITH_ATOMIC_FLUSH_COMMANDS="[
                'user':'root',
                $PARSER
            },
+            $UPLOAD_DB_DIR,
        ],
        $REPORT
    }
@@ -684,6 +719,7 @@ FORMAT_COMPATIBLE_COMMANDS="[
    {
        'name':'Rocksdb Format Compatible tests',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -717,6 +753,7 @@ NO_COMPRESSION_COMMANDS="[
    {
        'name':'Rocksdb No Compression tests',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -794,6 +831,7 @@ JAVA_BUILD_TEST_COMMANDS="[
    {
        'name':'Rocksdb Java Build',
        'oncall':'$ONCALL',
+        'executeLocal': 'true',
        'steps': [
            $CLEANUP_ENV,
            {
@@ -3,9 +3,6 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
 #ifndef GFLAGS
 #include <cstdio>
 int main() {
@@ -14,7 +11,7 @@ int main() {
 }
 #else

-#include <inttypes.h>
+#include <cinttypes>
 #include <sys/types.h>
 #include <stdio.h>

@@ -16,9 +16,9 @@
 #include <vector>
 #include "cache/clock_cache.h"
 #include "cache/lru_cache.h"
+#include "test_util/testharness.h"
 #include "util/coding.h"
 #include "util/string_util.h"
-#include "util/testharness.h"

 namespace rocksdb {

@@ -90,7 +90,7 @@ class CacheTest : public testing::TestWithParam<std::string> {
                                  bool strict_capacity_limit) {
    auto type = GetParam();
    if (type == kLRU) {
-      return NewLRUCache(capacity, num_shard_bits, strict_capacity_limit);
+      return NewLRUCache(capacity, num_shard_bits, strict_capacity_limit, 0.0);
    }
    if (type == kClock) {
      return NewClockCache(capacity, num_shard_bits, strict_capacity_limit);
@@ -562,6 +562,7 @@ TEST_P(CacheTest, SetStrictCapacityLimit) {
    ASSERT_OK(s);
    ASSERT_NE(nullptr, handles[i]);
  }
+  ASSERT_EQ(10, cache->GetUsage());

  // test2: set the flag to true. Insert and check if it fails.
  std::string extra_key = "extra";
@@ -571,6 +572,7 @@ TEST_P(CacheTest, SetStrictCapacityLimit) {
  s = cache->Insert(extra_key, extra_value, 1, &deleter, &handle);
  ASSERT_TRUE(s.IsIncomplete());
  ASSERT_EQ(nullptr, handle);
+  ASSERT_EQ(10, cache->GetUsage());

  for (size_t i = 0; i < 10; i++) {
    cache->Release(handles[i]);
@@ -591,7 +593,7 @@ TEST_P(CacheTest, SetStrictCapacityLimit) {
  s = cache2->Insert(extra_key, extra_value, 1, &deleter);
  // AS if the key have been inserted into cache but get evicted immediately.
  ASSERT_OK(s);
-  ASSERT_EQ(5, cache->GetUsage());
+  ASSERT_EQ(5, cache2->GetUsage());
  ASSERT_EQ(nullptr, cache2->Lookup(extra_key));

  for (size_t i = 0; i < 5; i++) {
@@ -686,6 +688,14 @@ TEST_P(CacheTest, DefaultShardBits) {
  ASSERT_EQ(6, sc->GetNumShardBits());
 }

+TEST_P(CacheTest, GetCharge) {
+  Insert(1, 2);
+  Cache::Handle* h1 = cache_->Lookup(EncodeKey(1));
+  ASSERT_EQ(2, DecodeValue(cache_->Value(h1)));
+  ASSERT_EQ(1, cache_->GetCharge(h1));
+  cache_->Release(h1);
+}
+
 #ifdef SUPPORT_CLOCK_CACHE
 std::shared_ptr<Cache> (*new_clock_cache_func)(size_t, int,
                                               bool) = NewClockCache;
@@ -7,10 +7,6 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-
 #include "cache/lru_cache.h"

 #include <assert.h>
@@ -28,7 +24,7 @@ LRUHandleTable::LRUHandleTable() : list_(nullptr), length_(0), elems_(0) {

 LRUHandleTable::~LRUHandleTable() {
  ApplyToAllCacheEntries([](LRUHandle* h) {
-    if (h->refs == 1) {
+    if (!h->HasRefs()) {
      h->Free();
    }
  });
@@ -117,29 +113,17 @@ LRUCacheShard::LRUCacheShard(size_t capacity, bool strict_capacity_limit,
  SetCapacity(capacity);
 }

-LRUCacheShard::~LRUCacheShard() {}
-
-bool LRUCacheShard::Unref(LRUHandle* e) {
-  assert(e->refs > 0);
-  e->refs--;
-  return e->refs == 0;
-}
-
-// Call deleter and free
-
 void LRUCacheShard::EraseUnRefEntries() {
  autovector<LRUHandle*> last_reference_list;
  {
    MutexLock l(&mutex_);
    while (lru_.next != &lru_) {
      LRUHandle* old = lru_.next;
-      assert(old->InCache());
-      assert(old->refs ==
-             1);  // LRU list contains elements which may be evicted
+      // LRU list contains only elements which can be evicted
+      assert(old->InCache() && !old->HasRefs());
      LRU_Remove(old);
      table_.Remove(old->key(), old->hash);
      old->SetInCache(false);
-      Unref(old);
      usage_ -= old->charge;
      last_reference_list.push_back(old);
    }
@@ -152,22 +136,27 @@ void LRUCacheShard::EraseUnRefEntries() {

 void LRUCacheShard::ApplyToAllCacheEntries(void (*callback)(void*, size_t),
                                           bool thread_safe) {
+  const auto applyCallback = [&]() {
+    table_.ApplyToAllCacheEntries(
+        [callback](LRUHandle* h) { callback(h->value, h->charge); });
+  };
+
  if (thread_safe) {
-    mutex_.Lock();
-  }
-  table_.ApplyToAllCacheEntries(
-      [callback](LRUHandle* h) { callback(h->value, h->charge); });
-  if (thread_safe) {
-    mutex_.Unlock();
+    MutexLock l(&mutex_);
+    applyCallback();
+  } else {
+    applyCallback();
  }
 }

 void LRUCacheShard::TEST_GetLRUList(LRUHandle** lru, LRUHandle** lru_low_pri) {
+  MutexLock l(&mutex_);
  *lru = &lru_;
  *lru_low_pri = lru_low_pri_;
 }

 size_t LRUCacheShard::TEST_GetLRUSize() {
+  MutexLock l(&mutex_);
  LRUHandle* lru_handle = lru_.next;
  size_t lru_size = 0;
  while (lru_handle != &lru_) {
@@ -235,14 +224,13 @@ void LRUCacheShard::MaintainPoolSize() {

 void LRUCacheShard::EvictFromLRU(size_t charge,
                                 autovector<LRUHandle*>* deleted) {
-  while (usage_ + charge > capacity_ && lru_.next != &lru_) {
+  while ((usage_ + charge) > capacity_ && lru_.next != &lru_) {
    LRUHandle* old = lru_.next;
-    assert(old->InCache());
-    assert(old->refs == 1);  // LRU list contains elements which may be evicted
+    // LRU list contains only elements which can be evicted
+    assert(old->InCache() && !old->HasRefs());
    LRU_Remove(old);
    table_.Remove(old->key(), old->hash);
    old->SetInCache(false);
-    Unref(old);
    usage_ -= old->charge;
    deleted->push_back(old);
  }
@@ -256,8 +244,8 @@ void LRUCacheShard::SetCapacity(size_t capacity) {
    high_pri_pool_capacity_ = capacity_ * high_pri_pool_ratio_;
    EvictFromLRU(0, &last_reference_list);
  }
-  // we free the entries here outside of mutex for
-  // performance reasons
+
+  // Free the entries outside of mutex for performance reasons
  for (auto entry : last_reference_list) {
    entry->Free();
  }
@@ -273,22 +261,22 @@ Cache::Handle* LRUCacheShard::Lookup(const Slice& key, uint32_t hash) {
  LRUHandle* e = table_.Lookup(key, hash);
  if (e != nullptr) {
    assert(e->InCache());
-    if (e->refs == 1) {
+    if (!e->HasRefs()) {
+      // The entry is in LRU since it's in hash and has no external references
      LRU_Remove(e);
    }
-    e->refs++;
+    e->Ref();
    e->SetHit();
  }
  return reinterpret_cast<Cache::Handle*>(e);
 }

 bool LRUCacheShard::Ref(Cache::Handle* h) {
-  LRUHandle* handle = reinterpret_cast<LRUHandle*>(h);
+  LRUHandle* e = reinterpret_cast<LRUHandle*>(h);
  MutexLock l(&mutex_);
-  if (handle->InCache() && handle->refs == 1) {
-    LRU_Remove(handle);
-  }
-  handle->refs++;
+  // To create another reference - entry must be already externally referenced
+  assert(e->HasRefs());
+  e->Ref();
  return true;
 }

@@ -307,30 +295,27 @@ bool LRUCacheShard::Release(Cache::Handle* handle, bool force_erase) {
  bool last_reference = false;
  {
    MutexLock l(&mutex_);
-    last_reference = Unref(e);
+    last_reference = e->Unref();
+    if (last_reference && e->InCache()) {
+      // The item is still in cache, and nobody else holds a reference to it
+      if (usage_ > capacity_ || force_erase) {
+        // The LRU list must be empty since the cache is full
+        assert(lru_.next == &lru_ || force_erase);
+        // Take this opportunity and remove the item
+        table_.Remove(e->key(), e->hash);
+        e->SetInCache(false);
+      } else {
+        // Put the item back on the LRU list, and don't free it
+        LRU_Insert(e);
+        last_reference = false;
+      }
+    }
    if (last_reference) {
      usage_ -= e->charge;
    }
-    if (e->refs == 1 && e->InCache()) {
-      // The item is still in cache, and nobody else holds a reference to it
-      if (usage_ > capacity_ || force_erase) {
-        // the cache is full
-        // The LRU list must be empty since the cache is full
-        assert(!(usage_ > capacity_) || lru_.next == &lru_);
-        // take this opportunity and remove the item
-        table_.Remove(e->key(), e->hash);
-        e->SetInCache(false);
-        Unref(e);
-        usage_ -= e->charge;
-        last_reference = true;
-      } else {
-        // put the item on the list to be potentially freed
-        LRU_Insert(e);
-      }
-    }
  }

-  // free outside of mutex
+  // Free the entry here outside of mutex for performance reasons
  if (last_reference) {
    e->Free();
  }
@@ -346,7 +331,7 @@ Status LRUCacheShard::Insert(const Slice& key, uint32_t hash, void* value,
  // It shouldn't happen very often though.
  LRUHandle* e = reinterpret_cast<LRUHandle*>(
      new char[sizeof(LRUHandle) - 1 + key.size()]);
-  Status s;
+  Status s = Status::OK();
  autovector<LRUHandle*> last_reference_list;

  e->value = value;
@@ -355,9 +340,7 @@ Status LRUCacheShard::Insert(const Slice& key, uint32_t hash, void* value,
  e->key_length = key.size();
  e->flags = 0;
  e->hash = hash;
-  e->refs = (handle == nullptr
-                 ? 1
-                 : 2);  // One from LRUCache, one for the returned handle
+  e->refs = 0;
  e->next = e->prev = nullptr;
  e->SetInCache(true);
  e->SetPriority(priority);
@@ -370,11 +353,12 @@ Status LRUCacheShard::Insert(const Slice& key, uint32_t hash, void* value,
    // is freed or the lru list is empty
    EvictFromLRU(charge, &last_reference_list);

-    if (usage_ - lru_usage_ + charge > capacity_ &&
+    if ((usage_ + charge) > capacity_ &&
        (strict_capacity_limit_ || handle == nullptr)) {
      if (handle == nullptr) {
        // Don't insert the entry but still return ok, as if the entry inserted
        // into cache and get evicted immediately.
+        e->SetInCache(false);
        last_reference_list.push_back(e);
      } else {
        delete[] reinterpret_cast<char*>(e);
@@ -382,32 +366,30 @@ Status LRUCacheShard::Insert(const Slice& key, uint32_t hash, void* value,
        s = Status::Incomplete("Insert failed due to LRU cache being full.");
      }
    } else {
-      // insert into the cache
-      // note that the cache might get larger than its capacity if not enough
-      // space was freed
+      // Insert into the cache. Note that the cache might get larger than its
+      // capacity if not enough space was freed up.
      LRUHandle* old = table_.Insert(e);
      usage_ += e->charge;
      if (old != nullptr) {
+        assert(old->InCache());
        old->SetInCache(false);
-        if (Unref(old)) {
-          usage_ -= old->charge;
-          // old is on LRU because it's in cache and its reference count
-          // was just 1 (Unref returned 0)
+        if (!old->HasRefs()) {
+          // old is on LRU because it's in cache and its reference count is 0
          LRU_Remove(old);
+          usage_ -= old->charge;
          last_reference_list.push_back(old);
        }
      }
      if (handle == nullptr) {
        LRU_Insert(e);
      } else {
+        e->Ref();
        *handle = reinterpret_cast<Cache::Handle*>(e);
      }
-      s = Status::OK();
    }
  }

-  // we free the entries here outside of mutex for
-  // performance reasons
+  // Free the entries here outside of mutex for performance reasons
  for (auto entry : last_reference_list) {
    entry->Free();
  }
@@ -422,18 +404,18 @@ void LRUCacheShard::Erase(const Slice& key, uint32_t hash) {
    MutexLock l(&mutex_);
    e = table_.Remove(key, hash);
    if (e != nullptr) {
-      last_reference = Unref(e);
-      if (last_reference) {
-        usage_ -= e->charge;
-      }
-      if (last_reference && e->InCache()) {
-        LRU_Remove(e);
-      }
+      assert(e->InCache());
      e->SetInCache(false);
+      if (!e->HasRefs()) {
+        // The entry is in LRU since it's in hash and has no external references
+        LRU_Remove(e);
+        usage_ -= e->charge;
+        last_reference = true;
+      }
    }
  }

-  // mutex not held here
+  // Free the entry here outside of mutex for performance reasons
  // last_reference will only be true if e != nullptr
  if (last_reference) {
    e->Free();
@@ -17,31 +17,34 @@

 namespace rocksdb {

-// LRU cache implementation
+// LRU cache implementation. This class is not thread-safe.

 // An entry is a variable length heap-allocated structure.
 // Entries are referenced by cache and/or by any external entity.
-// The cache keeps all its entries in table. Some elements
+// The cache keeps all its entries in a hash table. Some elements
 // are also stored on LRU list.
 //
 // LRUHandle can be in these states:
 // 1. Referenced externally AND in hash table.
-//  In that case the entry is *not* in the LRU. (refs > 1 && in_cache == true)
-// 2. Not referenced externally and in hash table. In that case the entry is
-// in the LRU and can be freed. (refs == 1 && in_cache == true)
-// 3. Referenced externally and not in hash table. In that case the entry is
-// in not on LRU and not in table. (refs >= 1 && in_cache == false)
+//    In that case the entry is *not* in the LRU list
+//    (refs >= 1 && in_cache == true)
+// 2. Not referenced externally AND in hash table.
+//    In that case the entry is in the LRU list and can be freed.
+//    (refs == 0 && in_cache == true)
+// 3. Referenced externally AND not in hash table.
+//    In that case the entry is not in the LRU list and not in hash table.
+//    The entry can be freed when refs becomes 0.
+//    (refs >= 1 && in_cache == false)
 //
 // All newly created LRUHandles are in state 1. If you call
-// LRUCacheShard::Release
-// on entry in state 1, it will go into state 2. To move from state 1 to
-// state 3, either call LRUCacheShard::Erase or LRUCacheShard::Insert with the
-// same key.
+// LRUCacheShard::Release on entry in state 1, it will go into state 2.
+// To move from state 1 to state 3, either call LRUCacheShard::Erase or
+// LRUCacheShard::Insert with the same key (but possibly different value).
 // To move from state 2 to state 1, use LRUCacheShard::Lookup.
 // Before destruction, make sure that no handles are in state 1. This means
 // that any successful LRUCacheShard::Lookup/LRUCacheShard::Insert have a
-// matching
-// RUCache::Release (to move into state 2) or LRUCacheShard::Erase (for state 3)
+// matching LRUCache::Release (to move into state 2) or LRUCacheShard::Erase
+// (to move into state 3).

 struct LRUHandle {
  void* value;
@@ -51,37 +54,42 @@ struct LRUHandle {
  LRUHandle* prev;
  size_t charge;  // TODO(opt): Only allow uint32_t?
  size_t key_length;
-  uint32_t refs;     // a number of refs to this entry
-                     // cache itself is counted as 1
+  // The hash of key(). Used for fast sharding and comparisons.
+  uint32_t hash;
+  // The number of external refs to this entry. The cache itself is not counted.
+  uint32_t refs;

-  // Include the following flags:
-  //   IN_CACHE:         whether this entry is referenced by the hash table.
-  //   IS_HIGH_PRI:      whether this entry is high priority entry.
-  //   IN_HIGH_PRI_POOL: whether this entry is in high-pri pool.
-  //   HAS_HIT:          whether this entry has had any lookups (hits).
  enum Flags : uint8_t {
+    // Whether this entry is referenced by the hash table.
    IN_CACHE = (1 << 0),
+    // Whether this entry is high priority entry.
    IS_HIGH_PRI = (1 << 1),
+    // Whether this entry is in high-pri pool.
    IN_HIGH_PRI_POOL = (1 << 2),
+    // Wwhether this entry has had any lookups (hits).
    HAS_HIT = (1 << 3),
  };

  uint8_t flags;

-  uint32_t hash;     // Hash of key(); used for fast sharding and comparisons
+  // Beginning of the key (MUST BE THE LAST FIELD IN THIS STRUCT!)
+  char key_data[1];

-  char key_data[1];  // Beginning of key
+  Slice key() const { return Slice(key_data, key_length); }

-  Slice key() const {
-    // For cheaper lookups, we allow a temporary Handle object
-    // to store a pointer to a key in "value".
-    if (next == this) {
-      return *(reinterpret_cast<Slice*>(value));
-    } else {
-      return Slice(key_data, key_length);
-    }
+  // Increase the reference count by 1.
+  void Ref() { refs++; }
+
+  // Just reduce the reference count by 1. Return true if it was last reference.
+  bool Unref() {
+    assert(refs > 0);
+    refs--;
+    return refs == 0;
  }

+  // Return true if there are external refs, false otherwise.
+  bool HasRefs() const { return refs > 0; }
+
  bool InCache() const { return flags & IN_CACHE; }
  bool IsHighPri() const { return flags & IS_HIGH_PRI; }
  bool InHighPriPool() const { return flags & IN_HIGH_PRI_POOL; }
@@ -114,7 +122,7 @@ struct LRUHandle {
  void SetHit() { flags |= HAS_HIT; }

  void Free() {
-    assert((refs == 1 && InCache()) || (refs == 0 && !InCache()));
+    assert(refs == 0);
    if (deleter) {
      (*deleter)(key(), value);
    }
@@ -169,7 +177,7 @@ class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShard {
 public:
  LRUCacheShard(size_t capacity, bool strict_capacity_limit,
                double high_pri_pool_ratio, bool use_adaptive_mutex);
-  virtual ~LRUCacheShard();
+  virtual ~LRUCacheShard() override = default;

  // Separate from constructor so caller can easily make an array of LRUCache
  // if current usage is more than new capacity, the function will attempt to
@@ -225,10 +233,6 @@ class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShard {
  // high-pri pool is no larger than the size specify by high_pri_pool_pct.
  void MaintainPoolSize();

-  // Just reduce the reference count by 1.
-  // Return true if last reference
-  bool Unref(LRUHandle* e);
-
  // Free some space following strict LRU policy until enough space
  // to hold (usage_ + charge) is freed or the lru list is empty
  // This function is not thread safe - it needs to be executed while
@@ -8,7 +8,7 @@
 #include <string>
 #include <vector>
 #include "port/port.h"
-#include "util/testharness.h"
+#include "test_util/testharness.h"

 namespace rocksdb {

@@ -7,10 +7,6 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-
 #include "cache/sharded_cache.h"

 #include <string>
@@ -54,7 +54,8 @@ class ShardedCache : public Cache {
  virtual CacheShard* GetShard(int shard) = 0;
  virtual const CacheShard* GetShard(int shard) const = 0;
  virtual void* Value(Handle* handle) override = 0;
-  virtual size_t GetCharge(Handle* handle) const = 0;
+  virtual size_t GetCharge(Handle* handle) const override = 0;
+
  virtual uint32_t GetHash(Handle* handle) const = 0;
  virtual void DisownData() override = 0;

@@ -1,21 +1,29 @@
 # - Find JeMalloc library
 # Find the native JeMalloc includes and library
 #
-# JEMALLOC_INCLUDE_DIR - where to find jemalloc.h, etc.
-# JEMALLOC_LIBRARIES - List of libraries when using jemalloc.
-# JEMALLOC_FOUND - True if jemalloc found.
+# JeMalloc_INCLUDE_DIRS - where to find jemalloc.h, etc.
+# JeMalloc_LIBRARIES - List of libraries when using jemalloc.
+# JeMalloc_FOUND - True if jemalloc found.

-find_path(JEMALLOC_INCLUDE_DIR
+find_path(JeMalloc_INCLUDE_DIRS
  NAMES jemalloc/jemalloc.h
  HINTS ${JEMALLOC_ROOT_DIR}/include)

-find_library(JEMALLOC_LIBRARIES
+find_library(JeMalloc_LIBRARIES
  NAMES jemalloc
  HINTS ${JEMALLOC_ROOT_DIR}/lib)

 include(FindPackageHandleStandardArgs)
-find_package_handle_standard_args(jemalloc DEFAULT_MSG JEMALLOC_LIBRARIES JEMALLOC_INCLUDE_DIR)
+find_package_handle_standard_args(JeMalloc DEFAULT_MSG JeMalloc_LIBRARIES JeMalloc_INCLUDE_DIRS)

 mark_as_advanced(
-  JEMALLOC_LIBRARIES
-  JEMALLOC_INCLUDE_DIR)
+  JeMalloc_LIBRARIES
+  JeMalloc_INCLUDE_DIRS)
+
+if(JeMalloc_FOUND AND NOT (TARGET JeMalloc::JeMalloc))
+  add_library (JeMalloc::JeMalloc UNKNOWN IMPORTED)
+  set_target_properties(JeMalloc::JeMalloc
+    PROPERTIES
+      IMPORTED_LOCATION ${JeMalloc_LIBRARIES}
+      INTERFACE_INCLUDE_DIRECTORIES ${JeMalloc_INCLUDE_DIRS})
+endif()
@@ -1,11 +1,11 @@
 # - Find NUMA
 # Find the NUMA library and includes
 #
-# NUMA_INCLUDE_DIR - where to find numa.h, etc.
+# NUMA_INCLUDE_DIRS - where to find numa.h, etc.
 # NUMA_LIBRARIES - List of libraries when using NUMA.
 # NUMA_FOUND - True if NUMA found.

-find_path(NUMA_INCLUDE_DIR
+find_path(NUMA_INCLUDE_DIRS
  NAMES numa.h numaif.h
  HINTS ${NUMA_ROOT_DIR}/include)

@@ -14,8 +14,16 @@ find_library(NUMA_LIBRARIES
  HINTS ${NUMA_ROOT_DIR}/lib)

 include(FindPackageHandleStandardArgs)
-find_package_handle_standard_args(NUMA DEFAULT_MSG NUMA_LIBRARIES NUMA_INCLUDE_DIR)
+find_package_handle_standard_args(NUMA DEFAULT_MSG NUMA_LIBRARIES NUMA_INCLUDE_DIRS)

 mark_as_advanced(
  NUMA_LIBRARIES
-  NUMA_INCLUDE_DIR)
+  NUMA_INCLUDE_DIRS)
+
+if(NUMA_FOUND AND NOT (TARGET NUMA::NUMA))
+  add_library (NUMA::NUMA UNKNOWN IMPORTED)
+  set_target_properties(NUMA::NUMA
+    PROPERTIES
+      IMPORTED_LOCATION ${NUMA_LIBRARIES}
+      INTERFACE_INCLUDE_DIRECTORIES ${NUMA_INCLUDE_DIRS})
+endif()
@@ -1,7 +1,7 @@
 # - Find TBB
 # Find the Thread Building Blocks library and includes
 #
-# TBB_INCLUDE_DIR - where to find tbb.h, etc.
+# TBB_INCLUDE_DIRS - where to find tbb.h, etc.
 # TBB_LIBRARIES - List of libraries when using TBB.
 # TBB_FOUND - True if TBB found.

@@ -9,17 +9,25 @@ if(NOT DEFINED TBB_ROOT_DIR)
  set(TBB_ROOT_DIR "$ENV{TBBROOT}")
 endif()

-find_path(TBB_INCLUDE_DIR
-NAMES tbb/tbb.h
-HINTS ${TBB_ROOT_DIR}/include)
+find_path(TBB_INCLUDE_DIRS
+  NAMES tbb/tbb.h
+  HINTS ${TBB_ROOT_DIR}/include)

 find_library(TBB_LIBRARIES
-NAMES tbb
-HINTS ${TBB_ROOT_DIR}/lib ENV LIBRARY_PATH)
+  NAMES tbb
+  HINTS ${TBB_ROOT_DIR}/lib ENV LIBRARY_PATH)

 include(FindPackageHandleStandardArgs)
-find_package_handle_standard_args(TBB DEFAULT_MSG TBB_LIBRARIES TBB_INCLUDE_DIR)
+find_package_handle_standard_args(TBB DEFAULT_MSG TBB_LIBRARIES TBB_INCLUDE_DIRS)

 mark_as_advanced(
-TBB_LIBRARIES
-TBB_INCLUDE_DIR)
+  TBB_LIBRARIES
+  TBB_INCLUDE_DIRS)
+
+if(TBB_FOUND AND NOT (TARGET TBB::TBB))
+  add_library (TBB::TBB UNKNOWN IMPORTED)
+  set_target_properties(TBB::TBB
+    PROPERTIES
+      IMPORTED_LOCATION ${TBB_LIBRARIES}
+      INTERFACE_INCLUDE_DIRECTORIES ${TBB_INCLUDE_DIRS})
+endif()
@@ -1,21 +0,0 @@
-# - Find Bzip2
-# Find the bzip2 compression library and includes
-#
-# BZIP2_INCLUDE_DIR - where to find bzlib.h, etc.
-# BZIP2_LIBRARIES - List of libraries when using bzip2.
-# BZIP2_FOUND - True if bzip2 found.
-
-find_path(BZIP2_INCLUDE_DIR
-  NAMES bzlib.h
-  HINTS ${BZIP2_ROOT_DIR}/include)
-
-find_library(BZIP2_LIBRARIES
-  NAMES bz2
-  HINTS ${BZIP2_ROOT_DIR}/lib)
-
-include(FindPackageHandleStandardArgs)
-find_package_handle_standard_args(bzip2 DEFAULT_MSG BZIP2_LIBRARIES BZIP2_INCLUDE_DIR)
-
-mark_as_advanced(
-  BZIP2_LIBRARIES
-  BZIP2_INCLUDE_DIR)
@@ -1,21 +1,29 @@
 # - Find Lz4
 # Find the lz4 compression library and includes
 #
-# LZ4_INCLUDE_DIR - where to find lz4.h, etc.
-# LZ4_LIBRARIES - List of libraries when using lz4.
-# LZ4_FOUND - True if lz4 found.
+# lz4_INCLUDE_DIRS - where to find lz4.h, etc.
+# lz4_LIBRARIES - List of libraries when using lz4.
+# lz4_FOUND - True if lz4 found.

-find_path(LZ4_INCLUDE_DIR
+find_path(lz4_INCLUDE_DIRS
  NAMES lz4.h
-  HINTS ${LZ4_ROOT_DIR}/include)
+  HINTS ${lz4_ROOT_DIR}/include)

-find_library(LZ4_LIBRARIES
+find_library(lz4_LIBRARIES
  NAMES lz4
-  HINTS ${LZ4_ROOT_DIR}/lib)
+  HINTS ${lz4_ROOT_DIR}/lib)

 include(FindPackageHandleStandardArgs)
-find_package_handle_standard_args(lz4 DEFAULT_MSG LZ4_LIBRARIES LZ4_INCLUDE_DIR)
+find_package_handle_standard_args(lz4 DEFAULT_MSG lz4_LIBRARIES lz4_INCLUDE_DIRS)

 mark_as_advanced(
-  LZ4_LIBRARIES
-  LZ4_INCLUDE_DIR)
+  lz4_LIBRARIES
+  lz4_INCLUDE_DIRS)
+
+if(lz4_FOUND AND NOT (TARGET lz4::lz4))
+  add_library(lz4::lz4 UNKNOWN IMPORTED)
+  set_target_properties(lz4::lz4
+    PROPERTIES
+      IMPORTED_LOCATION ${lz4_LIBRARIES}
+      INTERFACE_INCLUDE_DIRECTORIES ${lz4_INCLUDE_DIRS})
+endif()
@@ -1,21 +1,29 @@
 # - Find Snappy
 # Find the snappy compression library and includes
 #
-# SNAPPY_INCLUDE_DIR - where to find snappy.h, etc.
-# SNAPPY_LIBRARIES - List of libraries when using snappy.
-# SNAPPY_FOUND - True if snappy found.
+# snappy_INCLUDE_DIRS - where to find snappy.h, etc.
+# snappy_LIBRARIES - List of libraries when using snappy.
+# snappy_FOUND - True if snappy found.

-find_path(SNAPPY_INCLUDE_DIR
+find_path(snappy_INCLUDE_DIRS
  NAMES snappy.h
-  HINTS ${SNAPPY_ROOT_DIR}/include)
+  HINTS ${snappy_ROOT_DIR}/include)

-find_library(SNAPPY_LIBRARIES
+find_library(snappy_LIBRARIES
  NAMES snappy
-  HINTS ${SNAPPY_ROOT_DIR}/lib)
+  HINTS ${snappy_ROOT_DIR}/lib)

 include(FindPackageHandleStandardArgs)
-find_package_handle_standard_args(snappy DEFAULT_MSG SNAPPY_LIBRARIES SNAPPY_INCLUDE_DIR)
+find_package_handle_standard_args(snappy DEFAULT_MSG snappy_LIBRARIES snappy_INCLUDE_DIRS)

 mark_as_advanced(
-  SNAPPY_LIBRARIES
-  SNAPPY_INCLUDE_DIR)
+  snappy_LIBRARIES
+  snappy_INCLUDE_DIRS)
+
+if(snappy_FOUND AND NOT (TARGET snappy::snappy))
+  add_library (snappy::snappy UNKNOWN IMPORTED)
+  set_target_properties(snappy::snappy
+    PROPERTIES
+      IMPORTED_LOCATION ${snappy_LIBRARIES}
+      INTERFACE_INCLUDE_DIRECTORIES ${snappy_INCLUDE_DIRS})
+endif()
@@ -1,21 +1,29 @@
 # - Find zstd
 # Find the zstd compression library and includes
 #
-# ZSTD_INCLUDE_DIR - where to find zstd.h, etc.
-# ZSTD_LIBRARIES - List of libraries when using zstd.
-# ZSTD_FOUND - True if zstd found.
+# zstd_INCLUDE_DIRS - where to find zstd.h, etc.
+# zstd_LIBRARIES - List of libraries when using zstd.
+# zstd_FOUND - True if zstd found.

-find_path(ZSTD_INCLUDE_DIR
+find_path(zstd_INCLUDE_DIRS
  NAMES zstd.h
-  HINTS ${ZSTD_ROOT_DIR}/include)
+  HINTS ${zstd_ROOT_DIR}/include)

-find_library(ZSTD_LIBRARIES
+find_library(zstd_LIBRARIES
  NAMES zstd
-  HINTS ${ZSTD_ROOT_DIR}/lib)
+  HINTS ${zstd_ROOT_DIR}/lib)

 include(FindPackageHandleStandardArgs)
-find_package_handle_standard_args(zstd DEFAULT_MSG ZSTD_LIBRARIES ZSTD_INCLUDE_DIR)
+find_package_handle_standard_args(zstd DEFAULT_MSG zstd_LIBRARIES zstd_INCLUDE_DIRS)

 mark_as_advanced(
-  ZSTD_LIBRARIES
-  ZSTD_INCLUDE_DIR)
+  zstd_LIBRARIES
+  zstd_INCLUDE_DIRS)
+
+if(zstd_FOUND AND NOT (TARGET zstd::zstd))
+  add_library (zstd::zstd UNKNOWN IMPORTED)
+  set_target_properties(zstd::zstd
+    PROPERTIES
+      IMPORTED_LOCATION ${zstd_LIBRARIES}
+      INTERFACE_INCLUDE_DIRECTORIES ${zstd_INCLUDE_DIRS})
+endif()
@@ -0,0 +1,10 @@
+# Read rocksdb version from version.h header file.
+
+function(get_rocksdb_version version_var)
+  file(READ "${CMAKE_SOURCE_DIR}/include/rocksdb/version.h" version_header_file)
+  foreach(component MAJOR MINOR PATCH)
+    string(REGEX MATCH "#define ROCKSDB_${component} ([0-9]+)" _ ${version_header_file})
+    set(ROCKSDB_VERSION_${component} ${CMAKE_MATCH_1})
+  endforeach()
+  set(${version_var} "${ROCKSDB_VERSION_MAJOR}.${ROCKSDB_VERSION_MINOR}.${ROCKSDB_VERSION_PATCH}" PARENT_SCOPE)
+endfunction()
@@ -1,3 +1,4 @@
+#!/usr/bin/env python2
 # Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
 import re
 import sys
@@ -13,7 +13,7 @@
 #include <deque>
 #include <vector>

-#include "db/compaction_iterator.h"
+#include "db/compaction/compaction_iterator.h"
 #include "db/dbformat.h"
 #include "db/event_helpers.h"
 #include "db/internal_stats.h"
@@ -21,6 +21,7 @@
 #include "db/range_del_aggregator.h"
 #include "db/table_cache.h"
 #include "db/version_edit.h"
+#include "file/filename.h"
 #include "monitoring/iostats_context_imp.h"
 #include "monitoring/thread_status_util.h"
 #include "rocksdb/db.h"
@@ -28,13 +29,12 @@
 #include "rocksdb/iterator.h"
 #include "rocksdb/options.h"
 #include "rocksdb/table.h"
-#include "table/block_based_table_builder.h"
+#include "table/block_based/block_based_table_builder.h"
 #include "table/format.h"
 #include "table/internal_iterator.h"
+#include "test_util/sync_point.h"
 #include "util/file_reader_writer.h"
-#include "util/filename.h"
 #include "util/stop_watch.h"
-#include "util/sync_point.h"

 namespace rocksdb {

@@ -221,8 +221,9 @@ Status BuildTable(
          mutable_cf_options.prefix_extractor.get(), nullptr,
          (internal_stats == nullptr) ? nullptr
                                      : internal_stats->GetFileReadHist(0),
-          false /* for_compaction */, nullptr /* arena */,
-          false /* skip_filter */, level));
+          TableReaderCaller::kFlush, /*arena=*/nullptr,
+          /*skip_filter=*/false, level, /*smallest_compaction_key=*/nullptr,
+          /*largest_compaction_key*/ nullptr));
      s = it->status();
      if (s.ok() && paranoid_file_checks) {
        for (it->SeekToFirst(); it->Valid(); it->Next()) {
@@ -11,6 +11,7 @@
 #include <vector>
 #include "db/range_tombstone_fragmenter.h"
 #include "db/table_properties_collector.h"
+#include "logging/event_logger.h"
 #include "options/cf_options.h"
 #include "rocksdb/comparator.h"
 #include "rocksdb/env.h"
@@ -20,7 +21,6 @@
 #include "rocksdb/table_properties.h"
 #include "rocksdb/types.h"
 #include "table/scoped_arena_iterator.h"
-#include "util/event_logger.h"

 namespace rocksdb {

@@ -517,6 +517,21 @@ rocksdb_t* rocksdb_open_for_read_only(
  return result;
 }

+rocksdb_t* rocksdb_open_as_secondary(const rocksdb_options_t* options,
+                                     const char* name,
+                                     const char* secondary_path,
+                                     char** errptr) {
+  DB* db;
+  if (SaveError(errptr,
+                DB::OpenAsSecondary(options->rep, std::string(name),
+                                    std::string(secondary_path), &db))) {
+    return nullptr;
+  }
+  rocksdb_t* result = new rocksdb_t;
+  result->rep = db;
+  return result;
+}
+
 rocksdb_backup_engine_t* rocksdb_backup_engine_open(
    const rocksdb_options_t* options, const char* path, char** errptr) {
  BackupEngine* be;
@@ -717,6 +732,37 @@ rocksdb_t* rocksdb_open_for_read_only_column_families(
  return result;
 }

+rocksdb_t* rocksdb_open_as_secondary_column_families(
+    const rocksdb_options_t* db_options, const char* name,
+    const char* secondary_path, int num_column_families,
+    const char** column_family_names,
+    const rocksdb_options_t** column_family_options,
+    rocksdb_column_family_handle_t** column_family_handles, char** errptr) {
+  std::vector<ColumnFamilyDescriptor> column_families;
+  for (int i = 0; i != num_column_families; ++i) {
+    column_families.emplace_back(
+        std::string(column_family_names[i]),
+        ColumnFamilyOptions(column_family_options[i]->rep));
+  }
+  DB* db;
+  std::vector<ColumnFamilyHandle*> handles;
+  if (SaveError(errptr, DB::OpenAsSecondary(DBOptions(db_options->rep),
+                                            std::string(name),
+                                            std::string(secondary_path),
+                                            column_families, &handles, &db))) {
+    return nullptr;
+  }
+  for (size_t i = 0; i != handles.size(); ++i) {
+    rocksdb_column_family_handle_t* c_handle =
+        new rocksdb_column_family_handle_t;
+    c_handle->rep = handles[i];
+    column_family_handles[i] = c_handle;
+  }
+  rocksdb_t* result = new rocksdb_t;
+  result->rep = db;
+  return result;
+}
+
 char** rocksdb_list_column_families(
    const rocksdb_options_t* options,
    const char* name,
@@ -988,7 +1034,7 @@ void rocksdb_wal_iter_destroy (const rocksdb_wal_iterator_t* iter) {
 rocksdb_writebatch_t* rocksdb_wal_iter_get_batch (const rocksdb_wal_iterator_t* iter, uint64_t* seq) {
  rocksdb_writebatch_t* result = rocksdb_writebatch_create();
  BatchResult wal_batch = iter->rep->GetBatch();
-  result->rep = * wal_batch.writeBatchPtr.release();
+  result->rep = std::move(*wal_batch.writeBatchPtr);
  if (seq != nullptr) {
    *seq = wal_batch.sequence;
  }
@@ -2468,11 +2514,21 @@ void rocksdb_options_set_max_write_buffer_number_to_maintain(
  opt->rep.max_write_buffer_number_to_maintain = n;
 }

+void rocksdb_options_set_max_write_buffer_size_to_maintain(
+    rocksdb_options_t* opt, int64_t n) {
+  opt->rep.max_write_buffer_size_to_maintain = n;
+}
+
 void rocksdb_options_set_enable_pipelined_write(rocksdb_options_t* opt,
                                                unsigned char v) {
  opt->rep.enable_pipelined_write = v;
 }

+void rocksdb_options_set_unordered_write(rocksdb_options_t* opt,
+                                         unsigned char v) {
+  opt->rep.unordered_write = v;
+}
+
 void rocksdb_options_set_max_subcompactions(rocksdb_options_t* opt,
                                            uint32_t n) {
  opt->rep.max_subcompactions = n;
@@ -3168,6 +3224,11 @@ void rocksdb_writeoptions_set_low_pri(
  opt->rep.low_pri = v;
 }

+void rocksdb_writeoptions_set_memtable_insert_hint_per_batch(
+    rocksdb_writeoptions_t* opt, unsigned char v) {
+  opt->rep.memtable_insert_hint_per_batch = v;
+}
+
 rocksdb_compactoptions_t* rocksdb_compactoptions_create() {
  return new rocksdb_compactoptions_t;
 }
@@ -3268,6 +3329,22 @@ void rocksdb_env_join_all_threads(rocksdb_env_t* env) {
  env->rep->WaitForJoin();
 }

+void rocksdb_env_lower_thread_pool_io_priority(rocksdb_env_t* env) {
+  env->rep->LowerThreadPoolIOPriority();
+}
+
+void rocksdb_env_lower_high_priority_thread_pool_io_priority(rocksdb_env_t* env) {
+  env->rep->LowerThreadPoolIOPriority(Env::HIGH);
+}
+
+void rocksdb_env_lower_thread_pool_cpu_priority(rocksdb_env_t* env) {
+  env->rep->LowerThreadPoolCPUPriority();
+}
+
+void rocksdb_env_lower_high_priority_thread_pool_cpu_priority(rocksdb_env_t* env) {
+  env->rep->LowerThreadPoolCPUPriority(Env::HIGH);
+}
+
 void rocksdb_env_destroy(rocksdb_env_t* env) {
  if (!env->is_default) delete env->rep;
  delete env;
@@ -3402,6 +3479,10 @@ void rocksdb_ingest_external_file_cf(
  SaveError(errptr, db->rep->IngestExternalFile(handle->rep, files, opt->rep));
 }

+void rocksdb_try_catch_up_with_primary(rocksdb_t* db, char** errptr) {
+  SaveError(errptr, db->rep->TryCatchUpWithPrimary());
+}
+
 rocksdb_slicetransform_t* rocksdb_slicetransform_create(
    void* state,
    void (*destructor)(void*),
@@ -45,6 +45,7 @@ static char sstfilename[200];
 static char dbbackupname[200];
 static char dbcheckpointname[200];
 static char dbpathname[200];
+static char secondary_path[200];

 static void StartPhase(const char* name) {
  fprintf(stderr, "=== Test %s\n", name);
@@ -1722,6 +1723,59 @@ int main(int argc, char** argv) {
    CheckNoError(err);
  }

+  // Check that secondary instance works.
+  StartPhase("open_as_secondary");
+  {
+    rocksdb_close(db);
+    rocksdb_destroy_db(options, dbname, &err);
+
+    rocksdb_options_t* db_options = rocksdb_options_create();
+    rocksdb_options_set_create_if_missing(db_options, 1);
+    db = rocksdb_open(db_options, dbname, &err);
+    CheckNoError(err);
+    rocksdb_t* db1;
+    rocksdb_options_t* opts = rocksdb_options_create();
+    rocksdb_options_set_max_open_files(opts, -1);
+    rocksdb_options_set_create_if_missing(opts, 1);
+    snprintf(secondary_path, sizeof(secondary_path),
+             "%s/rocksdb_c_test_secondary-%d", GetTempDir(), ((int)geteuid()));
+    db1 = rocksdb_open_as_secondary(opts, dbname, secondary_path, &err);
+    CheckNoError(err);
+
+    rocksdb_writeoptions_set_sync(woptions, 0);
+    rocksdb_writeoptions_disable_WAL(woptions, 1);
+    rocksdb_put(db, woptions, "key0", 4, "value0", 6, &err);
+    CheckNoError(err);
+    rocksdb_flushoptions_t* flush_opts = rocksdb_flushoptions_create();
+    rocksdb_flushoptions_set_wait(flush_opts, 1);
+    rocksdb_flush(db, flush_opts, &err);
+    CheckNoError(err);
+    rocksdb_try_catch_up_with_primary(db1, &err);
+    CheckNoError(err);
+    rocksdb_readoptions_t* ropts = rocksdb_readoptions_create();
+    rocksdb_readoptions_set_verify_checksums(ropts, 1);
+    rocksdb_readoptions_set_snapshot(ropts, NULL);
+    CheckGet(db, ropts, "key0", "value0");
+    CheckGet(db1, ropts, "key0", "value0");
+
+    rocksdb_writeoptions_disable_WAL(woptions, 0);
+    rocksdb_put(db, woptions, "key1", 4, "value1", 6, &err);
+    CheckNoError(err);
+    rocksdb_try_catch_up_with_primary(db1, &err);
+    CheckNoError(err);
+    CheckGet(db1, ropts, "key0", "value0");
+    CheckGet(db1, ropts, "key1", "value1");
+
+    rocksdb_close(db1);
+    rocksdb_destroy_db(opts, secondary_path, &err);
+    CheckNoError(err);
+
+    rocksdb_options_destroy(db_options);
+    rocksdb_options_destroy(opts);
+    rocksdb_readoptions_destroy(ropts);
+    rocksdb_flushoptions_destroy(flush_opts);
+  }
+
  // Simple sanity check that options setting db_paths work.
  StartPhase("open_db_paths");
  {
@@ -9,34 +9,31 @@

 #include "db/column_family.h"

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-
-#include <inttypes.h>
+#include <cinttypes>
 #include <vector>
 #include <string>
 #include <algorithm>
 #include <limits>

-#include "db/compaction_picker.h"
-#include "db/compaction_picker_fifo.h"
-#include "db/compaction_picker_universal.h"
-#include "db/db_impl.h"
+#include "db/compaction/compaction_picker.h"
+#include "db/compaction/compaction_picker_fifo.h"
+#include "db/compaction/compaction_picker_level.h"
+#include "db/compaction/compaction_picker_universal.h"
+#include "db/db_impl/db_impl.h"
 #include "db/internal_stats.h"
 #include "db/job_context.h"
 #include "db/range_del_aggregator.h"
 #include "db/table_properties_collector.h"
 #include "db/version_set.h"
 #include "db/write_controller.h"
+#include "file/sst_file_manager_impl.h"
 #include "memtable/hash_skiplist_rep.h"
 #include "monitoring/thread_status_util.h"
 #include "options/options_helper.h"
-#include "table/block_based_table_factory.h"
+#include "table/block_based/block_based_table_factory.h"
 #include "table/merging_iterator.h"
 #include "util/autovector.h"
 #include "util/compression.h"
-#include "util/sst_file_manager_impl.h"

 namespace rocksdb {

@@ -230,7 +227,14 @@ ColumnFamilyOptions SanitizeOptions(const ImmutableDBOptions& db_options,
  if (result.max_write_buffer_number < 2) {
    result.max_write_buffer_number = 2;
  }
-  if (result.max_write_buffer_number_to_maintain < 0) {
+  // fall back max_write_buffer_number_to_maintain if
+  // max_write_buffer_size_to_maintain is not set
+  if (result.max_write_buffer_size_to_maintain < 0) {
+    result.max_write_buffer_size_to_maintain =
+        result.max_write_buffer_number *
+        static_cast<int64_t>(result.write_buffer_size);
+  } else if (result.max_write_buffer_size_to_maintain == 0 &&
+             result.max_write_buffer_number_to_maintain < 0) {
    result.max_write_buffer_number_to_maintain = result.max_write_buffer_number;
  }
  // bloom filter size shouldn't exceed 1/4 of memtable size.
@@ -408,7 +412,8 @@ 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 EnvOptions& env_options, ColumnFamilySet* column_family_set,
+    BlockCacheTracer* const block_cache_tracer)
    : id_(id),
      name_(name),
      dummy_versions_(_dummy_versions),
@@ -425,7 +430,8 @@ ColumnFamilyData::ColumnFamilyData(
      write_buffer_manager_(write_buffer_manager),
      mem_(nullptr),
      imm_(ioptions_.min_write_buffer_number_to_merge,
-           ioptions_.max_write_buffer_number_to_maintain),
+           ioptions_.max_write_buffer_number_to_maintain,
+           ioptions_.max_write_buffer_size_to_maintain),
      super_version_(nullptr),
      super_version_number_(0),
      local_sv_(new ThreadLocalPtr(&SuperVersionUnrefHandle)),
@@ -448,7 +454,8 @@ 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_, env_options, _table_cache,
+                                      block_cache_tracer));
    if (ioptions_.compaction_style == kCompactionStyleLevel) {
      compaction_picker_.reset(
          new LevelCompactionPicker(ioptions_, &internal_comparator_));
@@ -1147,13 +1154,60 @@ void ColumnFamilyData::ResetThreadLocalSuperVersions() {
  }
 }

+Status ColumnFamilyData::ValidateOptions(
+    const DBOptions& db_options, const ColumnFamilyOptions& cf_options) {
+  Status s;
+  s = CheckCompressionSupported(cf_options);
+  if (s.ok() && db_options.allow_concurrent_memtable_write) {
+    s = CheckConcurrentWritesSupported(cf_options);
+  }
+  if (s.ok()) {
+    s = CheckCFPathsSupported(db_options, cf_options);
+  }
+  if (!s.ok()) {
+    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.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.table_factory->Name() != BlockBasedTableFactory().Name()) {
+      return Status::NotSupported(
+          "Periodic Compaction is only supported in "
+          "Block-Based Table format. ");
+    }
+  }
+  return s;
+}
+
 #ifndef ROCKSDB_LITE
 Status ColumnFamilyData::SetOptions(
-      const std::unordered_map<std::string, std::string>& options_map) {
+    const DBOptions& db_options,
+    const std::unordered_map<std::string, std::string>& options_map) {
  MutableCFOptions new_mutable_cf_options;
  Status s =
      GetMutableOptionsFromStrings(mutable_cf_options_, options_map,
                                   ioptions_.info_log, &new_mutable_cf_options);
+  if (s.ok()) {
+    ColumnFamilyOptions cf_options =
+        BuildColumnFamilyOptions(initial_cf_options_, new_mutable_cf_options);
+    s = ValidateOptions(db_options, cf_options);
+  }
  if (s.ok()) {
    mutable_cf_options_ = new_mutable_cf_options;
    mutable_cf_options_.RefreshDerivedOptions(ioptions_);
@@ -1210,18 +1264,20 @@ ColumnFamilySet::ColumnFamilySet(const std::string& dbname,
                                 const EnvOptions& env_options,
                                 Cache* table_cache,
                                 WriteBufferManager* write_buffer_manager,
-                                 WriteController* write_controller)
+                                 WriteController* write_controller,
+                                 BlockCacheTracer* const block_cache_tracer)
    : max_column_family_(0),
-      dummy_cfd_(new ColumnFamilyData(0, "", nullptr, nullptr, nullptr,
-                                      ColumnFamilyOptions(), *db_options,
-                                      env_options, nullptr)),
+      dummy_cfd_(new ColumnFamilyData(
+          0, "", nullptr, nullptr, nullptr, ColumnFamilyOptions(), *db_options,
+          env_options, nullptr, block_cache_tracer)),
      default_cfd_cache_(nullptr),
      db_name_(dbname),
      db_options_(db_options),
      env_options_(env_options),
      table_cache_(table_cache),
      write_buffer_manager_(write_buffer_manager),
-      write_controller_(write_controller) {
+      write_controller_(write_controller),
+      block_cache_tracer_(block_cache_tracer) {
  // initialize linked list
  dummy_cfd_->prev_ = dummy_cfd_;
  dummy_cfd_->next_ = dummy_cfd_;
@@ -1289,7 +1345,7 @@ 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);
+      *db_options_, env_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);
@@ -24,6 +24,7 @@
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
 #include "rocksdb/options.h"
+#include "trace_replay/block_cache_tracer.h"
 #include "util/thread_local.h"

 namespace rocksdb {
@@ -45,6 +46,112 @@ class InstrumentedMutexLock;
 struct SuperVersionContext;

 extern const double kIncSlowdownRatio;
+// This file contains a list of data structures for managing column family
+// level metadata.
+//
+// The basic relationships among classes declared here are illustrated as
+// following:
+//
+//       +----------------------+    +----------------------+   +--------+
+//   +---+ ColumnFamilyHandle 1 | +--+ ColumnFamilyHandle 2 |   | DBImpl |
+//   |   +----------------------+ |  +----------------------+   +----+---+
+//   | +--------------------------+                                  |
+//   | |                               +-----------------------------+
+//   | |                               |
+//   | | +-----------------------------v-------------------------------+
+//   | | |                                                             |
+//   | | |                      ColumnFamilySet                        |
+//   | | |                                                             |
+//   | | +-------------+--------------------------+----------------+---+
+//   | |               |                          |                |
+//   | +-------------------------------------+    |                |
+//   |                 |                     |    |                v
+//   |   +-------------v-------------+ +-----v----v---------+
+//   |   |                           | |                    |
+//   |   |     ColumnFamilyData 1    | | ColumnFamilyData 2 |    ......
+//   |   |                           | |                    |
+//   +--->                           | |                    |
+//       |                 +---------+ |                    |
+//       |                 | MemTable| |                    |
+//       |                 |  List   | |                    |
+//       +--------+---+--+-+----+----+ +--------------------++
+//                |   |  |      |
+//                |   |  |      |
+//                |   |  |      +-----------------------+
+//                |   |  +-----------+                  |
+//                v   +--------+     |                  |
+//       +--------+--------+   |     |                  |
+//       |                 |   |     |       +----------v----------+
+// +---> |SuperVersion 1.a +----------------->                     |
+//       |                 +------+  |       | MemTableListVersion |
+//       +---+-------------+   |  |  |       |                     |
+//           |                 |  |  |       +----+------------+---+
+//           |      current    |  |  |            |            |
+//           |   +-------------+  |  |mem         |            |
+//           |   |                |  |            |            |
+//         +-v---v-------+    +---v--v---+  +-----v----+  +----v-----+
+//         |             |    |          |  |          |  |          |
+//         | Version 1.a |    | memtable |  | memtable |  | memtable |
+//         |             |    |   1.a    |  |   1.b    |  |   1.c    |
+//         +-------------+    |          |  |          |  |          |
+//                            +----------+  +----------+  +----------+
+//
+// DBImpl keeps a ColumnFamilySet, which references to all column families by
+// pointing to respective ColumnFamilyData object of each column family.
+// This is how DBImpl can list and operate on all the column families.
+// ColumnFamilyHandle also points to ColumnFamilyData directly, so that
+// when a user executes a query, it can directly find memtables and Version
+// as well as SuperVersion to the column family, without going through
+// ColumnFamilySet.
+//
+// ColumnFamilySet points to the latest view of the LSM-tree (list of memtables
+// and SST files) indirectly, while ongoing operations may hold references
+// to a current or an out-of-date SuperVersion, which in turn points to a
+// point-in-time view of the LSM-tree. This guarantees the memtables and SST
+// files being operated on will not go away, until the SuperVersion is
+// unreferenced to 0 and destoryed.
+//
+// The following graph illustrates a possible referencing relationships:
+//
+// Column       +--------------+      current       +-----------+
+// Family +---->+              +------------------->+           |
+//  Data        | SuperVersion +----------+         | Version A |
+//              |      3       |   imm    |         |           |
+// Iter2 +----->+              |  +-------v------+  +-----------+
+//              +-----+--------+  | MemtableList +----------------> Empty
+//                    |           |   Version r  |  +-----------+
+//                    |           +--------------+  |           |
+//                    +------------------+   current| Version B |
+//              +--------------+         |   +----->+           |
+//              |              |         |   |      +-----+-----+
+// Compaction +>+ SuperVersion +-------------+            ^
+//    Job       |      2       +------+  |                |current
+//              |              +----+ |  |     mem        |    +------------+
+//              +--------------+    | |  +--------------------->            |
+//                                  | +------------------------> MemTable a |
+//                                  |          mem        |    |            |
+//              +--------------+    |                     |    +------------+
+//              |              +--------------------------+
+//  Iter1 +-----> SuperVersion |    |                          +------------+
+//              |      1       +------------------------------>+            |
+//              |              +-+  |        mem               | MemTable b |
+//              +--------------+ |  |                          |            |
+//                               |  |    +--------------+      +-----^------+
+//                               |  |imm | MemtableList |            |
+//                               |  +--->+   Version s  +------------+
+//                               |       +--------------+
+//                               |       +--------------+
+//                               |       | MemtableList |
+//                               +------>+   Version t  +-------->  Empty
+//                                 imm   +--------------+
+//
+// In this example, even if the current LSM-tree consists of Version A and
+// memtable a, which is also referenced by SuperVersion, two older SuperVersion
+// SuperVersion2 and Superversion1 still exist, and are referenced by a
+// compaction job and an old iterator Iter1, respectively. SuperVersion2
+// contains Version B, memtable a and memtable b; SuperVersion1 contains
+// Version B and memtable b (mutable). As a result, Version B and memtable b
+// are prevented from being destroyed or deleted.

 // ColumnFamilyHandleImpl is the class that clients use to access different
 // column families. It has non-trivial destructor, which gets called when client
@@ -168,7 +275,7 @@ class ColumnFamilyData {
  // Ref() can only be called from a context where the caller can guarantee
  // that ColumnFamilyData is alive (while holding a non-zero ref already,
  // holding a DB mutex, or as the leader in a write batch group).
-  void Ref() { refs_.fetch_add(1, std::memory_order_relaxed); }
+  void Ref() { refs_.fetch_add(1); }

  // Unref decreases the reference count, but does not handle deletion
  // when the count goes to 0.  If this method returns true then the
@@ -176,7 +283,7 @@ class ColumnFamilyData {
  // FreeDeadColumnFamilies().  Unref() can only be called while holding
  // a DB mutex, or during single-threaded recovery.
  bool Unref() {
-    int old_refs = refs_.fetch_sub(1, std::memory_order_relaxed);
+    int old_refs = refs_.fetch_sub(1);
    assert(old_refs > 0);
    return old_refs == 1;
  }
@@ -232,9 +339,13 @@ class ColumnFamilyData {

  bool is_delete_range_supported() { return is_delete_range_supported_; }

+  // Validate CF options against DB options
+  static Status ValidateOptions(const DBOptions& db_options,
+                                const ColumnFamilyOptions& cf_options);
 #ifndef ROCKSDB_LITE
  // REQUIRES: DB mutex held
  Status SetOptions(
+      const DBOptions& db_options,
      const std::unordered_map<std::string, std::string>& options_map);
 #endif  // ROCKSDB_LITE

@@ -394,7 +505,8 @@ class ColumnFamilyData {
                   const ColumnFamilyOptions& options,
                   const ImmutableDBOptions& db_options,
                   const EnvOptions& env_options,
-                   ColumnFamilySet* column_family_set);
+                   ColumnFamilySet* column_family_set,
+                   BlockCacheTracer* const block_cache_tracer);

  uint32_t id_;
  const std::string name_;
@@ -522,7 +634,8 @@ class ColumnFamilySet {
                  const ImmutableDBOptions* db_options,
                  const EnvOptions& env_options, Cache* table_cache,
                  WriteBufferManager* write_buffer_manager,
-                  WriteController* write_controller);
+                  WriteController* write_controller,
+                  BlockCacheTracer* const block_cache_tracer);
  ~ColumnFamilySet();

  ColumnFamilyData* GetDefault() const;
@@ -581,6 +694,7 @@ class ColumnFamilySet {
  Cache* table_cache_;
  WriteBufferManager* write_buffer_manager_;
  WriteController* write_controller_;
+  BlockCacheTracer* const block_cache_tracer_;
 };

 // We use ColumnFamilyMemTablesImpl to provide WriteBatch a way to access
@@ -12,20 +12,22 @@
 #include <string>
 #include <thread>

-#include "db/db_impl.h"
+#include "db/db_impl/db_impl.h"
 #include "db/db_test_util.h"
 #include "memtable/hash_skiplist_rep.h"
 #include "options/options_parser.h"
 #include "port/port.h"
+#include "port/stack_trace.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
 #include "rocksdb/iterator.h"
+#include "rocksdb/utilities/object_registry.h"
+#include "test_util/fault_injection_test_env.h"
+#include "test_util/sync_point.h"
+#include "test_util/testharness.h"
+#include "test_util/testutil.h"
 #include "util/coding.h"
-#include "util/fault_injection_test_env.h"
 #include "util/string_util.h"
-#include "util/sync_point.h"
-#include "util/testharness.h"
-#include "util/testutil.h"
 #include "utilities/merge_operators.h"

 namespace rocksdb {
@@ -60,8 +62,20 @@ class EnvCounter : public EnvWrapper {

 class ColumnFamilyTestBase : public testing::Test {
 public:
-  ColumnFamilyTestBase(uint32_t format) : rnd_(139), format_(format) {
-    env_ = new EnvCounter(Env::Default());
+  explicit ColumnFamilyTestBase(uint32_t format) : rnd_(139), format_(format) {
+    Env* base_env = Env::Default();
+#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();
+      EXPECT_OK(s);
+      EXPECT_NE(Env::Default(), base_env);
+    }
+#endif  // !ROCKSDB_LITE
+    EXPECT_NE(nullptr, base_env);
+    env_ = new EnvCounter(base_env);
    dbname_ = test::PerThreadDBPath("column_family_test");
    db_options_.create_if_missing = true;
    db_options_.fail_if_options_file_error = true;
@@ -532,6 +546,7 @@ class ColumnFamilyTestBase : public testing::Test {
  std::string dbname_;
  DB* db_ = nullptr;
  EnvCounter* env_;
+  std::shared_ptr<Env> env_guard_;
  Random rnd_;
  uint32_t format_;
 };
@@ -1117,22 +1132,25 @@ TEST_P(ColumnFamilyTest, DifferentWriteBufferSizes) {
  default_cf.arena_block_size = 4 * 4096;
  default_cf.max_write_buffer_number = 10;
  default_cf.min_write_buffer_number_to_merge = 1;
-  default_cf.max_write_buffer_number_to_maintain = 0;
+  default_cf.max_write_buffer_size_to_maintain = 0;
  one.write_buffer_size = 200000;
  one.arena_block_size = 4 * 4096;
  one.max_write_buffer_number = 10;
  one.min_write_buffer_number_to_merge = 2;
-  one.max_write_buffer_number_to_maintain = 1;
+  one.max_write_buffer_size_to_maintain =
+      static_cast<int>(one.write_buffer_size);
  two.write_buffer_size = 1000000;
  two.arena_block_size = 4 * 4096;
  two.max_write_buffer_number = 10;
  two.min_write_buffer_number_to_merge = 3;
-  two.max_write_buffer_number_to_maintain = 2;
+  two.max_write_buffer_size_to_maintain =
+      static_cast<int>(two.write_buffer_size);
  three.write_buffer_size = 4096 * 22;
  three.arena_block_size = 4096;
  three.max_write_buffer_number = 10;
  three.min_write_buffer_number_to_merge = 4;
-  three.max_write_buffer_number_to_maintain = -1;
+  three.max_write_buffer_size_to_maintain =
+      static_cast<int>(three.write_buffer_size);

  Reopen({default_cf, one, two, three});

@@ -3312,7 +3330,17 @@ TEST_P(ColumnFamilyTest, MultipleCFPathsTest) {

 }  // namespace rocksdb

+#ifdef ROCKSDB_UNITTESTS_WITH_CUSTOM_OBJECTS_FROM_STATIC_LIBS
+extern "C" {
+void RegisterCustomObjects(int argc, char** argv);
+}
+#else
+void RegisterCustomObjects(int /*argc*/, char** /*argv*/) {}
+#endif  // !ROCKSDB_UNITTESTS_WITH_CUSTOM_OBJECTS_FROM_STATIC_LIBS
+
 int main(int argc, char** argv) {
+  rocksdb::port::InstallStackTraceHandler();
  ::testing::InitGoogleTest(&argc, argv);
+  RegisterCustomObjects(argc, argv);
  return RUN_ALL_TESTS();
 }
@@ -10,13 +10,13 @@
 #include <thread>
 #include <vector>

-#include "db/db_impl.h"
+#include "db/db_impl/db_impl.h"
 #include "port/port.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
+#include "test_util/sync_point.h"
+#include "test_util/testharness.h"
 #include "util/string_util.h"
-#include "util/sync_point.h"
-#include "util/testharness.h"

 namespace rocksdb {

@@ -5,7 +5,7 @@

 #ifndef ROCKSDB_LITE
 #include "db/compacted_db_impl.h"
-#include "db/db_impl.h"
+#include "db/db_impl/db_impl.h"
 #include "db/version_set.h"
 #include "table/get_context.h"

@@ -37,7 +37,7 @@ Status CompactedDBImpl::Get(const ReadOptions& options, ColumnFamilyHandle*,
                            const Slice& key, PinnableSlice* value) {
  GetContext get_context(user_comparator_, nullptr, nullptr, nullptr,
                         GetContext::kNotFound, key, value, nullptr, nullptr,
-                         nullptr, nullptr);
+                         true, nullptr, nullptr);
  LookupKey lkey(key, kMaxSequenceNumber);
  files_.files[FindFile(key)].fd.table_reader->Get(options, lkey.internal_key(),
                                                   &get_context, nullptr);
@@ -70,7 +70,7 @@ std::vector<Status> CompactedDBImpl::MultiGet(const ReadOptions& options,
      std::string& value = (*values)[idx];
      GetContext get_context(user_comparator_, nullptr, nullptr, nullptr,
                             GetContext::kNotFound, keys[idx], &pinnable_val,
-                             nullptr, nullptr, nullptr, nullptr);
+                             nullptr, nullptr, true, nullptr, nullptr);
      LookupKey lkey(keys[idx], kMaxSequenceNumber);
      r->Get(options, lkey.internal_key(), &get_context, nullptr);
      value.assign(pinnable_val.data(), pinnable_val.size());
@@ -5,15 +5,19 @@

 #pragma once
 #ifndef ROCKSDB_LITE
-#include "db/db_impl.h"
-#include <vector>
 #include <string>
+#include <vector>
+#include "db/db_impl/db_impl.h"

 namespace rocksdb {

 class CompactedDBImpl : public DBImpl {
 public:
  CompactedDBImpl(const DBOptions& options, const std::string& dbname);
+  // No copying allowed
+  CompactedDBImpl(const CompactedDBImpl&) = delete;
+  void operator=(const CompactedDBImpl&) = delete;
+
  virtual ~CompactedDBImpl();

  static Status Open(const Options& options, const std::string& dbname,
@@ -85,6 +89,15 @@ class CompactedDBImpl : public DBImpl {
      const IngestExternalFileOptions& /*ingestion_options*/) override {
    return Status::NotSupported("Not supported in compacted db mode.");
  }
+  using DB::CreateColumnFamilyWithImport;
+  virtual Status CreateColumnFamilyWithImport(
+      const ColumnFamilyOptions& /*options*/,
+      const std::string& /*column_family_name*/,
+      const ImportColumnFamilyOptions& /*import_options*/,
+      const ExportImportFilesMetaData& /*metadata*/,
+      ColumnFamilyHandle** /*handle*/) override {
+    return Status::NotSupported("Not supported in compacted db mode.");
+  }

 private:
  friend class DB;
@@ -95,10 +108,6 @@ class CompactedDBImpl : public DBImpl {
  Version* version_;
  const Comparator* user_comparator_;
  LevelFilesBrief files_;
-
-  // No copying allowed
-  CompactedDBImpl(const CompactedDBImpl&);
-  void operator=(const CompactedDBImpl&);
 };
 }
 #endif  // ROCKSDB_LITE
@@ -7,19 +7,14 @@
 // 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 "db/compaction.h"
-
-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-
-#include <inttypes.h>
+#include <cinttypes>
 #include <vector>

 #include "db/column_family.h"
+#include "db/compaction/compaction.h"
 #include "rocksdb/compaction_filter.h"
+#include "test_util/sync_point.h"
 #include "util/string_util.h"
-#include "util/sync_point.h"

 namespace rocksdb {

@@ -9,11 +9,13 @@

 #pragma once
 #include "db/version_set.h"
+#include "memory/arena.h"
 #include "options/cf_options.h"
-#include "util/arena.h"
 #include "util/autovector.h"

 namespace rocksdb {
+// The file contains class Compaction, as well as some helper functions
+// and data structures used by the class.

 // Utility for comparing sstable boundary keys. Returns -1 if either a or b is
 // null which provides the property that a==null indicates a key that is less
@@ -63,7 +65,7 @@ class ColumnFamilyData;
 class VersionStorageInfo;
 class CompactionFilter;

-// A Compaction encapsulates information about a compaction.
+// A Compaction encapsulates metadata about a compaction.
 class Compaction {
 public:
  Compaction(VersionStorageInfo* input_version,
@@ -376,7 +378,7 @@ class Compaction {
  CompactionReason compaction_reason_;
 };

-// Utility function
+// Return sum of sizes of all files in `files`.
 extern uint64_t TotalFileSize(const std::vector<FileMetaData*>& files);

 }  // namespace rocksdb
@@ -3,13 +3,12 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).

-#include "db/compaction_iterator.h"
-
+#include "db/compaction/compaction_iterator.h"
 #include "db/snapshot_checker.h"
 #include "port/likely.h"
 #include "rocksdb/listener.h"
 #include "table/internal_iterator.h"
-#include "util/sync_point.h"
+#include "test_util/sync_point.h"

 #define DEFINITELY_IN_SNAPSHOT(seq, snapshot)                       \
  ((seq) <= (snapshot) &&                                           \
@@ -493,20 +492,6 @@ void CompactionIterator::NextFromInput() {
      // in this snapshot.
      assert(last_sequence >= current_user_key_sequence_);

-      // Note2: if last_snapshot < current_user_key_snapshot, it can only
-      // mean last_snapshot is released between we process last value and
-      // this value, and findEarliestVisibleSnapshot returns the next snapshot
-      // as current_user_key_snapshot. In this case last value and current
-      // value are both in current_user_key_snapshot currently.
-      // Although last_snapshot is released we might still get a definitive
-      // response when key sequence number changes, e.g., when seq is determined
-      // too old and visible in all snapshots.
-      assert(last_snapshot == current_user_key_snapshot_ ||
-             (snapshot_checker_ != nullptr &&
-              snapshot_checker_->CheckInSnapshot(current_user_key_sequence_,
-                                                 last_snapshot) !=
-                  SnapshotCheckerResult::kNotInSnapshot));
-
      ++iter_stats_.num_record_drop_hidden;  // (A)
      input_->Next();
    } else if (compaction_ != nullptr && ikey_.type == kTypeDeletion &&
@@ -10,8 +10,8 @@
 #include <unordered_set>
 #include <vector>

-#include "db/compaction.h"
-#include "db/compaction_iteration_stats.h"
+#include "db/compaction/compaction.h"
+#include "db/compaction/compaction_iteration_stats.h"
 #include "db/merge_helper.h"
 #include "db/pinned_iterators_manager.h"
 #include "db/range_del_aggregator.h"
@@ -39,6 +39,7 @@ class SnapshotListFetchCallback {
  virtual void Refresh(std::vector<SequenceNumber>* snapshots,
                       SequenceNumber max) = 0;
  inline bool TimeToRefresh(const size_t key_index) {
+    assert(snap_refresh_nanos_ != 0);
    // skip the key if key_index % every_nth_key (which is of power 2) is not 0.
    if ((key_index & every_nth_key_minus_one_) != 0) {
      return false;
@@ -3,15 +3,15 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).

-#include "db/compaction_iterator.h"

 #include <string>
 #include <vector>

+#include "db/compaction/compaction_iterator.h"
 #include "port/port.h"
+#include "test_util/testharness.h"
+#include "test_util/testutil.h"
 #include "util/string_util.h"
-#include "util/testharness.h"
-#include "util/testutil.h"
 #include "utilities/merge_operators.h"

 namespace rocksdb {
@@ -7,13 +7,7 @@
 // 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 "db/compaction_job.h"
-
-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-
-#include <inttypes.h>
+#include <cinttypes>
 #include <algorithm>
 #include <functional>
 #include <list>
@@ -25,7 +19,8 @@
 #include <vector>

 #include "db/builder.h"
-#include "db/db_impl.h"
+#include "db/compaction/compaction_job.h"
+#include "db/db_impl/db_impl.h"
 #include "db/db_iter.h"
 #include "db/dbformat.h"
 #include "db/error_handler.h"
@@ -38,6 +33,10 @@
 #include "db/merge_helper.h"
 #include "db/range_del_aggregator.h"
 #include "db/version_set.h"
+#include "file/filename.h"
+#include "file/sst_file_manager_impl.h"
+#include "logging/log_buffer.h"
+#include "logging/logging.h"
 #include "monitoring/iostats_context_imp.h"
 #include "monitoring/perf_context_imp.h"
 #include "monitoring/thread_status_util.h"
@@ -47,21 +46,17 @@
 #include "rocksdb/statistics.h"
 #include "rocksdb/status.h"
 #include "rocksdb/table.h"
-#include "table/block.h"
-#include "table/block_based_table_factory.h"
+#include "table/block_based/block.h"
+#include "table/block_based/block_based_table_factory.h"
 #include "table/merging_iterator.h"
 #include "table/table_builder.h"
+#include "test_util/sync_point.h"
 #include "util/coding.h"
 #include "util/file_reader_writer.h"
-#include "util/filename.h"
-#include "util/log_buffer.h"
-#include "util/logging.h"
 #include "util/mutexlock.h"
 #include "util/random.h"
-#include "util/sst_file_manager_impl.h"
 #include "util/stop_watch.h"
 #include "util/string_util.h"
-#include "util/sync_point.h"

 namespace rocksdb {

@@ -324,7 +319,7 @@ CompactionJob::CompactionJob(
      db_options_(db_options),
      env_options_(env_options),
      env_(db_options.env),
-      env_optiosn_for_read_(
+      env_options_for_read_(
          env_->OptimizeForCompactionTableRead(env_options, db_options_)),
      versions_(versions),
      shutting_down_(shutting_down),
@@ -415,7 +410,6 @@ void CompactionJob::Prepare() {

  write_hint_ =
      c->column_family_data()->CalculateSSTWriteHint(c->output_level());
-  // Is this compaction producing files at the bottommost level?
  bottommost_level_ = c->bottommost_level();

  if (c->ShouldFormSubcompactions()) {
@@ -445,11 +439,6 @@ struct RangeWithSize {
      : range(a, b), size(s) {}
 };

-// Generates a histogram representing potential divisions of key ranges from
-// the input. It adds the starting and/or ending keys of certain input files
-// to the working set and then finds the approximate size of data in between
-// each consecutive pair of slices. Then it divides these ranges into
-// consecutive groups such that each group has a similar size.
 void CompactionJob::GenSubcompactionBoundaries() {
  auto* c = compact_->compaction;
  auto* cfd = c->column_family_data();
@@ -519,7 +508,7 @@ void CompactionJob::GenSubcompactionBoundaries() {
  auto* v = compact_->compaction->input_version();
  for (auto it = bounds.begin();;) {
    const Slice a = *it;
-    it++;
+    ++it;

    if (it == bounds.end()) {
      break;
@@ -531,7 +520,9 @@ void CompactionJob::GenSubcompactionBoundaries() {
    // to the index block and may incur I/O cost in the process. Unlock db
    // mutex to reduce contention
    db_mutex_->Unlock();
-    uint64_t size = versions_->ApproximateSize(v, a, b, start_lvl, out_lvl + 1);
+    uint64_t size = versions_->ApproximateSize(SizeApproximationOptions(), v, a,
+                                               b, start_lvl, out_lvl + 1,
+                                               TableReaderCaller::kCompaction);
    db_mutex_->Lock();
    ranges.emplace_back(a, b, size);
    sum += size;
@@ -656,12 +647,14 @@ Status CompactionJob::Run() {
        // to cache it here for further user reads
        InternalIterator* iter = cfd->table_cache()->NewIterator(
            ReadOptions(), env_options_, cfd->internal_comparator(),
-            *files_meta[file_idx], nullptr /* range_del_agg */,
-            prefix_extractor, nullptr,
+            *files_meta[file_idx], /*range_del_agg=*/nullptr, prefix_extractor,
+            /*table_reader_ptr=*/nullptr,
            cfd->internal_stats()->GetFileReadHist(
                compact_->compaction->output_level()),
-            false, nullptr /* arena */, false /* skip_filters */,
-            compact_->compaction->output_level());
+            TableReaderCaller::kCompactionRefill, /*arena=*/nullptr,
+            /*skip_filters=*/false, compact_->compaction->output_level(),
+            /*smallest_compaction_key=*/nullptr,
+            /*largest_compaction_key=*/nullptr);
        auto s = iter->status();

        if (s.ok() && paranoid_file_checks_) {
@@ -842,7 +835,7 @@ 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_optiosn_for_read_));
+      sub_compact->compaction, &range_del_agg, env_options_for_read_));

  AutoThreadOperationStageUpdater stage_updater(
      ThreadStatus::STAGE_COMPACTION_PROCESS_KV);
@@ -893,7 +886,9 @@ 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_, snap_list_callback_));
+      shutting_down_, preserve_deletes_seqnum_,
+      // Currently range_del_agg is incompatible with snapshot refresh feature.
+      range_del_agg.IsEmpty() ? snap_list_callback_ : nullptr));
  auto c_iter = sub_compact->c_iter.get();
  c_iter->SeekToFirst();
  if (c_iter->Valid() && sub_compact->compaction->output_level() != 0) {
@@ -1002,10 +997,13 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
  RecordDroppedKeys(c_iter_stats, &sub_compact->compaction_job_stats);
  RecordCompactionIOStats();

-  if (status.ok() &&
-      (shutting_down_->load(std::memory_order_relaxed) || cfd->IsDropped())) {
-    status = Status::ShutdownInProgress(
-        "Database shutdown or Column family drop during compaction");
+  if (status.ok() && cfd->IsDropped()) {
+    status =
+        Status::ColumnFamilyDropped("Column family dropped during compaction");
+  }
+  if ((status.ok() || status.IsColumnFamilyDropped()) &&
+      shutting_down_->load(std::memory_order_relaxed)) {
+    status = Status::ShutdownInProgress("Database shutdown");
  }
  if (status.ok()) {
    status = input->status();
@@ -18,7 +18,7 @@
 #include <vector>

 #include "db/column_family.h"
-#include "db/compaction_iterator.h"
+#include "db/compaction/compaction_iterator.h"
 #include "db/dbformat.h"
 #include "db/flush_scheduler.h"
 #include "db/internal_stats.h"
@@ -29,6 +29,7 @@
 #include "db/version_edit.h"
 #include "db/write_controller.h"
 #include "db/write_thread.h"
+#include "logging/event_logger.h"
 #include "options/cf_options.h"
 #include "options/db_options.h"
 #include "port/port.h"
@@ -40,7 +41,6 @@
 #include "rocksdb/transaction_log.h"
 #include "table/scoped_arena_iterator.h"
 #include "util/autovector.h"
-#include "util/event_logger.h"
 #include "util/stop_watch.h"
 #include "util/thread_local.h"

@@ -55,6 +55,11 @@ class Version;
 class VersionEdit;
 class VersionSet;

+// CompactionJob is responsible for executing the compaction. Each (manual or
+// automated) compaction corresponds to a CompactionJob object, and usually
+// goes through the stages of `Prepare()`->`Run()`->`Install()`. CompactionJob
+// will divide the compaction into subcompactions and execute them in parallel
+// if needed.
 class CompactionJob {
 public:
  CompactionJob(
@@ -80,17 +85,28 @@ class CompactionJob {
  CompactionJob& operator=(const CompactionJob& job) = delete;

  // REQUIRED: mutex held
+  // Prepare for the compaction by setting up boundaries for each subcompaction
  void Prepare();
  // REQUIRED mutex not held
+  // Launch threads for each subcompaction and wait for them to finish. After
+  // that, verify table is usable and finally do bookkeeping to unify
+  // subcompaction results
  Status Run();

  // REQUIRED: mutex held
+  // Add compaction input/output to the current version
  Status Install(const MutableCFOptions& mutable_cf_options);

 private:
  struct SubcompactionState;

  void AggregateStatistics();
+
+  // Generates a histogram representing potential divisions of key ranges from
+  // the input. It adds the starting and/or ending keys of certain input files
+  // to the working set and then finds the approximate size of data in between
+  // each consecutive pair of slices. Then it divides these ranges into
+  // consecutive groups such that each group has a similar size.
  void GenSubcompactionBoundaries();

  // update the thread status for starting a compaction.
@@ -135,7 +151,7 @@ class CompactionJob {

  Env* env_;
  // env_option optimized for compaction table reads
-  EnvOptions env_optiosn_for_read_;
+  EnvOptions env_options_for_read_;
  VersionSet* versions_;
  const std::atomic<bool>* shutting_down_;
  const SequenceNumber preserve_deletes_seqnum_;
@@ -163,6 +179,7 @@ class CompactionJob {

  EventLogger* event_logger_;

+  // Is this compaction creating a file in the bottom most level?
  bool bottommost_level_;
  bool paranoid_file_checks_;
  bool measure_io_stats_;
@@ -7,11 +7,7 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-
-#include <inttypes.h>
+#include <cinttypes>
 #include <algorithm>
 #include <iostream>
 #include <mutex>
@@ -21,12 +17,14 @@
 #include <unordered_set>
 #include <utility>

-#include "db/db_impl.h"
+#include "db/db_impl/db_impl.h"
 #include "db/dbformat.h"
 #include "db/job_context.h"
 #include "db/version_set.h"
 #include "db/write_batch_internal.h"
 #include "env/mock_env.h"
+#include "file/filename.h"
+#include "logging/logging.h"
 #include "memtable/hash_linklist_rep.h"
 #include "monitoring/statistics.h"
 #include "monitoring/thread_status_util.h"
@@ -47,20 +45,18 @@
 #include "rocksdb/thread_status.h"
 #include "rocksdb/utilities/checkpoint.h"
 #include "rocksdb/utilities/write_batch_with_index.h"
-#include "table/block_based_table_factory.h"
+#include "table/block_based/block_based_table_factory.h"
 #include "table/mock_table.h"
-#include "table/plain_table_factory.h"
+#include "table/plain/plain_table_factory.h"
 #include "table/scoped_arena_iterator.h"
+#include "test_util/sync_point.h"
+#include "test_util/testharness.h"
+#include "test_util/testutil.h"
 #include "util/compression.h"
-#include "util/filename.h"
 #include "util/hash.h"
-#include "util/logging.h"
 #include "util/mutexlock.h"
 #include "util/rate_limiter.h"
 #include "util/string_util.h"
-#include "util/sync_point.h"
-#include "util/testharness.h"
-#include "util/testutil.h"
 #include "utilities/merge_operators.h"

 #if !defined(IOS_CROSS_COMPILE)
@@ -5,11 +5,7 @@

 #ifndef ROCKSDB_LITE

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-
-#include <inttypes.h>
+#include <cinttypes>
 #include <algorithm>
 #include <array>
 #include <map>
@@ -17,7 +13,8 @@
 #include <tuple>

 #include "db/column_family.h"
-#include "db/compaction_job.h"
+#include "db/compaction/compaction_job.h"
+#include "db/db_impl/db_impl.h"
 #include "db/error_handler.h"
 #include "db/version_set.h"
 #include "rocksdb/cache.h"
@@ -25,10 +22,10 @@
 #include "rocksdb/options.h"
 #include "rocksdb/write_buffer_manager.h"
 #include "table/mock_table.h"
+#include "test_util/testharness.h"
+#include "test_util/testutil.h"
 #include "util/file_reader_writer.h"
 #include "util/string_util.h"
-#include "util/testharness.h"
-#include "util/testutil.h"
 #include "utilities/merge_operators.h"

 namespace rocksdb {
@@ -81,7 +78,8 @@ class CompactionJobTest : public testing::Test {
        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_)),
+                                 &write_controller_,
+                                 /*block_cache_tracer=*/nullptr)),
        shutting_down_(false),
        preserve_deletes_seqnum_(0),
        mock_table_factory_(new mock::MockTableFactory()),
@@ -204,10 +202,18 @@ class CompactionJobTest : public testing::Test {
    EXPECT_OK(env_->CreateDirIfMissing(dbname_));
    versions_.reset(new VersionSet(dbname_, &db_options_, env_options_,
                                   table_cache_.get(), &write_buffer_manager_,
-                                   &write_controller_));
+                                   &write_controller_,
+                                   /*block_cache_tracer=*/nullptr));
    compaction_job_stats_.Reset();
+    SetIdentityFile(env_, dbname_);

    VersionEdit new_db;
+    if (db_options_.write_dbid_to_manifest) {
+      DBImpl* impl = new DBImpl(DBOptions(), dbname_);
+      std::string db_id;
+      impl->GetDbIdentityFromIdentityFile(&db_id);
+      new_db.SetDBId(db_id);
+    }
    new_db.SetLogNumber(0);
    new_db.SetNextFile(2);
    new_db.SetLastSequence(0);
@@ -966,7 +972,7 @@ TEST_F(CompactionJobTest, SnapshotRefresh) {
   public:
    SnapshotListFetchCallbackTest(Env* env, Random64& rand,
                                  std::vector<SequenceNumber>* snapshots)
-        : SnapshotListFetchCallback(env, 0 /*no time delay*/,
+        : SnapshotListFetchCallback(env, 1 /*short time delay*/,
                                    1 /*fetch after each key*/),
          rand_(rand),
          snapshots_(snapshots) {}
@@ -7,25 +7,21 @@
 // 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 "db/compaction_picker.h"
+#include "db/compaction/compaction_picker.h"

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-
-#include <inttypes.h>
+#include <cinttypes>
 #include <limits>
 #include <queue>
 #include <string>
 #include <utility>
 #include <vector>
 #include "db/column_family.h"
+#include "file/filename.h"
+#include "logging/log_buffer.h"
 #include "monitoring/statistics.h"
-#include "util/filename.h"
-#include "util/log_buffer.h"
+#include "test_util/sync_point.h"
 #include "util/random.h"
 #include "util/string_util.h"
-#include "util/sync_point.h"

 namespace rocksdb {

@@ -42,19 +38,23 @@ uint64_t TotalCompensatedFileSize(const std::vector<FileMetaData*>& files) {
 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;
  size_t compact_bytes_per_del_file = port::kMaxSizet;
-  // compaction range will be [0, span_len).
+  // Compaction range will be [0, span_len).
  size_t span_len;
-  // pull in files until the amount of compaction work per deleted file begins
-  // increasing.
+  // 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 ||
-        new_compact_bytes_per_del_file > compact_bytes_per_del_file) {
+        new_compact_bytes_per_del_file > compact_bytes_per_del_file ||
+        compensated_compact_bytes > max_compaction_bytes) {
      break;
    }
    compact_bytes_per_del_file = new_compact_bytes_per_del_file;
@@ -1107,536 +1107,4 @@ bool CompactionPicker::GetOverlappingL0Files(
  return true;
 }

-bool LevelCompactionPicker::NeedsCompaction(
-    const VersionStorageInfo* vstorage) const {
-  if (!vstorage->ExpiredTtlFiles().empty()) {
-    return true;
-  }
-  if (!vstorage->FilesMarkedForPeriodicCompaction().empty()) {
-    return true;
-  }
-  if (!vstorage->BottommostFilesMarkedForCompaction().empty()) {
-    return true;
-  }
-  if (!vstorage->FilesMarkedForCompaction().empty()) {
-    return true;
-  }
-  for (int i = 0; i <= vstorage->MaxInputLevel(); i++) {
-    if (vstorage->CompactionScore(i) >= 1) {
-      return true;
-    }
-  }
-  return false;
-}
-
-namespace {
-// A class to build a leveled compaction step-by-step.
-class LevelCompactionBuilder {
- public:
-  LevelCompactionBuilder(const std::string& cf_name,
-                         VersionStorageInfo* vstorage,
-                         CompactionPicker* compaction_picker,
-                         LogBuffer* log_buffer,
-                         const MutableCFOptions& mutable_cf_options,
-                         const ImmutableCFOptions& ioptions)
-      : cf_name_(cf_name),
-        vstorage_(vstorage),
-        compaction_picker_(compaction_picker),
-        log_buffer_(log_buffer),
-        mutable_cf_options_(mutable_cf_options),
-        ioptions_(ioptions) {}
-
-  // Pick and return a compaction.
-  Compaction* PickCompaction();
-
-  // Pick the initial files to compact to the next level. (or together
-  // in Intra-L0 compactions)
-  void SetupInitialFiles();
-
-  // If the initial files are from L0 level, pick other L0
-  // files if needed.
-  bool SetupOtherL0FilesIfNeeded();
-
-  // Based on initial files, setup other files need to be compacted
-  // in this compaction, accordingly.
-  bool SetupOtherInputsIfNeeded();
-
-  Compaction* GetCompaction();
-
-  // For the specfied level, pick a file that we want to compact.
-  // Returns false if there is no file to compact.
-  // If it returns true, inputs->files.size() will be exactly one.
-  // If level is 0 and there is already a compaction on that level, this
-  // function will return false.
-  bool PickFileToCompact();
-
-  // For L0->L0, picks the longest span of files that aren't currently
-  // undergoing compaction for which work-per-deleted-file decreases. The span
-  // always starts from the newest L0 file.
-  //
-  // Intra-L0 compaction is independent of all other files, so it can be
-  // performed even when L0->base_level compactions are blocked.
-  //
-  // Returns true if `inputs` is populated with a span of files to be compacted;
-  // otherwise, returns false.
-  bool PickIntraL0Compaction();
-
-  void PickExpiredTtlFiles();
-
-  void PickFilesMarkedForPeriodicCompaction();
-
-  const std::string& cf_name_;
-  VersionStorageInfo* vstorage_;
-  CompactionPicker* compaction_picker_;
-  LogBuffer* log_buffer_;
-  int start_level_ = -1;
-  int output_level_ = -1;
-  int parent_index_ = -1;
-  int base_index_ = -1;
-  double start_level_score_ = 0;
-  bool is_manual_ = false;
-  CompactionInputFiles start_level_inputs_;
-  std::vector<CompactionInputFiles> compaction_inputs_;
-  CompactionInputFiles output_level_inputs_;
-  std::vector<FileMetaData*> grandparents_;
-  CompactionReason compaction_reason_ = CompactionReason::kUnknown;
-
-  const MutableCFOptions& mutable_cf_options_;
-  const ImmutableCFOptions& ioptions_;
-  // Pick a path ID to place a newly generated file, with its level
-  static uint32_t GetPathId(const ImmutableCFOptions& ioptions,
-                            const MutableCFOptions& mutable_cf_options,
-                            int level);
-
-  static const int kMinFilesForIntraL0Compaction = 4;
-};
-
-void LevelCompactionBuilder::PickExpiredTtlFiles() {
-  if (vstorage_->ExpiredTtlFiles().empty()) {
-    return;
-  }
-
-  auto continuation = [&](std::pair<int, FileMetaData*> level_file) {
-    // If it's being compacted it has nothing to do here.
-    // If this assert() fails that means that some function marked some
-    // files as being_compacted, but didn't call ComputeCompactionScore()
-    assert(!level_file.second->being_compacted);
-    start_level_ = level_file.first;
-    output_level_ =
-        (start_level_ == 0) ? vstorage_->base_level() : start_level_ + 1;
-
-    if ((start_level_ == vstorage_->num_non_empty_levels() - 1) ||
-        (start_level_ == 0 &&
-         !compaction_picker_->level0_compactions_in_progress()->empty())) {
-      return false;
-    }
-
-    start_level_inputs_.files = {level_file.second};
-    start_level_inputs_.level = start_level_;
-    return compaction_picker_->ExpandInputsToCleanCut(cf_name_, vstorage_,
-                                                      &start_level_inputs_);
-  };
-
-  for (auto& level_file : vstorage_->ExpiredTtlFiles()) {
-    if (continuation(level_file)) {
-      // found the compaction!
-      return;
-    }
-  }
-
-  start_level_inputs_.files.clear();
-}
-
-void LevelCompactionBuilder::PickFilesMarkedForPeriodicCompaction() {
-  if (vstorage_->FilesMarkedForPeriodicCompaction().empty()) {
-    return;
-  }
-
-  auto continuation = [&](std::pair<int, FileMetaData*> level_file) {
-    // If it's being compacted it has nothing to do here.
-    // If this assert() fails that means that some function marked some
-    // files as being_compacted, but didn't call ComputeCompactionScore()
-    assert(!level_file.second->being_compacted);
-    output_level_ = start_level_ = level_file.first;
-
-    if (start_level_ == 0 &&
-        !compaction_picker_->level0_compactions_in_progress()->empty()) {
-      return false;
-    }
-
-    start_level_inputs_.files = {level_file.second};
-    start_level_inputs_.level = start_level_;
-    return compaction_picker_->ExpandInputsToCleanCut(cf_name_, vstorage_,
-                                                      &start_level_inputs_);
-  };
-
-  for (auto& level_file : vstorage_->FilesMarkedForPeriodicCompaction()) {
-    if (continuation(level_file)) {
-      // found the compaction!
-      return;
-    }
-  }
-
-  start_level_inputs_.files.clear();
-}
-
-void LevelCompactionBuilder::SetupInitialFiles() {
-  // Find the compactions by size on all levels.
-  bool skipped_l0_to_base = false;
-  for (int i = 0; i < compaction_picker_->NumberLevels() - 1; i++) {
-    start_level_score_ = vstorage_->CompactionScore(i);
-    start_level_ = vstorage_->CompactionScoreLevel(i);
-    assert(i == 0 || start_level_score_ <= vstorage_->CompactionScore(i - 1));
-    if (start_level_score_ >= 1) {
-      if (skipped_l0_to_base && start_level_ == vstorage_->base_level()) {
-        // If L0->base_level compaction is pending, don't schedule further
-        // compaction from base level. Otherwise L0->base_level compaction
-        // may starve.
-        continue;
-      }
-      output_level_ =
-          (start_level_ == 0) ? vstorage_->base_level() : start_level_ + 1;
-      if (PickFileToCompact()) {
-        // found the compaction!
-        if (start_level_ == 0) {
-          // L0 score = `num L0 files` / `level0_file_num_compaction_trigger`
-          compaction_reason_ = CompactionReason::kLevelL0FilesNum;
-        } else {
-          // L1+ score = `Level files size` / `MaxBytesForLevel`
-          compaction_reason_ = CompactionReason::kLevelMaxLevelSize;
-        }
-        break;
-      } else {
-        // didn't find the compaction, clear the inputs
-        start_level_inputs_.clear();
-        if (start_level_ == 0) {
-          skipped_l0_to_base = true;
-          // L0->base_level may be blocked due to ongoing L0->base_level
-          // compactions. It may also be blocked by an ongoing compaction from
-          // base_level downwards.
-          //
-          // In these cases, to reduce L0 file count and thus reduce likelihood
-          // of write stalls, we can attempt compacting a span of files within
-          // L0.
-          if (PickIntraL0Compaction()) {
-            output_level_ = 0;
-            compaction_reason_ = CompactionReason::kLevelL0FilesNum;
-            break;
-          }
-        }
-      }
-    }
-  }
-
-  // if we didn't find a compaction, check if there are any files marked for
-  // compaction
-  if (start_level_inputs_.empty()) {
-    parent_index_ = base_index_ = -1;
-
-    compaction_picker_->PickFilesMarkedForCompaction(
-        cf_name_, vstorage_, &start_level_, &output_level_, &start_level_inputs_);
-    if (!start_level_inputs_.empty()) {
-      is_manual_ = true;
-      compaction_reason_ = CompactionReason::kFilesMarkedForCompaction;
-      return;
-    }
-  }
-
-  // Bottommost Files Compaction on deleting tombstones
-  if (start_level_inputs_.empty()) {
-    size_t i;
-    for (i = 0; i < vstorage_->BottommostFilesMarkedForCompaction().size();
-         ++i) {
-      auto& level_and_file = vstorage_->BottommostFilesMarkedForCompaction()[i];
-      assert(!level_and_file.second->being_compacted);
-      start_level_inputs_.level = output_level_ = start_level_ =
-          level_and_file.first;
-      start_level_inputs_.files = {level_and_file.second};
-      if (compaction_picker_->ExpandInputsToCleanCut(cf_name_, vstorage_,
-                                                     &start_level_inputs_)) {
-        break;
-      }
-    }
-    if (i == vstorage_->BottommostFilesMarkedForCompaction().size()) {
-      start_level_inputs_.clear();
-    } else {
-      assert(!start_level_inputs_.empty());
-      compaction_reason_ = CompactionReason::kBottommostFiles;
-      return;
-    }
-  }
-
-  // TTL Compaction
-  if (start_level_inputs_.empty()) {
-    PickExpiredTtlFiles();
-    if (!start_level_inputs_.empty()) {
-      compaction_reason_ = CompactionReason::kTtl;
-      return;
-    }
-  }
-
-  // Periodic Compaction
-  if (start_level_inputs_.empty()) {
-    PickFilesMarkedForPeriodicCompaction();
-    if (!start_level_inputs_.empty()) {
-      compaction_reason_ = CompactionReason::kPeriodicCompaction;
-      return;
-    }
-  }
-}
-
-bool LevelCompactionBuilder::SetupOtherL0FilesIfNeeded() {
-  if (start_level_ == 0 && output_level_ != 0) {
-    return compaction_picker_->GetOverlappingL0Files(
-        vstorage_, &start_level_inputs_, output_level_, &parent_index_);
-  }
-  return true;
-}
-
-bool LevelCompactionBuilder::SetupOtherInputsIfNeeded() {
-  // Setup input files from output level. For output to L0, we only compact
-  // spans of files that do not interact with any pending compactions, so don't
-  // need to consider other levels.
-  if (output_level_ != 0) {
-    output_level_inputs_.level = output_level_;
-    if (!compaction_picker_->SetupOtherInputs(
-            cf_name_, mutable_cf_options_, vstorage_, &start_level_inputs_,
-            &output_level_inputs_, &parent_index_, base_index_)) {
-      return false;
-    }
-
-    compaction_inputs_.push_back(start_level_inputs_);
-    if (!output_level_inputs_.empty()) {
-      compaction_inputs_.push_back(output_level_inputs_);
-    }
-
-    // In some edge cases we could pick a compaction that will be compacting
-    // a key range that overlap with another running compaction, and both
-    // of them have the same output level. This could happen if
-    // (1) we are running a non-exclusive manual compaction
-    // (2) AddFile ingest a new file into the LSM tree
-    // We need to disallow this from happening.
-    if (compaction_picker_->FilesRangeOverlapWithCompaction(compaction_inputs_,
-                                                            output_level_)) {
-      // This compaction output could potentially conflict with the output
-      // of a currently running compaction, we cannot run it.
-      return false;
-    }
-    compaction_picker_->GetGrandparents(vstorage_, start_level_inputs_,
-                                        output_level_inputs_, &grandparents_);
-  } else {
-    compaction_inputs_.push_back(start_level_inputs_);
-  }
-  return true;
-}
-
-Compaction* LevelCompactionBuilder::PickCompaction() {
-  // Pick up the first file to start compaction. It may have been extended
-  // to a clean cut.
-  SetupInitialFiles();
-  if (start_level_inputs_.empty()) {
-    return nullptr;
-  }
-  assert(start_level_ >= 0 && output_level_ >= 0);
-
-  // If it is a L0 -> base level compaction, we need to set up other L0
-  // files if needed.
-  if (!SetupOtherL0FilesIfNeeded()) {
-    return nullptr;
-  }
-
-  // Pick files in the output level and expand more files in the start level
-  // if needed.
-  if (!SetupOtherInputsIfNeeded()) {
-    return nullptr;
-  }
-
-  // Form a compaction object containing the files we picked.
-  Compaction* c = GetCompaction();
-
-  TEST_SYNC_POINT_CALLBACK("LevelCompactionPicker::PickCompaction:Return", c);
-
-  return c;
-}
-
-Compaction* LevelCompactionBuilder::GetCompaction() {
-  auto c = new Compaction(
-      vstorage_, ioptions_, mutable_cf_options_, std::move(compaction_inputs_),
-      output_level_,
-      MaxFileSizeForLevel(mutable_cf_options_, output_level_,
-                          ioptions_.compaction_style, vstorage_->base_level(),
-                          ioptions_.level_compaction_dynamic_level_bytes),
-      mutable_cf_options_.max_compaction_bytes,
-      GetPathId(ioptions_, mutable_cf_options_, output_level_),
-      GetCompressionType(ioptions_, vstorage_, mutable_cf_options_,
-                         output_level_, vstorage_->base_level()),
-      GetCompressionOptions(ioptions_, vstorage_, output_level_),
-      /* max_subcompactions */ 0, std::move(grandparents_), is_manual_,
-      start_level_score_, false /* deletion_compaction */, compaction_reason_);
-
-  // If it's level 0 compaction, make sure we don't execute any other level 0
-  // compactions in parallel
-  compaction_picker_->RegisterCompaction(c);
-
-  // Creating a compaction influences the compaction score because the score
-  // takes running compactions into account (by skipping files that are already
-  // being compacted). Since we just changed compaction score, we recalculate it
-  // here
-  vstorage_->ComputeCompactionScore(ioptions_, mutable_cf_options_);
-  return c;
-}
-
-/*
- * Find the optimal path to place a file
- * Given a level, finds the path where levels up to it will fit in levels
- * up to and including this path
- */
-uint32_t LevelCompactionBuilder::GetPathId(
-    const ImmutableCFOptions& ioptions,
-    const MutableCFOptions& mutable_cf_options, int level) {
-  uint32_t p = 0;
-  assert(!ioptions.cf_paths.empty());
-
-  // size remaining in the most recent path
-  uint64_t current_path_size = ioptions.cf_paths[0].target_size;
-
-  uint64_t level_size;
-  int cur_level = 0;
-
-  // max_bytes_for_level_base denotes L1 size.
-  // We estimate L0 size to be the same as L1.
-  level_size = mutable_cf_options.max_bytes_for_level_base;
-
-  // Last path is the fallback
-  while (p < ioptions.cf_paths.size() - 1) {
-    if (level_size <= current_path_size) {
-      if (cur_level == level) {
-        // Does desired level fit in this path?
-        return p;
-      } else {
-        current_path_size -= level_size;
-        if (cur_level > 0) {
-          if (ioptions.level_compaction_dynamic_level_bytes) {
-            // Currently, level_compaction_dynamic_level_bytes is ignored when
-            // multiple db paths are specified. https://github.com/facebook/
-            // rocksdb/blob/master/db/column_family.cc.
-            // Still, adding this check to avoid accidentally using
-            // max_bytes_for_level_multiplier_additional
-            level_size = static_cast<uint64_t>(
-                level_size * mutable_cf_options.max_bytes_for_level_multiplier);
-          } else {
-            level_size = static_cast<uint64_t>(
-                level_size * mutable_cf_options.max_bytes_for_level_multiplier *
-                mutable_cf_options.MaxBytesMultiplerAdditional(cur_level));
-          }
-        }
-        cur_level++;
-        continue;
-      }
-    }
-    p++;
-    current_path_size = ioptions.cf_paths[p].target_size;
-  }
-  return p;
-}
-
-bool LevelCompactionBuilder::PickFileToCompact() {
-  // level 0 files are overlapping. So we cannot pick more
-  // than one concurrent compactions at this level. This
-  // could be made better by looking at key-ranges that are
-  // being compacted at level 0.
-  if (start_level_ == 0 &&
-      !compaction_picker_->level0_compactions_in_progress()->empty()) {
-    TEST_SYNC_POINT("LevelCompactionPicker::PickCompactionBySize:0");
-    return false;
-  }
-
-  start_level_inputs_.clear();
-
-  assert(start_level_ >= 0);
-
-  // Pick the largest file in this level that is not already
-  // being compacted
-  const std::vector<int>& file_size =
-      vstorage_->FilesByCompactionPri(start_level_);
-  const std::vector<FileMetaData*>& level_files =
-      vstorage_->LevelFiles(start_level_);
-
-  unsigned int cmp_idx;
-  for (cmp_idx = vstorage_->NextCompactionIndex(start_level_);
-       cmp_idx < file_size.size(); cmp_idx++) {
-    int index = file_size[cmp_idx];
-    auto* f = level_files[index];
-
-    // do not pick a file to compact if it is being compacted
-    // from n-1 level.
-    if (f->being_compacted) {
-      continue;
-    }
-
-    start_level_inputs_.files.push_back(f);
-    start_level_inputs_.level = start_level_;
-    if (!compaction_picker_->ExpandInputsToCleanCut(cf_name_, vstorage_,
-                                                    &start_level_inputs_) ||
-        compaction_picker_->FilesRangeOverlapWithCompaction(
-            {start_level_inputs_}, output_level_)) {
-      // A locked (pending compaction) input-level file was pulled in due to
-      // user-key overlap.
-      start_level_inputs_.clear();
-      continue;
-    }
-
-    // Now that input level is fully expanded, we check whether any output files
-    // are locked due to pending compaction.
-    //
-    // Note we rely on ExpandInputsToCleanCut() to tell us whether any output-
-    // level files are locked, not just the extra ones pulled in for user-key
-    // overlap.
-    InternalKey smallest, largest;
-    compaction_picker_->GetRange(start_level_inputs_, &smallest, &largest);
-    CompactionInputFiles output_level_inputs;
-    output_level_inputs.level = output_level_;
-    vstorage_->GetOverlappingInputs(output_level_, &smallest, &largest,
-                                    &output_level_inputs.files);
-    if (!output_level_inputs.empty() &&
-        !compaction_picker_->ExpandInputsToCleanCut(cf_name_, vstorage_,
-                                                    &output_level_inputs)) {
-      start_level_inputs_.clear();
-      continue;
-    }
-    base_index_ = index;
-    break;
-  }
-
-  // store where to start the iteration in the next call to PickCompaction
-  vstorage_->SetNextCompactionIndex(start_level_, cmp_idx);
-
-  return start_level_inputs_.size() > 0;
-}
-
-bool LevelCompactionBuilder::PickIntraL0Compaction() {
-  start_level_inputs_.clear();
-  const std::vector<FileMetaData*>& level_files =
-      vstorage_->LevelFiles(0 /* level */);
-  if (level_files.size() <
-          static_cast<size_t>(
-              mutable_cf_options_.level0_file_num_compaction_trigger + 2) ||
-      level_files[0]->being_compacted) {
-    // If L0 isn't accumulating much files beyond the regular trigger, don't
-    // resort to L0->L0 compaction yet.
-    return false;
-  }
-  return FindIntraL0Compaction(level_files, kMinFilesForIntraL0Compaction,
-                               port::kMaxUint64, &start_level_inputs_);
-}
-}  // 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_);
-  return builder.PickCompaction();
-}
-
 }  // namespace rocksdb
@@ -15,7 +15,7 @@
 #include <unordered_set>
 #include <vector>

-#include "db/compaction.h"
+#include "db/compaction/compaction.h"
 #include "db/version_set.h"
 #include "options/cf_options.h"
 #include "rocksdb/env.h"
@@ -24,11 +24,26 @@

 namespace rocksdb {

+// The file contains an abstract class CompactionPicker, and its two
+// sub-classes LevelCompactionPicker and NullCompactionPicker, as
+// well as some helper functions used by them.
+
 class LogBuffer;
 class Compaction;
 class VersionStorageInfo;
 struct CompactionInputFiles;

+// An abstract class to pick compactions from an existing LSM-tree.
+//
+// Each compaction style inherits the class and implement the
+// interface to form automatic compactions. If NeedCompaction() is true,
+// then call PickCompaction() to find what files need to be compacted
+// and where to put the output files.
+//
+// Non-virtual functions CompactRange() and CompactFiles() are used to
+// pick files to compact based on users' DB::CompactRange() and
+// DB::CompactFiles() requests, respectively. There is little
+// compaction style specific logic for them.
 class CompactionPicker {
 public:
  CompactionPicker(const ImmutableCFOptions& ioptions,
@@ -221,21 +236,9 @@ class CompactionPicker {
  const InternalKeyComparator* const icmp_;
 };

-class LevelCompactionPicker : public CompactionPicker {
- public:
-  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 bool NeedsCompaction(
-      const VersionStorageInfo* vstorage) const override;
-};
-
 #ifndef ROCKSDB_LITE
+// A dummy compaction that never triggers any automatic
+// compaction.
 class NullCompactionPicker : public CompactionPicker {
 public:
  NullCompactionPicker(const ImmutableCFOptions& ioptions,
@@ -273,9 +276,26 @@ class NullCompactionPicker : public CompactionPicker {
 };
 #endif  // !ROCKSDB_LITE

+// Attempts to find an intra L0 compaction conforming to the given parameters.
+//
+// @param level_files                     Metadata for L0 files.
+// @param min_files_to_compact            Minimum number of files required to
+//                                        do the compaction.
+// @param max_compact_bytes_per_del_file  Maximum average size in bytes per
+//                                        file that is going to get deleted by
+//                                        the compaction.
+// @param max_compaction_bytes            Maximum total size in bytes (in terms
+//                                        of compensated file size) for files
+//                                        to be compacted.
+// @param [out] comp_inputs               If a compaction was found, will be
+//                                        initialized with corresponding input
+//                                        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);

 CompressionType GetCompressionType(const ImmutableCFOptions& ioptions,
@@ -7,18 +7,14 @@
 // 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 "db/compaction_picker_fifo.h"
+#include "db/compaction/compaction_picker_fifo.h"
 #ifndef ROCKSDB_LITE

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-
-#include <inttypes.h>
+#include <cinttypes>
 #include <string>
 #include <vector>
 #include "db/column_family.h"
-#include "util/log_buffer.h"
+#include "logging/log_buffer.h"
 #include "util/string_util.h"

 namespace rocksdb {
@@ -58,6 +54,15 @@ Compaction* FIFOCompactionPicker::PickTTLCompaction(
  }
  const uint64_t current_time = static_cast<uint64_t>(_current_time);

+  if (!level0_compactions_in_progress_.empty()) {
+    ROCKS_LOG_BUFFER(
+        log_buffer,
+        "[%s] FIFO compaction: Already executing compaction. No need "
+        "to run parallel compactions since compactions are very fast",
+        cf_name.c_str());
+    return nullptr;
+  }
+
  std::vector<CompactionInputFiles> inputs;
  inputs.emplace_back();
  inputs[0].level = 0;
@@ -134,7 +139,8 @@ Compaction* FIFOCompactionPicker::PickSizeCompaction(
              mutable_cf_options
                  .level0_file_num_compaction_trigger /* min_files_to_compact */
              ,
-              max_compact_bytes_per_del_file, &comp_inputs)) {
+              max_compact_bytes_per_del_file,
+              mutable_cf_options.max_compaction_bytes, &comp_inputs)) {
        Compaction* c = new Compaction(
            vstorage, ioptions_, mutable_cf_options, {comp_inputs}, 0,
            16 * 1024 * 1024 /* output file size limit */,
@@ -10,7 +10,7 @@
 #pragma once
 #ifndef ROCKSDB_LITE

-#include "db/compaction_picker.h"
+#include "db/compaction/compaction_picker.h"

 namespace rocksdb {
 class FIFOCompactionPicker : public CompactionPicker {
@@ -0,0 +1,553 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  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).
+//
+// 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 <string>
+#include <utility>
+#include <vector>
+
+#include "db/compaction/compaction_picker_level.h"
+#include "logging/log_buffer.h"
+#include "test_util/sync_point.h"
+
+namespace rocksdb {
+
+bool LevelCompactionPicker::NeedsCompaction(
+    const VersionStorageInfo* vstorage) const {
+  if (!vstorage->ExpiredTtlFiles().empty()) {
+    return true;
+  }
+  if (!vstorage->FilesMarkedForPeriodicCompaction().empty()) {
+    return true;
+  }
+  if (!vstorage->BottommostFilesMarkedForCompaction().empty()) {
+    return true;
+  }
+  if (!vstorage->FilesMarkedForCompaction().empty()) {
+    return true;
+  }
+  for (int i = 0; i <= vstorage->MaxInputLevel(); i++) {
+    if (vstorage->CompactionScore(i) >= 1) {
+      return true;
+    }
+  }
+  return false;
+}
+
+namespace {
+// A class to build a leveled compaction step-by-step.
+class LevelCompactionBuilder {
+ public:
+  LevelCompactionBuilder(const std::string& cf_name,
+                         VersionStorageInfo* vstorage,
+                         CompactionPicker* compaction_picker,
+                         LogBuffer* log_buffer,
+                         const MutableCFOptions& mutable_cf_options,
+                         const ImmutableCFOptions& ioptions)
+      : cf_name_(cf_name),
+        vstorage_(vstorage),
+        compaction_picker_(compaction_picker),
+        log_buffer_(log_buffer),
+        mutable_cf_options_(mutable_cf_options),
+        ioptions_(ioptions) {}
+
+  // Pick and return a compaction.
+  Compaction* PickCompaction();
+
+  // Pick the initial files to compact to the next level. (or together
+  // in Intra-L0 compactions)
+  void SetupInitialFiles();
+
+  // If the initial files are from L0 level, pick other L0
+  // files if needed.
+  bool SetupOtherL0FilesIfNeeded();
+
+  // Based on initial files, setup other files need to be compacted
+  // in this compaction, accordingly.
+  bool SetupOtherInputsIfNeeded();
+
+  Compaction* GetCompaction();
+
+  // For the specfied level, pick a file that we want to compact.
+  // Returns false if there is no file to compact.
+  // If it returns true, inputs->files.size() will be exactly one.
+  // If level is 0 and there is already a compaction on that level, this
+  // function will return false.
+  bool PickFileToCompact();
+
+  // For L0->L0, picks the longest span of files that aren't currently
+  // undergoing compaction for which work-per-deleted-file decreases. The span
+  // always starts from the newest L0 file.
+  //
+  // Intra-L0 compaction is independent of all other files, so it can be
+  // performed even when L0->base_level compactions are blocked.
+  //
+  // Returns true if `inputs` is populated with a span of files to be compacted;
+  // otherwise, returns false.
+  bool PickIntraL0Compaction();
+
+  void PickExpiredTtlFiles();
+
+  void PickFilesMarkedForPeriodicCompaction();
+
+  const std::string& cf_name_;
+  VersionStorageInfo* vstorage_;
+  CompactionPicker* compaction_picker_;
+  LogBuffer* log_buffer_;
+  int start_level_ = -1;
+  int output_level_ = -1;
+  int parent_index_ = -1;
+  int base_index_ = -1;
+  double start_level_score_ = 0;
+  bool is_manual_ = false;
+  CompactionInputFiles start_level_inputs_;
+  std::vector<CompactionInputFiles> compaction_inputs_;
+  CompactionInputFiles output_level_inputs_;
+  std::vector<FileMetaData*> grandparents_;
+  CompactionReason compaction_reason_ = CompactionReason::kUnknown;
+
+  const MutableCFOptions& mutable_cf_options_;
+  const ImmutableCFOptions& ioptions_;
+  // Pick a path ID to place a newly generated file, with its level
+  static uint32_t GetPathId(const ImmutableCFOptions& ioptions,
+                            const MutableCFOptions& mutable_cf_options,
+                            int level);
+
+  static const int kMinFilesForIntraL0Compaction = 4;
+};
+
+void LevelCompactionBuilder::PickExpiredTtlFiles() {
+  if (vstorage_->ExpiredTtlFiles().empty()) {
+    return;
+  }
+
+  auto continuation = [&](std::pair<int, FileMetaData*> level_file) {
+    // If it's being compacted it has nothing to do here.
+    // If this assert() fails that means that some function marked some
+    // files as being_compacted, but didn't call ComputeCompactionScore()
+    assert(!level_file.second->being_compacted);
+    start_level_ = level_file.first;
+    output_level_ =
+        (start_level_ == 0) ? vstorage_->base_level() : start_level_ + 1;
+
+    if ((start_level_ == vstorage_->num_non_empty_levels() - 1) ||
+        (start_level_ == 0 &&
+         !compaction_picker_->level0_compactions_in_progress()->empty())) {
+      return false;
+    }
+
+    start_level_inputs_.files = {level_file.second};
+    start_level_inputs_.level = start_level_;
+    return compaction_picker_->ExpandInputsToCleanCut(cf_name_, vstorage_,
+                                                      &start_level_inputs_);
+  };
+
+  for (auto& level_file : vstorage_->ExpiredTtlFiles()) {
+    if (continuation(level_file)) {
+      // found the compaction!
+      return;
+    }
+  }
+
+  start_level_inputs_.files.clear();
+}
+
+void LevelCompactionBuilder::PickFilesMarkedForPeriodicCompaction() {
+  if (vstorage_->FilesMarkedForPeriodicCompaction().empty()) {
+    return;
+  }
+
+  auto continuation = [&](std::pair<int, FileMetaData*> level_file) {
+    // If it's being compacted it has nothing to do here.
+    // If this assert() fails that means that some function marked some
+    // files as being_compacted, but didn't call ComputeCompactionScore()
+    assert(!level_file.second->being_compacted);
+    output_level_ = start_level_ = level_file.first;
+
+    if (start_level_ == 0 &&
+        !compaction_picker_->level0_compactions_in_progress()->empty()) {
+      return false;
+    }
+
+    start_level_inputs_.files = {level_file.second};
+    start_level_inputs_.level = start_level_;
+    return compaction_picker_->ExpandInputsToCleanCut(cf_name_, vstorage_,
+                                                      &start_level_inputs_);
+  };
+
+  for (auto& level_file : vstorage_->FilesMarkedForPeriodicCompaction()) {
+    if (continuation(level_file)) {
+      // found the compaction!
+      return;
+    }
+  }
+
+  start_level_inputs_.files.clear();
+}
+
+void LevelCompactionBuilder::SetupInitialFiles() {
+  // Find the compactions by size on all levels.
+  bool skipped_l0_to_base = false;
+  for (int i = 0; i < compaction_picker_->NumberLevels() - 1; i++) {
+    start_level_score_ = vstorage_->CompactionScore(i);
+    start_level_ = vstorage_->CompactionScoreLevel(i);
+    assert(i == 0 || start_level_score_ <= vstorage_->CompactionScore(i - 1));
+    if (start_level_score_ >= 1) {
+      if (skipped_l0_to_base && start_level_ == vstorage_->base_level()) {
+        // If L0->base_level compaction is pending, don't schedule further
+        // compaction from base level. Otherwise L0->base_level compaction
+        // may starve.
+        continue;
+      }
+      output_level_ =
+          (start_level_ == 0) ? vstorage_->base_level() : start_level_ + 1;
+      if (PickFileToCompact()) {
+        // found the compaction!
+        if (start_level_ == 0) {
+          // L0 score = `num L0 files` / `level0_file_num_compaction_trigger`
+          compaction_reason_ = CompactionReason::kLevelL0FilesNum;
+        } else {
+          // L1+ score = `Level files size` / `MaxBytesForLevel`
+          compaction_reason_ = CompactionReason::kLevelMaxLevelSize;
+        }
+        break;
+      } else {
+        // didn't find the compaction, clear the inputs
+        start_level_inputs_.clear();
+        if (start_level_ == 0) {
+          skipped_l0_to_base = true;
+          // L0->base_level may be blocked due to ongoing L0->base_level
+          // compactions. It may also be blocked by an ongoing compaction from
+          // base_level downwards.
+          //
+          // In these cases, to reduce L0 file count and thus reduce likelihood
+          // of write stalls, we can attempt compacting a span of files within
+          // L0.
+          if (PickIntraL0Compaction()) {
+            output_level_ = 0;
+            compaction_reason_ = CompactionReason::kLevelL0FilesNum;
+            break;
+          }
+        }
+      }
+    }
+  }
+
+  // if we didn't find a compaction, check if there are any files marked for
+  // compaction
+  if (start_level_inputs_.empty()) {
+    parent_index_ = base_index_ = -1;
+
+    compaction_picker_->PickFilesMarkedForCompaction(
+        cf_name_, vstorage_, &start_level_, &output_level_,
+        &start_level_inputs_);
+    if (!start_level_inputs_.empty()) {
+      is_manual_ = true;
+      compaction_reason_ = CompactionReason::kFilesMarkedForCompaction;
+      return;
+    }
+  }
+
+  // Bottommost Files Compaction on deleting tombstones
+  if (start_level_inputs_.empty()) {
+    size_t i;
+    for (i = 0; i < vstorage_->BottommostFilesMarkedForCompaction().size();
+         ++i) {
+      auto& level_and_file = vstorage_->BottommostFilesMarkedForCompaction()[i];
+      assert(!level_and_file.second->being_compacted);
+      start_level_inputs_.level = output_level_ = start_level_ =
+          level_and_file.first;
+      start_level_inputs_.files = {level_and_file.second};
+      if (compaction_picker_->ExpandInputsToCleanCut(cf_name_, vstorage_,
+                                                     &start_level_inputs_)) {
+        break;
+      }
+    }
+    if (i == vstorage_->BottommostFilesMarkedForCompaction().size()) {
+      start_level_inputs_.clear();
+    } else {
+      assert(!start_level_inputs_.empty());
+      compaction_reason_ = CompactionReason::kBottommostFiles;
+      return;
+    }
+  }
+
+  // TTL Compaction
+  if (start_level_inputs_.empty()) {
+    PickExpiredTtlFiles();
+    if (!start_level_inputs_.empty()) {
+      compaction_reason_ = CompactionReason::kTtl;
+      return;
+    }
+  }
+
+  // Periodic Compaction
+  if (start_level_inputs_.empty()) {
+    PickFilesMarkedForPeriodicCompaction();
+    if (!start_level_inputs_.empty()) {
+      compaction_reason_ = CompactionReason::kPeriodicCompaction;
+      return;
+    }
+  }
+}
+
+bool LevelCompactionBuilder::SetupOtherL0FilesIfNeeded() {
+  if (start_level_ == 0 && output_level_ != 0) {
+    return compaction_picker_->GetOverlappingL0Files(
+        vstorage_, &start_level_inputs_, output_level_, &parent_index_);
+  }
+  return true;
+}
+
+bool LevelCompactionBuilder::SetupOtherInputsIfNeeded() {
+  // Setup input files from output level. For output to L0, we only compact
+  // spans of files that do not interact with any pending compactions, so don't
+  // need to consider other levels.
+  if (output_level_ != 0) {
+    output_level_inputs_.level = output_level_;
+    if (!compaction_picker_->SetupOtherInputs(
+            cf_name_, mutable_cf_options_, vstorage_, &start_level_inputs_,
+            &output_level_inputs_, &parent_index_, base_index_)) {
+      return false;
+    }
+
+    compaction_inputs_.push_back(start_level_inputs_);
+    if (!output_level_inputs_.empty()) {
+      compaction_inputs_.push_back(output_level_inputs_);
+    }
+
+    // In some edge cases we could pick a compaction that will be compacting
+    // a key range that overlap with another running compaction, and both
+    // of them have the same output level. This could happen if
+    // (1) we are running a non-exclusive manual compaction
+    // (2) AddFile ingest a new file into the LSM tree
+    // We need to disallow this from happening.
+    if (compaction_picker_->FilesRangeOverlapWithCompaction(compaction_inputs_,
+                                                            output_level_)) {
+      // This compaction output could potentially conflict with the output
+      // of a currently running compaction, we cannot run it.
+      return false;
+    }
+    compaction_picker_->GetGrandparents(vstorage_, start_level_inputs_,
+                                        output_level_inputs_, &grandparents_);
+  } else {
+    compaction_inputs_.push_back(start_level_inputs_);
+  }
+  return true;
+}
+
+Compaction* LevelCompactionBuilder::PickCompaction() {
+  // Pick up the first file to start compaction. It may have been extended
+  // to a clean cut.
+  SetupInitialFiles();
+  if (start_level_inputs_.empty()) {
+    return nullptr;
+  }
+  assert(start_level_ >= 0 && output_level_ >= 0);
+
+  // If it is a L0 -> base level compaction, we need to set up other L0
+  // files if needed.
+  if (!SetupOtherL0FilesIfNeeded()) {
+    return nullptr;
+  }
+
+  // Pick files in the output level and expand more files in the start level
+  // if needed.
+  if (!SetupOtherInputsIfNeeded()) {
+    return nullptr;
+  }
+
+  // Form a compaction object containing the files we picked.
+  Compaction* c = GetCompaction();
+
+  TEST_SYNC_POINT_CALLBACK("LevelCompactionPicker::PickCompaction:Return", c);
+
+  return c;
+}
+
+Compaction* LevelCompactionBuilder::GetCompaction() {
+  auto c = new Compaction(
+      vstorage_, ioptions_, mutable_cf_options_, std::move(compaction_inputs_),
+      output_level_,
+      MaxFileSizeForLevel(mutable_cf_options_, output_level_,
+                          ioptions_.compaction_style, vstorage_->base_level(),
+                          ioptions_.level_compaction_dynamic_level_bytes),
+      mutable_cf_options_.max_compaction_bytes,
+      GetPathId(ioptions_, mutable_cf_options_, output_level_),
+      GetCompressionType(ioptions_, vstorage_, mutable_cf_options_,
+                         output_level_, vstorage_->base_level()),
+      GetCompressionOptions(ioptions_, vstorage_, output_level_),
+      /* max_subcompactions */ 0, std::move(grandparents_), is_manual_,
+      start_level_score_, false /* deletion_compaction */, compaction_reason_);
+
+  // If it's level 0 compaction, make sure we don't execute any other level 0
+  // compactions in parallel
+  compaction_picker_->RegisterCompaction(c);
+
+  // Creating a compaction influences the compaction score because the score
+  // takes running compactions into account (by skipping files that are already
+  // being compacted). Since we just changed compaction score, we recalculate it
+  // here
+  vstorage_->ComputeCompactionScore(ioptions_, mutable_cf_options_);
+  return c;
+}
+
+/*
+ * Find the optimal path to place a file
+ * Given a level, finds the path where levels up to it will fit in levels
+ * up to and including this path
+ */
+uint32_t LevelCompactionBuilder::GetPathId(
+    const ImmutableCFOptions& ioptions,
+    const MutableCFOptions& mutable_cf_options, int level) {
+  uint32_t p = 0;
+  assert(!ioptions.cf_paths.empty());
+
+  // size remaining in the most recent path
+  uint64_t current_path_size = ioptions.cf_paths[0].target_size;
+
+  uint64_t level_size;
+  int cur_level = 0;
+
+  // max_bytes_for_level_base denotes L1 size.
+  // We estimate L0 size to be the same as L1.
+  level_size = mutable_cf_options.max_bytes_for_level_base;
+
+  // Last path is the fallback
+  while (p < ioptions.cf_paths.size() - 1) {
+    if (level_size <= current_path_size) {
+      if (cur_level == level) {
+        // Does desired level fit in this path?
+        return p;
+      } else {
+        current_path_size -= level_size;
+        if (cur_level > 0) {
+          if (ioptions.level_compaction_dynamic_level_bytes) {
+            // Currently, level_compaction_dynamic_level_bytes is ignored when
+            // multiple db paths are specified. https://github.com/facebook/
+            // rocksdb/blob/master/db/column_family.cc.
+            // Still, adding this check to avoid accidentally using
+            // max_bytes_for_level_multiplier_additional
+            level_size = static_cast<uint64_t>(
+                level_size * mutable_cf_options.max_bytes_for_level_multiplier);
+          } else {
+            level_size = static_cast<uint64_t>(
+                level_size * mutable_cf_options.max_bytes_for_level_multiplier *
+                mutable_cf_options.MaxBytesMultiplerAdditional(cur_level));
+          }
+        }
+        cur_level++;
+        continue;
+      }
+    }
+    p++;
+    current_path_size = ioptions.cf_paths[p].target_size;
+  }
+  return p;
+}
+
+bool LevelCompactionBuilder::PickFileToCompact() {
+  // level 0 files are overlapping. So we cannot pick more
+  // than one concurrent compactions at this level. This
+  // could be made better by looking at key-ranges that are
+  // being compacted at level 0.
+  if (start_level_ == 0 &&
+      !compaction_picker_->level0_compactions_in_progress()->empty()) {
+    TEST_SYNC_POINT("LevelCompactionPicker::PickCompactionBySize:0");
+    return false;
+  }
+
+  start_level_inputs_.clear();
+
+  assert(start_level_ >= 0);
+
+  // Pick the largest file in this level that is not already
+  // being compacted
+  const std::vector<int>& file_size =
+      vstorage_->FilesByCompactionPri(start_level_);
+  const std::vector<FileMetaData*>& level_files =
+      vstorage_->LevelFiles(start_level_);
+
+  unsigned int cmp_idx;
+  for (cmp_idx = vstorage_->NextCompactionIndex(start_level_);
+       cmp_idx < file_size.size(); cmp_idx++) {
+    int index = file_size[cmp_idx];
+    auto* f = level_files[index];
+
+    // do not pick a file to compact if it is being compacted
+    // from n-1 level.
+    if (f->being_compacted) {
+      continue;
+    }
+
+    start_level_inputs_.files.push_back(f);
+    start_level_inputs_.level = start_level_;
+    if (!compaction_picker_->ExpandInputsToCleanCut(cf_name_, vstorage_,
+                                                    &start_level_inputs_) ||
+        compaction_picker_->FilesRangeOverlapWithCompaction(
+            {start_level_inputs_}, output_level_)) {
+      // A locked (pending compaction) input-level file was pulled in due to
+      // user-key overlap.
+      start_level_inputs_.clear();
+      continue;
+    }
+
+    // Now that input level is fully expanded, we check whether any output files
+    // are locked due to pending compaction.
+    //
+    // Note we rely on ExpandInputsToCleanCut() to tell us whether any output-
+    // level files are locked, not just the extra ones pulled in for user-key
+    // overlap.
+    InternalKey smallest, largest;
+    compaction_picker_->GetRange(start_level_inputs_, &smallest, &largest);
+    CompactionInputFiles output_level_inputs;
+    output_level_inputs.level = output_level_;
+    vstorage_->GetOverlappingInputs(output_level_, &smallest, &largest,
+                                    &output_level_inputs.files);
+    if (!output_level_inputs.empty() &&
+        !compaction_picker_->ExpandInputsToCleanCut(cf_name_, vstorage_,
+                                                    &output_level_inputs)) {
+      start_level_inputs_.clear();
+      continue;
+    }
+    base_index_ = index;
+    break;
+  }
+
+  // store where to start the iteration in the next call to PickCompaction
+  vstorage_->SetNextCompactionIndex(start_level_, cmp_idx);
+
+  return start_level_inputs_.size() > 0;
+}
+
+bool LevelCompactionBuilder::PickIntraL0Compaction() {
+  start_level_inputs_.clear();
+  const std::vector<FileMetaData*>& level_files =
+      vstorage_->LevelFiles(0 /* level */);
+  if (level_files.size() <
+          static_cast<size_t>(
+              mutable_cf_options_.level0_file_num_compaction_trigger + 2) ||
+      level_files[0]->being_compacted) {
+    // If L0 isn't accumulating much files beyond the regular trigger, don't
+    // resort to L0->L0 compaction yet.
+    return false;
+  }
+  return FindIntraL0Compaction(
+      level_files, kMinFilesForIntraL0Compaction, port::kMaxUint64,
+      mutable_cf_options_.max_compaction_bytes, &start_level_inputs_);
+}
+}  // 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_);
+  return builder.PickCompaction();
+}
+}  // namespace rocksdb
@@ -0,0 +1,32 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  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).
+//
+// 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.
+
+#pragma once
+
+#include "db/compaction/compaction_picker.h"
+
+namespace rocksdb {
+// Picking compactions for leveled compaction. See wiki page
+// https://github.com/facebook/rocksdb/wiki/Leveled-Compaction
+// for description of Leveled compaction.
+class LevelCompactionPicker : public CompactionPicker {
+ public:
+  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 bool NeedsCompaction(
+      const VersionStorageInfo* vstorage) const override;
+};
+
+}  // namespace rocksdb
@@ -3,19 +3,19 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).

-#include "db/compaction_picker.h"

 #include <limits>
 #include <string>
 #include <utility>
-#include "db/compaction.h"
-#include "db/compaction_picker_fifo.h"
-#include "db/compaction_picker_universal.h"
+#include "db/compaction/compaction.h"
+#include "db/compaction/compaction_picker_fifo.h"
+#include "db/compaction/compaction_picker_level.h"
+#include "db/compaction/compaction_picker_universal.h"

-#include "util/logging.h"
+#include "logging/logging.h"
+#include "test_util/testharness.h"
+#include "test_util/testutil.h"
 #include "util/string_util.h"
-#include "util/testharness.h"
-#include "util/testutil.h"

 namespace rocksdb {

@@ -1478,6 +1478,65 @@ TEST_F(CompactionPickerTest, CacheNextCompactionIndex) {
  ASSERT_EQ(4, vstorage_->NextCompactionIndex(1 /* level */));
 }

+TEST_F(CompactionPickerTest, IntraL0MaxCompactionBytesNotHit) {
+  // 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 = 1000000u;
+  NewVersionStorage(6, kCompactionStyleLevel);
+
+  // 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);
+  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_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(1U, compaction->num_input_levels());
+  ASSERT_EQ(5U, compaction->num_input_files(0));
+  ASSERT_EQ(CompactionReason::kLevelL0FilesNum,
+            compaction->compaction_reason());
+  ASSERT_EQ(0, compaction->output_level());
+}
+
+TEST_F(CompactionPickerTest, IntraL0MaxCompactionBytesHit) {
+  // 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 5 L0 files will be picked for intra L0 compaction due to
+  // 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);
+  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_));
+  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) {
@@ -7,25 +7,21 @@
 // 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 "db/compaction_picker_universal.h"
+#include "db/compaction/compaction_picker_universal.h"
 #ifndef ROCKSDB_LITE

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-
-#include <inttypes.h>
+#include <cinttypes>
 #include <limits>
 #include <queue>
 #include <string>
 #include <utility>
 #include "db/column_family.h"
+#include "file/filename.h"
+#include "logging/log_buffer.h"
 #include "monitoring/statistics.h"
-#include "util/filename.h"
-#include "util/log_buffer.h"
+#include "test_util/sync_point.h"
 #include "util/random.h"
 #include "util/string_util.h"
-#include "util/sync_point.h"

 namespace rocksdb {
 namespace {
@@ -10,7 +10,7 @@
 #pragma once
 #ifndef ROCKSDB_LITE

-#include "db/compaction_picker.h"
+#include "db/compaction/compaction_picker.h"

 namespace rocksdb {
 class UniversalCompactionPicker : public CompactionPicker {
@@ -9,11 +9,11 @@
 #include "memtable/stl_wrappers.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
+#include "test_util/testharness.h"
+#include "test_util/testutil.h"
 #include "util/hash.h"
 #include "util/kv_map.h"
 #include "util/string_util.h"
-#include "util/testharness.h"
-#include "util/testutil.h"
 #include "utilities/merge_operators.h"

 using std::unique_ptr;
@@ -8,7 +8,7 @@

 #include "rocksdb/convenience.h"

-#include "db/db_impl.h"
+#include "db/db_impl/db_impl.h"
 #include "util/cast_util.h"

 namespace rocksdb {
@@ -35,6 +35,12 @@ Status DeleteFilesInRanges(DB* db, ColumnFamilyHandle* column_family,
 Status VerifySstFileChecksum(const Options& options,
                             const EnvOptions& env_options,
                             const std::string& file_path) {
+  return VerifySstFileChecksum(options, env_options, ReadOptions(), file_path);
+}
+Status VerifySstFileChecksum(const Options& options,
+                             const EnvOptions& env_options,
+                             const ReadOptions& read_options,
+                             const std::string& file_path) {
  std::unique_ptr<RandomAccessFile> file;
  uint64_t file_size;
  InternalKeyComparator internal_comparator(options.comparator);
@@ -59,7 +65,8 @@ Status VerifySstFileChecksum(const Options& options,
  if (!s.ok()) {
    return s;
  }
-  s = table_reader->VerifyChecksum();
+  s = table_reader->VerifyChecksum(read_options,
+                                   TableReaderCaller::kUserVerifyChecksum);
  return s;
 }

@@ -13,23 +13,24 @@

 #include <errno.h>
 #include <fcntl.h>
-#include <inttypes.h>
 #include <sys/stat.h>
 #include <sys/types.h>
-#include "db/db_impl.h"
+#include <cinttypes>
+#include "db/db_impl/db_impl.h"
+#include "db/db_test_util.h"
 #include "db/log_format.h"
 #include "db/version_set.h"
+#include "file/filename.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/convenience.h"
 #include "rocksdb/env.h"
 #include "rocksdb/table.h"
 #include "rocksdb/write_batch.h"
-#include "table/block_based_table_builder.h"
+#include "table/block_based/block_based_table_builder.h"
 #include "table/meta_blocks.h"
-#include "util/filename.h"
+#include "test_util/testharness.h"
+#include "test_util/testutil.h"
 #include "util/string_util.h"
-#include "util/testharness.h"
-#include "util/testutil.h"

 namespace rocksdb {

@@ -76,7 +77,11 @@ class CorruptionTest : public testing::Test {
    delete db_;
    db_ = nullptr;
    Options opt = (options ? *options : options_);
-    opt.env = &env_;
+    if (opt.env == Options().env) {
+      // If env is not overridden, replace it with ErrorEnv.
+      // Otherwise, the test already uses a non-default Env.
+      opt.env = &env_;
+    }
    opt.arena_block_size = 4096;
    BlockBasedTableOptions table_options;
    table_options.block_cache = tiny_cache_;
@@ -321,6 +326,59 @@ TEST_F(CorruptionTest, TableFile) {
  ASSERT_NOK(dbi->VerifyChecksum());
 }

+TEST_F(CorruptionTest, VerifyChecksumReadahead) {
+  Options options;
+  SpecialEnv senv(Env::Default());
+  options.env = &senv;
+  // Disable block cache as we are going to check checksum for
+  // the same file twice and measure number of reads.
+  BlockBasedTableOptions table_options_no_bc;
+  table_options_no_bc.no_block_cache = true;
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options_no_bc));
+
+  Reopen(&options);
+
+  Build(10000);
+  DBImpl* dbi = reinterpret_cast<DBImpl*>(db_);
+  dbi->TEST_FlushMemTable();
+  dbi->TEST_CompactRange(0, nullptr, nullptr);
+  dbi->TEST_CompactRange(1, nullptr, nullptr);
+
+  senv.count_random_reads_ = true;
+  senv.random_read_counter_.Reset();
+  ASSERT_OK(dbi->VerifyChecksum());
+
+  // Make sure the counter is enabled.
+  ASSERT_GT(senv.random_read_counter_.Read(), 0);
+
+  // The SST file is about 10MB. Default readahead size is 256KB.
+  // Give a conservative 20 reads for metadata blocks, The number
+  // of random reads should be within 10 MB / 256KB + 20 = 60.
+  ASSERT_LT(senv.random_read_counter_.Read(), 60);
+
+  senv.random_read_bytes_counter_ = 0;
+  ReadOptions ro;
+  ro.readahead_size = size_t{32 * 1024};
+  ASSERT_OK(dbi->VerifyChecksum(ro));
+  // The SST file is about 10MB. We set readahead size to 32KB.
+  // Give 0 to 20 reads for metadata blocks, and allow real read
+  // to range from 24KB to 48KB. The lower bound would be:
+  //   10MB / 48KB + 0 = 213
+  // The higher bound is
+  //   10MB / 24KB + 20 = 447.
+  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();
+}
+
 TEST_F(CorruptionTest, TableFileIndexData) {
  Options options;
  // very big, we'll trigger flushes manually
@@ -5,15 +5,15 @@

 #ifndef ROCKSDB_LITE

-#include "db/db_impl.h"
+#include "db/db_impl/db_impl.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
-#include "table/cuckoo_table_factory.h"
-#include "table/cuckoo_table_reader.h"
+#include "table/cuckoo/cuckoo_table_factory.h"
+#include "table/cuckoo/cuckoo_table_reader.h"
 #include "table/meta_blocks.h"
+#include "test_util/testharness.h"
+#include "test_util/testutil.h"
 #include "util/string_util.h"
-#include "util/testharness.h"
-#include "util/testutil.h"

 namespace rocksdb {

@@ -10,9 +10,12 @@
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
 #include "rocksdb/perf_context.h"
-#include "util/fault_injection_test_env.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)
-#include "util/sync_point.h"
+#include "test_util/sync_point.h"
 #endif

 namespace rocksdb {
@@ -68,6 +71,44 @@ TEST_F(DBBasicTest, ReadOnlyDB) {
  ASSERT_TRUE(db_->SyncWAL().IsNotSupported());
 }

+TEST_F(DBBasicTest, ReadOnlyDBWithWriteDBIdToManifestSet) {
+  ASSERT_OK(Put("foo", "v1"));
+  ASSERT_OK(Put("bar", "v2"));
+  ASSERT_OK(Put("foo", "v3"));
+  Close();
+
+  auto options = CurrentOptions();
+  options.write_dbid_to_manifest = true;
+  assert(options.env == env_);
+  ASSERT_OK(ReadOnlyReopen(options));
+  std::string db_id1;
+  db_->GetDbIdentity(db_id1);
+  ASSERT_EQ("v3", Get("foo"));
+  ASSERT_EQ("v2", Get("bar"));
+  Iterator* iter = db_->NewIterator(ReadOptions());
+  int count = 0;
+  for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
+    ASSERT_OK(iter->status());
+    ++count;
+  }
+  ASSERT_EQ(count, 2);
+  delete iter;
+  Close();
+
+  // Reopen and flush memtable.
+  Reopen(options);
+  Flush();
+  Close();
+  // Now check keys in read only mode.
+  ASSERT_OK(ReadOnlyReopen(options));
+  ASSERT_EQ("v3", Get("foo"));
+  ASSERT_EQ("v2", Get("bar"));
+  ASSERT_TRUE(db_->SyncWAL().IsNotSupported());
+  std::string db_id2;
+  db_->GetDbIdentity(db_id2);
+  ASSERT_EQ(db_id1, db_id2);
+}
+
 TEST_F(DBBasicTest, CompactedDB) {
  const uint64_t kFileSize = 1 << 20;
  Options options = CurrentOptions();
@@ -295,7 +336,7 @@ TEST_F(DBBasicTest, FlushMultipleMemtable) {
    writeOpt.disableWAL = true;
    options.max_write_buffer_number = 4;
    options.min_write_buffer_number_to_merge = 3;
-    options.max_write_buffer_number_to_maintain = -1;
+    options.max_write_buffer_size_to_maintain = -1;
    CreateAndReopenWithCF({"pikachu"}, options);
    ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v1"));
    ASSERT_OK(Flush(1));
@@ -325,7 +366,8 @@ TEST_F(DBBasicTest, FlushEmptyColumnFamily) {
  writeOpt.disableWAL = true;
  options.max_write_buffer_number = 2;
  options.min_write_buffer_number_to_merge = 1;
-  options.max_write_buffer_number_to_maintain = 1;
+  options.max_write_buffer_size_to_maintain =
+      static_cast<int64_t>(options.write_buffer_size);
  CreateAndReopenWithCF({"pikachu"}, options);

  // Compaction can still go through even if no thread can flush the
@@ -421,7 +463,7 @@ TEST_F(DBBasicTest, ManifestRollOver) {
  } while (ChangeCompactOptions());
 }

-TEST_F(DBBasicTest, IdentityAcrossRestarts) {
+TEST_F(DBBasicTest, IdentityAcrossRestarts1) {
  do {
    std::string id1;
    ASSERT_OK(db_->GetDbIdentity(id1));
@@ -433,13 +475,40 @@ TEST_F(DBBasicTest, IdentityAcrossRestarts) {
    // id1 should match id2 because identity was not regenerated
    ASSERT_EQ(id1.compare(id2), 0);

+    std::string idfilename = IdentityFileName(dbname_);
+    ASSERT_OK(env_->DeleteFile(idfilename));
+    Reopen(options);
+    std::string id3;
+    ASSERT_OK(db_->GetDbIdentity(id3));
+    if (options.write_dbid_to_manifest) {
+      ASSERT_EQ(id1.compare(id3), 0);
+    } else {
+      // id1 should NOT match id3 because identity was regenerated
+      ASSERT_NE(id1.compare(id3), 0);
+    }
+  } while (ChangeCompactOptions());
+}
+
+TEST_F(DBBasicTest, IdentityAcrossRestarts2) {
+  do {
+    std::string id1;
+    ASSERT_OK(db_->GetDbIdentity(id1));
+
+    Options options = CurrentOptions();
+    options.write_dbid_to_manifest = true;
+    Reopen(options);
+    std::string id2;
+    ASSERT_OK(db_->GetDbIdentity(id2));
+    // id1 should match id2 because identity was not regenerated
+    ASSERT_EQ(id1.compare(id2), 0);
+
    std::string idfilename = IdentityFileName(dbname_);
    ASSERT_OK(env_->DeleteFile(idfilename));
    Reopen(options);
    std::string id3;
    ASSERT_OK(db_->GetDbIdentity(id3));
    // id1 should NOT match id3 because identity was regenerated
-    ASSERT_NE(id1.compare(id3), 0);
+    ASSERT_EQ(id1, id3);
  } while (ChangeCompactOptions());
 }

@@ -1284,10 +1353,697 @@ TEST_F(DBBasicTest, MultiGetBatchedMultiLevel) {
    }
  }
 }
+
+// Test class for batched MultiGet with prefix extractor
+// Param bool - If true, use partitioned filters
+//              If false, use full filter block
+class MultiGetPrefixExtractorTest
+    : public DBBasicTest,
+      public ::testing::WithParamInterface<bool> {};
+
+TEST_P(MultiGetPrefixExtractorTest, Batched) {
+  Options options = CurrentOptions();
+  options.prefix_extractor.reset(NewFixedPrefixTransform(2));
+  BlockBasedTableOptions bbto;
+  if (GetParam()) {
+    bbto.index_type = BlockBasedTableOptions::IndexType::kTwoLevelIndexSearch;
+    bbto.partition_filters = true;
+  }
+  bbto.filter_policy.reset(NewBloomFilterPolicy(10, false));
+  bbto.whole_key_filtering = false;
+  bbto.cache_index_and_filter_blocks = false;
+  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
+  Reopen(options);
+
+  // 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"));
+  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");
+  ASSERT_EQ(values[1], "v1");
+  ASSERT_EQ(values[2], "v2");
+  ASSERT_EQ(values[3], "v3");
+  ASSERT_EQ(values[4], "v4");
+  // Filter hits for 4 in-domain keys
+  ASSERT_EQ(get_perf_context()->bloom_sst_hit_count, 4);
+}
+
+INSTANTIATE_TEST_CASE_P(
+    MultiGetPrefix, MultiGetPrefixExtractorTest,
+    ::testing::Bool());
+
+#ifndef ROCKSDB_LITE
+class DBMultiGetRowCacheTest
+    : public DBBasicTest,
+      public ::testing::WithParamInterface<bool> {};
+
+TEST_P(DBMultiGetRowCacheTest, MultiGetBatched) {
+  do {
+    option_config_ = kRowCache;
+    Options options = CurrentOptions();
+    options.statistics = rocksdb::CreateDBStatistics();
+    CreateAndReopenWithCF({"pikachu"}, options);
+    SetPerfLevel(kEnableCount);
+    ASSERT_OK(Put(1, "k1", "v1"));
+    ASSERT_OK(Put(1, "k2", "v2"));
+    ASSERT_OK(Put(1, "k3", "v3"));
+    ASSERT_OK(Put(1, "k4", "v4"));
+    Flush(1);
+    ASSERT_OK(Put(1, "k5", "v5"));
+    const Snapshot* snap1 = dbfull()->GetSnapshot();
+    ASSERT_OK(Delete(1, "k4"));
+    Flush(1);
+    const Snapshot* snap2 = dbfull()->GetSnapshot();
+
+    get_perf_context()->Reset();
+
+    std::vector<Slice> keys({"no_key", "k5", "k4", "k3", "k1"});
+    std::vector<PinnableSlice> values(keys.size());
+    std::vector<ColumnFamilyHandle*> cfs(keys.size(), handles_[1]);
+    std::vector<Status> s(keys.size());
+
+    ReadOptions ro;
+    bool use_snapshots = GetParam();
+    if (use_snapshots) {
+      ro.snapshot = snap2;
+    }
+    db_->MultiGet(ro, handles_[1], keys.size(), keys.data(), values.data(),
+                  s.data(), false);
+
+    ASSERT_EQ(values.size(), keys.size());
+    ASSERT_EQ(std::string(values[4].data(), values[4].size()), "v1");
+    ASSERT_EQ(std::string(values[3].data(), values[3].size()), "v3");
+    ASSERT_EQ(std::string(values[1].data(), values[1].size()), "v5");
+    // four kv pairs * two bytes per value
+    ASSERT_EQ(6, (int)get_perf_context()->multiget_read_bytes);
+
+    ASSERT_TRUE(s[0].IsNotFound());
+    ASSERT_OK(s[1]);
+    ASSERT_TRUE(s[2].IsNotFound());
+    ASSERT_OK(s[3]);
+    ASSERT_OK(s[4]);
+
+    // Call MultiGet() again with some intersection with the previous set of
+    // keys. Those should already be in the row cache.
+    keys.assign({"no_key", "k5", "k3", "k2"});
+    for (size_t i = 0; i < keys.size(); ++i) {
+      values[i].Reset();
+      s[i] = Status::OK();
+    }
+    get_perf_context()->Reset();
+
+    if (use_snapshots) {
+      ro.snapshot = snap1;
+    }
+    db_->MultiGet(ReadOptions(), handles_[1], keys.size(), keys.data(),
+                  values.data(), s.data(), false);
+
+    ASSERT_EQ(std::string(values[3].data(), values[3].size()), "v2");
+    ASSERT_EQ(std::string(values[2].data(), values[2].size()), "v3");
+    ASSERT_EQ(std::string(values[1].data(), values[1].size()), "v5");
+    // four kv pairs * two bytes per value
+    ASSERT_EQ(6, (int)get_perf_context()->multiget_read_bytes);
+
+    ASSERT_TRUE(s[0].IsNotFound());
+    ASSERT_OK(s[1]);
+    ASSERT_OK(s[2]);
+    ASSERT_OK(s[3]);
+    if (use_snapshots) {
+      // Only reads from the first SST file would have been cached, since
+      // snapshot seq no is > fd.largest_seqno
+      ASSERT_EQ(1, TestGetTickerCount(options, ROW_CACHE_HIT));
+    } else {
+      ASSERT_EQ(2, TestGetTickerCount(options, ROW_CACHE_HIT));
+    }
+
+    SetPerfLevel(kDisable);
+    dbfull()->ReleaseSnapshot(snap1);
+    dbfull()->ReleaseSnapshot(snap2);
+  } while (ChangeCompactOptions());
+}
+
+INSTANTIATE_TEST_CASE_P(DBMultiGetRowCacheTest, DBMultiGetRowCacheTest,
+                        testing::Values(true, false));
+
+TEST_F(DBBasicTest, GetAllKeyVersions) {
+  Options options = CurrentOptions();
+  options.env = env_;
+  options.create_if_missing = true;
+  options.disable_auto_compactions = true;
+  CreateAndReopenWithCF({"pikachu"}, options);
+  ASSERT_EQ(2, handles_.size());
+  const size_t kNumInserts = 4;
+  const size_t kNumDeletes = 4;
+  const size_t kNumUpdates = 4;
+
+  // Check default column family
+  for (size_t i = 0; i != kNumInserts; ++i) {
+    ASSERT_OK(Put(std::to_string(i), "value"));
+  }
+  for (size_t i = 0; i != kNumUpdates; ++i) {
+    ASSERT_OK(Put(std::to_string(i), "value1"));
+  }
+  for (size_t i = 0; i != kNumDeletes; ++i) {
+    ASSERT_OK(Delete(std::to_string(i)));
+  }
+  std::vector<KeyVersion> key_versions;
+  ASSERT_OK(rocksdb::GetAllKeyVersions(db_, Slice(), Slice(),
+                                       std::numeric_limits<size_t>::max(),
+                                       &key_versions));
+  ASSERT_EQ(kNumInserts + kNumDeletes + kNumUpdates, key_versions.size());
+  ASSERT_OK(rocksdb::GetAllKeyVersions(db_, handles_[0], Slice(), Slice(),
+                                       std::numeric_limits<size_t>::max(),
+                                       &key_versions));
+  ASSERT_EQ(kNumInserts + kNumDeletes + kNumUpdates, key_versions.size());
+
+  // Check non-default column family
+  for (size_t i = 0; i != kNumInserts - 1; ++i) {
+    ASSERT_OK(Put(1, std::to_string(i), "value"));
+  }
+  for (size_t i = 0; i != kNumUpdates - 1; ++i) {
+    ASSERT_OK(Put(1, std::to_string(i), "value1"));
+  }
+  for (size_t i = 0; i != kNumDeletes - 1; ++i) {
+    ASSERT_OK(Delete(1, std::to_string(i)));
+  }
+  ASSERT_OK(rocksdb::GetAllKeyVersions(db_, handles_[1], Slice(), Slice(),
+                                       std::numeric_limits<size_t>::max(),
+                                       &key_versions));
+  ASSERT_EQ(kNumInserts + kNumDeletes + kNumUpdates - 3, key_versions.size());
+}
+#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>> {
+ public:
+  DBBasicTestWithParallelIO()
+      : DBTestBase("/db_basic_test_with_parallel_io") {
+    bool compressed_cache = std::get<0>(GetParam());
+    bool uncompressed_cache = std::get<1>(GetParam());
+    compression_enabled_ = std::get<2>(GetParam());
+    fill_cache_ = std::get<3>(GetParam());
+
+    if (compressed_cache) {
+      std::shared_ptr<Cache> cache = NewLRUCache(1048576);
+      compressed_cache_ = std::make_shared<MyBlockCache>(cache);
+    }
+    if (uncompressed_cache) {
+      std::shared_ptr<Cache> cache = NewLRUCache(1048576);
+      uncompressed_cache_ = std::make_shared<MyBlockCache>(cache);
+    }
+
+    env_->count_random_reads_ = true;
+
+    Options options = CurrentOptions();
+    Random rnd(301);
+    BlockBasedTableOptions table_options;
+    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());
+    options.table_factory.reset(new BlockBasedTableFactory(table_options));
+    if (!compression_enabled_) {
+      options.compression = kNoCompression;
+    }
+    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
+      values_.emplace_back(RandomString(&rnd, 128) + zero_str);
+      assert(Put(Key(i), values_[i]) == Status::OK());
+    }
+    Flush();
+  }
+
+  bool CheckValue(int i, const std::string& value) {
+    if (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(); }
+
+  int num_lookups_compressed() {
+    return compressed_cache_->num_lookups();
+  }
+  int num_found_compressed() {
+    return compressed_cache_->num_found();
+  }
+  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() {}
+
+ private:
+  class MyFlushBlockPolicyFactory
+    : public FlushBlockPolicyFactory {
+   public:
+    MyFlushBlockPolicyFactory() {}
+
+    virtual const char* Name() const override {
+      return "MyFlushBlockPolicyFactory";
+    }
+
+    virtual FlushBlockPolicy* NewFlushBlockPolicy(
+        const BlockBasedTableOptions& /*table_options*/,
+        const BlockBuilder& data_block_builder) const override {
+      return new MyFlushBlockPolicy(data_block_builder);
+    }
+  };
+
+  class MyFlushBlockPolicy
+    : public FlushBlockPolicy {
+   public:
+    explicit MyFlushBlockPolicy(const BlockBuilder& data_block_builder)
+      : num_keys_(0), data_block_builder_(data_block_builder) {}
+
+    bool Update(const Slice& /*key*/, const Slice& /*value*/) override {
+      if (data_block_builder_.empty()) {
+        // First key in this block
+        num_keys_ = 1;
+        return false;
+      }
+      // Flush every 10 keys
+      if (num_keys_ == 10) {
+        num_keys_ = 1;
+        return true;
+      }
+      num_keys_++;
+      return false;
+    }
+
+   private:
+    int num_keys_;
+    const BlockBuilder& data_block_builder_;
+  };
+
+  class MyBlockCache
+    : public Cache {
+   public:
+    explicit MyBlockCache(std::shared_ptr<Cache>& target)
+      : target_(target), num_lookups_(0), num_found_(0), num_inserts_(0) {}
+
+    virtual const char* Name() const override { return "MyBlockCache"; }
+
+    virtual Status Insert(const Slice& key, void* value, size_t charge,
+                          void (*deleter)(const Slice& key, void* value),
+                          Handle** handle = nullptr,
+                          Priority priority = Priority::LOW) override {
+      num_inserts_++;
+      return target_->Insert(key, value, charge, deleter, handle, priority);
+    }
+
+    virtual Handle* Lookup(const Slice& key,
+                           Statistics* stats = nullptr) override {
+      num_lookups_++;
+      Handle* handle = target_->Lookup(key, stats);
+      if (handle != nullptr) {
+        num_found_++;
+      }
+      return handle;
+    }
+
+    virtual bool Ref(Handle* handle) override {
+      return target_->Ref(handle);
+    }
+
+    virtual bool Release(Handle* handle, bool force_erase = false) override {
+      return target_->Release(handle, force_erase);
+    }
+
+    virtual void* Value(Handle* handle) override {
+      return target_->Value(handle);
+    }
+
+    virtual void Erase(const Slice& key) override {
+      target_->Erase(key);
+    }
+    virtual uint64_t NewId() override {
+      return target_->NewId();
+    }
+
+    virtual void SetCapacity(size_t capacity) override {
+      target_->SetCapacity(capacity);
+    }
+
+    virtual void SetStrictCapacityLimit(bool strict_capacity_limit) override {
+      target_->SetStrictCapacityLimit(strict_capacity_limit);
+    }
+
+    virtual bool HasStrictCapacityLimit() const override {
+      return target_->HasStrictCapacityLimit();
+    }
+
+    virtual size_t GetCapacity() const override {
+      return target_->GetCapacity();
+    }
+
+    virtual size_t GetUsage() const override {
+      return target_->GetUsage();
+    }
+
+    virtual size_t GetUsage(Handle* handle) const override {
+      return target_->GetUsage(handle);
+    }
+
+    virtual size_t GetPinnedUsage() const override {
+      return target_->GetPinnedUsage();
+    }
+
+    virtual size_t GetCharge(Handle* /*handle*/) const override { return 0; }
+
+    virtual void ApplyToAllCacheEntries(void (*callback)(void*, size_t),
+                                        bool thread_safe) override {
+      return target_->ApplyToAllCacheEntries(callback, thread_safe);
+    }
+
+    virtual void EraseUnRefEntries() override {
+      return target_->EraseUnRefEntries();
+    }
+
+    int num_lookups() { return num_lookups_; }
+
+    int num_found() { return num_found_; }
+
+    int num_inserts() { return num_inserts_; }
+   private:
+    std::shared_ptr<Cache> target_;
+    int num_lookups_;
+    int num_found_;
+    int num_inserts_;
+  };
+
+  std::shared_ptr<MyBlockCache> compressed_cache_;
+  std::shared_ptr<MyBlockCache> uncompressed_cache_;
+  bool compression_enabled_;
+  std::vector<std::string> values_;
+  bool fill_cache_;
+};
+
+TEST_P(DBBasicTestWithParallelIO, MultiGet) {
+  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;
+  ro.fill_cache = fill_cache();
+
+  // 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);
+  ASSERT_TRUE(CheckValue(0, values[0].ToString()));
+  ASSERT_TRUE(CheckValue(50, values[1].ToString()));
+
+  int random_reads = env_->random_read_counter_.Read();
+  key_data[0] = Key(1);
+  key_data[1] = Key(51);
+  keys[0] = Slice(key_data[0]);
+  keys[1] = Slice(key_data[1]);
+  values[0].Reset();
+  values[1].Reset();
+  dbfull()->MultiGet(ro, dbfull()->DefaultColumnFamily(), keys.size(),
+           keys.data(), values.data(), statuses.data(), true);
+  ASSERT_TRUE(CheckValue(1, values[0].ToString()));
+  ASSERT_TRUE(CheckValue(51, values[1].ToString()));
+
+  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);
+  statuses.resize(10);
+  std::vector<int> key_ints{1,2,15,16,55,81,82,83,84,85};
+  for (size_t i = 0; i < key_ints.size(); ++i) {
+    key_data[i] = Key(key_ints[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_ints.size(); ++i) {
+    ASSERT_OK(statuses[i]);
+    ASSERT_TRUE(CheckValue(key_ints[i], values[i].ToString()));
+  }
+  expected_reads += (read_from_cache ? 2 : 4);
+  ASSERT_EQ(env_->random_read_counter_.Read(), expected_reads);
+}
+
+INSTANTIATE_TEST_CASE_P(
+    ParallelIO, DBBasicTestWithParallelIO,
+    // Params are as follows -
+    // Param 0 - Compressed cache enabled
+    // Param 1 - Uncompressed cache enabled
+    // Param 2 - Data compression enabled
+    // Param 3 - ReadOptions::fill_cache
+    ::testing::Combine(::testing::Bool(), ::testing::Bool(),
+                       ::testing::Bool(), ::testing::Bool()));
+
+class DBBasicTestWithTimestampWithParam
+    : public DBTestBase,
+      public testing::WithParamInterface<bool> {
+ public:
+  DBBasicTestWithTimestampWithParam()
+      : DBTestBase("/db_basic_test_with_timestamp") {}
+
+ protected:
+  class TestComparator : public Comparator {
+   private:
+    const Comparator* cmp_without_ts_;
+
+   public:
+    explicit TestComparator(size_t ts_sz)
+        : Comparator(ts_sz), cmp_without_ts_(nullptr) {
+      cmp_without_ts_ = BytewiseComparator();
+    }
+
+    const char* Name() const override { return "TestComparator"; }
+
+    void FindShortSuccessor(std::string*) const override {}
+
+    void FindShortestSeparator(std::string*, const Slice&) const override {}
+
+    int Compare(const Slice& a, const Slice& b) const override {
+      int r = CompareWithoutTimestamp(a, b);
+      if (r != 0 || 0 == timestamp_size()) {
+        return r;
+      }
+      return CompareTimestamp(
+          Slice(a.data() + a.size() - timestamp_size(), timestamp_size()),
+          Slice(b.data() + b.size() - timestamp_size(), timestamp_size()));
+    }
+
+    int CompareWithoutTimestamp(const Slice& a, const Slice& b) const override {
+      assert(a.size() >= timestamp_size());
+      assert(b.size() >= timestamp_size());
+      Slice k1 = StripTimestampFromUserKey(a, timestamp_size());
+      Slice k2 = StripTimestampFromUserKey(b, timestamp_size());
+
+      return cmp_without_ts_->Compare(k1, k2);
+    }
+
+    int CompareTimestamp(const Slice& ts1, const Slice& ts2) const override {
+      if (!ts1.data() && !ts2.data()) {
+        return 0;
+      } else if (ts1.data() && !ts2.data()) {
+        return 1;
+      } else if (!ts1.data() && ts2.data()) {
+        return -1;
+      }
+      assert(ts1.size() == ts2.size());
+      uint64_t low1 = 0;
+      uint64_t low2 = 0;
+      uint64_t high1 = 0;
+      uint64_t high2 = 0;
+      auto* ptr1 = const_cast<Slice*>(&ts1);
+      auto* ptr2 = const_cast<Slice*>(&ts2);
+      if (!GetFixed64(ptr1, &low1) || !GetFixed64(ptr1, &high1) ||
+          !GetFixed64(ptr2, &low2) || !GetFixed64(ptr2, &high2)) {
+        assert(false);
+      }
+      if (high1 < high2) {
+        return 1;
+      } else if (high1 > high2) {
+        return -1;
+      }
+      if (low1 < low2) {
+        return 1;
+      } else if (low1 > low2) {
+        return -1;
+      }
+      return 0;
+    }
+  };
+
+  Slice EncodeTimestamp(uint64_t low, uint64_t high, std::string* ts) {
+    assert(nullptr != ts);
+    ts->clear();
+    PutFixed64(ts, low);
+    PutFixed64(ts, high);
+    assert(ts->size() == sizeof(low) + sizeof(high));
+    return Slice(*ts);
+  }
+};
+
+TEST_P(DBBasicTestWithTimestampWithParam, PutAndGet) {
+  const int kNumKeysPerFile = 8192;
+  const size_t kNumTimestamps = 6;
+  bool memtable_only = GetParam();
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.env = env_;
+  options.memtable_factory.reset(new SpecialSkipListFactory(kNumKeysPerFile));
+  std::string tmp;
+  size_t ts_sz = EncodeTimestamp(0, 0, &tmp).size();
+  TestComparator test_cmp(ts_sz);
+  options.comparator = &test_cmp;
+  BlockBasedTableOptions bbto;
+  bbto.filter_policy.reset(NewBloomFilterPolicy(
+      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));
+      }
+    }
+  }
+  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();
+}
+
+INSTANTIATE_TEST_CASE_P(Timestamp, DBBasicTestWithTimestampWithParam,
+                        ::testing::Bool());
+
 }  // namespace rocksdb

+#ifdef ROCKSDB_UNITTESTS_WITH_CUSTOM_OBJECTS_FROM_STATIC_LIBS
+extern "C" {
+void RegisterCustomObjects(int argc, char** argv);
+}
+#else
+void RegisterCustomObjects(int /*argc*/, char** /*argv*/) {}
+#endif  // !ROCKSDB_UNITTESTS_WITH_CUSTOM_OBJECTS_FROM_STATIC_LIBS
+
 int main(int argc, char** argv) {
  rocksdb::port::InstallStackTraceHandler();
  ::testing::InitGoogleTest(&argc, argv);
+  RegisterCustomObjects(argc, argv);
  return RUN_ALL_TESTS();
 }
@@ -63,9 +63,11 @@ class DBBlobIndexTest : public DBTestBase {
    ReadOptions read_options;
    read_options.snapshot = snapshot;
    PinnableSlice value;
-    auto s = dbfull()->GetImpl(read_options, cfh(), key, &value,
-                               nullptr /*value_found*/, nullptr /*callback*/,
-                               is_blob_index);
+    DBImpl::GetImplOptions get_impl_options;
+    get_impl_options.column_family = cfh();
+    get_impl_options.value = &value;
+    get_impl_options.is_blob_index = is_blob_index;
+    auto s = dbfull()->GetImpl(read_options, key, get_impl_options);
    if (s.IsNotFound()) {
      return "NOT_FOUND";
    }
@@ -19,6 +19,9 @@ class DBBlockCacheTest : public DBTestBase {
  size_t hit_count_ = 0;
  size_t insert_count_ = 0;
  size_t failure_count_ = 0;
+  size_t compression_dict_miss_count_ = 0;
+  size_t compression_dict_hit_count_ = 0;
+  size_t compression_dict_insert_count_ = 0;
  size_t compressed_miss_count_ = 0;
  size_t compressed_hit_count_ = 0;
  size_t compressed_insert_count_ = 0;
@@ -69,6 +72,15 @@ class DBBlockCacheTest : public DBTestBase {
        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_ADD_FAILURES);
  }

+  void RecordCacheCountersForCompressionDict(const Options& options) {
+    compression_dict_miss_count_ =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_MISS);
+    compression_dict_hit_count_ =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_HIT);
+    compression_dict_insert_count_ =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_ADD);
+  }
+
  void CheckCacheCounters(const Options& options, size_t expected_misses,
                          size_t expected_hits, size_t expected_inserts,
                          size_t expected_failures) {
@@ -87,6 +99,28 @@ class DBBlockCacheTest : public DBTestBase {
    failure_count_ = new_failure_count;
  }

+  void CheckCacheCountersForCompressionDict(
+      const Options& options, size_t expected_compression_dict_misses,
+      size_t expected_compression_dict_hits,
+      size_t expected_compression_dict_inserts) {
+    size_t new_compression_dict_miss_count =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_MISS);
+    size_t new_compression_dict_hit_count =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_HIT);
+    size_t new_compression_dict_insert_count =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_ADD);
+    ASSERT_EQ(compression_dict_miss_count_ + expected_compression_dict_misses,
+              new_compression_dict_miss_count);
+    ASSERT_EQ(compression_dict_hit_count_ + expected_compression_dict_hits,
+              new_compression_dict_hit_count);
+    ASSERT_EQ(
+        compression_dict_insert_count_ + expected_compression_dict_inserts,
+        new_compression_dict_insert_count);
+    compression_dict_miss_count_ = new_compression_dict_miss_count;
+    compression_dict_hit_count_ = new_compression_dict_hit_count;
+    compression_dict_insert_count_ = new_compression_dict_insert_count;
+  }
+
  void CheckCompressedCacheCounters(const Options& options,
                                    size_t expected_misses,
                                    size_t expected_hits,
@@ -365,8 +399,11 @@ TEST_F(DBBlockCacheTest, IndexAndFilterBlocksStats) {
  ASSERT_EQ(cache->GetUsage(), index_bytes_insert + filter_bytes_insert);
  // set the cache capacity to the current usage
  cache->SetCapacity(index_bytes_insert + filter_bytes_insert);
-  ASSERT_EQ(TestGetTickerCount(options, BLOCK_CACHE_INDEX_BYTES_EVICT), 0);
-  ASSERT_EQ(TestGetTickerCount(options, BLOCK_CACHE_FILTER_BYTES_EVICT), 0);
+  // The index and filter eviction statistics were broken by the refactoring
+  // that moved the readers out of the block cache. Disabling these until we can
+  // bring the stats back.
+  // ASSERT_EQ(TestGetTickerCount(options, BLOCK_CACHE_INDEX_BYTES_EVICT), 0);
+  // ASSERT_EQ(TestGetTickerCount(options, BLOCK_CACHE_FILTER_BYTES_EVICT), 0);
  // Note that the second key needs to be no longer than the first one.
  // Otherwise the second index block may not fit in cache.
  ASSERT_OK(Put(1, "key", "val"));
@@ -377,10 +414,13 @@ TEST_F(DBBlockCacheTest, IndexAndFilterBlocksStats) {
            index_bytes_insert);
  ASSERT_GT(TestGetTickerCount(options, BLOCK_CACHE_FILTER_BYTES_INSERT),
            filter_bytes_insert);
-  ASSERT_EQ(TestGetTickerCount(options, BLOCK_CACHE_INDEX_BYTES_EVICT),
-            index_bytes_insert);
-  ASSERT_EQ(TestGetTickerCount(options, BLOCK_CACHE_FILTER_BYTES_EVICT),
-            filter_bytes_insert);
+  // The index and filter eviction statistics were broken by the refactoring
+  // that moved the readers out of the block cache. Disabling these until we can
+  // bring the stats back.
+  // ASSERT_EQ(TestGetTickerCount(options, BLOCK_CACHE_INDEX_BYTES_EVICT),
+  //           index_bytes_insert);
+  // ASSERT_EQ(TestGetTickerCount(options, BLOCK_CACHE_FILTER_BYTES_EVICT),
+  //           filter_bytes_insert);
 }

 namespace {
@@ -444,11 +484,11 @@ TEST_F(DBBlockCacheTest, IndexAndFilterBlocksCachePriority) {
              TestGetTickerCount(options, BLOCK_CACHE_ADD));
    ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_DATA_MISS));
    if (priority == Cache::Priority::LOW) {
-      ASSERT_EQ(0, MockCache::high_pri_insert_count);
-      ASSERT_EQ(2, MockCache::low_pri_insert_count);
+      ASSERT_EQ(0u, MockCache::high_pri_insert_count);
+      ASSERT_EQ(2u, MockCache::low_pri_insert_count);
    } else {
-      ASSERT_EQ(2, MockCache::high_pri_insert_count);
-      ASSERT_EQ(0, MockCache::low_pri_insert_count);
+      ASSERT_EQ(2u, MockCache::high_pri_insert_count);
+      ASSERT_EQ(0u, MockCache::low_pri_insert_count);
    }

    // Access data block.
@@ -462,11 +502,11 @@ TEST_F(DBBlockCacheTest, IndexAndFilterBlocksCachePriority) {

    // Data block should be inserted with low priority.
    if (priority == Cache::Priority::LOW) {
-      ASSERT_EQ(0, MockCache::high_pri_insert_count);
-      ASSERT_EQ(3, MockCache::low_pri_insert_count);
+      ASSERT_EQ(0u, MockCache::high_pri_insert_count);
+      ASSERT_EQ(3u, MockCache::low_pri_insert_count);
    } else {
-      ASSERT_EQ(2, MockCache::high_pri_insert_count);
-      ASSERT_EQ(1, MockCache::low_pri_insert_count);
+      ASSERT_EQ(2u, MockCache::high_pri_insert_count);
+      ASSERT_EQ(1u, MockCache::low_pri_insert_count);
    }
  }
 }
@@ -665,6 +705,8 @@ TEST_F(DBBlockCacheTest, CacheCompressionDict) {
    options.table_factory.reset(new BlockBasedTableFactory(table_options));
    DestroyAndReopen(options);

+    RecordCacheCountersForCompressionDict(options);
+
    for (int i = 0; i < kNumFiles; ++i) {
      ASSERT_EQ(i, NumTableFilesAtLevel(0, 0));
      for (int j = 0; j < kNumEntriesPerFile; ++j) {
@@ -677,27 +719,26 @@ TEST_F(DBBlockCacheTest, CacheCompressionDict) {
    ASSERT_EQ(0, NumTableFilesAtLevel(0));
    ASSERT_EQ(kNumFiles, NumTableFilesAtLevel(1));

+    // Compression dictionary blocks are preloaded.
+    CheckCacheCountersForCompressionDict(
+        options, kNumFiles /* expected_compression_dict_misses */,
+        0 /* expected_compression_dict_hits */,
+        kNumFiles /* expected_compression_dict_inserts */);
+
    // Seek to a key in a file. It should cause the SST's dictionary meta-block
    // to be read.
    RecordCacheCounters(options);
-    ASSERT_EQ(0,
-              TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_MISS));
-    ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_ADD));
-    ASSERT_EQ(
-        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_BYTES_INSERT),
-        0);
+    RecordCacheCountersForCompressionDict(options);
    ReadOptions read_options;
    ASSERT_NE("NOT_FOUND", Get(Key(kNumFiles * kNumEntriesPerFile - 1)));
-    // Two blocks missed/added: dictionary and data block
-    // One block hit: index since it's prefetched
-    CheckCacheCounters(options, 2 /* expected_misses */, 1 /* expected_hits */,
-                       2 /* expected_inserts */, 0 /* expected_failures */);
-    ASSERT_EQ(1,
-              TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_MISS));
-    ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_ADD));
-    ASSERT_GT(
-        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSION_DICT_BYTES_INSERT),
-        0);
+    // Two block hits: index and dictionary since they are prefetched
+    // One block missed/added: data block
+    CheckCacheCounters(options, 1 /* expected_misses */, 2 /* expected_hits */,
+                       1 /* expected_inserts */, 0 /* expected_failures */);
+    CheckCacheCountersForCompressionDict(
+        options, 0 /* expected_compression_dict_misses */,
+        1 /* expected_compression_dict_hits */,
+        0 /* expected_compression_dict_inserts */);
  }
 }

@@ -1095,6 +1095,8 @@ TEST_F(DBBloomFilterTest, PrefixScan) {
    options.max_background_compactions = 2;
    options.create_if_missing = true;
    options.memtable_factory.reset(NewHashSkipListRepFactory(16));
+    assert(!options.unordered_write);
+    // It is incompatible with allow_concurrent_memtable_write=false
    options.allow_concurrent_memtable_write = false;

    BlockBasedTableOptions table_options;
@@ -720,7 +720,7 @@ TEST_F(DBTestCompactionFilter, CompactionFilterIgnoreSnapshot) {
  cfilter_count = 0;
  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
  // The filter should delete 40 records.
-  ASSERT_EQ(40U, cfilter_count);
+  ASSERT_EQ(40, cfilter_count);

  {
    // Scan the entire database as of the snapshot to ensure
@@ -13,9 +13,9 @@
 #include "rocksdb/concurrent_task_limiter.h"
 #include "rocksdb/experimental.h"
 #include "rocksdb/utilities/convenience.h"
+#include "test_util/fault_injection_test_env.h"
+#include "test_util/sync_point.h"
 #include "util/concurrent_task_limiter_impl.h"
-#include "util/fault_injection_test_env.h"
-#include "util/sync_point.h"

 namespace rocksdb {

@@ -141,7 +141,7 @@ Options DeletionTriggerOptions(Options options) {
  options.compression = kNoCompression;
  options.write_buffer_size = kCDTKeysPerBuffer * (kCDTValueSize + 24);
  options.min_write_buffer_number_to_merge = 1;
-  options.max_write_buffer_number_to_maintain = 0;
+  options.max_write_buffer_size_to_maintain = 0;
  options.num_levels = kCDTNumLevels;
  options.level0_file_num_compaction_trigger = 1;
  options.target_file_size_base = options.write_buffer_size * 2;
@@ -497,14 +497,14 @@ TEST_F(DBCompactionTest, TestTableReaderForCompaction) {

  // Create new iterator for:
  // (1) 1 for verifying flush results
-  // (2) 3 for compaction input files
-  // (3) 1 for verifying compaction results.
-  ASSERT_EQ(num_new_table_reader, 5);
+  // (2) 1 for verifying compaction results.
+  // (3) New TableReaders will not be created for compaction inputs
+  ASSERT_EQ(num_new_table_reader, 2);

  num_table_cache_lookup = 0;
  num_new_table_reader = 0;
  ASSERT_EQ(Key(1), Get(Key(1)));
-  ASSERT_EQ(num_table_cache_lookup + old_num_table_cache_lookup2, 3);
+  ASSERT_EQ(num_table_cache_lookup + old_num_table_cache_lookup2, 5);
  ASSERT_EQ(num_new_table_reader, 0);

  num_table_cache_lookup = 0;
@@ -519,13 +519,14 @@ TEST_F(DBCompactionTest, TestTableReaderForCompaction) {
  // May preload table cache too.
  ASSERT_GE(num_table_cache_lookup, 1);
  old_num_table_cache_lookup2 = num_table_cache_lookup;
-  // One for compaction input, one for verifying compaction results.
-  ASSERT_EQ(num_new_table_reader, 2);
+  // One for verifying compaction results.
+  // No new iterator created for compaction.
+  ASSERT_EQ(num_new_table_reader, 1);

  num_table_cache_lookup = 0;
  num_new_table_reader = 0;
  ASSERT_EQ(Key(1), Get(Key(1)));
-  ASSERT_EQ(num_table_cache_lookup + old_num_table_cache_lookup2, 2);
+  ASSERT_EQ(num_table_cache_lookup + old_num_table_cache_lookup2, 3);
  ASSERT_EQ(num_new_table_reader, 0);

  rocksdb::SyncPoint::GetInstance()->ClearAllCallBacks();
@@ -731,7 +732,7 @@ TEST_F(DBCompactionTest, BGCompactionsAllowed) {

  // Now all column families qualify compaction but only one should be
  // scheduled, because no column family hits speed up condition.
-  ASSERT_EQ(1, env_->GetThreadPoolQueueLen(Env::Priority::LOW));
+  ASSERT_EQ(1u, env_->GetThreadPoolQueueLen(Env::Priority::LOW));

  // Create two more files for one column family, which triggers speed up
  // condition, three compactions will be scheduled.
@@ -745,7 +746,7 @@ TEST_F(DBCompactionTest, BGCompactionsAllowed) {
    ASSERT_EQ(options.level0_file_num_compaction_trigger + num + 1,
              NumTableFilesAtLevel(0, 2));
  }
-  ASSERT_EQ(3, env_->GetThreadPoolQueueLen(Env::Priority::LOW));
+  ASSERT_EQ(3U, env_->GetThreadPoolQueueLen(Env::Priority::LOW));

  // Unblock all threads to unblock all compactions.
  for (size_t i = 0; i < kTotalTasks; i++) {
@@ -776,7 +777,7 @@ TEST_F(DBCompactionTest, BGCompactionsAllowed) {

  // Now all column families qualify compaction but only one should be
  // scheduled, because no column family hits speed up condition.
-  ASSERT_EQ(1, env_->GetThreadPoolQueueLen(Env::Priority::LOW));
+  ASSERT_EQ(1U, env_->GetThreadPoolQueueLen(Env::Priority::LOW));

  for (size_t i = 0; i < kTotalTasks; i++) {
    sleeping_tasks[i].WakeUp();
@@ -3890,11 +3891,17 @@ TEST_F(DBCompactionTest, CompactRangeShutdownWhileDelayed) {
      }
      Flush(1);
    }
-    auto manual_compaction_thread = port::Thread([this]() {
+    auto manual_compaction_thread = port::Thread([this, i]() {
      CompactRangeOptions cro;
      cro.allow_write_stall = false;
-      ASSERT_TRUE(db_->CompactRange(cro, handles_[1], nullptr, nullptr)
-                      .IsShutdownInProgress());
+      Status s = db_->CompactRange(cro, handles_[1], nullptr, nullptr);
+      if (i == 0) {
+        ASSERT_TRUE(db_->CompactRange(cro, handles_[1], nullptr, nullptr)
+                        .IsColumnFamilyDropped());
+      } else {
+        ASSERT_TRUE(db_->CompactRange(cro, handles_[1], nullptr, nullptr)
+                        .IsShutdownInProgress());
+      }
    });

    TEST_SYNC_POINT(
@@ -4158,7 +4165,7 @@ TEST_F(DBCompactionTest, CompactionLimiter) {

  const char* cf_names[] = {"default", "0", "1", "2", "3", "4", "5",
    "6", "7", "8", "9", "a", "b", "c", "d", "e", "f" };
-  const int cf_count = sizeof cf_names / sizeof cf_names[0];
+  const unsigned int cf_count = sizeof cf_names / sizeof cf_names[0];

  std::unordered_map<std::string, CompactionLimiter*> cf_to_limiter;

@@ -4177,7 +4184,7 @@ TEST_F(DBCompactionTest, CompactionLimiter) {
  std::vector<Options> option_vector;
  option_vector.reserve(cf_count);

-  for (int cf = 0; cf < cf_count; cf++) {
+  for (unsigned int cf = 0; cf < cf_count; cf++) {
    ColumnFamilyOptions cf_opt(options);
    if (cf == 0) {
      // "Default" CF does't use compaction limiter
@@ -4195,7 +4202,7 @@ TEST_F(DBCompactionTest, CompactionLimiter) {
    option_vector.emplace_back(DBOptions(options), cf_opt);
  }

-  for (int cf = 1; cf < cf_count; cf++) {
+  for (unsigned int cf = 1; cf < cf_count; cf++) {
    CreateColumnFamilies({cf_names[cf]}, option_vector[cf]);
  }

@@ -4247,7 +4254,7 @@ TEST_F(DBCompactionTest, CompactionLimiter) {
  int keyIndex = 0;

  for (int n = 0; n < options.level0_file_num_compaction_trigger; n++) {
-    for (int cf = 0; cf < cf_count; cf++) {
+    for (unsigned int cf = 0; cf < cf_count; cf++) {
      for (int i = 0; i < kNumKeysPerFile; i++) {
        ASSERT_OK(Put(cf, Key(keyIndex++), ""));
      }
@@ -4255,13 +4262,13 @@ TEST_F(DBCompactionTest, CompactionLimiter) {
      ASSERT_OK(Put(cf, "", ""));
    }

-    for (int cf = 0; cf < cf_count; cf++) {
+    for (unsigned int cf = 0; cf < cf_count; cf++) {
      dbfull()->TEST_WaitForFlushMemTable(handles_[cf]);
    }
  }

  // Enough L0 files to trigger compaction
-  for (int cf = 0; cf < cf_count; cf++) {
+  for (unsigned int cf = 0; cf < cf_count; cf++) {
    ASSERT_EQ(NumTableFilesAtLevel(0, cf),
      options.level0_file_num_compaction_trigger);
  }
@@ -4288,7 +4295,7 @@ TEST_F(DBCompactionTest, CompactionLimiter) {
    sleeping_compact_tasks[i].WaitUntilDone();
  }

-  for (int cf = 0; cf < cf_count; cf++) {
+  for (unsigned int cf = 0; cf < cf_count; cf++) {
    dbfull()->TEST_WaitForFlushMemTable(handles_[cf]);
  }

@@ -4333,12 +4340,6 @@ TEST_P(DBCompactionDirectIOTest, DirectIO) {
  options.env = new MockEnv(Env::Default());
  Reopen(options);
  bool readahead = false;
-  SyncPoint::GetInstance()->SetCallBack(
-      "TableCache::NewIterator:for_compaction", [&](void* arg) {
-        bool* use_direct_reads = static_cast<bool*>(arg);
-        ASSERT_EQ(*use_direct_reads,
-                  options.use_direct_reads);
-      });
  SyncPoint::GetInstance()->SetCallBack(
      "CompactionJob::OpenCompactionOutputFile", [&](void* arg) {
        bool* use_direct_writes = static_cast<bool*>(arg);
@@ -4551,6 +4552,36 @@ TEST_F(DBCompactionTest, ManualCompactionBottomLevelOptimized) {
  ASSERT_EQ(num, 0);
 }

+TEST_F(DBCompactionTest, CompactionDuringShutdown) {
+  Options opts = CurrentOptions();
+  opts.level0_file_num_compaction_trigger = 2;
+  opts.disable_auto_compactions = true;
+  DestroyAndReopen(opts);
+  ColumnFamilyHandleImpl* cfh =
+      static_cast<ColumnFamilyHandleImpl*>(dbfull()->DefaultColumnFamily());
+  ColumnFamilyData* cfd = cfh->cfd();
+  InternalStats* internal_stats_ptr = cfd->internal_stats();
+  ASSERT_NE(internal_stats_ptr, nullptr);
+
+  Random rnd(301);
+  for (auto i = 0; i < 2; ++i) {
+    for (auto j = 0; j < 10; ++j) {
+      ASSERT_OK(
+          Put("foo" + std::to_string(i * 10 + j), RandomString(&rnd, 1024)));
+    }
+    Flush();
+  }
+
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::BackgroundCompaction:NonTrivial:BeforeRun",
+      [&](void* /*arg*/) {
+    dbfull()->shutting_down_.store(true);
+  });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+  dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+  ASSERT_OK(dbfull()->error_handler_.GetBGError());
+}
+
 // FixFileIngestionCompactionDeadlock tests and verifies that compaction and
 // file ingestion do not cause deadlock in the event of write stall triggered
 // by number of L0 files reaching level0_stop_writes_trigger.
@@ -4627,7 +4658,31 @@ TEST_P(DBCompactionTestWithParam, FixFileIngestionCompactionDeadlock) {
  ASSERT_OK(dbfull()->TEST_WaitForCompact());
  Close();
 }
+TEST_F(DBCompactionTest, ConsistencyFailTest) {
+  Options options = CurrentOptions();
+  DestroyAndReopen(options);

+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "VersionBuilder::CheckConsistency", [&](void* arg) {
+      auto p =
+            reinterpret_cast<std::pair<FileMetaData**, FileMetaData**>*>(arg);
+        // just swap the two FileMetaData so that we hit error
+        // in CheckConsistency funcion
+        FileMetaData* temp = *(p->first);
+        *(p->first) = *(p->second);
+        *(p->second) = temp;
+      });
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  for (int k = 0; k < 2; ++k) {
+    ASSERT_OK(Put("foo", "bar"));
+    Flush();
+  }
+
+  ASSERT_NOK(Put("foo", "bar"));
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+}
 #endif // !defined(ROCKSDB_LITE)
 }  // namespace rocksdb

@@ -7,7 +7,7 @@
 #include "port/stack_trace.h"
 #include "rocksdb/perf_context.h"
 #if !defined(ROCKSDB_LITE)
-#include "util/sync_point.h"
+#include "test_util/sync_point.h"
 #endif
 #include <iostream>
 #include <string>
@@ -83,6 +83,34 @@ TEST_F(DBEncryptionTest, CheckEncrypted) {
  }
 }

+TEST_F(DBEncryptionTest, ReadEmptyFile) {
+  auto defaultEnv = Env::Default();
+
+  // create empty file for reading it back in later
+  auto envOptions = EnvOptions(CurrentOptions());
+  auto filePath = dbname_ + "/empty.empty";
+
+  Status status;
+  {
+    std::unique_ptr<WritableFile> writableFile;
+    status = defaultEnv->NewWritableFile(filePath, &writableFile, envOptions);
+    ASSERT_OK(status);
+  }
+
+  std::unique_ptr<SequentialFile> seqFile;
+  status = defaultEnv->NewSequentialFile(filePath, &seqFile, envOptions);
+  ASSERT_OK(status);
+
+  std::string scratch;
+  Slice data;
+  // reading back 16 bytes from the empty file shouldn't trigger an assertion.
+  // it should just work and return an empty string
+  status = seqFile->Read(16, &data, (char*)scratch.data());
+  ASSERT_OK(status);
+
+  ASSERT_TRUE(data.empty());
+}
+
 #endif // ROCKSDB_LITE

 }  // namespace rocksdb
@@ -6,24 +6,20 @@

 #ifndef ROCKSDB_LITE

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-
-#include <inttypes.h>
+#include <cinttypes>
 #include <stdint.h>
 #include <algorithm>
 #include <string>
-#include "db/db_impl.h"
+#include "db/db_impl/db_impl.h"
 #include "db/job_context.h"
 #include "db/version_set.h"
+#include "file/file_util.h"
+#include "file/filename.h"
 #include "port/port.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
-#include "util/file_util.h"
-#include "util/filename.h"
+#include "test_util/sync_point.h"
 #include "util/mutexlock.h"
-#include "util/sync_point.h"

 namespace rocksdb {

@@ -61,7 +57,9 @@ Status DBImpl::EnableFileDeletions(bool force) {
  }
  if (file_deletion_enabled) {
    ROCKS_LOG_INFO(immutable_db_options_.info_log, "File Deletions Enabled");
-    PurgeObsoleteFiles(job_context);
+    if (job_context.HaveSomethingToDelete()) {
+      PurgeObsoleteFiles(job_context);
+    }
  } else {
    ROCKS_LOG_WARN(immutable_db_options_.info_log,
                   "File Deletions Enable, but not really enabled. Counter: %d",
@@ -165,6 +163,16 @@ Status DBImpl::GetSortedWalFiles(VectorLogPtr& files) {
  return wal_manager_.GetSortedWalFiles(files);
 }

+Status DBImpl::GetCurrentWalFile(std::unique_ptr<LogFile>* current_log_file) {
+  uint64_t current_logfile_number;
+  {
+    InstrumentedMutexLock l(&mutex_);
+    current_logfile_number = logfile_number_;
+  }
+
+  return wal_manager_.GetLiveWalFile(current_logfile_number, current_log_file);
+}
+
 }

 #endif  // ROCKSDB_LITE
@@ -9,8 +9,8 @@

 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
-#include "util/fault_injection_test_env.h"
-#include "util/sync_point.h"
+#include "test_util/fault_injection_test_env.h"
+#include "test_util/sync_point.h"

 namespace rocksdb {

@@ -290,6 +290,39 @@ TEST_F(DBFlushTest, ManualFlushFailsInReadOnlyMode) {
  Close();
 }

+TEST_F(DBFlushTest, CFDropRaceWithWaitForFlushMemTables) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  CreateAndReopenWithCF({"pikachu"}, options);
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->LoadDependency(
+      {{"DBImpl::FlushMemTable:AfterScheduleFlush",
+        "DBFlushTest::CFDropRaceWithWaitForFlushMemTables:BeforeDrop"},
+       {"DBFlushTest::CFDropRaceWithWaitForFlushMemTables:AfterFree",
+        "DBImpl::BackgroundCallFlush:start"},
+       {"DBImpl::BackgroundCallFlush:start",
+        "DBImpl::FlushMemTable:BeforeWaitForBgFlush"}});
+  SyncPoint::GetInstance()->EnableProcessing();
+  ASSERT_EQ(2, handles_.size());
+  ASSERT_OK(Put(1, "key", "value"));
+  auto* cfd = static_cast<ColumnFamilyHandleImpl*>(handles_[1])->cfd();
+  port::Thread drop_cf_thr([&]() {
+    TEST_SYNC_POINT(
+        "DBFlushTest::CFDropRaceWithWaitForFlushMemTables:BeforeDrop");
+    ASSERT_OK(dbfull()->DropColumnFamily(handles_[1]));
+    ASSERT_OK(dbfull()->DestroyColumnFamilyHandle(handles_[1]));
+    handles_.resize(1);
+    TEST_SYNC_POINT(
+        "DBFlushTest::CFDropRaceWithWaitForFlushMemTables:AfterFree");
+  });
+  FlushOptions flush_opts;
+  flush_opts.allow_write_stall = true;
+  ASSERT_NOK(dbfull()->TEST_FlushMemTable(cfd, flush_opts));
+  drop_cf_thr.join();
+  Close();
+  SyncPoint::GetInstance()->DisableProcessing();
+}
+
 TEST_P(DBAtomicFlushTest, ManualAtomicFlush) {
  Options options = CurrentOptions();
  options.create_if_missing = true;
@@ -431,7 +464,7 @@ TEST_P(DBAtomicFlushTest, FlushMultipleCFs_DropSomeBeforeRequestFlush) {
    cf_ids.push_back(cf_id);
  }
  ASSERT_OK(dbfull()->DropColumnFamily(handles_[1]));
-  ASSERT_TRUE(Flush(cf_ids).IsShutdownInProgress());
+  ASSERT_TRUE(Flush(cf_ids).IsColumnFamilyDropped());
  Destroy(options);
 }

@@ -514,6 +547,80 @@ TEST_P(DBAtomicFlushTest, TriggerFlushAndClose) {
  ASSERT_EQ("value", Get(0, "key"));
 }

+TEST_P(DBAtomicFlushTest, PickMemtablesRaceWithBackgroundFlush) {
+  bool atomic_flush = GetParam();
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.atomic_flush = atomic_flush;
+  options.max_write_buffer_number = 4;
+  // Set min_write_buffer_number_to_merge to be greater than 1, so that
+  // a column family with one memtable in the imm will not cause IsFlushPending
+  // to return true when flush_requested_ is false.
+  options.min_write_buffer_number_to_merge = 2;
+  CreateAndReopenWithCF({"pikachu"}, options);
+  ASSERT_EQ(2, handles_.size());
+  ASSERT_OK(dbfull()->PauseBackgroundWork());
+  ASSERT_OK(Put(0, "key00", "value00"));
+  ASSERT_OK(Put(1, "key10", "value10"));
+  FlushOptions flush_opts;
+  flush_opts.wait = false;
+  ASSERT_OK(dbfull()->Flush(flush_opts, handles_));
+  ASSERT_OK(Put(0, "key01", "value01"));
+  // Since max_write_buffer_number is 4, the following flush won't cause write
+  // stall.
+  ASSERT_OK(dbfull()->Flush(flush_opts));
+  ASSERT_OK(dbfull()->DropColumnFamily(handles_[1]));
+  ASSERT_OK(dbfull()->DestroyColumnFamilyHandle(handles_[1]));
+  handles_[1] = nullptr;
+  ASSERT_OK(dbfull()->ContinueBackgroundWork());
+  ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable(handles_[0]));
+  delete handles_[0];
+  handles_.clear();
+}
+
+TEST_P(DBAtomicFlushTest, CFDropRaceWithWaitForFlushMemTables) {
+  bool atomic_flush = GetParam();
+  if (!atomic_flush) {
+    return;
+  }
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.atomic_flush = atomic_flush;
+  CreateAndReopenWithCF({"pikachu"}, options);
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->LoadDependency(
+      {{"DBImpl::AtomicFlushMemTables:AfterScheduleFlush",
+        "DBAtomicFlushTest::CFDropRaceWithWaitForFlushMemTables:BeforeDrop"},
+       {"DBAtomicFlushTest::CFDropRaceWithWaitForFlushMemTables:AfterFree",
+        "DBImpl::BackgroundCallFlush:start"},
+       {"DBImpl::BackgroundCallFlush:start",
+        "DBImpl::AtomicFlushMemTables:BeforeWaitForBgFlush"}});
+  SyncPoint::GetInstance()->EnableProcessing();
+  ASSERT_EQ(2, handles_.size());
+  ASSERT_OK(Put(0, "key", "value"));
+  ASSERT_OK(Put(1, "key", "value"));
+  auto* cfd_default =
+      static_cast<ColumnFamilyHandleImpl*>(dbfull()->DefaultColumnFamily())
+          ->cfd();
+  auto* cfd_pikachu = static_cast<ColumnFamilyHandleImpl*>(handles_[1])->cfd();
+  port::Thread drop_cf_thr([&]() {
+    TEST_SYNC_POINT(
+        "DBAtomicFlushTest::CFDropRaceWithWaitForFlushMemTables:BeforeDrop");
+    ASSERT_OK(dbfull()->DropColumnFamily(handles_[1]));
+    delete handles_[1];
+    handles_.resize(1);
+    TEST_SYNC_POINT(
+        "DBAtomicFlushTest::CFDropRaceWithWaitForFlushMemTables:AfterFree");
+  });
+  FlushOptions flush_opts;
+  flush_opts.allow_write_stall = true;
+  ASSERT_OK(dbfull()->TEST_AtomicFlushMemTables({cfd_default, cfd_pikachu},
+                                                flush_opts));
+  drop_cf_thr.join();
+  Close();
+  SyncPoint::GetInstance()->DisableProcessing();
+}
+
 INSTANTIATE_TEST_CASE_P(DBFlushDirectIOTest, DBFlushDirectIOTest,
                        testing::Bool());

@@ -6,23 +6,21 @@
 // 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 "db/db_impl.h"
+#include "db/db_impl/db_impl.h"

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-#include <inttypes.h>
+#include <cinttypes>

 #include "db/builder.h"
 #include "db/error_handler.h"
 #include "db/event_helpers.h"
+#include "file/sst_file_manager_impl.h"
 #include "monitoring/iostats_context_imp.h"
 #include "monitoring/perf_context_imp.h"
 #include "monitoring/thread_status_updater.h"
 #include "monitoring/thread_status_util.h"
+#include "test_util/sync_point.h"
+#include "util/cast_util.h"
 #include "util/concurrent_task_limiter_impl.h"
-#include "util/sst_file_manager_impl.h"
-#include "util/sync_point.h"

 namespace rocksdb {

@@ -110,6 +108,13 @@ Status DBImpl::SyncClosedLogs(JobContext* job_context) {
      if (!s.ok()) {
        break;
      }
+
+      if (immutable_db_options_.recycle_log_file_num > 0) {
+        s = log->Close();
+        if (!s.ok()) {
+          break;
+        }
+      }
    }
    if (s.ok()) {
      s = directories_.GetWalDir()->Fsync();
@@ -201,7 +206,7 @@ Status DBImpl::FlushMemTableToOutputFile(
                     cfd->current()->storage_info()->LevelSummary(&tmp));
  }

-  if (!s.ok() && !s.IsShutdownInProgress()) {
+  if (!s.ok() && !s.IsShutdownInProgress() && !s.IsColumnFamilyDropped()) {
    Status new_bg_error = s;
    error_handler_.SetBGError(new_bg_error, BackgroundErrorReason::kFlush);
  }
@@ -254,7 +259,7 @@ Status DBImpl::FlushMemTablesToOutputFiles(
        snapshot_checker, log_buffer, thread_pri);
    if (!s.ok()) {
      status = s;
-      if (!s.IsShutdownInProgress()) {
+      if (!s.IsShutdownInProgress() && !s.IsColumnFamilyDropped()) {
        // At this point, DB is not shutting down, nor is cfd dropped.
        // Something is wrong, thus we break out of the loop.
        break;
@@ -385,7 +390,8 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
    for (const auto& e : exec_status) {
      if (!e.second.ok()) {
        s = e.second;
-        if (!e.second.IsShutdownInProgress()) {
+        if (!e.second.IsShutdownInProgress() &&
+            !e.second.IsColumnFamilyDropped()) {
          // If a flush job did not return OK, and the CF is not dropped, and
          // the DB is not shutting down, then we have to return this result to
          // caller later.
@@ -397,15 +403,11 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(
    s = error_status.ok() ? s : error_status;
  }

-  // If db is NOT shutting down, and one or more column families have been
-  // dropped.
-  // TODO: use separate status code for db shutdown and column family dropped.
-  if (s.IsShutdownInProgress() &&
-      !shutting_down_.load(std::memory_order_acquire)) {
+  if (s.IsColumnFamilyDropped()) {
    s = Status::OK();
  }

-  if (s.ok() || s.IsShutdownInProgress()) {
+  if (s.ok() || s.IsShutdownInProgress() || s.IsColumnFamilyDropped()) {
    // Sync on all distinct output directories.
    for (auto dir : distinct_output_dirs) {
      if (dir != nullptr) {
@@ -523,7 +525,7 @@ Status DBImpl::AtomicFlushMemTablesToOutputFiles(

  // Need to undo atomic flush if something went wrong, i.e. s is not OK and
  // it is not because of CF drop.
-  if (!s.ok() && !s.IsShutdownInProgress()) {
+  if (!s.ok() && !s.IsColumnFamilyDropped()) {
    // Have to cancel the flush jobs that have NOT executed because we need to
    // unref the versions.
    for (int i = 0; i != num_cfs; ++i) {
@@ -798,6 +800,7 @@ Status DBImpl::CompactRange(const CompactRangeOptions& options,
  return s;
 }

+namespace {
 class SnapshotListFetchCallbackImpl : public SnapshotListFetchCallback {
 public:
  SnapshotListFetchCallbackImpl(DBImpl* db_impl, Env* env,
@@ -820,6 +823,7 @@ class SnapshotListFetchCallbackImpl : public SnapshotListFetchCallback {
  DBImpl* db_impl_;
  Logger* info_log_;
 };
+}  // namespace

 Status DBImpl::CompactFiles(const CompactionOptions& compact_options,
                            ColumnFamilyHandle* column_family,
@@ -1005,8 +1009,10 @@ Status DBImpl::CompactFilesImpl(
      c->mutable_cf_options()->paranoid_file_checks,
      c->mutable_cf_options()->report_bg_io_stats, dbname_,
      &compaction_job_stats, Env::Priority::USER,
-      immutable_db_options_.max_subcompactions <= 1 ? &fetch_callback
-                                                    : nullptr);
+      immutable_db_options_.max_subcompactions <= 1 &&
+              c->mutable_cf_options()->snap_refresh_nanos > 0
+          ? &fetch_callback
+          : nullptr);

  // Creating a compaction influences the compaction score because the score
  // takes running compactions into account (by skipping files that are already
@@ -1050,7 +1056,7 @@ Status DBImpl::CompactFilesImpl(

  if (status.ok()) {
    // Done
-  } else if (status.IsShutdownInProgress()) {
+  } else if (status.IsColumnFamilyDropped() || status.IsShutdownInProgress()) {
    // Ignore compaction errors found during shutting down
  } else {
    ROCKS_LOG_WARN(immutable_db_options_.info_log,
@@ -1548,13 +1554,39 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
    if (!cfd->mem()->IsEmpty() || !cached_recoverable_state_empty_.load()) {
      s = SwitchMemtable(cfd, &context);
    }
-
    if (s.ok()) {
      if (cfd->imm()->NumNotFlushed() != 0 || !cfd->mem()->IsEmpty() ||
          !cached_recoverable_state_empty_.load()) {
        flush_memtable_id = cfd->imm()->GetLatestMemTableID();
        flush_req.emplace_back(cfd, flush_memtable_id);
      }
+      if (immutable_db_options_.persist_stats_to_disk) {
+        ColumnFamilyData* cfd_stats =
+            versions_->GetColumnFamilySet()->GetColumnFamily(
+                kPersistentStatsColumnFamilyName);
+        if (cfd_stats != nullptr && cfd_stats != cfd &&
+            !cfd_stats->mem()->IsEmpty()) {
+          // only force flush stats CF when it will be the only CF lagging
+          // behind after the current flush
+          bool stats_cf_flush_needed = true;
+          for (auto* loop_cfd : *versions_->GetColumnFamilySet()) {
+            if (loop_cfd == cfd_stats || loop_cfd == cfd) {
+              continue;
+            }
+            if (loop_cfd->GetLogNumber() <= cfd_stats->GetLogNumber()) {
+              stats_cf_flush_needed = false;
+            }
+          }
+          if (stats_cf_flush_needed) {
+            ROCKS_LOG_INFO(immutable_db_options_.info_log,
+                           "Force flushing stats CF with manual flush of %s "
+                           "to avoid holding old logs", cfd->GetName().c_str());
+            s = SwitchMemtable(cfd_stats, &context);
+            flush_memtable_id = cfd_stats->imm()->GetLatestMemTableID();
+            flush_req.emplace_back(cfd_stats, flush_memtable_id);
+          }
+        }
+      }
    }

    if (s.ok() && !flush_req.empty()) {
@@ -1562,6 +1594,16 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
        ColumnFamilyData* loop_cfd = elem.first;
        loop_cfd->imm()->FlushRequested();
      }
+      // If the caller wants to wait for this flush to complete, it indicates
+      // that the caller expects the ColumnFamilyData not to be free'ed by
+      // other threads which may drop the column family concurrently.
+      // Therefore, we increase the cfd's ref count.
+      if (flush_options.wait) {
+        for (auto& elem : flush_req) {
+          ColumnFamilyData* loop_cfd = elem.first;
+          loop_cfd->Ref();
+        }
+      }
      SchedulePendingFlush(flush_req, flush_reason);
      MaybeScheduleFlushOrCompaction();
    }
@@ -1570,7 +1612,8 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
      write_thread_.ExitUnbatched(&w);
    }
  }
-
+  TEST_SYNC_POINT("DBImpl::FlushMemTable:AfterScheduleFlush");
+  TEST_SYNC_POINT("DBImpl::FlushMemTable:BeforeWaitForBgFlush");
  if (s.ok() && flush_options.wait) {
    autovector<ColumnFamilyData*> cfds;
    autovector<const uint64_t*> flush_memtable_ids;
@@ -1580,6 +1623,13 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
    }
    s = WaitForFlushMemTables(cfds, flush_memtable_ids,
                              (flush_reason == FlushReason::kErrorRecovery));
+    for (auto* tmp_cfd : cfds) {
+      if (tmp_cfd->Unref()) {
+        // Only one thread can reach here.
+        InstrumentedMutexLock lock_guard(&mutex_);
+        delete tmp_cfd;
+      }
+    }
  }
  TEST_SYNC_POINT("FlushMemTableFinished");
  return s;
@@ -1643,6 +1693,15 @@ Status DBImpl::AtomicFlushMemTables(
      for (auto cfd : cfds) {
        cfd->imm()->FlushRequested();
      }
+      // If the caller wants to wait for this flush to complete, it indicates
+      // that the caller expects the ColumnFamilyData not to be free'ed by
+      // other threads which may drop the column family concurrently.
+      // Therefore, we increase the cfd's ref count.
+      if (flush_options.wait) {
+        for (auto cfd : cfds) {
+          cfd->Ref();
+        }
+      }
      GenerateFlushRequest(cfds, &flush_req);
      SchedulePendingFlush(flush_req, flush_reason);
      MaybeScheduleFlushOrCompaction();
@@ -1653,7 +1712,7 @@ Status DBImpl::AtomicFlushMemTables(
    }
  }
  TEST_SYNC_POINT("DBImpl::AtomicFlushMemTables:AfterScheduleFlush");
-
+  TEST_SYNC_POINT("DBImpl::AtomicFlushMemTables:BeforeWaitForBgFlush");
  if (s.ok() && flush_options.wait) {
    autovector<const uint64_t*> flush_memtable_ids;
    for (auto& iter : flush_req) {
@@ -1661,6 +1720,13 @@ Status DBImpl::AtomicFlushMemTables(
    }
    s = WaitForFlushMemTables(cfds, flush_memtable_ids,
                              (flush_reason == FlushReason::kErrorRecovery));
+    for (auto* cfd : cfds) {
+      if (cfd->Unref()) {
+        // Only one thread can reach here.
+        InstrumentedMutexLock lock_guard(&mutex_);
+        delete cfd;
+      }
+    }
  }
  return s;
 }
@@ -1695,7 +1761,10 @@ Status DBImpl::WaitUntilFlushWouldNotStallWrites(ColumnFamilyData* cfd,
                       cfd->GetName().c_str());
        bg_cv_.Wait();
      }
-      if (cfd->IsDropped() || shutting_down_.load(std::memory_order_acquire)) {
+      if (cfd->IsDropped()) {
+        return Status::ColumnFamilyDropped();
+      }
+      if (shutting_down_.load(std::memory_order_acquire)) {
        return Status::ShutdownInProgress();
      }

@@ -2004,7 +2073,8 @@ void DBImpl::BGWorkFlush(void* arg) {

  IOSTATS_SET_THREAD_POOL_ID(fta.thread_pri_);
  TEST_SYNC_POINT("DBImpl::BGWorkFlush");
-  reinterpret_cast<DBImpl*>(fta.db_)->BackgroundCallFlush(fta.thread_pri_);
+  static_cast_with_check<DBImpl, DB>(fta.db_)->BackgroundCallFlush(
+      fta.thread_pri_);
  TEST_SYNC_POINT("DBImpl::BGWorkFlush:done");
 }

@@ -2015,7 +2085,7 @@ void DBImpl::BGWorkCompaction(void* arg) {
  TEST_SYNC_POINT("DBImpl::BGWorkCompaction");
  auto prepicked_compaction =
      static_cast<PrepickedCompaction*>(ca.prepicked_compaction);
-  reinterpret_cast<DBImpl*>(ca.db)->BackgroundCallCompaction(
+  static_cast_with_check<DBImpl, DB>(ca.db)->BackgroundCallCompaction(
      prepicked_compaction, Env::Priority::LOW);
  delete prepicked_compaction;
 }
@@ -2080,6 +2150,7 @@ Status DBImpl::BackgroundFlush(bool* made_progress, JobContext* job_context,
  autovector<BGFlushArg> bg_flush_args;
  std::vector<SuperVersionContext>& superversion_contexts =
      job_context->superversion_contexts;
+  autovector<ColumnFamilyData*> column_families_not_to_flush;
  while (!flush_queue_.empty()) {
    // This cfd is already referenced
    const FlushRequest& flush_req = PopFirstFromFlushQueue();
@@ -2090,9 +2161,7 @@ Status DBImpl::BackgroundFlush(bool* made_progress, JobContext* job_context,
      ColumnFamilyData* cfd = iter.first;
      if (cfd->IsDropped() || !cfd->imm()->IsFlushPending()) {
        // can't flush this CF, try next one
-        if (cfd->Unref()) {
-          delete cfd;
-        }
+        column_families_not_to_flush.push_back(cfd);
        continue;
      }
      superversion_contexts.emplace_back(SuperVersionContext(true));
@@ -2120,6 +2189,7 @@ Status DBImpl::BackgroundFlush(bool* made_progress, JobContext* job_context,
    }
    status = FlushMemTablesToOutputFiles(bg_flush_args, made_progress,
                                         job_context, log_buffer, thread_pri);
+    TEST_SYNC_POINT("DBImpl::BackgroundFlush:BeforeFlush");
    // All the CFDs in the FlushReq must have the same flush reason, so just
    // grab the first one
    *reason = bg_flush_args[0].cfd_->GetFlushReason();
@@ -2131,6 +2201,11 @@ Status DBImpl::BackgroundFlush(bool* made_progress, JobContext* job_context,
      }
    }
  }
+  for (auto cfd : column_families_not_to_flush) {
+    if (cfd->Unref()) {
+      delete cfd;
+    }
+  }
  return status;
 }

@@ -2153,7 +2228,7 @@ void DBImpl::BackgroundCallFlush(Env::Priority thread_pri) {

    Status s = BackgroundFlush(&made_progress, &job_context, &log_buffer,
                               &reason, thread_pri);
-    if (!s.ok() && !s.IsShutdownInProgress() &&
+    if (!s.ok() && !s.IsShutdownInProgress() && !s.IsColumnFamilyDropped() &&
        reason != FlushReason::kErrorRecovery) {
      // Wait a little bit before retrying background flush in
      // case this is an environmental problem and we do not want to
@@ -2178,7 +2253,8 @@ void DBImpl::BackgroundCallFlush(Env::Priority thread_pri) {

    // If flush failed, we want to delete all temporary files that we might have
    // created. Thus, we force full scan in FindObsoleteFiles()
-    FindObsoleteFiles(&job_context, !s.ok() && !s.IsShutdownInProgress());
+    FindObsoleteFiles(&job_context, !s.ok() && !s.IsShutdownInProgress() &&
+                                        !s.IsColumnFamilyDropped());
    // delete unnecessary files if any, this is done outside the mutex
    if (job_context.HaveSomethingToClean() ||
        job_context.HaveSomethingToDelete() || !log_buffer.IsEmpty()) {
@@ -2242,7 +2318,8 @@ void DBImpl::BackgroundCallCompaction(PrepickedCompaction* prepicked_compaction,
      mutex_.Unlock();
      env_->SleepForMicroseconds(10000);  // prevent hot loop
      mutex_.Lock();
-    } else if (!s.ok() && !s.IsShutdownInProgress()) {
+    } else if (!s.ok() && !s.IsShutdownInProgress() &&
+               !s.IsColumnFamilyDropped()) {
      // Wait a little bit before retrying background compaction in
      // case this is an environmental problem and we do not want to
      // chew up resources for failed compactions for the duration of
@@ -2266,7 +2343,8 @@ void DBImpl::BackgroundCallCompaction(PrepickedCompaction* prepicked_compaction,
    // If compaction failed, we want to delete all temporary files that we might
    // have created (they might not be all recorded in job_context in case of a
    // failure). Thus, we force full scan in FindObsoleteFiles()
-    FindObsoleteFiles(&job_context, !s.ok() && !s.IsShutdownInProgress());
+    FindObsoleteFiles(&job_context, !s.ok() && !s.IsShutdownInProgress() &&
+                                        !s.IsColumnFamilyDropped());
    TEST_SYNC_POINT("DBImpl::BackgroundCallCompaction:FoundObsoleteFiles");

    // delete unnecessary files if any, this is done outside the mutex
@@ -2663,14 +2741,18 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
        &event_logger_, c->mutable_cf_options()->paranoid_file_checks,
        c->mutable_cf_options()->report_bg_io_stats, dbname_,
        &compaction_job_stats, thread_pri,
-        immutable_db_options_.max_subcompactions <= 1 ? &fetch_callback
-                                                      : nullptr);
+        immutable_db_options_.max_subcompactions <= 1 &&
+                c->mutable_cf_options()->snap_refresh_nanos > 0
+            ? &fetch_callback
+            : nullptr);
    compaction_job.Prepare();

    NotifyOnCompactionBegin(c->column_family_data(), c.get(), status,
                            compaction_job_stats, job_context->job_id);

    mutex_.Unlock();
+    TEST_SYNC_POINT_CALLBACK(
+        "DBImpl::BackgroundCompaction:NonTrivial:BeforeRun", nullptr);
    compaction_job.Run();
    TEST_SYNC_POINT("DBImpl::BackgroundCompaction:NonTrivial:AfterRun");
    mutex_.Lock();
@@ -2704,7 +2786,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,

  if (status.ok() || status.IsCompactionTooLarge()) {
    // Done
-  } else if (status.IsShutdownInProgress()) {
+  } else if (status.IsColumnFamilyDropped() || status.IsShutdownInProgress()) {
    // Ignore compaction errors found during shutting down
  } else {
    ROCKS_LOG_WARN(immutable_db_options_.info_log, "Compaction error: %s",
@@ -2788,7 +2870,7 @@ void DBImpl::RemoveManualCompaction(DBImpl::ManualCompactionState* m) {
      it = manual_compaction_dequeue_.erase(it);
      return;
    }
-    it++;
+    ++it;
  }
  assert(false);
  return;
@@ -2809,7 +2891,7 @@ bool DBImpl::ShouldntRunManualCompaction(ManualCompactionState* m) {
  bool seen = false;
  while (it != manual_compaction_dequeue_.end()) {
    if (m == (*it)) {
-      it++;
+      ++it;
      seen = true;
      continue;
    } else if (MCOverlap(m, (*it)) && (!seen && !(*it)->in_progress)) {
@@ -2818,7 +2900,7 @@ bool DBImpl::ShouldntRunManualCompaction(ManualCompactionState* m) {
      // and (*it) is ahead in the queue and is not yet in progress
      return true;
    }
-    it++;
+    ++it;
  }
  return false;
 }
@@ -2836,7 +2918,7 @@ bool DBImpl::HaveManualCompaction(ColumnFamilyData* cfd) {
      // in progress
      return true;
    }
-    it++;
+    ++it;
  }
  return false;
 }
@@ -2849,7 +2931,7 @@ bool DBImpl::HasExclusiveManualCompaction() {
    if ((*it)->exclusive) {
      return true;
    }
-    it++;
+    ++it;
  }
  return false;
 }
@@ -9,12 +9,13 @@

 #ifndef NDEBUG

-#include "db/db_impl.h"
+#include "db/column_family.h"
+#include "db/db_impl/db_impl.h"
 #include "db/error_handler.h"
 #include "monitoring/thread_status_updater.h"
+#include "util/cast_util.h"

 namespace rocksdb {
-
 uint64_t DBImpl::TEST_GetLevel0TotalSize() {
  InstrumentedMutexLock l(&mutex_);
  return default_cf_handle_->cfd()->current()->storage_info()->NumLevelBytes(0);
@@ -122,6 +123,16 @@ Status DBImpl::TEST_FlushMemTable(bool wait, bool allow_write_stall,
  return FlushMemTable(cfd, fo, FlushReason::kTest);
 }

+Status DBImpl::TEST_FlushMemTable(ColumnFamilyData* cfd,
+                                  const FlushOptions& flush_opts) {
+  return FlushMemTable(cfd, flush_opts, FlushReason::kTest);
+}
+
+Status DBImpl::TEST_AtomicFlushMemTables(
+    const autovector<ColumnFamilyData*>& cfds, const FlushOptions& flush_opts) {
+  return AtomicFlushMemTables(cfds, flush_opts, FlushReason::kTest);
+}
+
 Status DBImpl::TEST_WaitForFlushMemTable(ColumnFamilyHandle* column_family) {
  ColumnFamilyData* cfd;
  if (column_family == nullptr) {
@@ -262,8 +273,8 @@ bool DBImpl::TEST_IsPersistentStatsEnabled() const {
  return thread_persist_stats_ && thread_persist_stats_->IsRunning();
 }

-size_t DBImpl::TEST_EstiamteStatsHistorySize() const {
-  return EstiamteStatsHistorySize();
+size_t DBImpl::TEST_EstimateInMemoryStatsHistorySize() const {
+  return EstimateInMemoryStatsHistorySize();
 }
 }  // namespace rocksdb
 #endif  // NDEBUG
@@ -7,13 +7,9 @@
 // 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 "db/db_impl.h"
+#include "db/db_impl/db_impl.h"

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-
-#include <inttypes.h>
+#include <cinttypes>
 #include <vector>

 #include "db/column_family.h"
@@ -6,18 +6,15 @@
 // 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 "db/db_impl.h"
+#include "db/db_impl/db_impl.h"

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-#include <inttypes.h>
+#include <cinttypes>
 #include <set>
 #include <unordered_set>
 #include "db/event_helpers.h"
 #include "db/memtable_list.h"
-#include "util/file_util.h"
-#include "util/sst_file_manager_impl.h"
+#include "file/file_util.h"
+#include "file/sst_file_manager_impl.h"

 namespace rocksdb {

@@ -261,7 +258,8 @@ void DBImpl::DeleteObsoleteFileImpl(int job_id, const std::string& fname,
  Status file_deletion_status;
  if (type == kTableFile || type == kLogFile) {
    file_deletion_status =
-        DeleteDBFile(&immutable_db_options_, fname, path_to_sync);
+        DeleteDBFile(&immutable_db_options_, fname, path_to_sync,
+                     /*force_bg=*/false, /*force_fg=*/!wal_in_db_path_);
  } else {
    file_deletion_status = env_->DeleteFile(fname);
  }
@@ -318,11 +316,9 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
      candidate_files.size() + state.sst_delete_files.size() +
      state.log_delete_files.size() + state.manifest_delete_files.size());
  // We may ignore the dbname when generating the file names.
-  const char* kDumbDbName = "";
  for (auto& file : state.sst_delete_files) {
    candidate_files.emplace_back(
-        MakeTableFileName(kDumbDbName, file.metadata->fd.GetNumber()),
-        file.path);
+        MakeTableFileName(file.metadata->fd.GetNumber()), file.path);
    if (file.metadata->table_reader_handle) {
      table_cache_->Release(file.metadata->table_reader_handle);
    }
@@ -331,7 +327,7 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {

  for (auto file_num : state.log_delete_files) {
    if (file_num > 0) {
-      candidate_files.emplace_back(LogFileName(kDumbDbName, file_num),
+      candidate_files.emplace_back(LogFileName(file_num),
                                   immutable_db_options_.wal_dir);
    }
  }
@@ -6,21 +6,19 @@
 // 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 "db/db_impl.h"
+#include "db/db_impl/db_impl.h"

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-#include <inttypes.h>
+#include <cinttypes>

 #include "db/builder.h"
 #include "db/error_handler.h"
+#include "file/sst_file_manager_impl.h"
+#include "monitoring/persistent_stats_history.h"
 #include "options/options_helper.h"
 #include "rocksdb/wal_filter.h"
-#include "table/block_based_table_factory.h"
+#include "table/block_based/block_based_table_factory.h"
+#include "test_util/sync_point.h"
 #include "util/rate_limiter.h"
-#include "util/sst_file_manager_impl.h"
-#include "util/sync_point.h"

 namespace rocksdb {
 Options SanitizeOptions(const std::string& dbname, const Options& src) {
@@ -124,6 +122,25 @@ DBOptions SanitizeOptions(const std::string& dbname, const DBOptions& src) {
  }

 #ifndef ROCKSDB_LITE
+  ImmutableDBOptions immutable_db_options(result);
+  if (!IsWalDirSameAsDBPath(&immutable_db_options)) {
+    // Either the WAL dir and db_paths[0]/db_name are not the same, or we
+    // cannot tell for sure. In either case, assume they're different and
+    // explicitly cleanup the trash log files (bypass DeleteScheduler)
+    // Do this first so even if we end up calling
+    // DeleteScheduler::CleanupDirectory on the same dir later, it will be
+    // safe
+    std::vector<std::string> filenames;
+    result.env->GetChildren(result.wal_dir, &filenames);
+    for (std::string& filename : filenames) {
+      if (filename.find(".log.trash",
+                  filename.length() - std::string(".log.trash").length()) !=
+                  std::string::npos) {
+        std::string trash_file = result.wal_dir + "/" + filename;
+        result.env->DeleteFile(trash_file);
+      }
+    }
+  }
  // When the DB is stopped, it's possible that there are some .trash files that
  // were not deleted yet, when we open the DB we will find these .trash files
  // and schedule them to be deleted (or delete immediately if SstFileManager
@@ -145,7 +162,6 @@ DBOptions SanitizeOptions(const std::string& dbname, const DBOptions& src) {
 }

 namespace {
-
 Status SanitizeOptionsByTable(
    const DBOptions& db_opts,
    const std::vector<ColumnFamilyDescriptor>& column_families) {
@@ -158,52 +174,23 @@ Status SanitizeOptionsByTable(
  }
  return Status::OK();
 }
+}  // namespace

-static Status ValidateOptions(
+Status DBImpl::ValidateOptions(
    const DBOptions& db_options,
    const std::vector<ColumnFamilyDescriptor>& column_families) {
  Status s;
-
  for (auto& cfd : column_families) {
-    s = CheckCompressionSupported(cfd.options);
-    if (s.ok() && db_options.allow_concurrent_memtable_write) {
-      s = CheckConcurrentWritesSupported(cfd.options);
-    }
-    if (s.ok()) {
-      s = CheckCFPathsSupported(db_options, cfd.options);
-    }
+    s = ColumnFamilyData::ValidateOptions(db_options, cfd.options);
    if (!s.ok()) {
      return s;
    }
-
-    if (cfd.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 (cfd.options.table_factory->Name() !=
-          BlockBasedTableFactory().Name()) {
-        return Status::NotSupported(
-            "TTL is only supported in Block-Based Table format. ");
-      }
-    }
-
-    if (cfd.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 (cfd.options.table_factory->Name() !=
-          BlockBasedTableFactory().Name()) {
-        return Status::NotSupported(
-            "Periodic Compaction is only supported in "
-            "Block-Based Table format. ");
-      }
-    }
  }
+  s = ValidateOptions(db_options);
+  return s;
+}

+Status DBImpl::ValidateOptions(const DBOptions& db_options) {
  if (db_options.db_paths.size() > 4) {
    return Status::NotSupported(
        "More than four DB paths are not supported yet. ");
@@ -228,17 +215,35 @@ static Status ValidateOptions(
    return Status::InvalidArgument("keep_log_file_num must be greater than 0");
  }

+  if (db_options.unordered_write &&
+      !db_options.allow_concurrent_memtable_write) {
+    return Status::InvalidArgument(
+        "unordered_write is incompatible with !allow_concurrent_memtable_write");
+  }
+
+  if (db_options.unordered_write && db_options.enable_pipelined_write) {
+    return Status::InvalidArgument(
+        "unordered_write is incompatible with enable_pipelined_write");
+  }
+
  return Status::OK();
 }
-}  // namespace
+
 Status DBImpl::NewDB() {
  VersionEdit new_db;
+  Status s = SetIdentityFile(env_, dbname_);
+  if (!s.ok()) {
+    return s;
+  }
+  if (immutable_db_options_.write_dbid_to_manifest) {
+    std::string temp_db_id;
+    GetDbIdentityFromIdentityFile(&temp_db_id);
+    new_db.SetDBId(temp_db_id);
+  }
  new_db.SetLogNumber(0);
  new_db.SetNextFile(2);
  new_db.SetLastSequence(0);

-  Status s;
-
  ROCKS_LOG_INFO(immutable_db_options_.info_log, "Creating manifest 1 \n");
  const std::string manifest = DescriptorFileName(dbname_, 1);
  {
@@ -286,9 +291,9 @@ Status DBImpl::CreateAndNewDirectory(Env* env, const std::string& dirname,
  return env->NewDirectory(dirname, directory);
 }

-Status DBImpl::Directories::SetDirectories(
-    Env* env, const std::string& dbname, const std::string& wal_dir,
-    const std::vector<DbPath>& data_paths) {
+Status Directories::SetDirectories(Env* env, const std::string& dbname,
+                                   const std::string& wal_dir,
+                                   const std::vector<DbPath>& data_paths) {
  Status s = DBImpl::CreateAndNewDirectory(env, dbname, &db_dir_);
  if (!s.ok()) {
    return s;
@@ -341,6 +346,9 @@ Status DBImpl::Recover(
    s = env_->FileExists(CurrentFileName(dbname_));
    if (s.IsNotFound()) {
      if (immutable_db_options_.create_if_missing) {
+        // Has to be called only after Identity File creation is successful
+        // because DB ID is stored in Manifest if
+        // immutable_db_options_.write_dbid_to_manifest = true
        s = NewDB();
        is_new_db = true;
        if (!s.ok()) {
@@ -360,30 +368,19 @@ Status DBImpl::Recover(
      assert(s.IsIOError());
      return s;
    }
-    // Check for the IDENTITY file and create it if not there
-    s = env_->FileExists(IdentityFileName(dbname_));
-    if (s.IsNotFound()) {
-      s = SetIdentityFile(env_, dbname_);
-      if (!s.ok()) {
-        return s;
-      }
-    } else if (!s.ok()) {
-      assert(s.IsIOError());
-      return s;
-    }
    // Verify compatibility of env_options_ and filesystem
    {
      std::unique_ptr<RandomAccessFile> idfile;
      EnvOptions customized_env(env_options_);
      customized_env.use_direct_reads |=
          immutable_db_options_.use_direct_io_for_flush_and_compaction;
-      s = env_->NewRandomAccessFile(IdentityFileName(dbname_), &idfile,
+      s = env_->NewRandomAccessFile(CurrentFileName(dbname_), &idfile,
                                    customized_env);
      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(IdentityFileName(dbname_), &idfile,
+        s = env_->NewRandomAccessFile(CurrentFileName(dbname_), &idfile,
                                      customized_env);
        if (s.ok()) {
          return Status::InvalidArgument(
@@ -395,8 +392,43 @@ Status DBImpl::Recover(
      }
    }
  }
+  assert(db_id_.empty());
+  Status s = versions_->Recover(column_families, read_only, &db_id_);
+  if (!s.ok()) {
+    return s;
+  }
+  // Happens when immutable_db_options_.write_dbid_to_manifest is set to true
+  // the very first time.
+  if (db_id_.empty()) {
+    // Check for the IDENTITY file and create it if not there.
+    s = env_->FileExists(IdentityFileName(dbname_));
+    // 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.
+    if (s.IsNotFound()) {
+      s = SetIdentityFile(env_, dbname_);
+      if (!s.ok()) {
+        return s;
+      }
+    } else if (!s.ok()) {
+      assert(s.IsIOError());
+      return s;
+    }
+    GetDbIdentityFromIdentityFile(&db_id_);
+    if (immutable_db_options_.write_dbid_to_manifest) {
+      VersionEdit edit;
+      edit.SetDBId(db_id_);
+      Options options;
+      MutableCFOptions mutable_cf_options(options);
+      versions_->db_id_ = db_id_;
+      versions_->LogAndApply(versions_->GetColumnFamilySet()->GetDefault(),
+                             mutable_cf_options, &edit, &mutex_, nullptr,
+                             false);
+    }
+  } else {
+    SetIdentityFile(env_, dbname_, db_id_);
+  }

-  Status s = versions_->Recover(column_families, read_only);
  if (immutable_db_options_.paranoid_checks && s.ok()) {
    s = CheckConsistency();
  }
@@ -408,6 +440,10 @@ Status DBImpl::Recover(
      }
    }
  }
+  // DB mutex is already held
+  if (s.ok() && immutable_db_options_.persist_stats_to_disk) {
+    s = InitPersistStatsColumnFamily();
+  }

  // Initial max_total_in_memory_state_ before recovery logs. Log recovery
  // may check this value to decide whether to flush.
@@ -423,6 +459,8 @@ Status DBImpl::Recover(
    default_cf_handle_ = new ColumnFamilyHandleImpl(
        versions_->GetColumnFamilySet()->GetDefault(), this, &mutex_);
    default_cf_internal_stats_ = default_cf_handle_->cfd()->internal_stats();
+    // TODO(Zhongyi): handle single_column_family_mode_ when
+    // persistent_stats is enabled
    single_column_family_mode_ =
        versions_->GetColumnFamilySet()->NumberOfColumnFamilies() == 1;

@@ -518,6 +556,98 @@ Status DBImpl::Recover(
  return s;
 }

+Status DBImpl::PersistentStatsProcessFormatVersion() {
+  mutex_.AssertHeld();
+  Status s;
+  // persist version when stats CF doesn't exist
+  bool should_persist_format_version = !persistent_stats_cfd_exists_;
+  mutex_.Unlock();
+  if (persistent_stats_cfd_exists_) {
+    // Check persistent stats format version compatibility. Drop and recreate
+    // persistent stats CF if format version is incompatible
+    uint64_t format_version_recovered = 0;
+    Status s_format = DecodePersistentStatsVersionNumber(
+        this, StatsVersionKeyType::kFormatVersion, &format_version_recovered);
+    uint64_t compatible_version_recovered = 0;
+    Status s_compatible = DecodePersistentStatsVersionNumber(
+        this, StatsVersionKeyType::kCompatibleVersion,
+        &compatible_version_recovered);
+    // abort reading from existing stats CF if any of following is true:
+    // 1. failed to read format version or compatible version from disk
+    // 2. sst's format version is greater than current format version, meaning
+    // this sst is encoded with a newer RocksDB release, and current compatible
+    // version is below the sst's compatible version
+    if (!s_format.ok() || !s_compatible.ok() ||
+        (kStatsCFCurrentFormatVersion < format_version_recovered &&
+         kStatsCFCompatibleFormatVersion < compatible_version_recovered)) {
+      if (!s_format.ok() || !s_compatible.ok()) {
+        ROCKS_LOG_INFO(
+            immutable_db_options_.info_log,
+            "Reading persistent stats version key failed. Format key: %s, "
+            "compatible key: %s",
+            s_format.ToString().c_str(), s_compatible.ToString().c_str());
+      } else {
+        ROCKS_LOG_INFO(
+            immutable_db_options_.info_log,
+            "Disable persistent stats due to corrupted or incompatible format "
+            "version\n");
+      }
+      DropColumnFamily(persist_stats_cf_handle_);
+      DestroyColumnFamilyHandle(persist_stats_cf_handle_);
+      ColumnFamilyHandle* handle = nullptr;
+      ColumnFamilyOptions cfo;
+      OptimizeForPersistentStats(&cfo);
+      s = CreateColumnFamily(cfo, kPersistentStatsColumnFamilyName, &handle);
+      persist_stats_cf_handle_ = static_cast<ColumnFamilyHandleImpl*>(handle);
+      // should also persist version here because old stats CF is discarded
+      should_persist_format_version = true;
+    }
+  }
+  if (s.ok() && should_persist_format_version) {
+    // Persistent stats CF being created for the first time, need to write
+    // format version key
+    WriteBatch batch;
+    batch.Put(persist_stats_cf_handle_, kFormatVersionKeyString,
+              ToString(kStatsCFCurrentFormatVersion));
+    batch.Put(persist_stats_cf_handle_, kCompatibleVersionKeyString,
+              ToString(kStatsCFCompatibleFormatVersion));
+    WriteOptions wo;
+    wo.low_pri = true;
+    wo.no_slowdown = true;
+    wo.sync = false;
+    s = Write(wo, &batch);
+  }
+  mutex_.Lock();
+  return s;
+}
+
+Status DBImpl::InitPersistStatsColumnFamily() {
+  mutex_.AssertHeld();
+  assert(!persist_stats_cf_handle_);
+  ColumnFamilyData* persistent_stats_cfd =
+      versions_->GetColumnFamilySet()->GetColumnFamily(
+          kPersistentStatsColumnFamilyName);
+  persistent_stats_cfd_exists_ = persistent_stats_cfd != nullptr;
+
+  Status s;
+  if (persistent_stats_cfd != nullptr) {
+    // We are recovering from a DB which already contains persistent stats CF,
+    // the CF is already created in VersionSet::ApplyOneVersionEdit, but
+    // column family handle was not. Need to explicitly create handle here.
+    persist_stats_cf_handle_ =
+        new ColumnFamilyHandleImpl(persistent_stats_cfd, this, &mutex_);
+  } else {
+    mutex_.Unlock();
+    ColumnFamilyHandle* handle = nullptr;
+    ColumnFamilyOptions cfo;
+    OptimizeForPersistentStats(&cfo);
+    s = CreateColumnFamily(cfo, kPersistentStatsColumnFamilyName, &handle);
+    persist_stats_cf_handle_ = static_cast<ColumnFamilyHandleImpl*>(handle);
+    mutex_.Lock();
+  }
+  return s;
+}
+
 // REQUIRES: log_numbers are sorted in ascending order
 Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
                               SequenceNumber* next_sequence, bool read_only) {
@@ -577,12 +707,13 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
  bool stop_replay_for_corruption = false;
  bool flushed = false;
  uint64_t corrupted_log_number = kMaxSequenceNumber;
+  uint64_t min_log_number = MinLogNumberToKeep();
  for (auto log_number : log_numbers) {
-    if (log_number < versions_->min_log_number_to_keep_2pc()) {
+    if (log_number < min_log_number) {
      ROCKS_LOG_INFO(immutable_db_options_.info_log,
                     "Skipping log #%" PRIu64
                     " since it is older than min log to keep #%" PRIu64,
-                     log_number, versions_->min_log_number_to_keep_2pc());
+                     log_number, min_log_number);
      continue;
    }
    // The previous incarnation may not have written any MANIFEST
@@ -623,7 +754,8 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
          continue;
        }
      }
-      file_reader.reset(new SequentialFileReader(std::move(file), fname));
+      file_reader.reset(new SequentialFileReader(
+          std::move(file), fname, immutable_db_options_.log_readahead_size));
    }

    // Create the log reader.
@@ -763,9 +895,10 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
      // That's why we set ignore missing column families to true
      bool has_valid_writes = false;
      status = WriteBatchInternal::InsertInto(
-          &batch, column_family_memtables_.get(), &flush_scheduler_, true,
-          log_number, this, false /* concurrent_memtable_writes */,
-          next_sequence, &has_valid_writes, seq_per_batch_, batch_per_txn_);
+          &batch, column_family_memtables_.get(), &flush_scheduler_,
+          &trim_history_scheduler_, true, log_number, this,
+          false /* concurrent_memtable_writes */, next_sequence,
+          &has_valid_writes, seq_per_batch_, batch_per_txn_);
      MaybeIgnoreError(&status);
      if (!status.ok()) {
        // We are treating this as a failure while reading since we read valid
@@ -832,6 +965,7 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
    }

    flush_scheduler_.Clear();
+    trim_history_scheduler_.Clear();
    auto last_sequence = *next_sequence - 1;
    if ((*next_sequence != kMaxSequenceNumber) &&
        (versions_->LastSequence() <= last_sequence)) {
@@ -1086,12 +1220,23 @@ Status DB::Open(const Options& options, const std::string& dbname, DB** dbptr) {
  std::vector<ColumnFamilyDescriptor> column_families;
  column_families.push_back(
      ColumnFamilyDescriptor(kDefaultColumnFamilyName, cf_options));
+  if (db_options.persist_stats_to_disk) {
+    column_families.push_back(
+        ColumnFamilyDescriptor(kPersistentStatsColumnFamilyName, cf_options));
+  }
  std::vector<ColumnFamilyHandle*> handles;
  Status s = DB::Open(db_options, dbname, column_families, &handles, dbptr);
  if (s.ok()) {
-    assert(handles.size() == 1);
+    if (db_options.persist_stats_to_disk) {
+      assert(handles.size() == 2);
+    } else {
+      assert(handles.size() == 1);
+    }
    // i can delete the handle since DBImpl is always holding a reference to
    // default column family
+    if (db_options.persist_stats_to_disk && handles[1] != nullptr) {
+      delete handles[1];
+    }
    delete handles[0];
  }
  return s;
@@ -1204,6 +1349,10 @@ Status DBImpl::Open(const DBOptions& db_options, const std::string& dbname,
    delete impl;
    return s;
  }
+
+  impl->wal_in_db_path_ =
+      IsWalDirSameAsDBPath(&impl->immutable_db_options_);
+
  impl->mutex_.Lock();
  // Handles create_if_missing, error_if_exists
  s = impl->Recover(column_families);
@@ -1268,6 +1417,10 @@ Status DBImpl::Open(const DBOptions& db_options, const std::string& dbname,
      s = impl->directories_.GetDbDir()->Fsync();
    }
  }
+  if (s.ok() && impl->immutable_db_options_.persist_stats_to_disk) {
+    // try to read format version but no need to fail Open() even if it fails
+    s = impl->PersistentStatsProcessFormatVersion();
+  }

  if (s.ok()) {
    for (auto cfd : *impl->versions_->GetColumnFamilySet()) {
@@ -3,10 +3,10 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).

-#include "db/db_impl_readonly.h"
+#include "db/db_impl/db_impl_readonly.h"

 #include "db/compacted_db_impl.h"
-#include "db/db_impl.h"
+#include "db/db_impl/db_impl.h"
 #include "db/db_iter.h"
 #include "db/merge_context.h"
 #include "monitoring/perf_context_imp.h"
@@ -9,13 +9,17 @@

 #include <string>
 #include <vector>
-#include "db/db_impl.h"
+#include "db/db_impl/db_impl.h"

 namespace rocksdb {

 class DBImplReadOnly : public DBImpl {
 public:
  DBImplReadOnly(const DBOptions& options, const std::string& dbname);
+  // No copying allowed
+  DBImplReadOnly(const DBImplReadOnly&) = delete;
+  void operator=(const DBImplReadOnly&) = delete;
+
  virtual ~DBImplReadOnly();

  // Implementations of the DB interface
@@ -115,12 +119,18 @@ class DBImplReadOnly : public DBImpl {
    return Status::NotSupported("Not supported operation in read only mode.");
  }

+  using DB::CreateColumnFamilyWithImport;
+  virtual Status CreateColumnFamilyWithImport(
+      const ColumnFamilyOptions& /*options*/,
+      const std::string& /*column_family_name*/,
+      const ImportColumnFamilyOptions& /*import_options*/,
+      const ExportImportFilesMetaData& /*metadata*/,
+      ColumnFamilyHandle** /*handle*/) override {
+    return Status::NotSupported("Not supported operation in read only mode.");
+  }
+
 private:
  friend class DB;
-
-  // No copying allowed
-  DBImplReadOnly(const DBImplReadOnly&);
-  void operator=(const DBImplReadOnly&);
 };
 }  // namespace rocksdb

@@ -3,22 +3,18 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).

-#include "db/db_impl_secondary.h"
+#include "db/db_impl/db_impl_secondary.h"

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-#include <inttypes.h>
+#include <cinttypes>

 #include "db/db_iter.h"
 #include "db/merge_context.h"
+#include "logging/auto_roll_logger.h"
 #include "monitoring/perf_context_imp.h"
-#include "util/auto_roll_logger.h"

 namespace rocksdb {

 #ifndef ROCKSDB_LITE
-
 DBImplSecondary::DBImplSecondary(const DBOptions& db_options,
                                 const std::string& dbname)
    : DBImpl(db_options, dbname) {
@@ -35,6 +31,7 @@ Status DBImplSecondary::Recover(
    bool /*error_if_data_exists_in_logs*/) {
  mutex_.AssertHeld();

+  JobContext job_context(0);
  Status s;
  s = static_cast<ReactiveVersionSet*>(versions_.get())
          ->Recover(column_families, &manifest_reader_, &manifest_reporter_,
@@ -59,49 +56,73 @@ Status DBImplSecondary::Recover(
    single_column_family_mode_ =
        versions_->GetColumnFamilySet()->NumberOfColumnFamilies() == 1;

-    // Recover from all newer log files than the ones named in the
-    // descriptor.
-    std::vector<std::string> filenames;
-    s = env_->GetChildren(immutable_db_options_.wal_dir, &filenames);
-    if (s.IsNotFound()) {
-      return Status::InvalidArgument("Failed to open wal_dir",
-                                     immutable_db_options_.wal_dir);
-    } else if (!s.ok()) {
-      return s;
-    }
-
-    std::vector<uint64_t> logs;
-    // if log_readers_ is non-empty, it means we have applied all logs with log
-    // numbers smaller than the smallest log in log_readers_, so there is no
-    // need to pass these logs to RecoverLogFiles
-    uint64_t log_number_min = 0;
-    if (log_readers_.size() > 0) {
-      log_number_min = log_readers_.begin()->first;
-    }
-    for (size_t i = 0; i < filenames.size(); i++) {
-      uint64_t number;
-      FileType type;
-      if (ParseFileName(filenames[i], &number, &type) && type == kLogFile &&
-          number >= log_number_min) {
-        logs.push_back(number);
-      }
-    }
-
-    if (!logs.empty()) {
-      // Recover in the order in which the logs were generated
-      std::sort(logs.begin(), logs.end());
-      SequenceNumber next_sequence(kMaxSequenceNumber);
-      s = RecoverLogFiles(logs, &next_sequence, true /*read_only*/);
-    }
+    std::unordered_set<ColumnFamilyData*> cfds_changed;
+    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();
+  }
  // TODO: update options_file_number_ needed?

+  job_context.Clean();
+  return s;
+}
+
+// find new WAL and apply them in order to the secondary instance
+Status DBImplSecondary::FindAndRecoverLogFiles(
+    std::unordered_set<ColumnFamilyData*>* cfds_changed,
+    JobContext* job_context) {
+  assert(nullptr != cfds_changed);
+  assert(nullptr != job_context);
+  Status s;
+  std::vector<uint64_t> logs;
+  s = FindNewLogNumbers(&logs);
+  if (s.ok() && !logs.empty()) {
+    SequenceNumber next_sequence(kMaxSequenceNumber);
+    s = RecoverLogFiles(logs, &next_sequence, cfds_changed, job_context);
+  }
+  return s;
+}
+
+// List wal_dir and find all new WALs, return these log numbers
+Status DBImplSecondary::FindNewLogNumbers(std::vector<uint64_t>* logs) {
+  assert(logs != nullptr);
+  std::vector<std::string> filenames;
+  Status s;
+  s = env_->GetChildren(immutable_db_options_.wal_dir, &filenames);
+  if (s.IsNotFound()) {
+    return Status::InvalidArgument("Failed to open wal_dir",
+                                   immutable_db_options_.wal_dir);
+  } else if (!s.ok()) {
+    return s;
+  }
+
+  // if log_readers_ is non-empty, it means we have applied all logs with log
+  // numbers smaller than the smallest log in log_readers_, so there is no
+  // need to pass these logs to RecoverLogFiles
+  uint64_t log_number_min = 0;
+  if (!log_readers_.empty()) {
+    log_number_min = log_readers_.begin()->first;
+  }
+  for (size_t i = 0; i < filenames.size(); i++) {
+    uint64_t number;
+    FileType type;
+    if (ParseFileName(filenames[i], &number, &type) && type == kLogFile &&
+        number >= log_number_min) {
+      logs->push_back(number);
+    }
+  }
+  // Recover logs in the order that they were generated
+  if (!logs->empty()) {
+    std::sort(logs->begin(), logs->end());
+  }
  return s;
 }

-// try to find log reader using log_number from log_readers_ map, initialize
-// if it doesn't exist
 Status DBImplSecondary::MaybeInitLogReader(
    uint64_t log_number, log::FragmentBufferedReader** log_reader) {
  auto iter = log_readers_.find(log_number);
@@ -129,7 +150,8 @@ Status DBImplSecondary::MaybeInitLogReader(
        *log_reader = nullptr;
        return status;
      }
-      file_reader.reset(new SequentialFileReader(std::move(file), fname));
+      file_reader.reset(new SequentialFileReader(
+          std::move(file), fname, immutable_db_options_.log_readahead_size));
    }

    // Create the log reader.
@@ -149,7 +171,10 @@ Status DBImplSecondary::MaybeInitLogReader(
 // REQUIRES: log_numbers are sorted in ascending order
 Status DBImplSecondary::RecoverLogFiles(
    const std::vector<uint64_t>& log_numbers, SequenceNumber* next_sequence,
-    bool /*read_only*/) {
+    std::unordered_set<ColumnFamilyData*>* cfds_changed,
+    JobContext* job_context) {
+  assert(nullptr != cfds_changed);
+  assert(nullptr != job_context);
  mutex_.AssertHeld();
  Status status;
  for (auto log_number : log_numbers) {
@@ -182,9 +207,56 @@ Status DBImplSecondary::RecoverLogFiles(
        continue;
      }
      WriteBatchInternal::SetContents(&batch, record);
-      // do not check sequence number because user may toggle disableWAL
-      // between writes which breaks sequence number continuity guarantee
-
+      SequenceNumber seq_of_batch = WriteBatchInternal::Sequence(&batch);
+      std::vector<uint32_t> column_family_ids;
+      status = CollectColumnFamilyIdsFromWriteBatch(batch, &column_family_ids);
+      if (status.ok()) {
+        for (const auto id : column_family_ids) {
+          ColumnFamilyData* cfd =
+              versions_->GetColumnFamilySet()->GetColumnFamily(id);
+          if (cfd == nullptr) {
+            continue;
+          }
+          if (cfds_changed->count(cfd) == 0) {
+            cfds_changed->insert(cfd);
+          }
+          const std::vector<FileMetaData*>& l0_files =
+              cfd->current()->storage_info()->LevelFiles(0);
+          SequenceNumber seq =
+              l0_files.empty() ? 0 : l0_files.back()->fd.largest_seqno;
+          // If the write batch's sequence number is smaller than the last
+          // sequence number of the largest sequence persisted for this column
+          // family, then its data must reside in an SST that has already been
+          // added in the prior MANIFEST replay.
+          if (seq_of_batch <= seq) {
+            continue;
+          }
+          auto curr_log_num = port::kMaxUint64;
+          if (cfd_to_current_log_.count(cfd) > 0) {
+            curr_log_num = cfd_to_current_log_[cfd];
+          }
+          // If the active memtable contains records added by replaying an
+          // earlier WAL, then we need to seal the memtable, add it to the
+          // immutable memtable list and create a new active memtable.
+          if (!cfd->mem()->IsEmpty() && (curr_log_num == port::kMaxUint64 ||
+                                         curr_log_num != log_number)) {
+            const MutableCFOptions mutable_cf_options =
+                *cfd->GetLatestMutableCFOptions();
+            MemTable* new_mem =
+                cfd->ConstructNewMemtable(mutable_cf_options, seq_of_batch);
+            cfd->mem()->SetNextLogNumber(log_number);
+            cfd->imm()->Add(cfd->mem(), &job_context->memtables_to_free);
+            new_mem->Ref();
+            cfd->SetMemtable(new_mem);
+          }
+        }
+        bool has_valid_writes = false;
+        status = WriteBatchInternal::InsertInto(
+            &batch, column_family_memtables_.get(),
+            nullptr /* flush_scheduler */, nullptr /* trim_history_scheduler*/,
+            true, log_number, this, false /* concurrent_memtable_writes */,
+            next_sequence, &has_valid_writes, seq_per_batch_, batch_per_txn_);
+      }
      // If column family was not found, it might mean that the WAL write
      // batch references to the column family that was dropped after the
      // insert. We don't want to fail the whole write batch in that case --
@@ -192,36 +264,43 @@ Status DBImplSecondary::RecoverLogFiles(
      // That's why we set ignore missing column families to true
      // passing null flush_scheduler will disable memtable flushing which is
      // needed for secondary instances
-      bool has_valid_writes = false;
-      status = WriteBatchInternal::InsertInto(
-          &batch, column_family_memtables_.get(), nullptr /* flush_scheduler */,
-          true, log_number, this, false /* concurrent_memtable_writes */,
-          next_sequence, &has_valid_writes, seq_per_batch_, batch_per_txn_);
-      if (!status.ok()) {
+      if (status.ok()) {
+        for (const auto id : column_family_ids) {
+          ColumnFamilyData* cfd =
+              versions_->GetColumnFamilySet()->GetColumnFamily(id);
+          if (cfd == nullptr) {
+            continue;
+          }
+          std::unordered_map<ColumnFamilyData*, uint64_t>::iterator iter =
+              cfd_to_current_log_.find(cfd);
+          if (iter == cfd_to_current_log_.end()) {
+            cfd_to_current_log_.insert({cfd, log_number});
+          } else if (log_number > iter->second) {
+            iter->second = log_number;
+          }
+        }
+        auto last_sequence = *next_sequence - 1;
+        if ((*next_sequence != kMaxSequenceNumber) &&
+            (versions_->LastSequence() <= last_sequence)) {
+          versions_->SetLastAllocatedSequence(last_sequence);
+          versions_->SetLastPublishedSequence(last_sequence);
+          versions_->SetLastSequence(last_sequence);
+        }
+      } else {
        // We are treating this as a failure while reading since we read valid
        // blocks that do not form coherent data
        reader->GetReporter()->Corruption(record.size(), status);
-        continue;
      }
    }
-
    if (!status.ok()) {
      return status;
    }
-
-    auto last_sequence = *next_sequence - 1;
-    if ((*next_sequence != kMaxSequenceNumber) &&
-        (versions_->LastSequence() <= last_sequence)) {
-      versions_->SetLastAllocatedSequence(last_sequence);
-      versions_->SetLastPublishedSequence(last_sequence);
-      versions_->SetLastSequence(last_sequence);
-    }
  }
  // remove logreaders from map after successfully recovering the WAL
  if (log_readers_.size() > 1) {
-    auto eraseIter = log_readers_.begin();
-    std::advance(eraseIter, log_readers_.size() - 1);
-    log_readers_.erase(log_readers_.begin(), eraseIter);
+    auto erase_iter = log_readers_.begin();
+    std::advance(erase_iter, log_readers_.size() - 1);
+    log_readers_.erase(log_readers_.begin(), erase_iter);
  }
  return status;
 }
@@ -373,21 +452,89 @@ Status DBImplSecondary::NewIterators(
  return Status::OK();
 }

+Status DBImplSecondary::CheckConsistency() {
+  mutex_.AssertHeld();
+  Status s = DBImpl::CheckConsistency();
+  // If DBImpl::CheckConsistency() which is stricter returns success, then we
+  // do not need to give a second chance.
+  if (s.ok()) {
+    return s;
+  }
+  // It's possible that DBImpl::CheckConssitency() can fail because the primary
+  // may have removed certain files, causing the GetFileSize(name) call to
+  // fail and returning a PathNotFound. In this case, we take a best-effort
+  // approach and just proceed.
+  TEST_SYNC_POINT_CALLBACK(
+      "DBImplSecondary::CheckConsistency:AfterFirstAttempt", &s);
+  std::vector<LiveFileMetaData> metadata;
+  versions_->GetLiveFilesMetaData(&metadata);
+
+  std::string corruption_messages;
+  for (const auto& md : metadata) {
+    // md.name has a leading "/".
+    std::string file_path = md.db_path + md.name;
+
+    uint64_t fsize = 0;
+    s = env_->GetFileSize(file_path, &fsize);
+    if (!s.ok() &&
+        (env_->GetFileSize(Rocks2LevelTableFileName(file_path), &fsize).ok() ||
+         s.IsPathNotFound())) {
+      s = Status::OK();
+    }
+    if (!s.ok()) {
+      corruption_messages +=
+          "Can't access " + md.name + ": " + s.ToString() + "\n";
+    }
+  }
+  return corruption_messages.empty() ? Status::OK()
+                                     : Status::Corruption(corruption_messages);
+}
+
 Status DBImplSecondary::TryCatchUpWithPrimary() {
  assert(versions_.get() != nullptr);
  assert(manifest_reader_.get() != nullptr);
  Status s;
+  // read the manifest and apply new changes to the secondary instance
  std::unordered_set<ColumnFamilyData*> cfds_changed;
+  JobContext job_context(0, true /*create_superversion*/);
  InstrumentedMutexLock lock_guard(&mutex_);
  s = static_cast<ReactiveVersionSet*>(versions_.get())
          ->ReadAndApply(&mutex_, &manifest_reader_, &cfds_changed);
-  if (s.ok()) {
-    SuperVersionContext sv_context(true /* create_superversion */);
-    for (auto cfd : cfds_changed) {
-      sv_context.NewSuperVersion();
-      cfd->InstallSuperVersion(&sv_context, &mutex_);
+
+  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;
    }
-    sv_context.Clean();
+    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();
+    }
+    job_context.Clean();
  }
  return s;
 }
@@ -425,39 +572,14 @@ Status DB::OpenAsSecondary(
  }

  DBOptions tmp_opts(db_options);
+  Status s;
  if (nullptr == tmp_opts.info_log) {
-    Env* env = tmp_opts.env;
-    assert(env != nullptr);
-    std::string secondary_abs_path;
-    env->GetAbsolutePath(secondary_path, &secondary_abs_path);
-    std::string fname = InfoLogFileName(secondary_path, secondary_abs_path,
-                                        tmp_opts.db_log_dir);
-
-    env->CreateDirIfMissing(secondary_path);
-    if (tmp_opts.log_file_time_to_roll > 0 || tmp_opts.max_log_file_size > 0) {
-      AutoRollLogger* result = new AutoRollLogger(
-          env, secondary_path, tmp_opts.db_log_dir, tmp_opts.max_log_file_size,
-          tmp_opts.log_file_time_to_roll, tmp_opts.info_log_level);
-      Status s = result->GetStatus();
-      if (!s.ok()) {
-        delete result;
-      } else {
-        tmp_opts.info_log.reset(result);
-      }
-    }
-    if (nullptr == tmp_opts.info_log) {
-      env->RenameFile(
-          fname, OldInfoLogFileName(secondary_path, env->NowMicros(),
-                                    secondary_abs_path, tmp_opts.db_log_dir));
-      Status s = env->NewLogger(fname, &(tmp_opts.info_log));
-      if (tmp_opts.info_log != nullptr) {
-        tmp_opts.info_log->SetInfoLogLevel(tmp_opts.info_log_level);
-      }
+    s = CreateLoggerFromOptions(secondary_path, tmp_opts, &tmp_opts.info_log);
+    if (!s.ok()) {
+      tmp_opts.info_log = nullptr;
    }
  }

-  assert(tmp_opts.info_log != nullptr);
-
  handles->clear();
  DBImplSecondary* impl = new DBImplSecondary(tmp_opts, dbname);
  impl->versions_.reset(new ReactiveVersionSet(
@@ -466,8 +588,11 @@ Status DB::OpenAsSecondary(
      &impl->write_controller_));
  impl->column_family_memtables_.reset(
      new ColumnFamilyMemTablesImpl(impl->versions_->GetColumnFamilySet()));
+  impl->wal_in_db_path_ =
+      IsWalDirSameAsDBPath(&impl->immutable_db_options_);
+
  impl->mutex_.Lock();
-  Status s = impl->Recover(column_families, true, false, false);
+  s = impl->Recover(column_families, true, false, false);
  if (s.ok()) {
    for (auto cf : column_families) {
      auto cfd =
@@ -9,10 +9,11 @@

 #include <string>
 #include <vector>
-#include "db/db_impl.h"
+#include "db/db_impl/db_impl.h"

 namespace rocksdb {

+// A wrapper class to hold log reader, log reporter, log status.
 class LogReaderContainer {
 public:
  LogReaderContainer()
@@ -62,11 +63,19 @@ class LogReaderContainer {
  };
 };

+// The secondary instance shares access to the storage as the primary.
+// The secondary is able to read and replay changes described in both the
+// MANIFEST and the WAL files without coordination with the primary.
+// The secondary instance can be opened using `DB::OpenAsSecondary`. After
+// that, it can call `DBImplSecondary::TryCatchUpWithPrimary` to make best
+// effort attempts to catch up with the primary.
 class DBImplSecondary : public DBImpl {
 public:
  DBImplSecondary(const DBOptions& options, const std::string& dbname);
  ~DBImplSecondary() override;

+  // Recover by replaying MANIFEST and WAL. Also initialize manifest_reader_
+  // and log_readers_ to facilitate future operations.
  Status Recover(const std::vector<ColumnFamilyDescriptor>& column_families,
                 bool read_only, bool error_if_log_file_exist,
                 bool error_if_data_exists_in_logs) override;
@@ -96,40 +105,40 @@ class DBImplSecondary : public DBImpl {
  Status Put(const WriteOptions& /*options*/,
             ColumnFamilyHandle* /*column_family*/, const Slice& /*key*/,
             const Slice& /*value*/) override {
-    return Status::NotSupported("Not supported operation in read only mode.");
+    return Status::NotSupported("Not supported operation in secondary mode.");
  }

  using DBImpl::Merge;
  Status Merge(const WriteOptions& /*options*/,
               ColumnFamilyHandle* /*column_family*/, const Slice& /*key*/,
               const Slice& /*value*/) override {
-    return Status::NotSupported("Not supported operation in read only mode.");
+    return Status::NotSupported("Not supported operation in secondary mode.");
  }

  using DBImpl::Delete;
  Status Delete(const WriteOptions& /*options*/,
                ColumnFamilyHandle* /*column_family*/,
                const Slice& /*key*/) override {
-    return Status::NotSupported("Not supported operation in read only mode.");
+    return Status::NotSupported("Not supported operation in secondary mode.");
  }

  using DBImpl::SingleDelete;
  Status SingleDelete(const WriteOptions& /*options*/,
                      ColumnFamilyHandle* /*column_family*/,
                      const Slice& /*key*/) override {
-    return Status::NotSupported("Not supported operation in read only mode.");
+    return Status::NotSupported("Not supported operation in secondary mode.");
  }

  Status Write(const WriteOptions& /*options*/,
               WriteBatch* /*updates*/) override {
-    return Status::NotSupported("Not supported operation in read only mode.");
+    return Status::NotSupported("Not supported operation in secondary mode.");
  }

  using DBImpl::CompactRange;
  Status CompactRange(const CompactRangeOptions& /*options*/,
                      ColumnFamilyHandle* /*column_family*/,
                      const Slice* /*begin*/, const Slice* /*end*/) override {
-    return Status::NotSupported("Not supported operation in read only mode.");
+    return Status::NotSupported("Not supported operation in secondary mode.");
  }

  using DBImpl::CompactFiles;
@@ -140,32 +149,32 @@ class DBImplSecondary : public DBImpl {
      const int /*output_level*/, const int /*output_path_id*/ = -1,
      std::vector<std::string>* const /*output_file_names*/ = nullptr,
      CompactionJobInfo* /*compaction_job_info*/ = nullptr) override {
-    return Status::NotSupported("Not supported operation in read only mode.");
+    return Status::NotSupported("Not supported operation in secondary mode.");
  }

  Status DisableFileDeletions() override {
-    return Status::NotSupported("Not supported operation in read only mode.");
+    return Status::NotSupported("Not supported operation in secondary mode.");
  }

  Status EnableFileDeletions(bool /*force*/) override {
-    return Status::NotSupported("Not supported operation in read only mode.");
+    return Status::NotSupported("Not supported operation in secondary mode.");
  }

  Status GetLiveFiles(std::vector<std::string>&,
                      uint64_t* /*manifest_file_size*/,
                      bool /*flush_memtable*/ = true) override {
-    return Status::NotSupported("Not supported operation in read only mode.");
+    return Status::NotSupported("Not supported operation in secondary mode.");
  }

  using DBImpl::Flush;
  Status Flush(const FlushOptions& /*options*/,
               ColumnFamilyHandle* /*column_family*/) override {
-    return Status::NotSupported("Not supported operation in read only mode.");
+    return Status::NotSupported("Not supported operation in secondary mode.");
  }

  using DBImpl::SyncWAL;
  Status SyncWAL() override {
-    return Status::NotSupported("Not supported operation in read only mode.");
+    return Status::NotSupported("Not supported operation in secondary mode.");
  }

  using DB::IngestExternalFile;
@@ -173,7 +182,7 @@ class DBImplSecondary : public DBImpl {
      ColumnFamilyHandle* /*column_family*/,
      const std::vector<std::string>& /*external_files*/,
      const IngestExternalFileOptions& /*ingestion_options*/) override {
-    return Status::NotSupported("Not supported operation in read only mode.");
+    return Status::NotSupported("Not supported operation in secondary mode.");
  }

  // Try to catch up with the primary by reading as much as possible from the
@@ -182,9 +191,84 @@ class DBImplSecondary : public DBImpl {
  // method can take long time due to all the I/O and CPU costs.
  Status TryCatchUpWithPrimary() override;

+
+  // Try to find log reader using log_number from log_readers_ map, initialize
+  // if it doesn't exist
  Status MaybeInitLogReader(uint64_t log_number,
                            log::FragmentBufferedReader** log_reader);

+  // Check if all live files exist on file system and that their file sizes
+  // matche to the in-memory records. It is possible that some live files may
+  // have been deleted by the primary. In this case, CheckConsistency() does
+  // not flag the missing file as inconsistency.
+  Status CheckConsistency() override;
+
+ protected:
+  // ColumnFamilyCollector is a write batch handler which does nothing
+  // except recording unique column family IDs
+  class ColumnFamilyCollector : public WriteBatch::Handler {
+    std::unordered_set<uint32_t> column_family_ids_;
+
+    Status AddColumnFamilyId(uint32_t column_family_id) {
+      if (column_family_ids_.find(column_family_id) ==
+          column_family_ids_.end()) {
+        column_family_ids_.insert(column_family_id);
+      }
+      return Status::OK();
+    }
+
+   public:
+    explicit ColumnFamilyCollector() {}
+
+    ~ColumnFamilyCollector() override {}
+
+    Status PutCF(uint32_t column_family_id, const Slice&,
+                 const Slice&) override {
+      return AddColumnFamilyId(column_family_id);
+    }
+
+    Status DeleteCF(uint32_t column_family_id, const Slice&) override {
+      return AddColumnFamilyId(column_family_id);
+    }
+
+    Status SingleDeleteCF(uint32_t column_family_id, const Slice&) override {
+      return AddColumnFamilyId(column_family_id);
+    }
+
+    Status DeleteRangeCF(uint32_t column_family_id, const Slice&,
+                         const Slice&) override {
+      return AddColumnFamilyId(column_family_id);
+    }
+
+    Status MergeCF(uint32_t column_family_id, const Slice&,
+                   const Slice&) override {
+      return AddColumnFamilyId(column_family_id);
+    }
+
+    Status PutBlobIndexCF(uint32_t column_family_id, const Slice&,
+                          const Slice&) override {
+      return AddColumnFamilyId(column_family_id);
+    }
+
+    const std::unordered_set<uint32_t>& column_families() const {
+      return column_family_ids_;
+    }
+  };
+
+  Status CollectColumnFamilyIdsFromWriteBatch(
+      const WriteBatch& batch, std::vector<uint32_t>* column_family_ids) {
+    assert(column_family_ids != nullptr);
+    column_family_ids->clear();
+    ColumnFamilyCollector handler;
+    Status s = batch.Iterate(&handler);
+    if (s.ok()) {
+      for (const auto& cf : handler.column_families()) {
+        column_family_ids->push_back(cf);
+      }
+    }
+    return s;
+  }
+
 private:
  friend class DB;

@@ -194,17 +278,27 @@ class DBImplSecondary : public DBImpl {

  using DBImpl::Recover;

+  Status FindAndRecoverLogFiles(
+      std::unordered_set<ColumnFamilyData*>* cfds_changed,
+      JobContext* job_context);
+  Status FindNewLogNumbers(std::vector<uint64_t>* logs);
+  // After manifest recovery, replay WALs and refresh log_readers_ if necessary
+  // REQUIRES: log_numbers are sorted in ascending order
  Status RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
                         SequenceNumber* next_sequence,
-                         bool read_only) override;
+                         std::unordered_set<ColumnFamilyData*>* cfds_changed,
+                         JobContext* job_context);

  std::unique_ptr<log::FragmentBufferedReader> manifest_reader_;
  std::unique_ptr<log::Reader::Reporter> manifest_reporter_;
  std::unique_ptr<Status> manifest_reader_status_;

-  // cache log readers for each log number, used for continue WAL replay
+  // Cache log readers for each log number, used for continue WAL replay
  // after recovery
  std::map<uint64_t, std::unique_ptr<LogReaderContainer>> log_readers_;
+
+  // Current WAL number replayed for each column family.
+  std::unordered_map<ColumnFamilyData*, uint64_t> cfd_to_current_log_;
 };

 }  // namespace rocksdb
@@ -6,17 +6,14 @@
 // 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 "db/db_impl.h"
+#include "db/db_impl/db_impl.h"

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-#include <inttypes.h>
+#include <cinttypes>
 #include "db/error_handler.h"
 #include "db/event_helpers.h"
 #include "monitoring/perf_context_imp.h"
 #include "options/options_helper.h"
-#include "util/sync_point.h"
+#include "test_util/sync_point.h"

 namespace rocksdb {
 // Convenience methods
@@ -94,6 +91,11 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
    return Status::NotSupported(
        "pipelined_writes is not compatible with seq_per_batch");
  }
+  if (immutable_db_options_.unordered_write &&
+      immutable_db_options_.enable_pipelined_write) {
+    return Status::NotSupported(
+        "pipelined_writes is not compatible with unordered_write");
+  }
  // Otherwise IsLatestPersistentState optimization does not make sense
  assert(!WriteBatchInternal::IsLatestPersistentState(my_batch) ||
         disable_memtable);
@@ -107,8 +109,39 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
  }

  if (two_write_queues_ && disable_memtable) {
-    return WriteImplWALOnly(write_options, my_batch, callback, log_used,
-                            log_ref, seq_used, batch_cnt, pre_release_callback);
+    AssignOrder assign_order =
+        seq_per_batch_ ? kDoAssignOrder : kDontAssignOrder;
+    // Otherwise it is WAL-only Prepare batches in WriteCommitted policy and
+    // they don't consume sequence.
+    return WriteImplWALOnly(&nonmem_write_thread_, write_options, my_batch,
+                            callback, log_used, log_ref, seq_used, batch_cnt,
+                            pre_release_callback, assign_order,
+                            kDontPublishLastSeq, disable_memtable);
+  }
+
+  if (immutable_db_options_.unordered_write) {
+    const size_t sub_batch_cnt = batch_cnt != 0
+                                     ? batch_cnt
+                                     // every key is a sub-batch consuming a seq
+                                     : WriteBatchInternal::Count(my_batch);
+    uint64_t seq;
+    // Use a write thread to i) optimize for WAL write, ii) publish last
+    // sequence in in increasing order, iii) call pre_release_callback serially
+    status = WriteImplWALOnly(&write_thread_, write_options, my_batch, callback,
+                              log_used, log_ref, &seq, sub_batch_cnt,
+                              pre_release_callback, kDoAssignOrder,
+                              kDoPublishLastSeq, disable_memtable);
+    if (!status.ok()) {
+      return status;
+    }
+    if (seq_used) {
+      *seq_used = seq;
+    }
+    if (!disable_memtable) {
+      status = UnorderedWriteMemtable(write_options, my_batch, callback,
+                                      log_ref, seq, sub_batch_cnt);
+    }
+    return status;
  }

  if (immutable_db_options_.enable_pipelined_write) {
@@ -138,8 +171,10 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
          versions_->GetColumnFamilySet());
      w.status = WriteBatchInternal::InsertInto(
          &w, w.sequence, &column_family_memtables, &flush_scheduler_,
+          &trim_history_scheduler_,
          write_options.ignore_missing_column_families, 0 /*log_number*/, this,
-          true /*concurrent_memtable_writes*/, seq_per_batch_, w.batch_cnt);
+          true /*concurrent_memtable_writes*/, seq_per_batch_, w.batch_cnt,
+          batch_per_txn_, write_options.memtable_insert_hint_per_batch);

      PERF_TIMER_START(write_pre_and_post_process_time);
    }
@@ -178,9 +213,6 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
  WriteThread::WriteGroup write_group;
  bool in_parallel_group = false;
  uint64_t last_sequence = kMaxSequenceNumber;
-  if (!two_write_queues_) {
-    last_sequence = versions_->LastSequence();
-  }

  mutex_.Lock();

@@ -195,6 +227,11 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
    PERF_TIMER_STOP(write_pre_and_post_process_time);

    status = PreprocessWrite(write_options, &need_log_sync, &write_context);
+    if (!two_write_queues_) {
+      // Assign it after ::PreprocessWrite since the sequence might advance
+      // inside it by WriteRecoverableState
+      last_sequence = versions_->LastSequence();
+    }

    PERF_TIMER_START(write_pre_and_post_process_time);
  }
@@ -228,6 +265,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
    size_t total_count = 0;
    size_t valid_batches = 0;
    size_t total_byte_size = 0;
+    size_t pre_release_callback_cnt = 0;
    for (auto* writer : write_group) {
      if (writer->CheckCallback(this)) {
        valid_batches += writer->batch_cnt;
@@ -235,9 +273,11 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
          total_count += WriteBatchInternal::Count(writer->batch);
          parallel = parallel && !writer->batch->HasMerge();
        }
-
        total_byte_size = WriteBatchInternal::AppendedByteSize(
            total_byte_size, WriteBatchInternal::ByteSize(writer->batch));
+        if (writer->pre_release_callback) {
+          pre_release_callback_cnt++;
+        }
      }
    }
    // Note about seq_per_batch_: either disableWAL is set for the entire write
@@ -254,18 +294,19 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
    // We're optimistic, updating the stats before we successfully
    // commit.  That lets us release our leader status early.
    auto stats = default_cf_internal_stats_;
-    stats->AddDBStats(InternalStats::NUMBER_KEYS_WRITTEN, total_count,
+    stats->AddDBStats(InternalStats::kIntStatsNumKeysWritten, total_count,
                      concurrent_update);
    RecordTick(stats_, NUMBER_KEYS_WRITTEN, total_count);
-    stats->AddDBStats(InternalStats::BYTES_WRITTEN, total_byte_size,
+    stats->AddDBStats(InternalStats::kIntStatsBytesWritten, total_byte_size,
                      concurrent_update);
    RecordTick(stats_, BYTES_WRITTEN, total_byte_size);
-    stats->AddDBStats(InternalStats::WRITE_DONE_BY_SELF, 1, concurrent_update);
+    stats->AddDBStats(InternalStats::kIntStatsWriteDoneBySelf, 1,
+                      concurrent_update);
    RecordTick(stats_, WRITE_DONE_BY_SELF);
    auto write_done_by_other = write_group.size - 1;
    if (write_done_by_other > 0) {
-      stats->AddDBStats(InternalStats::WRITE_DONE_BY_OTHER, write_done_by_other,
-                        concurrent_update);
+      stats->AddDBStats(InternalStats::kIntStatsWriteDoneByOther,
+                        write_done_by_other, concurrent_update);
      RecordTick(stats_, WRITE_DONE_BY_OTHER, write_done_by_other);
    }
    RecordInHistogram(stats_, BYTES_PER_WRITE, total_byte_size);
@@ -301,6 +342,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
    // PreReleaseCallback is called after WAL write and before memtable write
    if (status.ok()) {
      SequenceNumber next_sequence = current_sequence;
+      size_t index = 0;
      // Note: the logic for advancing seq here must be consistent with the
      // logic in WriteBatchInternal::InsertInto(write_group...) as well as
      // with WriteBatchInternal::InsertInto(write_batch...) that is called on
@@ -312,7 +354,8 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
        writer->sequence = next_sequence;
        if (writer->pre_release_callback) {
          Status ws = writer->pre_release_callback->Callback(
-              writer->sequence, disable_memtable, writer->log_used);
+              writer->sequence, disable_memtable, writer->log_used, index++,
+              pre_release_callback_cnt);
          if (!ws.ok()) {
            status = ws;
            break;
@@ -334,7 +377,8 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
        // w.sequence will be set inside InsertInto
        w.status = WriteBatchInternal::InsertInto(
            write_group, current_sequence, column_family_memtables_.get(),
-            &flush_scheduler_, write_options.ignore_missing_column_families,
+            &flush_scheduler_, &trim_history_scheduler_,
+            write_options.ignore_missing_column_families,
            0 /*recovery_log_number*/, this, parallel, seq_per_batch_,
            batch_per_txn_);
      } else {
@@ -350,9 +394,11 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
          assert(w.sequence == current_sequence);
          w.status = WriteBatchInternal::InsertInto(
              &w, w.sequence, &column_family_memtables, &flush_scheduler_,
+              &trim_history_scheduler_,
              write_options.ignore_missing_column_families, 0 /*log_number*/,
              this, true /*concurrent_memtable_writes*/, seq_per_batch_,
-              w.batch_cnt, batch_per_txn_);
+              w.batch_cnt, batch_per_txn_,
+              write_options.memtable_insert_hint_per_batch);
        }
      }
      if (seq_used != nullptr) {
@@ -459,9 +505,9 @@ Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,
    }

    auto stats = default_cf_internal_stats_;
-    stats->AddDBStats(InternalStats::NUMBER_KEYS_WRITTEN, total_count);
+    stats->AddDBStats(InternalStats::kIntStatsNumKeysWritten, total_count);
    RecordTick(stats_, NUMBER_KEYS_WRITTEN, total_count);
-    stats->AddDBStats(InternalStats::BYTES_WRITTEN, total_byte_size);
+    stats->AddDBStats(InternalStats::kIntStatsBytesWritten, total_byte_size);
    RecordTick(stats_, BYTES_WRITTEN, total_byte_size);
    RecordInHistogram(stats_, BYTES_PER_WRITE, total_byte_size);

@@ -469,10 +515,10 @@ Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,

    if (w.status.ok() && !write_options.disableWAL) {
      PERF_TIMER_GUARD(write_wal_time);
-      stats->AddDBStats(InternalStats::WRITE_DONE_BY_SELF, 1);
+      stats->AddDBStats(InternalStats::kIntStatsWriteDoneBySelf, 1);
      RecordTick(stats_, WRITE_DONE_BY_SELF, 1);
      if (wal_write_group.size > 1) {
-        stats->AddDBStats(InternalStats::WRITE_DONE_BY_OTHER,
+        stats->AddDBStats(InternalStats::kIntStatsWriteDoneByOther,
                          wal_write_group.size - 1);
        RecordTick(stats_, WRITE_DONE_BY_OTHER, wal_write_group.size - 1);
      }
@@ -504,9 +550,9 @@ Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,
    } else {
      memtable_write_group.status = WriteBatchInternal::InsertInto(
          memtable_write_group, w.sequence, column_family_memtables_.get(),
-          &flush_scheduler_, write_options.ignore_missing_column_families,
-          0 /*log_number*/, this, false /*concurrent_memtable_writes*/,
-          seq_per_batch_, batch_per_txn_);
+          &flush_scheduler_, &trim_history_scheduler_,
+          write_options.ignore_missing_column_families, 0 /*log_number*/, this,
+          false /*concurrent_memtable_writes*/, seq_per_batch_, batch_per_txn_);
      versions_->SetLastSequence(memtable_write_group.last_sequence);
      write_thread_.ExitAsMemTableWriter(&w, memtable_write_group);
    }
@@ -518,8 +564,10 @@ Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,
        versions_->GetColumnFamilySet());
    w.status = WriteBatchInternal::InsertInto(
        &w, w.sequence, &column_family_memtables, &flush_scheduler_,
-        write_options.ignore_missing_column_families, 0 /*log_number*/, this,
-        true /*concurrent_memtable_writes*/);
+        &trim_history_scheduler_, write_options.ignore_missing_column_families,
+        0 /*log_number*/, this, true /*concurrent_memtable_writes*/,
+        false /*seq_per_batch*/, 0 /*batch_cnt*/, true /*batch_per_txn*/,
+        write_options.memtable_insert_hint_per_batch);
    if (write_thread_.CompleteParallelMemTableWriter(&w)) {
      MemTableInsertStatusCheck(w.status);
      versions_->SetLastSequence(w.write_group->last_sequence);
@@ -534,23 +582,72 @@ Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,
  return w.FinalStatus();
 }

+Status DBImpl::UnorderedWriteMemtable(const WriteOptions& write_options,
+                                      WriteBatch* my_batch,
+                                      WriteCallback* callback, uint64_t log_ref,
+                                      SequenceNumber seq,
+                                      const size_t sub_batch_cnt) {
+  PERF_TIMER_GUARD(write_pre_and_post_process_time);
+  StopWatch write_sw(env_, immutable_db_options_.statistics.get(), DB_WRITE);
+
+  WriteThread::Writer w(write_options, my_batch, callback, log_ref,
+                        false /*disable_memtable*/);
+
+  if (w.CheckCallback(this) && w.ShouldWriteToMemtable()) {
+    w.sequence = seq;
+    size_t total_count = WriteBatchInternal::Count(my_batch);
+    InternalStats* stats = default_cf_internal_stats_;
+    stats->AddDBStats(InternalStats::kIntStatsNumKeysWritten, total_count);
+    RecordTick(stats_, NUMBER_KEYS_WRITTEN, total_count);
+
+    ColumnFamilyMemTablesImpl column_family_memtables(
+        versions_->GetColumnFamilySet());
+    w.status = WriteBatchInternal::InsertInto(
+        &w, w.sequence, &column_family_memtables, &flush_scheduler_,
+        &trim_history_scheduler_, write_options.ignore_missing_column_families,
+        0 /*log_number*/, this, true /*concurrent_memtable_writes*/,
+        seq_per_batch_, sub_batch_cnt, true /*batch_per_txn*/,
+        write_options.memtable_insert_hint_per_batch);
+
+    WriteStatusCheck(w.status);
+    if (write_options.disableWAL) {
+      has_unpersisted_data_.store(true, std::memory_order_relaxed);
+    }
+  }
+
+  size_t pending_cnt = pending_memtable_writes_.fetch_sub(1) - 1;
+  if (pending_cnt == 0) {
+    // switch_cv_ waits until pending_memtable_writes_ = 0. Locking its mutex
+    // before notify ensures that cv is in waiting state when it is notified
+    // thus not missing the update to pending_memtable_writes_ even though it is
+    // not modified under the mutex.
+    std::lock_guard<std::mutex> lck(switch_mutex_);
+    switch_cv_.notify_all();
+  }
+
+  if (!w.FinalStatus().ok()) {
+    return w.FinalStatus();
+  }
+  return Status::OK();
+}
+
 // The 2nd write queue. If enabled it will be used only for WAL-only writes.
 // This is the only queue that updates LastPublishedSequence which is only
 // applicable in a two-queue setting.
-Status DBImpl::WriteImplWALOnly(const WriteOptions& write_options,
-                                WriteBatch* my_batch, WriteCallback* callback,
-                                uint64_t* log_used, uint64_t log_ref,
-                                uint64_t* seq_used, size_t batch_cnt,
-                                PreReleaseCallback* pre_release_callback) {
+Status DBImpl::WriteImplWALOnly(
+    WriteThread* write_thread, const WriteOptions& write_options,
+    WriteBatch* my_batch, WriteCallback* callback, uint64_t* log_used,
+    const uint64_t log_ref, uint64_t* seq_used, const size_t sub_batch_cnt,
+    PreReleaseCallback* pre_release_callback, const AssignOrder assign_order,
+    const PublishLastSeq publish_last_seq, const bool disable_memtable) {
  Status status;
  PERF_TIMER_GUARD(write_pre_and_post_process_time);
  WriteThread::Writer w(write_options, my_batch, callback, log_ref,
-                        true /* disable_memtable */, batch_cnt,
-                        pre_release_callback);
+                        disable_memtable, sub_batch_cnt, pre_release_callback);
  RecordTick(stats_, WRITE_WITH_WAL);
  StopWatch write_sw(env_, immutable_db_options_.statistics.get(), DB_WRITE);

-  nonmem_write_thread_.JoinBatchGroup(&w);
+  write_thread->JoinBatchGroup(&w);
  assert(w.state != WriteThread::STATE_PARALLEL_MEMTABLE_WRITER);
  if (w.state == WriteThread::STATE_COMPLETED) {
    if (log_used != nullptr) {
@@ -563,17 +660,45 @@ Status DBImpl::WriteImplWALOnly(const WriteOptions& write_options,
  }
  // else we are the leader of the write batch group
  assert(w.state == WriteThread::STATE_GROUP_LEADER);
+
+  if (publish_last_seq == kDoPublishLastSeq) {
+    // Currently we only use kDoPublishLastSeq in unordered_write
+    assert(immutable_db_options_.unordered_write);
+    WriteContext write_context;
+    if (error_handler_.IsDBStopped()) {
+      status = error_handler_.GetBGError();
+    }
+    // TODO(myabandeh): Make preliminary checks thread-safe so we could do them
+    // without paying the cost of obtaining the mutex.
+    if (status.ok()) {
+      InstrumentedMutexLock l(&mutex_);
+      bool need_log_sync = false;
+      status = PreprocessWrite(write_options, &need_log_sync, &write_context);
+      WriteStatusCheck(status);
+    }
+    if (!status.ok()) {
+      WriteThread::WriteGroup write_group;
+      write_thread->EnterAsBatchGroupLeader(&w, &write_group);
+      write_thread->ExitAsBatchGroupLeader(write_group, status);
+      return status;
+    }
+  }
+
  WriteThread::WriteGroup write_group;
  uint64_t last_sequence;
-  nonmem_write_thread_.EnterAsBatchGroupLeader(&w, &write_group);
+  write_thread->EnterAsBatchGroupLeader(&w, &write_group);
  // Note: no need to update last_batch_group_size_ here since the batch writes
  // to WAL only

+  size_t pre_release_callback_cnt = 0;
  size_t total_byte_size = 0;
  for (auto* writer : write_group) {
    if (writer->CheckCallback(this)) {
      total_byte_size = WriteBatchInternal::AppendedByteSize(
          total_byte_size, WriteBatchInternal::ByteSize(writer->batch));
+      if (writer->pre_release_callback) {
+        pre_release_callback_cnt++;
+      }
    }
  }

@@ -583,15 +708,16 @@ Status DBImpl::WriteImplWALOnly(const WriteOptions& write_options,
  // We're optimistic, updating the stats before we successfully
  // commit.  That lets us release our leader status early.
  auto stats = default_cf_internal_stats_;
-  stats->AddDBStats(InternalStats::BYTES_WRITTEN, total_byte_size,
+  stats->AddDBStats(InternalStats::kIntStatsBytesWritten, total_byte_size,
                    concurrent_update);
  RecordTick(stats_, BYTES_WRITTEN, total_byte_size);
-  stats->AddDBStats(InternalStats::WRITE_DONE_BY_SELF, 1, concurrent_update);
+  stats->AddDBStats(InternalStats::kIntStatsWriteDoneBySelf, 1,
+                    concurrent_update);
  RecordTick(stats_, WRITE_DONE_BY_SELF);
  auto write_done_by_other = write_group.size - 1;
  if (write_done_by_other > 0) {
-    stats->AddDBStats(InternalStats::WRITE_DONE_BY_OTHER, write_done_by_other,
-                      concurrent_update);
+    stats->AddDBStats(InternalStats::kIntStatsWriteDoneByOther,
+                      write_done_by_other, concurrent_update);
    RecordTick(stats_, WRITE_DONE_BY_OTHER, write_done_by_other);
  }
  RecordInHistogram(stats_, BYTES_PER_WRITE, total_byte_size);
@@ -602,11 +728,13 @@ Status DBImpl::WriteImplWALOnly(const WriteOptions& write_options,
  // LastAllocatedSequence is increased inside WriteToWAL under
  // wal_write_mutex_ to ensure ordered events in WAL
  size_t seq_inc = 0 /* total_count */;
-  if (seq_per_batch_) {
+  if (assign_order == kDoAssignOrder) {
    size_t total_batch_cnt = 0;
    for (auto* writer : write_group) {
-      assert(writer->batch_cnt);
-      total_batch_cnt += writer->batch_cnt;
+      assert(writer->batch_cnt || !seq_per_batch_);
+      if (!writer->CallbackFailed()) {
+        total_batch_cnt += writer->batch_cnt;
+      }
    }
    seq_inc = total_batch_cnt;
  }
@@ -617,16 +745,21 @@ Status DBImpl::WriteImplWALOnly(const WriteOptions& write_options,
    // Otherwise we inc seq number to do solely the seq allocation
    last_sequence = versions_->FetchAddLastAllocatedSequence(seq_inc);
  }
+
+  size_t memtable_write_cnt = 0;
  auto curr_seq = last_sequence + 1;
  for (auto* writer : write_group) {
    if (writer->CallbackFailed()) {
      continue;
    }
    writer->sequence = curr_seq;
-    if (seq_per_batch_) {
-      assert(writer->batch_cnt);
+    if (assign_order == kDoAssignOrder) {
+      assert(writer->batch_cnt || !seq_per_batch_);
      curr_seq += writer->batch_cnt;
    }
+    if (!writer->disable_memtable) {
+      memtable_write_cnt++;
+    }
    // else seq advances only by memtable writes
  }
  if (status.ok() && write_options.sync) {
@@ -645,12 +778,13 @@ Status DBImpl::WriteImplWALOnly(const WriteOptions& write_options,
    WriteStatusCheck(status);
  }
  if (status.ok()) {
+    size_t index = 0;
    for (auto* writer : write_group) {
      if (!writer->CallbackFailed() && writer->pre_release_callback) {
        assert(writer->sequence != kMaxSequenceNumber);
-        const bool DISABLE_MEMTABLE = true;
        Status ws = writer->pre_release_callback->Callback(
-            writer->sequence, DISABLE_MEMTABLE, writer->log_used);
+            writer->sequence, disable_memtable, writer->log_used, index++,
+            pre_release_callback_cnt);
        if (!ws.ok()) {
          status = ws;
          break;
@@ -658,7 +792,15 @@ Status DBImpl::WriteImplWALOnly(const WriteOptions& write_options,
      }
    }
  }
-  nonmem_write_thread_.ExitAsBatchGroupLeader(write_group, status);
+  if (publish_last_seq == kDoPublishLastSeq) {
+    versions_->SetLastSequence(last_sequence + seq_inc);
+    // Currently we only use kDoPublishLastSeq in unordered_write
+    assert(immutable_db_options_.unordered_write);
+  }
+  if (immutable_db_options_.unordered_write && status.ok()) {
+    pending_memtable_writes_ += memtable_write_cnt;
+  }
+  write_thread->ExitAsBatchGroupLeader(write_group, status);
  if (status.ok()) {
    status = w.FinalStatus();
  }
@@ -710,6 +852,7 @@ Status DBImpl::PreprocessWrite(const WriteOptions& write_options,
         versions_->GetColumnFamilySet()->NumberOfColumnFamilies() == 1);
  if (UNLIKELY(status.ok() && !single_column_family_mode_ &&
               total_log_size_ > GetMaxTotalWalSize())) {
+    WaitForPendingWrites();
    status = SwitchWAL(write_context);
  }

@@ -719,10 +862,16 @@ Status DBImpl::PreprocessWrite(const WriteOptions& write_options,
    // thread is writing to another DB with the same write buffer, they may also
    // be flushed. We may end up with flushing much more DBs than needed. It's
    // suboptimal but still correct.
+    WaitForPendingWrites();
    status = HandleWriteBufferFull(write_context);
  }

+  if (UNLIKELY(status.ok() && !trim_history_scheduler_.Empty())) {
+    status = TrimMemtableHistory(write_context);
+  }
+
  if (UNLIKELY(status.ok() && !flush_scheduler_.Empty())) {
+    WaitForPendingWrites();
    status = ScheduleFlushes(write_context);
  }

@@ -900,12 +1049,12 @@ Status DBImpl::WriteToWAL(const WriteThread::WriteGroup& write_group,
  if (status.ok()) {
    auto stats = default_cf_internal_stats_;
    if (need_log_sync) {
-      stats->AddDBStats(InternalStats::WAL_FILE_SYNCED, 1);
+      stats->AddDBStats(InternalStats::kIntStatsWalFileSynced, 1);
      RecordTick(stats_, WAL_FILE_SYNCED);
    }
-    stats->AddDBStats(InternalStats::WAL_FILE_BYTES, log_size);
+    stats->AddDBStats(InternalStats::kIntStatsWalFileBytes, log_size);
    RecordTick(stats_, WAL_FILE_BYTES, log_size);
-    stats->AddDBStats(InternalStats::WRITE_WITH_WAL, write_with_wal);
+    stats->AddDBStats(InternalStats::kIntStatsWriteWithWal, write_with_wal);
    RecordTick(stats_, WRITE_WITH_WAL, write_with_wal);
  }
  return status;
@@ -951,9 +1100,10 @@ Status DBImpl::ConcurrentWriteToWAL(const WriteThread::WriteGroup& write_group,
  if (status.ok()) {
    const bool concurrent = true;
    auto stats = default_cf_internal_stats_;
-    stats->AddDBStats(InternalStats::WAL_FILE_BYTES, log_size, concurrent);
+    stats->AddDBStats(InternalStats::kIntStatsWalFileBytes, log_size,
+                      concurrent);
    RecordTick(stats_, WAL_FILE_BYTES, log_size);
-    stats->AddDBStats(InternalStats::WRITE_WITH_WAL, write_with_wal,
+    stats->AddDBStats(InternalStats::kIntStatsWriteWithWal, write_with_wal,
                      concurrent);
    RecordTick(stats_, WRITE_WITH_WAL, write_with_wal);
  }
@@ -977,9 +1127,9 @@ Status DBImpl::WriteRecoverableState() {
    WriteBatchInternal::SetSequence(&cached_recoverable_state_, seq + 1);
    auto status = WriteBatchInternal::InsertInto(
        &cached_recoverable_state_, column_family_memtables_.get(),
-        &flush_scheduler_, true, 0 /*recovery_log_number*/, this,
-        false /* concurrent_memtable_writes */, &next_seq, &dont_care_bool,
-        seq_per_batch_);
+        &flush_scheduler_, &trim_history_scheduler_, true,
+        0 /*recovery_log_number*/, this, false /* concurrent_memtable_writes */,
+        &next_seq, &dont_care_bool, seq_per_batch_);
    auto last_seq = next_seq - 1;
    if (two_write_queues_) {
      versions_->FetchAddLastAllocatedSequence(last_seq - seq);
@@ -994,8 +1144,12 @@ Status DBImpl::WriteRecoverableState() {
      for (uint64_t sub_batch_seq = seq + 1;
           sub_batch_seq < next_seq && status.ok(); sub_batch_seq++) {
        uint64_t const no_log_num = 0;
+        // Unlock it since the callback might end up locking mutex. e.g.,
+        // AddCommitted -> AdvanceMaxEvictedSeq -> GetSnapshotListFromDB
+        mutex_.Unlock();
        status = recoverable_state_pre_release_callback_->Callback(
-            sub_batch_seq, !DISABLE_MEMTABLE, no_log_num);
+            sub_batch_seq, !DISABLE_MEMTABLE, no_log_num, 0, 1);
+        mutex_.Lock();
      }
    }
    if (status.ok()) {
@@ -1090,6 +1244,7 @@ Status DBImpl::SwitchWAL(WriteContext* write_context) {
        cfds.push_back(cfd);
      }
    }
+    MaybeFlushStatsCF(&cfds);
  }
  for (const auto cfd : cfds) {
    cfd->Ref();
@@ -1156,6 +1311,7 @@ Status DBImpl::HandleWriteBufferFull(WriteContext* write_context) {
    if (cfd_picked != nullptr) {
      cfds.push_back(cfd_picked);
    }
+    MaybeFlushStatsCF(&cfds);
  }

  for (const auto cfd : cfds) {
@@ -1249,8 +1405,8 @@ Status DBImpl::DelayWrite(uint64_t num_bytes,
  }
  assert(!delayed || !write_options.no_slowdown);
  if (delayed) {
-    default_cf_internal_stats_->AddDBStats(InternalStats::WRITE_STALL_MICROS,
-                                           time_delayed);
+    default_cf_internal_stats_->AddDBStats(
+        InternalStats::kIntStatsWriteStallMicros, time_delayed);
    RecordTick(stats_, STALL_MICROS, time_delayed);
  }

@@ -1299,6 +1455,67 @@ Status DBImpl::ThrottleLowPriWritesIfNeeded(const WriteOptions& write_options,
  return Status::OK();
 }

+void DBImpl::MaybeFlushStatsCF(autovector<ColumnFamilyData*>* cfds) {
+  assert(cfds != nullptr);
+  if (!cfds->empty() && immutable_db_options_.persist_stats_to_disk) {
+    ColumnFamilyData* cfd_stats =
+        versions_->GetColumnFamilySet()->GetColumnFamily(
+            kPersistentStatsColumnFamilyName);
+    if (cfd_stats != nullptr && !cfd_stats->mem()->IsEmpty()) {
+      for (ColumnFamilyData* cfd : *cfds) {
+        if (cfd == cfd_stats) {
+          // stats CF already included in cfds
+          return;
+        }
+      }
+      // force flush stats CF when its log number is less than all other CF's
+      // log numbers
+      bool force_flush_stats_cf = true;
+      for (auto* loop_cfd : *versions_->GetColumnFamilySet()) {
+        if (loop_cfd == cfd_stats) {
+          continue;
+        }
+        if (loop_cfd->GetLogNumber() <= cfd_stats->GetLogNumber()) {
+          force_flush_stats_cf = false;
+        }
+      }
+      if (force_flush_stats_cf) {
+        cfds->push_back(cfd_stats);
+        ROCKS_LOG_INFO(immutable_db_options_.info_log,
+                       "Force flushing stats CF with automated flush "
+                       "to avoid holding old logs");
+      }
+    }
+  }
+}
+
+Status DBImpl::TrimMemtableHistory(WriteContext* context) {
+  autovector<ColumnFamilyData*> cfds;
+  ColumnFamilyData* tmp_cfd;
+  while ((tmp_cfd = trim_history_scheduler_.TakeNextColumnFamily()) !=
+         nullptr) {
+    cfds.push_back(tmp_cfd);
+  }
+  for (auto& cfd : cfds) {
+    autovector<MemTable*> to_delete;
+    cfd->imm()->TrimHistory(&to_delete, cfd->mem()->ApproximateMemoryUsage());
+    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_);
+    }
+
+    if (cfd->Unref()) {
+      delete cfd;
+      cfd = nullptr;
+    }
+  }
+  return Status::OK();
+}
+
 Status DBImpl::ScheduleFlushes(WriteContext* context) {
  autovector<ColumnFamilyData*> cfds;
  if (immutable_db_options_.atomic_flush) {
@@ -1312,6 +1529,7 @@ Status DBImpl::ScheduleFlushes(WriteContext* context) {
    while ((tmp_cfd = flush_scheduler_.TakeNextColumnFamily()) != nullptr) {
      cfds.push_back(tmp_cfd);
    }
+    MaybeFlushStatsCF(&cfds);
  }
  Status status;
  for (auto& cfd : cfds) {
@@ -1547,14 +1765,30 @@ size_t DBImpl::GetWalPreallocateBlockSize(uint64_t write_buffer_size) const {
 // can call if they wish
 Status DB::Put(const WriteOptions& opt, ColumnFamilyHandle* column_family,
               const Slice& key, const Slice& value) {
-  // Pre-allocate size of write batch conservatively.
-  // 8 bytes are taken by header, 4 bytes for count, 1 byte for type,
-  // and we allocate 11 extra bytes for key length, as well as value length.
-  WriteBatch batch(key.size() + value.size() + 24);
+  if (nullptr == opt.timestamp) {
+    // Pre-allocate size of write batch conservatively.
+    // 8 bytes are taken by header, 4 bytes for count, 1 byte for type,
+    // and we allocate 11 extra bytes for key length, as well as value length.
+    WriteBatch batch(key.size() + value.size() + 24);
+    Status s = batch.Put(column_family, key, value);
+    if (!s.ok()) {
+      return s;
+    }
+    return Write(opt, &batch);
+  }
+  const Slice* ts = opt.timestamp;
+  assert(nullptr != ts);
+  size_t ts_sz = ts->size();
+  WriteBatch batch(key.size() + ts_sz + value.size() + 24, /*max_bytes=*/0,
+                   ts_sz);
  Status s = batch.Put(column_family, key, value);
  if (!s.ok()) {
    return s;
  }
+  s = batch.AssignTimestamp(*ts);
+  if (!s.ok()) {
+    return s;
+  }
  return Write(opt, &batch);
 }

@@ -7,11 +7,11 @@
 // 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 "db/db_impl_secondary.h"
+#include "db/db_impl/db_impl_secondary.h"
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
-#include "util/fault_injection_test_env.h"
-#include "util/sync_point.h"
+#include "test_util/fault_injection_test_env.h"
+#include "test_util/sync_point.h"

 namespace rocksdb {

@@ -237,6 +237,17 @@ TEST_F(DBSecondaryTest, OpenAsSecondaryWALTailing) {
  };

  verify_db_func("foo_value2", "bar_value2");
+
+  ASSERT_OK(Put("foo", "new_foo_value"));
+  ASSERT_OK(Put("bar", "new_bar_value"));
+
+  ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
+  verify_db_func("new_foo_value", "new_bar_value");
+
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("foo", "new_foo_value_1"));
+  ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
+  verify_db_func("new_foo_value_1", "new_bar_value");
 }

 TEST_F(DBSecondaryTest, OpenWithNonExistColumnFamily) {
@@ -362,7 +373,7 @@ TEST_F(DBSecondaryTest, MissingTableFile) {
  SyncPoint::GetInstance()->DisableProcessing();
  SyncPoint::GetInstance()->ClearAllCallBacks();
  SyncPoint::GetInstance()->SetCallBack(
-      "ReactiveVersionSet::ReadAndApply:AfterLoadTableHandlers",
+      "ReactiveVersionSet::ApplyOneVersionEditToBuilder:AfterLoadTableHandlers",
      [&](void* arg) {
        Status s = *reinterpret_cast<Status*>(arg);
        if (s.IsPathNotFound()) {
@@ -513,6 +524,255 @@ TEST_F(DBSecondaryTest, SwitchManifest) {
  ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
  range_scan_db();
 }
+
+// Here, "Snapshot" refers to the version edits written by
+// VersionSet::WriteSnapshot() at the beginning of the new MANIFEST after
+// switching from the old one.
+TEST_F(DBSecondaryTest, SkipSnapshotAfterManifestSwitch) {
+  Options options;
+  options.env = env_;
+  options.disable_auto_compactions = true;
+  Reopen(options);
+
+  Options options1;
+  options1.env = env_;
+  options1.max_open_files = -1;
+  OpenSecondary(options1);
+
+  ASSERT_OK(Put("0", "value0"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
+  std::string value;
+  ReadOptions ropts;
+  ropts.verify_checksums = true;
+  ASSERT_OK(db_secondary_->Get(ropts, "0", &value));
+  ASSERT_EQ("value0", value);
+
+  Reopen(options);
+  ASSERT_OK(dbfull()->SetOptions({{"disable_auto_compactions", "false"}}));
+  ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
+}
+
+TEST_F(DBSecondaryTest, SwitchWAL) {
+  const int kNumKeysPerMemtable = 1;
+  Options options;
+  options.env = env_;
+  options.max_write_buffer_number = 4;
+  options.min_write_buffer_number_to_merge = 2;
+  options.memtable_factory.reset(
+      new SpecialSkipListFactory(kNumKeysPerMemtable));
+  Reopen(options);
+
+  Options options1;
+  options1.env = env_;
+  options1.max_open_files = -1;
+  OpenSecondary(options1);
+
+  const auto& verify_db = [](DB* db1, DB* db2) {
+    ASSERT_NE(nullptr, db1);
+    ASSERT_NE(nullptr, db2);
+    ReadOptions read_opts;
+    read_opts.verify_checksums = true;
+    std::unique_ptr<Iterator> it1(db1->NewIterator(read_opts));
+    std::unique_ptr<Iterator> it2(db2->NewIterator(read_opts));
+    it1->SeekToFirst();
+    it2->SeekToFirst();
+    for (; it1->Valid() && it2->Valid(); it1->Next(), it2->Next()) {
+      ASSERT_EQ(it1->key(), it2->key());
+      ASSERT_EQ(it1->value(), it2->value());
+    }
+    ASSERT_FALSE(it1->Valid());
+    ASSERT_FALSE(it2->Valid());
+
+    for (it1->SeekToFirst(); it1->Valid(); it1->Next()) {
+      std::string value;
+      ASSERT_OK(db2->Get(read_opts, it1->key(), &value));
+      ASSERT_EQ(it1->value(), value);
+    }
+    for (it2->SeekToFirst(); it2->Valid(); it2->Next()) {
+      std::string value;
+      ASSERT_OK(db1->Get(read_opts, it2->key(), &value));
+      ASSERT_EQ(it2->value(), value);
+    }
+  };
+  for (int k = 0; k != 16; ++k) {
+    ASSERT_OK(Put("key" + std::to_string(k), "value" + std::to_string(k)));
+    ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
+    verify_db(dbfull(), db_secondary_);
+  }
+}
+
+TEST_F(DBSecondaryTest, SwitchWALMultiColumnFamilies) {
+  const int kNumKeysPerMemtable = 1;
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->LoadDependency({
+      {"DBImpl::BackgroundCallFlush:ContextCleanedUp",
+       "DBSecondaryTest::SwitchWALMultipleColumnFamilies:BeforeCatchUp"}});
+  SyncPoint::GetInstance()->EnableProcessing();
+  const std::string kCFName1 = "pikachu";
+  Options options;
+  options.env = env_;
+  options.max_write_buffer_number = 4;
+  options.min_write_buffer_number_to_merge = 2;
+  options.memtable_factory.reset(
+      new SpecialSkipListFactory(kNumKeysPerMemtable));
+  CreateAndReopenWithCF({kCFName1}, options);
+
+  Options options1;
+  options1.env = env_;
+  options1.max_open_files = -1;
+  OpenSecondaryWithColumnFamilies({kCFName1}, options1);
+  ASSERT_EQ(2, handles_secondary_.size());
+
+  const auto& verify_db = [](DB* db1,
+                             const std::vector<ColumnFamilyHandle*>& handles1,
+                             DB* db2,
+                             const std::vector<ColumnFamilyHandle*>& handles2) {
+    ASSERT_NE(nullptr, db1);
+    ASSERT_NE(nullptr, db2);
+    ReadOptions read_opts;
+    read_opts.verify_checksums = true;
+    ASSERT_EQ(handles1.size(), handles2.size());
+    for (size_t i = 0; i != handles1.size(); ++i) {
+      std::unique_ptr<Iterator> it1(db1->NewIterator(read_opts, handles1[i]));
+      std::unique_ptr<Iterator> it2(db2->NewIterator(read_opts, handles2[i]));
+      it1->SeekToFirst();
+      it2->SeekToFirst();
+      for (; it1->Valid() && it2->Valid(); it1->Next(), it2->Next()) {
+        ASSERT_EQ(it1->key(), it2->key());
+        ASSERT_EQ(it1->value(), it2->value());
+      }
+      ASSERT_FALSE(it1->Valid());
+      ASSERT_FALSE(it2->Valid());
+
+      for (it1->SeekToFirst(); it1->Valid(); it1->Next()) {
+        std::string value;
+        ASSERT_OK(db2->Get(read_opts, handles2[i], it1->key(), &value));
+        ASSERT_EQ(it1->value(), value);
+      }
+      for (it2->SeekToFirst(); it2->Valid(); it2->Next()) {
+        std::string value;
+        ASSERT_OK(db1->Get(read_opts, handles1[i], it2->key(), &value));
+        ASSERT_EQ(it2->value(), value);
+      }
+    }
+  };
+  for (int k = 0; k != 8; ++k) {
+    ASSERT_OK(
+        Put(0 /*cf*/, "key" + std::to_string(k), "value" + std::to_string(k)));
+    ASSERT_OK(
+        Put(1 /*cf*/, "key" + std::to_string(k), "value" + std::to_string(k)));
+    TEST_SYNC_POINT("DBSecondaryTest::SwitchWALMultipleColumnFamilies:BeforeCatchUp");
+    ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
+    verify_db(dbfull(), handles_, db_secondary_, handles_secondary_);
+    SyncPoint::GetInstance()->ClearTrace();
+  }
+}
+
+TEST_F(DBSecondaryTest, CatchUpAfterFlush) {
+  const int kNumKeysPerMemtable = 16;
+  Options options;
+  options.env = env_;
+  options.max_write_buffer_number = 4;
+  options.min_write_buffer_number_to_merge = 2;
+  options.memtable_factory.reset(
+      new SpecialSkipListFactory(kNumKeysPerMemtable));
+  Reopen(options);
+
+  Options options1;
+  options1.env = env_;
+  options1.max_open_files = -1;
+  OpenSecondary(options1);
+
+  WriteOptions write_opts;
+  WriteBatch wb;
+  wb.Put("key0", "value0");
+  wb.Put("key1", "value1");
+  ASSERT_OK(dbfull()->Write(write_opts, &wb));
+  ReadOptions read_opts;
+  std::unique_ptr<Iterator> iter1(db_secondary_->NewIterator(read_opts));
+  iter1->Seek("key0");
+  ASSERT_FALSE(iter1->Valid());
+  iter1->Seek("key1");
+  ASSERT_FALSE(iter1->Valid());
+  ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
+  iter1->Seek("key0");
+  ASSERT_FALSE(iter1->Valid());
+  iter1->Seek("key1");
+  ASSERT_FALSE(iter1->Valid());
+  std::unique_ptr<Iterator> iter2(db_secondary_->NewIterator(read_opts));
+  iter2->Seek("key0");
+  ASSERT_TRUE(iter2->Valid());
+  ASSERT_EQ("value0", iter2->value());
+  iter2->Seek("key1");
+  ASSERT_TRUE(iter2->Valid());
+  ASSERT_EQ("value1", iter2->value());
+
+  {
+    WriteBatch wb1;
+    wb1.Put("key0", "value01");
+    wb1.Put("key1", "value11");
+    ASSERT_OK(dbfull()->Write(write_opts, &wb1));
+  }
+
+  {
+    WriteBatch wb2;
+    wb2.Put("key0", "new_value0");
+    wb2.Delete("key1");
+    ASSERT_OK(dbfull()->Write(write_opts, &wb2));
+  }
+
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(db_secondary_->TryCatchUpWithPrimary());
+  std::unique_ptr<Iterator> iter3(db_secondary_->NewIterator(read_opts));
+  // iter3 should not see value01 and value11 at all.
+  iter3->Seek("key0");
+  ASSERT_TRUE(iter3->Valid());
+  ASSERT_EQ("new_value0", iter3->value());
+  iter3->Seek("key1");
+  ASSERT_FALSE(iter3->Valid());
+}
+
+TEST_F(DBSecondaryTest, CheckConsistencyWhenOpen) {
+  bool called = false;
+  Options options;
+  options.env = env_;
+  options.disable_auto_compactions = true;
+  Reopen(options);
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
+  SyncPoint::GetInstance()->SetCallBack(
+      "DBImplSecondary::CheckConsistency:AfterFirstAttempt", [&](void* arg) {
+        ASSERT_NE(nullptr, arg);
+        called = true;
+        auto* s = reinterpret_cast<Status*>(arg);
+        ASSERT_NOK(*s);
+      });
+  SyncPoint::GetInstance()->LoadDependency(
+      {{"DBImpl::CheckConsistency:AfterGetLiveFilesMetaData",
+        "BackgroundCallCompaction:0"},
+       {"DBImpl::BackgroundCallCompaction:PurgedObsoleteFiles",
+        "DBImpl::CheckConsistency:BeforeGetFileSize"}});
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  ASSERT_OK(Put("a", "value0"));
+  ASSERT_OK(Put("c", "value0"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("b", "value1"));
+  ASSERT_OK(Put("d", "value1"));
+  ASSERT_OK(Flush());
+  port::Thread thread([this]() {
+    Options opts;
+    opts.env = env_;
+    opts.max_open_files = -1;
+    OpenSecondary(opts);
+  });
+  ASSERT_OK(dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  ASSERT_OK(dbfull()->TEST_WaitForCompact());
+  thread.join();
+  ASSERT_TRUE(called);
+}
 #endif  //! ROCKSDB_LITE

 }  // namespace rocksdb
@@ -3,20 +3,16 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).

-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-
 #include "db/db_info_dumper.h"

-#include <inttypes.h>
+#include <cinttypes>
 #include <stdio.h>
 #include <string>
 #include <algorithm>
 #include <vector>

+#include "file/filename.h"
 #include "rocksdb/env.h"
-#include "util/filename.h"

 namespace rocksdb {

@@ -16,6 +16,9 @@
 #include "db/merge_context.h"
 #include "db/merge_helper.h"
 #include "db/pinned_iterators_manager.h"
+#include "file/filename.h"
+#include "logging/logging.h"
+#include "memory/arena.h"
 #include "monitoring/perf_context_imp.h"
 #include "rocksdb/env.h"
 #include "rocksdb/iterator.h"
@@ -23,12 +26,9 @@
 #include "rocksdb/options.h"
 #include "table/internal_iterator.h"
 #include "table/iterator_wrapper.h"
-#include "util/arena.h"
-#include "util/filename.h"
-#include "util/logging.h"
+#include "trace_replay/trace_replay.h"
 #include "util/mutexlock.h"
 #include "util/string_util.h"
-#include "util/trace_replay.h"
 #include "util/user_comparator_wrapper.h"

 namespace rocksdb {
@@ -133,6 +133,7 @@ class DBIter final: public Iterator {
        direction_(kForward),
        valid_(false),
        current_entry_is_merged_(false),
+        is_key_seqnum_zero_(false),
        prefix_same_as_start_(read_options.prefix_same_as_start),
        pin_thru_lifetime_(read_options.pin_data),
        total_order_seek_(read_options.total_order_seek),
@@ -154,6 +155,10 @@ class DBIter final: public Iterator {
      iter_.iter()->SetPinnedItersMgr(&pinned_iters_mgr_);
    }
  }
+  // No copying allowed
+  DBIter(const DBIter&) = delete;
+  void operator=(const DBIter&) = delete;
+
  ~DBIter() override {
    // Release pinned data if any
    if (pinned_iters_mgr_.PinningEnabled()) {
@@ -262,12 +267,6 @@ class DBIter final: public Iterator {
  bool TooManyInternalKeysSkipped(bool increment = true);
  inline bool IsVisible(SequenceNumber sequence);

-  // CanReseekToSkip() returns whether the iterator can use the optimization
-  // where it reseek by sequence number to get the next key when there are too
-  // many versions. This is disabled for write unprepared because seeking to
-  // sequence number does not guarantee that it is visible.
-  inline bool CanReseekToSkip();
-
  // Temporarily pin the blocks that we encounter until ReleaseTempPinnedData()
  // is called
  void TempPinData() {
@@ -333,6 +332,10 @@ class DBIter final: public Iterator {
  Direction direction_;
  bool valid_;
  bool current_entry_is_merged_;
+  // True if we know that the current entry's seqnum is 0.
+  // This information is used as that the next entry will be for another
+  // user key.
+  bool is_key_seqnum_zero_;
  const bool prefix_same_as_start_;
  // Means that we will pin all data blocks we read as long the Iterator
  // is not deleted, will be true if ReadOptions::pin_data is true
@@ -346,15 +349,17 @@ class DBIter final: public Iterator {
  ReadRangeDelAggregator range_del_agg_;
  LocalStatistics local_stats_;
  PinnedIteratorsManager pinned_iters_mgr_;
+#ifdef ROCKSDB_LITE
+  ROCKSDB_FIELD_UNUSED
+#endif
  DBImpl* db_impl_;
+#ifdef ROCKSDB_LITE
+  ROCKSDB_FIELD_UNUSED
+#endif
  ColumnFamilyData* cfd_;
  // for diff snapshots we want the lower bound on the seqnum;
  // if this value > 0 iterator will return internal keys
  SequenceNumber start_seqnum_;
-
-  // No copying allowed
-  DBIter(const DBIter&);
-  void operator=(const DBIter&);
 };

 inline bool DBIter::ParseKey(ParsedInternalKey* ikey) {
@@ -381,6 +386,7 @@ void DBIter::Next() {
  num_internal_keys_skipped_ = 0;
  bool ok = true;
  if (direction_ == kReverse) {
+    is_key_seqnum_zero_ = false;
    if (!ReverseToForward()) {
      ok = false;
    }
@@ -400,6 +406,7 @@ void DBIter::Next() {
    FindNextUserEntry(true /* skipping the current user key */,
                      prefix_same_as_start_);
  } else {
+    is_key_seqnum_zero_ = false;
    valid_ = false;
  }
  if (statistics_ != nullptr && valid_) {
@@ -446,15 +453,28 @@ inline bool DBIter::FindNextUserEntryInternal(bool skipping, bool prefix_check)
  //                         greater than that,
  //  - none of the above  : saved_key_ can contain anything, it doesn't matter.
  uint64_t num_skipped = 0;
+  // For write unprepared, the target sequence number in reseek could be larger
+  // than the snapshot, and thus needs to be skipped again. This could result in
+  // an infinite loop of reseeks. To avoid that, we limit the number of reseeks
+  // to one.
+  bool reseek_done = false;

  is_blob_ = false;

  do {
+    // Will update is_key_seqnum_zero_ as soon as we parsed the current key
+    // but we need to save the previous value to be used in the loop.
+    bool is_prev_key_seqnum_zero = is_key_seqnum_zero_;
    if (!ParseKey(&ikey_)) {
+      is_key_seqnum_zero_ = false;
      return false;
    }

-    if (iterate_upper_bound_ != nullptr &&
+    is_key_seqnum_zero_ = (ikey_.sequence == 0);
+
+    assert(iterate_upper_bound_ == nullptr || iter_.MayBeOutOfUpperBound() ||
+           user_comparator_.Compare(ikey_.user_key, *iterate_upper_bound_) < 0);
+    if (iterate_upper_bound_ != nullptr && iter_.MayBeOutOfUpperBound() &&
        user_comparator_.Compare(ikey_.user_key, *iterate_upper_bound_) >= 0) {
      break;
    }
@@ -470,12 +490,20 @@ inline bool DBIter::FindNextUserEntryInternal(bool skipping, bool prefix_check)
    }

    if (IsVisible(ikey_.sequence)) {
-      if (skipping && user_comparator_.Compare(ikey_.user_key,
-                                               saved_key_.GetUserKey()) <= 0) {
+      // If the previous entry is of seqnum 0, the current entry will not
+      // possibly be skipped. This condition can potentially be relaxed to
+      // prev_key.seq <= ikey_.sequence. We are cautious because it will be more
+      // prone to bugs causing the same user key with the same sequence number.
+      if (!is_prev_key_seqnum_zero && skipping &&
+          user_comparator_.Compare(ikey_.user_key, saved_key_.GetUserKey()) <=
+              0) {
        num_skipped++;  // skip this entry
        PERF_COUNTER_ADD(internal_key_skipped_count, 1);
      } else {
+        assert(!skipping || user_comparator_.Compare(
+                                ikey_.user_key, saved_key_.GetUserKey()) > 0);
        num_skipped = 0;
+        reseek_done = false;
        switch (ikey_.type) {
          case kTypeDeletion:
          case kTypeSingleDeletion:
@@ -529,6 +557,7 @@ inline bool DBIter::FindNextUserEntryInternal(bool skipping, bool prefix_check)
                // they are hidden by this deletion.
                skipping = true;
                num_skipped = 0;
+                reseek_done = false;
                PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
              } else if (ikey_.type == kTypeBlobIndex) {
                if (!allow_blob_) {
@@ -559,6 +588,7 @@ inline bool DBIter::FindNextUserEntryInternal(bool skipping, bool prefix_check)
              // they are hidden by this deletion.
              skipping = true;
              num_skipped = 0;
+              reseek_done = false;
              PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
            } else {
              // By now, we are sure the current ikey is going to yield a
@@ -589,13 +619,23 @@ inline bool DBIter::FindNextUserEntryInternal(bool skipping, bool prefix_check)
            !iter_.iter()->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
        skipping = false;
        num_skipped = 0;
+        reseek_done = false;
      }
    }

    // If we have sequentially iterated via numerous equal keys, then it's
    // better to seek so that we can avoid too many key comparisons.
-    if (num_skipped > max_skip_ && CanReseekToSkip()) {
+    //
+    // To avoid infinite loops, do not reseek if we have already attempted to
+    // reseek previously.
+    //
+    // TODO(lth): If we reseek to sequence number greater than ikey_.sequence,
+    // than it does not make sense to reseek as we would actually land further
+    // away from the desired key. There is opportunity for optimization here.
+    if (num_skipped > max_skip_ && !reseek_done) {
+      is_key_seqnum_zero_ = false;
      num_skipped = 0;
+      reseek_done = true;
      std::string last_key;
      if (skipping) {
        // We're looking for the next user-key but all we see are the same
@@ -838,7 +878,10 @@ void DBIter::PrevInternal() {
      return;
    }

-    if (iterate_lower_bound_ != nullptr &&
+    assert(iterate_lower_bound_ == nullptr || iter_.MayBeOutOfLowerBound() ||
+           user_comparator_.Compare(saved_key_.GetUserKey(),
+                                    *iterate_lower_bound_) >= 0);
+    if (iterate_lower_bound_ != nullptr && iter_.MayBeOutOfLowerBound() &&
        user_comparator_.Compare(saved_key_.GetUserKey(),
                                 *iterate_lower_bound_) < 0) {
      // We've iterated earlier than the user-specified lower bound.
@@ -911,7 +954,7 @@ bool DBIter::FindValueForCurrentKey() {
    // This user key has lots of entries.
    // We're going from old to new, and it's taking too long. Let's do a Seek()
    // and go from new to old. This helps when a key was overwritten many times.
-    if (num_skipped >= max_skip_ && CanReseekToSkip()) {
+    if (num_skipped >= max_skip_) {
      return FindValueForCurrentKeyUsingSeek();
    }

@@ -1208,7 +1251,7 @@ bool DBIter::FindUserKeyBeforeSavedKey() {
      PERF_COUNTER_ADD(internal_key_skipped_count, 1);
    }

-    if (num_skipped >= max_skip_ && CanReseekToSkip()) {
+    if (num_skipped >= max_skip_) {
      num_skipped = 0;
      IterKey last_key;
      last_key.SetInternalKey(ParsedInternalKey(
@@ -1255,16 +1298,13 @@ bool DBIter::IsVisible(SequenceNumber sequence) {
  }
 }

-bool DBIter::CanReseekToSkip() {
-  return read_callback_ == nullptr || read_callback_->CanReseekToSkip();
-}
-
 void DBIter::Seek(const Slice& target) {
  PERF_CPU_TIMER_GUARD(iter_seek_cpu_nanos, env_);
  StopWatch sw(env_, statistics_, DB_SEEK);
  status_ = Status::OK();
  ReleaseTempPinnedData();
  ResetInternalKeysSkippedCounter();
+  is_key_seqnum_zero_ = false;

  SequenceNumber seq = sequence_;
  saved_key_.Clear();
@@ -1323,6 +1363,7 @@ void DBIter::SeekForPrev(const Slice& target) {
  status_ = Status::OK();
  ReleaseTempPinnedData();
  ResetInternalKeysSkippedCounter();
+  is_key_seqnum_zero_ = false;
  saved_key_.Clear();
  // now saved_key is used to store internal key.
  saved_key_.SetInternalKey(target, 0 /* sequence_number */,
@@ -1390,6 +1431,7 @@ void DBIter::SeekToFirst() {
  ReleaseTempPinnedData();
  ResetInternalKeysSkippedCounter();
  ClearSavedValue();
+  is_key_seqnum_zero_ = false;

  {
    PERF_TIMER_GUARD(seek_internal_seek_time);
@@ -1442,6 +1484,7 @@ void DBIter::SeekToLast() {
  ReleaseTempPinnedData();
  ResetInternalKeysSkippedCounter();
  ClearSavedValue();
+  is_key_seqnum_zero_ = false;

  {
    PERF_TIMER_GUARD(seek_internal_seek_time);
@@ -10,22 +10,47 @@
 #pragma once
 #include <stdint.h>
 #include <string>
-#include "db/db_impl.h"
+#include "db/db_impl/db_impl.h"
 #include "db/dbformat.h"
 #include "db/range_del_aggregator.h"
+#include "memory/arena.h"
 #include "options/cf_options.h"
 #include "rocksdb/db.h"
 #include "rocksdb/iterator.h"
-#include "util/arena.h"
 #include "util/autovector.h"

 namespace rocksdb {

+// This file declares the factory functions of DBIter, in its original form
+// or a wrapped form with class ArenaWrappedDBIter, which is defined here.
+// Class DBIter, which is declared and implemented inside db_iter.cc, is
+// a iterator that converts internal keys (yielded by an InternalIterator)
+// that were live at the specified sequence number into appropriate user
+// keys.
+// Each internal key is consist of a user key, a sequence number, and a value
+// type. DBIter deals with multiple key versions, tombstones, merge operands,
+// etc, and exposes an Iterator.
+// For example, DBIter may wrap following InternalIterator:
+//    user key: AAA  value: v3   seqno: 100    type: Put
+//    user key: AAA  value: v2   seqno: 97     type: Put
+//    user key: AAA  value: v1   seqno: 95     type: Put
+//    user key: BBB  value: v1   seqno: 90     type: Put
+//    user key: BBC  value: N/A  seqno: 98     type: Delete
+//    user key: BBC  value: v1   seqno: 95     type: Put
+// If the snapshot passed in is 102, then the DBIter is expected to
+// expose the following iterator:
+//    key: AAA  value: v3
+//    key: BBB  value: v1
+// If the snapshot passed in is 96, then it should expose:
+//    key: AAA  value: v1
+//    key: BBB  value: v1
+//    key: BBC  value: v1
+//
 class Arena;
 class DBIter;

 // Return a new iterator that converts internal keys (yielded by
-// "*internal_iter") that were live at the specified "sequence" number
+// "*internal_iter") that were live at the specified `sequence` number
 // into appropriate user keys.
 extern Iterator* NewDBIterator(
    Env* env, const ReadOptions& read_options,
@@ -41,6 +66,8 @@ extern Iterator* NewDBIterator(
 // a iterator hierarchy whose memory can be allocated inline. In that way,
 // accessing the iterator tree can be more cache friendly. It is also faster
 // to allocate.
+// When using the class's Iterator interface, the behavior is exactly
+// the same as the inner DBIter.
 class ArenaWrappedDBIter : public Iterator {
 public:
  virtual ~ArenaWrappedDBIter();
@@ -8,9 +8,9 @@
 #include "rocksdb/comparator.h"
 #include "rocksdb/options.h"
 #include "rocksdb/slice.h"
+#include "test_util/testharness.h"
 #include "util/random.h"
 #include "util/string_util.h"
-#include "util/testharness.h"
 #include "utilities/merge_operators.h"

 #ifdef GFLAGS
@@ -17,9 +17,9 @@
 #include "rocksdb/statistics.h"
 #include "table/iterator_wrapper.h"
 #include "table/merging_iterator.h"
+#include "test_util/sync_point.h"
+#include "test_util/testharness.h"
 #include "util/string_util.h"
-#include "util/sync_point.h"
-#include "util/testharness.h"
 #include "utilities/merge_operators.h"

 namespace rocksdb {
@@ -15,7 +15,7 @@
 #include "port/stack_trace.h"
 #include "rocksdb/iostats_context.h"
 #include "rocksdb/perf_context.h"
-#include "table/flush_block_policy.h"
+#include "table/block_based/flush_block_policy.h"

 namespace rocksdb {

@@ -182,6 +182,33 @@ TEST_P(DBIteratorTest, IterSeekBeforePrev) {
  delete iter;
 }

+TEST_P(DBIteratorTest, IterReseekNewUpperBound) {
+  Random rnd(301);
+  Options options = CurrentOptions();
+  BlockBasedTableOptions table_options;
+  table_options.block_size = 1024;
+  table_options.block_size_deviation = 50;
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  options.compression = kNoCompression;
+  Reopen(options);
+
+  ASSERT_OK(Put("a", RandomString(&rnd, 400)));
+  ASSERT_OK(Put("aabb", RandomString(&rnd, 400)));
+  ASSERT_OK(Put("aaef", RandomString(&rnd, 400)));
+  ASSERT_OK(Put("b", RandomString(&rnd, 400)));
+  dbfull()->Flush(FlushOptions());
+  ReadOptions opts;
+  Slice ub = Slice("aa");
+  opts.iterate_upper_bound = &ub;
+  auto iter = NewIterator(opts);
+  iter->Seek(Slice("a"));
+  ub = Slice("b");
+  iter->Seek(Slice("aabc"));
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ(iter->key().ToString(), "aaef");
+  delete iter;
+}
+
 TEST_P(DBIteratorTest, IterSeekForPrevBeforeNext) {
  ASSERT_OK(Put("a", "b"));
  ASSERT_OK(Put("c", "d"));
@@ -1049,6 +1076,148 @@ TEST_P(DBIteratorTest, DBIteratorBoundOptimizationTest) {
    ASSERT_EQ(upper_bound_hits, 1);
  }
 }
+
+// Enable kBinarySearchWithFirstKey, do some iterator operations and check that
+// they don't do unnecessary block reads.
+TEST_P(DBIteratorTest, IndexWithFirstKey) {
+  for (int tailing = 0; tailing < 2; ++tailing) {
+    SCOPED_TRACE("tailing = " + std::to_string(tailing));
+    Options options = CurrentOptions();
+    options.env = env_;
+    options.create_if_missing = true;
+    options.prefix_extractor = nullptr;
+    options.merge_operator = MergeOperators::CreateStringAppendOperator();
+    options.statistics = rocksdb::CreateDBStatistics();
+    Statistics* stats = options.statistics.get();
+    BlockBasedTableOptions table_options;
+    table_options.index_type =
+        BlockBasedTableOptions::IndexType::kBinarySearchWithFirstKey;
+    table_options.index_shortening =
+        BlockBasedTableOptions::IndexShorteningMode::kNoShortening;
+    table_options.flush_block_policy_factory =
+        std::make_shared<FlushBlockEveryKeyPolicyFactory>();
+    table_options.block_cache = NewLRUCache(1000);  // fits all blocks
+    options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+
+    DestroyAndReopen(options);
+    ASSERT_OK(Merge("a1", "x1"));
+    ASSERT_OK(Merge("b1", "y1"));
+    ASSERT_OK(Merge("c0", "z1"));
+    ASSERT_OK(Flush());
+    ASSERT_OK(Merge("a2", "x2"));
+    ASSERT_OK(Merge("b2", "y2"));
+    ASSERT_OK(Merge("c0", "z2"));
+    ASSERT_OK(Flush());
+    ASSERT_OK(Merge("a3", "x3"));
+    ASSERT_OK(Merge("b3", "y3"));
+    ASSERT_OK(Merge("c3", "z3"));
+    ASSERT_OK(Flush());
+
+    // Block cache is not important for this test.
+    // We use BLOCK_CACHE_DATA_* counters just because they're the most readily
+    // available way of counting block accesses.
+
+    ReadOptions ropt;
+    ropt.tailing = tailing;
+    std::unique_ptr<Iterator> iter(NewIterator(ropt));
+
+    iter->Seek("b10");
+    ASSERT_TRUE(iter->Valid());
+    EXPECT_EQ("b2", iter->key().ToString());
+    EXPECT_EQ("y2", iter->value().ToString());
+    EXPECT_EQ(1, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
+
+    iter->Next();
+    ASSERT_TRUE(iter->Valid());
+    EXPECT_EQ("b3", iter->key().ToString());
+    EXPECT_EQ("y3", iter->value().ToString());
+    EXPECT_EQ(2, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
+    EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
+
+    iter->Seek("c0");
+    ASSERT_TRUE(iter->Valid());
+    EXPECT_EQ("c0", iter->key().ToString());
+    EXPECT_EQ("z1,z2", iter->value().ToString());
+    EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
+    EXPECT_EQ(4, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
+
+    iter->Next();
+    ASSERT_TRUE(iter->Valid());
+    EXPECT_EQ("c3", iter->key().ToString());
+    EXPECT_EQ("z3", iter->value().ToString());
+    EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
+    EXPECT_EQ(5, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
+
+    iter.reset();
+
+    // Enable iterate_upper_bound and check that iterator is not trying to read
+    // blocks that are fully above upper bound.
+    std::string ub = "b3";
+    Slice ub_slice(ub);
+    ropt.iterate_upper_bound = &ub_slice;
+    iter.reset(NewIterator(ropt));
+
+    iter->Seek("b2");
+    ASSERT_TRUE(iter->Valid());
+    EXPECT_EQ("b2", iter->key().ToString());
+    EXPECT_EQ("y2", iter->value().ToString());
+    EXPECT_EQ(1, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
+    EXPECT_EQ(5, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
+
+    iter->Next();
+    ASSERT_FALSE(iter->Valid());
+    EXPECT_EQ(1, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
+    EXPECT_EQ(5, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
+  }
+}
+
+TEST_P(DBIteratorTest, IndexWithFirstKeyGet) {
+  Options options = CurrentOptions();
+  options.env = env_;
+  options.create_if_missing = true;
+  options.prefix_extractor = nullptr;
+  options.merge_operator = MergeOperators::CreateStringAppendOperator();
+  options.statistics = rocksdb::CreateDBStatistics();
+  Statistics* stats = options.statistics.get();
+  BlockBasedTableOptions table_options;
+  table_options.index_type =
+      BlockBasedTableOptions::IndexType::kBinarySearchWithFirstKey;
+  table_options.index_shortening =
+      BlockBasedTableOptions::IndexShorteningMode::kNoShortening;
+  table_options.flush_block_policy_factory =
+      std::make_shared<FlushBlockEveryKeyPolicyFactory>();
+  table_options.block_cache = NewLRUCache(1000);  // fits all blocks
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+
+  DestroyAndReopen(options);
+  ASSERT_OK(Merge("a", "x1"));
+  ASSERT_OK(Merge("c", "y1"));
+  ASSERT_OK(Merge("e", "z1"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Merge("c", "y2"));
+  ASSERT_OK(Merge("e", "z2"));
+  ASSERT_OK(Flush());
+
+  // Get() between blocks shouldn't read any blocks.
+  ASSERT_EQ("NOT_FOUND", Get("b"));
+  EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
+  EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
+
+  // Get() of an existing key shouldn't read any unnecessary blocks when there's
+  // only one key per block.
+
+  ASSERT_EQ("y1,y2", Get("c"));
+  EXPECT_EQ(2, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
+  EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
+
+  ASSERT_EQ("x1", Get("a"));
+  EXPECT_EQ(3, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
+  EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
+
+  EXPECT_EQ(std::vector<std::string>({"NOT_FOUND", "z1,z2"}),
+            MultiGet({"b", "e"}));
+}
+
 // TODO(3.13): fix the issue of Seek() + Prev() which might not necessary
 //             return the biggest key which is smaller than the seek key.
 TEST_P(DBIteratorTest, PrevAfterAndNextAfterMerge) {
@@ -2450,6 +2619,192 @@ TEST_P(DBIteratorTest, SeekBackwardAfterOutOfUpperBound) {
  ASSERT_EQ("a", it->key().ToString());
 }

+TEST_P(DBIteratorTest, AvoidReseekLevelIterator) {
+  Options options = CurrentOptions();
+  options.compression = CompressionType::kNoCompression;
+  BlockBasedTableOptions table_options;
+  table_options.block_size = 800;
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  Reopen(options);
+
+  Random rnd(301);
+  std::string random_str = RandomString(&rnd, 180);
+
+  ASSERT_OK(Put("1", random_str));
+  ASSERT_OK(Put("2", random_str));
+  ASSERT_OK(Put("3", random_str));
+  ASSERT_OK(Put("4", random_str));
+  // A new block
+  ASSERT_OK(Put("5", random_str));
+  ASSERT_OK(Put("6", random_str));
+  ASSERT_OK(Put("7", random_str));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("8", random_str));
+  ASSERT_OK(Put("9", random_str));
+  ASSERT_OK(Flush());
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+
+  int num_find_file_in_level = 0;
+  int num_idx_blk_seek = 0;
+  SyncPoint::GetInstance()->SetCallBack(
+      "LevelIterator::Seek:BeforeFindFile",
+      [&](void* /*arg*/) { num_find_file_in_level++; });
+  SyncPoint::GetInstance()->SetCallBack(
+      "IndexBlockIter::Seek:0", [&](void* /*arg*/) { num_idx_blk_seek++; });
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  {
+    std::unique_ptr<Iterator> iter(NewIterator(ReadOptions()));
+    iter->Seek("1");
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ(1, num_find_file_in_level);
+    ASSERT_EQ(1, num_idx_blk_seek);
+
+    iter->Seek("2");
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ(1, num_find_file_in_level);
+    ASSERT_EQ(1, num_idx_blk_seek);
+
+    iter->Seek("3");
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ(1, num_find_file_in_level);
+    ASSERT_EQ(1, num_idx_blk_seek);
+
+    iter->Next();
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ(1, num_find_file_in_level);
+    ASSERT_EQ(1, num_idx_blk_seek);
+
+    iter->Seek("5");
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ(1, num_find_file_in_level);
+    ASSERT_EQ(2, num_idx_blk_seek);
+
+    iter->Seek("6");
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ(1, num_find_file_in_level);
+    ASSERT_EQ(2, num_idx_blk_seek);
+
+    iter->Seek("7");
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ(1, num_find_file_in_level);
+    ASSERT_EQ(3, num_idx_blk_seek);
+
+    iter->Seek("8");
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ(2, num_find_file_in_level);
+    // Still re-seek because "8" is the boundary key, which has
+    // the same user key as the seek key.
+    ASSERT_EQ(4, num_idx_blk_seek);
+
+    iter->Seek("5");
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ(3, num_find_file_in_level);
+    ASSERT_EQ(5, num_idx_blk_seek);
+
+    iter->Next();
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ(3, num_find_file_in_level);
+    ASSERT_EQ(5, num_idx_blk_seek);
+
+    // Seek backward never triggers the index block seek to be skipped
+    iter->Seek("5");
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ(3, num_find_file_in_level);
+    ASSERT_EQ(6, num_idx_blk_seek);
+  }
+
+  SyncPoint::GetInstance()->DisableProcessing();
+}
+
+// MyRocks may change iterate bounds before seek. Simply test to make sure such
+// usage doesn't break iterator.
+TEST_P(DBIteratorTest, IterateBoundChangedBeforeSeek) {
+  Options options = CurrentOptions();
+  options.compression = CompressionType::kNoCompression;
+  BlockBasedTableOptions table_options;
+  table_options.block_size = 100;
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  std::string value(50, 'v');
+  Reopen(options);
+  ASSERT_OK(Put("aaa", value));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("bbb", "v"));
+  ASSERT_OK(Put("ccc", "v"));
+  ASSERT_OK(Put("ddd", "v"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("eee", "v"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+
+  std::string ub1 = "e";
+  std::string ub2 = "c";
+  Slice ub(ub1);
+  ReadOptions read_opts1;
+  read_opts1.iterate_upper_bound = &ub;
+  Iterator* iter = NewIterator(read_opts1);
+  // Seek and iterate accross block boundary.
+  iter->Seek("b");
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_OK(iter->status());
+  ASSERT_EQ("bbb", iter->key());
+  ub = Slice(ub2);
+  iter->Seek("b");
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_OK(iter->status());
+  ASSERT_EQ("bbb", iter->key());
+  iter->Next();
+  ASSERT_FALSE(iter->Valid());
+  ASSERT_OK(iter->status());
+  delete iter;
+
+  std::string lb1 = "a";
+  std::string lb2 = "c";
+  Slice lb(lb1);
+  ReadOptions read_opts2;
+  read_opts2.iterate_lower_bound = &lb;
+  iter = NewIterator(read_opts2);
+  iter->SeekForPrev("d");
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_OK(iter->status());
+  ASSERT_EQ("ccc", iter->key());
+  lb = Slice(lb2);
+  iter->SeekForPrev("d");
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_OK(iter->status());
+  ASSERT_EQ("ccc", iter->key());
+  iter->Prev();
+  ASSERT_FALSE(iter->Valid());
+  ASSERT_OK(iter->status());
+  delete iter;
+}
+
+TEST_P(DBIteratorTest, IterateWithLowerBoundAcrossFileBoundary) {
+  ASSERT_OK(Put("aaa", "v"));
+  ASSERT_OK(Put("bbb", "v"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Put("ccc", "v"));
+  ASSERT_OK(Put("ddd", "v"));
+  ASSERT_OK(Flush());
+  // Move both files to bottom level.
+  ASSERT_OK(dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  Slice lower_bound("b");
+  ReadOptions read_opts;
+  read_opts.iterate_lower_bound = &lower_bound;
+  std::unique_ptr<Iterator> iter(NewIterator(read_opts));
+  iter->SeekForPrev("d");
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_OK(iter->status());
+  ASSERT_EQ("ccc", iter->key());
+  iter->Prev();
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_OK(iter->status());
+  ASSERT_EQ("bbb", iter->key());
+  iter->Prev();
+  ASSERT_FALSE(iter->Valid());
+  ASSERT_OK(iter->status());
+}
+
 INSTANTIATE_TEST_CASE_P(DBIteratorTestInstance, DBIteratorTest,
                        testing::Values(true, false));

@@ -204,6 +204,76 @@ TEST_F(DBMemTableTest, DuplicateSeq) {
  delete mem;
 }

+// A simple test to verify that the concurrent merge writes is functional
+TEST_F(DBMemTableTest, ConcurrentMergeWrite) {
+  int num_ops = 1000;
+  std::string value;
+  Status s;
+  MergeContext merge_context;
+  Options options;
+  // A merge operator that is not sensitive to concurrent writes since in this
+  // test we don't order the writes.
+  options.merge_operator = MergeOperators::CreateUInt64AddOperator();
+
+  // Create a MemTable
+  InternalKeyComparator cmp(BytewiseComparator());
+  auto factory = std::make_shared<SkipListFactory>();
+  options.memtable_factory = factory;
+  options.allow_concurrent_memtable_write = true;
+  ImmutableCFOptions ioptions(options);
+  WriteBufferManager wb(options.db_write_buffer_size);
+  MemTable* mem = new MemTable(cmp, ioptions, MutableCFOptions(options), &wb,
+                               kMaxSequenceNumber, 0 /* column_family_id */);
+
+  // Put 0 as the base
+  PutFixed64(&value, static_cast<uint64_t>(0));
+  bool res = mem->Add(0, kTypeValue, "key", value);
+  ASSERT_TRUE(res);
+  value.clear();
+
+  // Write Merge concurrently
+  rocksdb::port::Thread write_thread1([&]() {
+  MemTablePostProcessInfo post_process_info1;
+    std::string v1;
+    for (int seq = 1; seq < num_ops / 2; seq++) {
+      PutFixed64(&v1, seq);
+      bool res1 =
+          mem->Add(seq, kTypeMerge, "key", v1, true, &post_process_info1);
+      ASSERT_TRUE(res1);
+      v1.clear();
+    }
+  });
+  rocksdb::port::Thread write_thread2([&]() {
+  MemTablePostProcessInfo post_process_info2;
+    std::string v2;
+    for (int seq = num_ops / 2; seq < num_ops; seq++) {
+      PutFixed64(&v2, seq);
+      bool res2 =
+          mem->Add(seq, kTypeMerge, "key", v2, true, &post_process_info2);
+      ASSERT_TRUE(res2);
+      v2.clear();
+    }
+  });
+  write_thread1.join();
+  write_thread2.join();
+
+  Status status;
+  ReadOptions roptions;
+  SequenceNumber max_covering_tombstone_seq = 0;
+  LookupKey lkey("key", kMaxSequenceNumber);
+  res = mem->Get(lkey, &value, &status, &merge_context,
+                 &max_covering_tombstone_seq, roptions);
+  ASSERT_TRUE(res);
+  uint64_t ivalue = DecodeFixed64(Slice(value).data());
+  uint64_t sum = 0;
+  for (int seq = 0; seq < num_ops; seq++) {
+    sum += seq;
+  }
+  ASSERT_EQ(ivalue, sum);
+
+  delete mem;
+}
+
 TEST_F(DBMemTableTest, InsertWithHint) {
  Options options;
  options.allow_concurrent_memtable_write = false;
@@ -252,7 +322,7 @@ TEST_F(DBMemTableTest, ColumnFamilyId) {
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);

-  for (int cf = 0; cf < 2; ++cf) {
+  for (uint32_t cf = 0; cf < 2; ++cf) {
    ASSERT_OK(Put(cf, "key", "val"));
    ASSERT_OK(Flush(cf));
    ASSERT_EQ(
@@ -0,0 +1,240 @@
+//  Copyright (c) 2018-present, Facebook, Inc.  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).
+
+#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_builder.h"
+#include "test_util/fault_injection_test_env.h"
+#if !defined(ROCKSDB_LITE)
+#include "test_util/sync_point.h"
+#endif
+#include "rocksdb/merge_operator.h"
+#include "utilities/merge_operators.h"
+#include "utilities/merge_operators/sortlist.h"
+#include "utilities/merge_operators/string_append/stringappend2.h"
+
+namespace rocksdb {
+
+class DBMergeOperandTest : public DBTestBase {
+ public:
+  DBMergeOperandTest() : DBTestBase("/db_merge_operand_test") {}
+};
+
+TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {
+  class LimitedStringAppendMergeOp : public StringAppendTESTOperator {
+   public:
+    LimitedStringAppendMergeOp(int limit, char delim)
+        : StringAppendTESTOperator(delim), limit_(limit) {}
+
+    const char* Name() const override {
+      return "DBMergeOperatorTest::LimitedStringAppendMergeOp";
+    }
+
+    bool ShouldMerge(const std::vector<Slice>& operands) const override {
+      if (operands.size() > 0 && limit_ > 0 && operands.size() >= limit_) {
+        return true;
+      }
+      return false;
+    }
+
+   private:
+    size_t limit_ = 0;
+  };
+
+  Options options;
+  options.create_if_missing = true;
+  // Use only the latest two merge operands.
+  options.merge_operator = std::make_shared<LimitedStringAppendMergeOp>(2, ',');
+  options.env = env_;
+  Reopen(options);
+  int num_records = 4;
+  int number_of_operands = 0;
+  std::vector<PinnableSlice> values(num_records);
+  GetMergeOperandsOptions merge_operands_info;
+  merge_operands_info.expected_max_number_of_operands = num_records;
+
+  // k0 value in memtable
+  Put("k0", "PutARock");
+  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k0",
+                        values.data(), &merge_operands_info,
+                        &number_of_operands);
+  ASSERT_EQ(values[0], "PutARock");
+
+  // k0.1 value in SST
+  Put("k0.1", "RockInSST");
+  ASSERT_OK(Flush());
+  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k0.1",
+                        values.data(), &merge_operands_info,
+                        &number_of_operands);
+  ASSERT_EQ(values[0], "RockInSST");
+
+  // All k1 values are in memtable.
+  ASSERT_OK(Merge("k1", "a"));
+  Put("k1", "x");
+  ASSERT_OK(Merge("k1", "b"));
+  ASSERT_OK(Merge("k1", "c"));
+  ASSERT_OK(Merge("k1", "d"));
+  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k1",
+                        values.data(), &merge_operands_info,
+                        &number_of_operands);
+  ASSERT_EQ(values[0], "x");
+  ASSERT_EQ(values[1], "b");
+  ASSERT_EQ(values[2], "c");
+  ASSERT_EQ(values[3], "d");
+
+  // expected_max_number_of_operands is less than number of merge operands so
+  // status should be Incomplete.
+  merge_operands_info.expected_max_number_of_operands = num_records - 1;
+  Status status = db_->GetMergeOperands(
+      ReadOptions(), db_->DefaultColumnFamily(), "k1", values.data(),
+      &merge_operands_info, &number_of_operands);
+  ASSERT_EQ(status.IsIncomplete(), true);
+  merge_operands_info.expected_max_number_of_operands = num_records;
+
+  // All k1.1 values are in memtable.
+  ASSERT_OK(Merge("k1.1", "r"));
+  Delete("k1.1");
+  ASSERT_OK(Merge("k1.1", "c"));
+  ASSERT_OK(Merge("k1.1", "k"));
+  ASSERT_OK(Merge("k1.1", "s"));
+  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k1.1",
+                        values.data(), &merge_operands_info,
+                        &number_of_operands);
+  ASSERT_EQ(values[0], "c");
+  ASSERT_EQ(values[1], "k");
+  ASSERT_EQ(values[2], "s");
+
+  // All k2 values are flushed to L0 into a single file.
+  ASSERT_OK(Merge("k2", "q"));
+  ASSERT_OK(Merge("k2", "w"));
+  ASSERT_OK(Merge("k2", "e"));
+  ASSERT_OK(Merge("k2", "r"));
+  ASSERT_OK(Flush());
+  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k2",
+                        values.data(), &merge_operands_info,
+                        &number_of_operands);
+  ASSERT_EQ(values[0], "q");
+  ASSERT_EQ(values[1], "w");
+  ASSERT_EQ(values[2], "e");
+  ASSERT_EQ(values[3], "r");
+
+  // All k2.1 values are flushed to L0 into a single file.
+  ASSERT_OK(Merge("k2.1", "m"));
+  Put("k2.1", "l");
+  ASSERT_OK(Merge("k2.1", "n"));
+  ASSERT_OK(Merge("k2.1", "o"));
+  ASSERT_OK(Flush());
+  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k2.1",
+                        values.data(), &merge_operands_info,
+                        &number_of_operands);
+  ASSERT_EQ(values[0], "l,n,o");
+
+  // All k2.2 values are flushed to L0 into a single file.
+  ASSERT_OK(Merge("k2.2", "g"));
+  Delete("k2.2");
+  ASSERT_OK(Merge("k2.2", "o"));
+  ASSERT_OK(Merge("k2.2", "t"));
+  ASSERT_OK(Flush());
+  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k2.2",
+                        values.data(), &merge_operands_info,
+                        &number_of_operands);
+  ASSERT_EQ(values[0], "o,t");
+
+  // Do some compaction that will make the following tests more predictable
+  //  Slice start("PutARock");
+  //  Slice end("t");
+  db_->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+
+  // All k3 values are flushed and are in different files.
+  ASSERT_OK(Merge("k3", "ab"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Merge("k3", "bc"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Merge("k3", "cd"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Merge("k3", "de"));
+  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k3",
+                        values.data(), &merge_operands_info,
+                        &number_of_operands);
+  ASSERT_EQ(values[0], "ab");
+  ASSERT_EQ(values[1], "bc");
+  ASSERT_EQ(values[2], "cd");
+  ASSERT_EQ(values[3], "de");
+
+  // All k3.1 values are flushed and are in different files.
+  ASSERT_OK(Merge("k3.1", "ab"));
+  ASSERT_OK(Flush());
+  Put("k3.1", "bc");
+  ASSERT_OK(Flush());
+  ASSERT_OK(Merge("k3.1", "cd"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Merge("k3.1", "de"));
+  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k3.1",
+                        values.data(), &merge_operands_info,
+                        &number_of_operands);
+  ASSERT_EQ(values[0], "bc");
+  ASSERT_EQ(values[1], "cd");
+  ASSERT_EQ(values[2], "de");
+
+  // All k3.2 values are flushed and are in different files.
+  ASSERT_OK(Merge("k3.2", "ab"));
+  ASSERT_OK(Flush());
+  Delete("k3.2");
+  ASSERT_OK(Flush());
+  ASSERT_OK(Merge("k3.2", "cd"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(Merge("k3.2", "de"));
+  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k3.2",
+                        values.data(), &merge_operands_info,
+                        &number_of_operands);
+  ASSERT_EQ(values[0], "cd");
+  ASSERT_EQ(values[1], "de");
+
+  // All K4 values are in different levels
+  ASSERT_OK(Merge("k4", "ba"));
+  ASSERT_OK(Flush());
+  MoveFilesToLevel(4);
+  ASSERT_OK(Merge("k4", "cb"));
+  ASSERT_OK(Flush());
+  MoveFilesToLevel(3);
+  ASSERT_OK(Merge("k4", "dc"));
+  ASSERT_OK(Flush());
+  MoveFilesToLevel(1);
+  ASSERT_OK(Merge("k4", "ed"));
+  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k4",
+                        values.data(), &merge_operands_info,
+                        &number_of_operands);
+  ASSERT_EQ(values[0], "ba");
+  ASSERT_EQ(values[1], "cb");
+  ASSERT_EQ(values[2], "dc");
+  ASSERT_EQ(values[3], "ed");
+
+  // First 3 k5 values are in SST and next 4 k5 values are in Immutable Memtable
+  ASSERT_OK(Merge("k5", "who"));
+  ASSERT_OK(Merge("k5", "am"));
+  ASSERT_OK(Merge("k5", "i"));
+  ASSERT_OK(Flush());
+  Put("k5", "remember");
+  ASSERT_OK(Merge("k5", "i"));
+  ASSERT_OK(Merge("k5", "am"));
+  ASSERT_OK(Merge("k5", "rocks"));
+  dbfull()->TEST_SwitchMemtable();
+  db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k5",
+                        values.data(), &merge_operands_info,
+                        &number_of_operands);
+  ASSERT_EQ(values[0], "remember");
+  ASSERT_EQ(values[1], "i");
+  ASSERT_EQ(values[2], "am");
+}
+
+}  // namespace rocksdb
+
+int main(int argc, char** argv) {
+  rocksdb::port::InstallStackTraceHandler();
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}
@@ -46,9 +46,11 @@ class DBMergeOperatorTest : public DBTestBase {
    ReadOptions read_opt;
    read_opt.snapshot = snapshot;
    PinnableSlice value;
-    Status s =
-        dbfull()->GetImpl(read_opt, db_->DefaultColumnFamily(), key, &value,
-                          nullptr /*value_found*/, &read_callback);
+    DBImpl::GetImplOptions get_impl_options;
+    get_impl_options.column_family = db_->DefaultColumnFamily();
+    get_impl_options.value = &value;
+    get_impl_options.callback = &read_callback;
+    Status s = dbfull()->GetImpl(read_opt, key, get_impl_options);
    if (!s.ok()) {
      return s.ToString();
    }
@@ -275,68 +277,6 @@ TEST_P(MergeOperatorPinningTest, OperandsMultiBlocks) {
  VerifyDBFromMap(true_data);
 }

-TEST_P(MergeOperatorPinningTest, Randomized) {
-  do {
-    Options options = CurrentOptions();
-    options.merge_operator = MergeOperators::CreateMaxOperator();
-    BlockBasedTableOptions table_options;
-    table_options.no_block_cache = disable_block_cache_;
-    options.table_factory.reset(NewBlockBasedTableFactory(table_options));
-    DestroyAndReopen(options);
-
-    Random rnd(301);
-    std::map<std::string, std::string> true_data;
-
-    const int kTotalMerges = 5000;
-    // Every key gets ~10 operands
-    const int kKeyRange = kTotalMerges / 10;
-    const int kOperandSize = 20;
-    const int kNumPutBefore = kKeyRange / 10;  // 10% value
-    const int kNumPutAfter = kKeyRange / 10;   // 10% overwrite
-    const int kNumDelete = kKeyRange / 10;     // 10% delete
-
-    // kNumPutBefore keys will have base values
-    for (int i = 0; i < kNumPutBefore; i++) {
-      std::string key = Key(rnd.Next() % kKeyRange);
-      std::string value = RandomString(&rnd, kOperandSize);
-      ASSERT_OK(db_->Put(WriteOptions(), key, value));
-
-      true_data[key] = value;
-    }
-
-    // Do kTotalMerges merges
-    for (int i = 0; i < kTotalMerges; i++) {
-      std::string key = Key(rnd.Next() % kKeyRange);
-      std::string value = RandomString(&rnd, kOperandSize);
-      ASSERT_OK(db_->Merge(WriteOptions(), key, value));
-
-      if (true_data[key] < value) {
-        true_data[key] = value;
-      }
-    }
-
-    // Overwrite random kNumPutAfter keys
-    for (int i = 0; i < kNumPutAfter; i++) {
-      std::string key = Key(rnd.Next() % kKeyRange);
-      std::string value = RandomString(&rnd, kOperandSize);
-      ASSERT_OK(db_->Put(WriteOptions(), key, value));
-
-      true_data[key] = value;
-    }
-
-    // Delete random kNumDelete keys
-    for (int i = 0; i < kNumDelete; i++) {
-      std::string key = Key(rnd.Next() % kKeyRange);
-      ASSERT_OK(db_->Delete(WriteOptions(), key));
-
-      true_data.erase(key);
-    }
-
-    VerifyDBFromMap(true_data);
-
-  } while (ChangeOptions(kSkipMergePut));
-}
-
 class MergeOperatorHook : public MergeOperator {
 public:
  explicit MergeOperatorHook(std::shared_ptr<MergeOperator> _merge_op)
@@ -637,6 +577,86 @@ TEST_F(DBMergeOperatorTest, SnapshotCheckerAndReadCallback) {
  db_->ReleaseSnapshot(snapshot2);
 }

+class PerConfigMergeOperatorPinningTest
+    : public DBMergeOperatorTest,
+      public testing::WithParamInterface<std::tuple<bool, int>> {
+ public:
+  PerConfigMergeOperatorPinningTest() {
+    std::tie(disable_block_cache_, option_config_) = GetParam();
+  }
+
+  bool disable_block_cache_;
+};
+
+INSTANTIATE_TEST_CASE_P(
+    MergeOperatorPinningTest, PerConfigMergeOperatorPinningTest,
+    ::testing::Combine(::testing::Bool(),
+                       ::testing::Range(static_cast<int>(DBTestBase::kDefault),
+                                        static_cast<int>(DBTestBase::kEnd))));
+
+TEST_P(PerConfigMergeOperatorPinningTest, Randomized) {
+  if (ShouldSkipOptions(option_config_, kSkipMergePut)) {
+    return;
+  }
+
+  Options options = CurrentOptions();
+  options.merge_operator = MergeOperators::CreateMaxOperator();
+  BlockBasedTableOptions table_options;
+  table_options.no_block_cache = disable_block_cache_;
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  DestroyAndReopen(options);
+
+  Random rnd(301);
+  std::map<std::string, std::string> true_data;
+
+  const int kTotalMerges = 5000;
+  // Every key gets ~10 operands
+  const int kKeyRange = kTotalMerges / 10;
+  const int kOperandSize = 20;
+  const int kNumPutBefore = kKeyRange / 10;  // 10% value
+  const int kNumPutAfter = kKeyRange / 10;   // 10% overwrite
+  const int kNumDelete = kKeyRange / 10;     // 10% delete
+
+  // kNumPutBefore keys will have base values
+  for (int i = 0; i < kNumPutBefore; i++) {
+    std::string key = Key(rnd.Next() % kKeyRange);
+    std::string value = RandomString(&rnd, kOperandSize);
+    ASSERT_OK(db_->Put(WriteOptions(), key, value));
+
+    true_data[key] = value;
+  }
+
+  // Do kTotalMerges merges
+  for (int i = 0; i < kTotalMerges; i++) {
+    std::string key = Key(rnd.Next() % kKeyRange);
+    std::string value = RandomString(&rnd, kOperandSize);
+    ASSERT_OK(db_->Merge(WriteOptions(), key, value));
+
+    if (true_data[key] < value) {
+      true_data[key] = value;
+    }
+  }
+
+  // Overwrite random kNumPutAfter keys
+  for (int i = 0; i < kNumPutAfter; i++) {
+    std::string key = Key(rnd.Next() % kKeyRange);
+    std::string value = RandomString(&rnd, kOperandSize);
+    ASSERT_OK(db_->Put(WriteOptions(), key, value));
+
+    true_data[key] = value;
+  }
+
+  // Delete random kNumDelete keys
+  for (int i = 0; i < kNumDelete; i++) {
+    std::string key = Key(rnd.Next() % kKeyRange);
+    ASSERT_OK(db_->Delete(WriteOptions(), key));
+
+    true_data.erase(key);
+  }
+
+  VerifyDBFromMap(true_data);
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -11,7 +11,7 @@
 #include <unordered_map>

 #include "db/column_family.h"
-#include "db/db_impl.h"
+#include "db/db_impl/db_impl.h"
 #include "db/db_test_util.h"
 #include "options/options_helper.h"
 #include "port/stack_trace.h"
@@ -19,14 +19,12 @@
 #include "rocksdb/convenience.h"
 #include "rocksdb/rate_limiter.h"
 #include "rocksdb/stats_history.h"
+#include "test_util/sync_point.h"
+#include "test_util/testutil.h"
 #include "util/random.h"
-#include "util/sync_point.h"
-#include "util/testutil.h"

 namespace rocksdb {

-const int kMicrosInSec = 1000000;
-
 class DBOptionsTest : public DBTestBase {
 public:
  DBOptionsTest() : DBTestBase("/db_options_test") {}
@@ -66,10 +64,10 @@ class DBOptionsTest : public DBTestBase {

  std::unordered_map<std::string, std::string> GetRandomizedMutableCFOptionsMap(
      Random* rnd) {
-    Options options;
+    Options options = CurrentOptions();
    options.env = env_;
    ImmutableDBOptions db_options(options);
-    test::RandomInitCFOptions(&options, rnd);
+    test::RandomInitCFOptions(&options, options, rnd);
    auto sanitized_options = SanitizeOptions(db_options, options);
    auto opt_map = GetMutableCFOptionsMap(sanitized_options);
    delete options.compaction_filter;
@@ -507,10 +505,10 @@ TEST_F(DBOptionsTest, SetStatsDumpPeriodSec) {
  options.stats_dump_period_sec = 5;
  options.env = env_;
  Reopen(options);
-  ASSERT_EQ(5, dbfull()->GetDBOptions().stats_dump_period_sec);
+  ASSERT_EQ(5u, dbfull()->GetDBOptions().stats_dump_period_sec);

  for (int i = 0; i < 20; i++) {
-    int num = rand() % 5000 + 1;
+    unsigned int num = rand() % 5000 + 1;
    ASSERT_OK(
        dbfull()->SetDBOptions({{"stats_dump_period_sec", ToString(num)}}));
    ASSERT_EQ(num, dbfull()->GetDBOptions().stats_dump_period_sec);
@@ -518,283 +516,18 @@ TEST_F(DBOptionsTest, SetStatsDumpPeriodSec) {
  Close();
 }

-TEST_F(DBOptionsTest, RunStatsDumpPeriodSec) {
-  Options options;
-  options.create_if_missing = true;
-  options.stats_dump_period_sec = 5;
-  std::unique_ptr<rocksdb::MockTimeEnv> mock_env;
-  mock_env.reset(new rocksdb::MockTimeEnv(env_));
-  mock_env->set_current_time(0); // in seconds
-  options.env = mock_env.get();
-  int counter = 0;
-  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
-  rocksdb::SyncPoint::GetInstance()->ClearAllCallBacks();
-#if defined(OS_MACOSX) && !defined(NDEBUG)
-  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "InstrumentedCondVar::TimedWaitInternal", [&](void* arg) {
-        uint64_t time_us = *reinterpret_cast<uint64_t*>(arg);
-        if (time_us < mock_env->RealNowMicros()) {
-          *reinterpret_cast<uint64_t*>(arg) = mock_env->RealNowMicros() + 1000;
-        }
-      });
-#endif  // OS_MACOSX && !NDEBUG
-  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "DBImpl::DumpStats:1", [&](void* /*arg*/) {
-        counter++;
-      });
-  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-  Reopen(options);
-  ASSERT_EQ(5, dbfull()->GetDBOptions().stats_dump_period_sec);
-  dbfull()->TEST_WaitForDumpStatsRun([&] { mock_env->set_current_time(5); });
-  ASSERT_GE(counter, 1);
-
-  // Test cacel job through SetOptions
-  ASSERT_OK(dbfull()->SetDBOptions({{"stats_dump_period_sec", "0"}}));
-  int old_val = counter;
-  for (int i = 6; i < 20; ++i) {
-    dbfull()->TEST_WaitForDumpStatsRun([&] { mock_env->set_current_time(i); });
-  }
-  ASSERT_EQ(counter, old_val);
-  Close();
-}
-
-// Test persistent stats background thread scheduling and cancelling
-TEST_F(DBOptionsTest, StatsPersistScheduling) {
-  Options options;
-  options.create_if_missing = true;
-  options.stats_persist_period_sec = 5;
-  std::unique_ptr<rocksdb::MockTimeEnv> mock_env;
-  mock_env.reset(new rocksdb::MockTimeEnv(env_));
-  mock_env->set_current_time(0);  // in seconds
-  options.env = mock_env.get();
-  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
-  rocksdb::SyncPoint::GetInstance()->ClearAllCallBacks();
-#if defined(OS_MACOSX) && !defined(NDEBUG)
-  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "InstrumentedCondVar::TimedWaitInternal", [&](void* arg) {
-        uint64_t time_us = *reinterpret_cast<uint64_t*>(arg);
-        if (time_us < mock_env->RealNowMicros()) {
-          *reinterpret_cast<uint64_t*>(arg) = mock_env->RealNowMicros() + 1000;
-        }
-      });
-#endif  // OS_MACOSX && !NDEBUG
-  int counter = 0;
-  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "DBImpl::PersistStats:Entry", [&](void* /*arg*/) { counter++; });
-  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-  Reopen(options);
-  ASSERT_EQ(5, dbfull()->GetDBOptions().stats_persist_period_sec);
-  dbfull()->TEST_WaitForPersistStatsRun([&] { mock_env->set_current_time(5); });
-  ASSERT_GE(counter, 1);
-
-  // Test cacel job through SetOptions
-  ASSERT_TRUE(dbfull()->TEST_IsPersistentStatsEnabled());
-  ASSERT_OK(dbfull()->SetDBOptions({{"stats_persist_period_sec", "0"}}));
-  ASSERT_FALSE(dbfull()->TEST_IsPersistentStatsEnabled());
-  Close();
-}
-
-// Test enabling persistent stats for the first time
-TEST_F(DBOptionsTest, PersistentStatsFreshInstall) {
-  Options options;
-  options.create_if_missing = true;
-  options.stats_persist_period_sec = 0;
-  std::unique_ptr<rocksdb::MockTimeEnv> mock_env;
-  mock_env.reset(new rocksdb::MockTimeEnv(env_));
-  mock_env->set_current_time(0);  // in seconds
-  options.env = mock_env.get();
-  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
-  rocksdb::SyncPoint::GetInstance()->ClearAllCallBacks();
-#if defined(OS_MACOSX) && !defined(NDEBUG)
-  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "InstrumentedCondVar::TimedWaitInternal", [&](void* arg) {
-        uint64_t time_us = *reinterpret_cast<uint64_t*>(arg);
-        if (time_us < mock_env->RealNowMicros()) {
-          *reinterpret_cast<uint64_t*>(arg) = mock_env->RealNowMicros() + 1000;
-        }
-      });
-#endif  // OS_MACOSX && !NDEBUG
-  int counter = 0;
-  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "DBImpl::PersistStats:Entry", [&](void* /*arg*/) { counter++; });
-  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-  Reopen(options);
-  ASSERT_OK(dbfull()->SetDBOptions({{"stats_persist_period_sec", "5"}}));
-  ASSERT_EQ(5, dbfull()->GetDBOptions().stats_persist_period_sec);
-  dbfull()->TEST_WaitForPersistStatsRun([&] { mock_env->set_current_time(5); });
-  ASSERT_GE(counter, 1);
-  Close();
-}
-
 TEST_F(DBOptionsTest, SetOptionsStatsPersistPeriodSec) {
  Options options;
  options.create_if_missing = true;
  options.stats_persist_period_sec = 5;
  options.env = env_;
  Reopen(options);
-  ASSERT_EQ(5, dbfull()->GetDBOptions().stats_persist_period_sec);
+  ASSERT_EQ(5u, dbfull()->GetDBOptions().stats_persist_period_sec);

  ASSERT_OK(dbfull()->SetDBOptions({{"stats_persist_period_sec", "12345"}}));
-  ASSERT_EQ(12345, dbfull()->GetDBOptions().stats_persist_period_sec);
+  ASSERT_EQ(12345u, dbfull()->GetDBOptions().stats_persist_period_sec);
  ASSERT_NOK(dbfull()->SetDBOptions({{"stats_persist_period_sec", "abcde"}}));
-  ASSERT_EQ(12345, dbfull()->GetDBOptions().stats_persist_period_sec);
-}
-
-TEST_F(DBOptionsTest, GetStatsHistory) {
-  Options options;
-  options.create_if_missing = true;
-  options.stats_persist_period_sec = 5;
-  options.statistics = rocksdb::CreateDBStatistics();
-  std::unique_ptr<rocksdb::MockTimeEnv> mock_env;
-  mock_env.reset(new rocksdb::MockTimeEnv(env_));
-  mock_env->set_current_time(0);  // in seconds
-  options.env = mock_env.get();
-#if defined(OS_MACOSX) && !defined(NDEBUG)
-  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
-  rocksdb::SyncPoint::GetInstance()->ClearAllCallBacks();
-  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "InstrumentedCondVar::TimedWaitInternal", [&](void* arg) {
-        uint64_t time_us = *reinterpret_cast<uint64_t*>(arg);
-        if (time_us < mock_env->RealNowMicros()) {
-          *reinterpret_cast<uint64_t*>(arg) = mock_env->RealNowMicros() + 1000;
-        }
-      });
-  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-#endif  // OS_MACOSX && !NDEBUG
-
-  CreateColumnFamilies({"pikachu"}, options);
-  ASSERT_OK(Put("foo", "bar"));
-  ReopenWithColumnFamilies({"default", "pikachu"}, options);
-
-  int mock_time = 1;
-  // Wait for stats persist to finish
-  dbfull()->TEST_WaitForPersistStatsRun([&] { mock_env->set_current_time(5); });
-  std::unique_ptr<StatsHistoryIterator> stats_iter;
-  db_->GetStatsHistory(0, 6 * kMicrosInSec, &stats_iter);
-  ASSERT_TRUE(stats_iter != nullptr);
-  // disabled stats snapshots
-  ASSERT_OK(dbfull()->SetDBOptions({{"stats_persist_period_sec", "0"}}));
-  size_t stats_count = 0;
-  for (; stats_iter->Valid(); stats_iter->Next()) {
-    auto stats_map = stats_iter->GetStatsMap();
-    stats_count += stats_map.size();
-  }
-  ASSERT_GT(stats_count, 0);
-  // Wait a bit and verify no more stats are found
-  for (mock_time = 6; mock_time < 20; ++mock_time) {
-    dbfull()->TEST_WaitForPersistStatsRun(
-        [&] { mock_env->set_current_time(mock_time); });
-  }
-  db_->GetStatsHistory(0, 20 * kMicrosInSec, &stats_iter);
-  ASSERT_TRUE(stats_iter != nullptr);
-  size_t stats_count_new = 0;
-  for (; stats_iter->Valid(); stats_iter->Next()) {
-    stats_count_new += stats_iter->GetStatsMap().size();
-  }
-  ASSERT_EQ(stats_count_new, stats_count);
-  Close();
-}
-
-TEST_F(DBOptionsTest, InMemoryStatsHistoryPurging) {
-  Options options;
-  options.create_if_missing = true;
-  options.statistics = rocksdb::CreateDBStatistics();
-  options.stats_persist_period_sec = 1;
-  std::unique_ptr<rocksdb::MockTimeEnv> mock_env;
-  mock_env.reset(new rocksdb::MockTimeEnv(env_));
-  mock_env->set_current_time(0);  // in seconds
-  options.env = mock_env.get();
-#if defined(OS_MACOSX) && !defined(NDEBUG)
-  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
-  rocksdb::SyncPoint::GetInstance()->ClearAllCallBacks();
-  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "InstrumentedCondVar::TimedWaitInternal", [&](void* arg) {
-        uint64_t time_us = *reinterpret_cast<uint64_t*>(arg);
-        if (time_us < mock_env->RealNowMicros()) {
-          *reinterpret_cast<uint64_t*>(arg) = mock_env->RealNowMicros() + 1000;
-        }
-      });
-  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-#endif  // OS_MACOSX && !NDEBUG
-
-  CreateColumnFamilies({"pikachu"}, options);
-  ASSERT_OK(Put("foo", "bar"));
-  ReopenWithColumnFamilies({"default", "pikachu"}, options);
-  // some random operation to populate statistics
-  ASSERT_OK(Delete("foo"));
-  ASSERT_OK(Put("sol", "sol"));
-  ASSERT_OK(Put("epic", "epic"));
-  ASSERT_OK(Put("ltd", "ltd"));
-  ASSERT_EQ("sol", Get("sol"));
-  ASSERT_EQ("epic", Get("epic"));
-  ASSERT_EQ("ltd", Get("ltd"));
-  Iterator* iterator = db_->NewIterator(ReadOptions());
-  for (iterator->SeekToFirst(); iterator->Valid(); iterator->Next()) {
-    ASSERT_TRUE(iterator->key() == iterator->value());
-  }
-  delete iterator;
-  ASSERT_OK(Flush());
-  ASSERT_OK(Delete("sol"));
-  db_->CompactRange(CompactRangeOptions(), nullptr, nullptr);
-  int mock_time = 1;
-  // Wait for stats persist to finish
-  for (; mock_time < 5; ++mock_time) {
-    dbfull()->TEST_WaitForPersistStatsRun(
-        [&] { mock_env->set_current_time(mock_time); });
-  }
-
-  // second round of ops
-  ASSERT_OK(Put("saigon", "saigon"));
-  ASSERT_OK(Put("noodle talk", "noodle talk"));
-  ASSERT_OK(Put("ping bistro", "ping bistro"));
-  iterator = db_->NewIterator(ReadOptions());
-  for (iterator->SeekToFirst(); iterator->Valid(); iterator->Next()) {
-    ASSERT_TRUE(iterator->key() == iterator->value());
-  }
-  delete iterator;
-  ASSERT_OK(Flush());
-  db_->CompactRange(CompactRangeOptions(), nullptr, nullptr);
-  for (; mock_time < 10; ++mock_time) {
-    dbfull()->TEST_WaitForPersistStatsRun(
-        [&] { mock_env->set_current_time(mock_time); });
-  }
-  std::unique_ptr<StatsHistoryIterator> stats_iter;
-  db_->GetStatsHistory(0, 10 * kMicrosInSec, &stats_iter);
-  ASSERT_TRUE(stats_iter != nullptr);
-  size_t stats_count = 0;
-  int slice_count = 0;
-  for (; stats_iter->Valid(); stats_iter->Next()) {
-    slice_count++;
-    auto stats_map = stats_iter->GetStatsMap();
-    stats_count += stats_map.size();
-  }
-  size_t stats_history_size = dbfull()->TEST_EstiamteStatsHistorySize();
-  ASSERT_GE(slice_count, 9);
-  ASSERT_GE(stats_history_size, 12000);
-  // capping memory cost at 12000 bytes since one slice is around 10000~12000
-  ASSERT_OK(dbfull()->SetDBOptions({{"stats_history_buffer_size", "12000"}}));
-  ASSERT_EQ(12000, dbfull()->GetDBOptions().stats_history_buffer_size);
-  // Wait for stats persist to finish
-  for (; mock_time < 20; ++mock_time) {
-    dbfull()->TEST_WaitForPersistStatsRun(
-        [&] { mock_env->set_current_time(mock_time); });
-  }
-  db_->GetStatsHistory(0, 20 * kMicrosInSec, &stats_iter);
-  ASSERT_TRUE(stats_iter != nullptr);
-  size_t stats_count_reopen = 0;
-  slice_count = 0;
-  for (; stats_iter->Valid(); stats_iter->Next()) {
-    slice_count++;
-    auto stats_map = stats_iter->GetStatsMap();
-    stats_count_reopen += stats_map.size();
-  }
-  size_t stats_history_size_reopen = dbfull()->TEST_EstiamteStatsHistorySize();
-  // only one slice can fit under the new stats_history_buffer_size
-  ASSERT_LT(slice_count, 2);
-  ASSERT_TRUE(stats_history_size_reopen < 12000 &&
-              stats_history_size_reopen > 0);
-  ASSERT_TRUE(stats_count_reopen < stats_count && stats_count_reopen > 0);
-  Close();
+  ASSERT_EQ(12345u, dbfull()->GetDBOptions().stats_persist_period_sec);
 }

 static void assert_candidate_files_empty(DBImpl* dbfull, const bool empty) {
@@ -1007,6 +740,53 @@ TEST_F(DBOptionsTest, CompactionReadaheadSizeChange) {
  ASSERT_EQ(256, env_->compaction_readahead_size_);
  Close();
 }
+
+TEST_F(DBOptionsTest, FIFOTtlBackwardCompatible) {
+  Options options;
+  options.compaction_style = kCompactionStyleFIFO;
+  options.write_buffer_size = 10 << 10;  // 10KB
+  options.create_if_missing = true;
+
+  ASSERT_OK(TryReopen(options));
+
+  Random rnd(301);
+  for (int i = 0; i < 10; i++) {
+    // Generate and flush a file about 10KB.
+    for (int j = 0; j < 10; j++) {
+      ASSERT_OK(Put(ToString(i * 20 + j), RandomString(&rnd, 980)));
+    }
+    Flush();
+  }
+  ASSERT_OK(dbfull()->TEST_WaitForCompact());
+  ASSERT_EQ(NumTableFilesAtLevel(0), 10);
+
+  // In release 6.0, ttl was promoted from a secondary level option under
+  // compaction_options_fifo to a top level option under ColumnFamilyOptions.
+  // We still need to handle old SetOptions calls but should ignore
+  // ttl under compaction_options_fifo.
+  ASSERT_OK(dbfull()->SetOptions(
+      {{"compaction_options_fifo",
+        "{allow_compaction=true;max_table_files_size=1024;ttl=731;}"},
+       {"ttl", "60"}}));
+  ASSERT_EQ(dbfull()->GetOptions().compaction_options_fifo.allow_compaction,
+            true);
+  ASSERT_EQ(dbfull()->GetOptions().compaction_options_fifo.max_table_files_size,
+            1024);
+  ASSERT_EQ(dbfull()->GetOptions().ttl, 60);
+
+  // Put ttl as the first option inside compaction_options_fifo. That works as
+  // it doesn't overwrite any other option.
+  ASSERT_OK(dbfull()->SetOptions(
+      {{"compaction_options_fifo",
+        "{ttl=985;allow_compaction=true;max_table_files_size=1024;}"},
+       {"ttl", "191"}}));
+  ASSERT_EQ(dbfull()->GetOptions().compaction_options_fifo.allow_compaction,
+            true);
+  ASSERT_EQ(dbfull()->GetOptions().compaction_options_fifo.max_table_files_size,
+            1024);
+  ASSERT_EQ(dbfull()->GetOptions().ttl, 191);
+}
+
 #endif  // ROCKSDB_LITE

 }  // namespace rocksdb
@@ -212,7 +212,8 @@ void VerifySimilar(uint64_t a, uint64_t b, double bias) {

 void VerifyTableProperties(const TableProperties& base_tp,
                           const TableProperties& new_tp,
-                           double filter_size_bias = 0.1,
+                           double filter_size_bias =
+                               CACHE_LINE_SIZE >= 256 ? 0.15 : 0.1,
                           double index_size_bias = 0.1,
                           double data_size_bias = 0.1,
                           double num_data_blocks_bias = 0.05) {
@@ -266,7 +267,8 @@ void GetExpectedTableProperties(
       // discount 1 byte as value size is not encoded in value delta encoding
       (value_delta_encoding ? 1 : 0));
  expected_tp->filter_size =
-      kTableCount * (kKeysPerTable * kBloomBitsPerKey / 8);
+      kTableCount * ((kKeysPerTable * kBloomBitsPerKey + 7) / 8 +
+                     /*average-ish overhead*/ CACHE_LINE_SIZE / 2);
 }
 }  // anonymous namespace

@@ -615,8 +617,9 @@ TEST_F(DBPropertiesTest, NumImmutableMemTable) {
    writeOpt.disableWAL = true;
    options.max_write_buffer_number = 4;
    options.min_write_buffer_number_to_merge = 3;
-    options.max_write_buffer_number_to_maintain = 4;
    options.write_buffer_size = 1000000;
+    options.max_write_buffer_size_to_maintain =
+        5 * static_cast<int64_t>(options.write_buffer_size);
    CreateAndReopenWithCF({"pikachu"}, options);

    std::string big_value(1000000 * 2, 'x');
@@ -747,7 +750,7 @@ TEST_F(DBPropertiesTest, DISABLED_GetProperty) {
  options.max_background_flushes = 1;
  options.max_write_buffer_number = 10;
  options.min_write_buffer_number_to_merge = 1;
-  options.max_write_buffer_number_to_maintain = 0;
+  options.max_write_buffer_size_to_maintain = 0;
  options.write_buffer_size = 1000000;
  Reopen(options);

@@ -997,7 +1000,7 @@ TEST_F(DBPropertiesTest, EstimatePendingCompBytes) {
  options.max_background_flushes = 1;
  options.max_write_buffer_number = 10;
  options.min_write_buffer_number_to_merge = 1;
-  options.max_write_buffer_number_to_maintain = 0;
+  options.max_write_buffer_size_to_maintain = 0;
  options.write_buffer_size = 1000000;
  Reopen(options);