Facility for cross-building RocksJava using Docker

Summary: As an alternative to Vagrant, we can now also use Docker to cross-build RocksDB. The advantages are: 1. The Docker images are fixed; they include all the latest updates and build tools. 2. The Vagrant image, required scripts that ran for every build that would update CentOS and install the buildtools. This lead to slow repeatable builds, we don't need to do this with Docker as they are already in the provided images. The Docker images I have used have their Docker build files here: https://github.com/evolvedbinary/docker-rocksjava and the images themselves are available from Docker hub: https://hub.docker.com/r/evolvedbinary/rocksjava/ I have added the following targets to the `Makefile`: 1. `rocksdbjavastaticreleasedocker` this uses Docker to perform the cross-builds. It is basically the Docker version of the existing Vagrant `rocksdbjavastaticrelease` target. 2. `rocksdbjavastaticpublishdocker` delegates to `rocksdbjavastaticreleasedocker` and then `rocksdbjavastaticpublishcentral` to upload the artiacts to Maven Central. Equivalent to the existing Vagrant target: `rocksdbjavastaticpublish` Closes https://github.com/facebook/rocksdb/pull/2278 Differential Revision: D5048206 Pulled By: yiwu-arbug fbshipit-source-id: 78fa96ef9d966fe09638ed01de282cd4e31961a9
Bump version to 5.3.5
2026-11-07 03:22:34 -05:00 · 2017-05-17 10:47:09 -07:00 · 2017-05-15 23:44:26 -07:00 · 2017-05-15 23:30:29 -07:00 · 2017-05-15 23:24:15 -07:00 · 2017-05-15 15:44:19 -07:00
489 changed files with 25670 additions and 10430 deletions
@@ -1,6 +1,6 @@
 {
  "project_id" : "rocksdb",
-  "conduit_uri" : "https://reviews.facebook.net/",
+  "conduit_uri" : "https://phabricator.fb.com/api/",
  "copyright_holder" : "Facebook",
  "load" : [
    "arcanist_util"
@@ -37,6 +37,7 @@ util/build_version.cc
 build_tools/VALGRIND_LOGS/
 coverage/COVERAGE_REPORT
 .gdbhistory
+.gdb_history
 package/
 .phutil_module_cache
 unity.a
@@ -69,3 +70,4 @@ LOG
 db_logs/
 tp2/
 fbcode/
+fbcode
@@ -1,37 +1,56 @@
 sudo: false
+dist: trusty
 language: cpp
 os:
  - linux
  - osx
 compiler:
  - clang
+  - gcc
+jdk:
+  - oraclejdk7
+cache:
+  - ccache
+  - apt

 addons:
   apt:
      sources: ['ubuntu-toolchain-r-test', 'llvm-toolchain-precise-3.6']
-      packages: ['clang-3.6' , 'zlib1g-dev', 'libbz2-dev', 'libsnappy-dev', 'curl', 'gflags']
+      packages: ['clang-3.6' , 'g++-6', 'zlib1g-dev', 'libbz2-dev', 'libsnappy-dev', 'curl', 'libgflags-dev']
 env:
-  # Run all tests before db_block_cache_test (db_test, db_test2)
-  - JOB_NAME=unittests ROCKSDBTESTS_END=db_block_cache_test
-  # Run all tests starting from db_block_cache_test (db_block_cache_test, db_iter_test, ...)
-  - JOB_NAME=unittests ROCKSDBTESTS_START=db_block_cache_test
+  - TEST_GROUP=platform_dependent1
+  - TEST_GROUP=platform_dependent2
+  - TEST_GROUP=platform_dependent3
+  - TEST_GROUP=1
+  - TEST_GROUP=2
  # Run java tests
  - JOB_NAME=java_test
  # Build ROCKSDB_LITE
  - JOB_NAME=lite_build

-install:
-  # Build gflags
-  # TODO(noetzli): Remove when gflags available through Travis
-  - pushd /tmp/ && curl -L https://github.com/gflags/gflags/archive/v2.1.2.tar.gz -o gflags.tar.gz && tar xfz gflags.tar.gz && cd gflags-2.1.2 && cmake . && make && popd
+matrix:
+  exclude:
+  - os: osx
+    compiler: gcc
+  - os: osx
+    env: TEST_GROUP=1
+  - os: osx
+    env: TEST_GROUP=2

 before_script:
-  - if [[ "${TRAVIS_OS_NAME}" == 'linux' ]]; then CXX=clang++-3.6; fi
+  - if [[ "${TRAVIS_OS_NAME}" == 'linux' && "${CXX}" == 'clang++' ]]; then CXX=clang++-3.6; fi
+  # test one linux g++ build with g++-6
+  - if [[ "${TRAVIS_OS_NAME}" == 'linux' && "${CXX}" == 'g++' && "${JOB_NAME}" == 'unittests' ]]; then CXX=g++-6; fi
  # Limit the maximum number of open file descriptors to 8192
  - ulimit -n 8192 || true

 script:
-  - if [[ "${JOB_NAME}" == 'unittests' ]]; then OPT=-DTRAVIS V=1 make -j4 check_some; fi
+  - ${CXX} --version
+  - if [[ "${TEST_GROUP}" == 'platform_dependent1' ]]; then OPT=-DTRAVIS V=1 ROCKSDBTESTS_END=external_sst_file_basic_test make -j4 check_some; fi
+  - if [[ "${TEST_GROUP}" == 'platform_dependent2' ]]; then OPT=-DTRAVIS V=1 ROCKSDBTESTS_START=external_sst_file_basic_test ROCKSDBTESTS_END=checkpoint_test make -j4 check_some; fi
+  - if [[ "${TEST_GROUP}" == 'platform_dependent3' ]]; then OPT=-DTRAVIS V=1 ROCKSDBTESTS_START=checkpoint_test ROCKSDBTESTS_END=db_block_cache_test make -j4 check_some; fi
+  - if [[ "${TEST_GROUP}" == '1' ]]; then OPT=-DTRAVIS V=1 ROCKSDBTESTS_START=db_block_cache_test ROCKSDBTESTS_END=comparator_db_test make -j4 check_some; fi
+  - if [[ "${TEST_GROUP}" == '2' ]]; then OPT=-DTRAVIS V=1 ROCKSDBTESTS_START=comparator_db_test make -j4 check_some; fi
  - if [[ "${JOB_NAME}" == 'java_test' ]]; then OPT=-DTRAVIS V=1 make clean jclean rocksdbjava jtest; fi
  - if [[ "${JOB_NAME}" == 'lite_build' ]]; then OPT="-DTRAVIS -DROCKSDB_LITE" V=1 make -j4 static_lib; fi
 notifications:
@@ -1,4 +1,5 @@
-# This cmake build is for Windows 64-bit only.
+# Prerequisites for Windows:
+#     This cmake build is for Windows 64-bit only.
 #
 # Prerequisites:
 #     You must have at least Visual Studio 2015 Update 3. Start the Developer Command Prompt window that is a part of Visual Studio installation.
@@ -24,6 +25,12 @@
 # 5. And release mode (/m[:<N>] is also supported)
 #        msbuild rocksdb.sln /p:Configuration=Release
 #
+# Linux:
+#
+# 1. Install a recent toolchain such as devtoolset-3 if you're on a older distro. C++11 required.
+# 2. mkdir build; cd build
+# 3. cmake ..
+# 4. make -j

 cmake_minimum_required(VERSION 2.6)
 project(rocksdb)
@@ -43,6 +50,11 @@ else()
    add_definitions(-DROCKSDB_JEMALLOC)
    include_directories(${JEMALLOC_INCLUDE_DIR})
  endif()
+  if(CMAKE_SYSTEM_NAME STREQUAL "FreeBSD")
+    # FreeBSD has jemaloc as default malloc
+    add_definitions(-DROCKSDB_JEMALLOC)
+    set(WITH_JEMALLOC ON)
+  endif()
  option(WITH_SNAPPY "build with SNAPPY" OFF)
  if(WITH_SNAPPY)
    find_package(snappy REQUIRED)
@@ -78,25 +90,51 @@ endif()

 string(REGEX REPLACE "[^0-9a-f]+" "" GIT_SHA "${GIT_SHA}")

+if (NOT WIN32)
+  execute_process(COMMAND
+      "./build_tools/version.sh" "full"
+      WORKING_DIRECTORY ${PROJECT_SOURCE_DIR}
+      OUTPUT_VARIABLE ROCKSDB_VERSION
+  )
+  string(STRIP "${ROCKSDB_VERSION}" ROCKSDB_VERSION)
+  execute_process(COMMAND
+      "./build_tools/version.sh" "major"
+      WORKING_DIRECTORY ${PROJECT_SOURCE_DIR}
+      OUTPUT_VARIABLE ROCKSDB_VERSION_MAJOR
+  )
+  string(STRIP "${ROCKSDB_VERSION_MAJOR}" ROCKSDB_VERSION_MAJOR)
+endif()
+
 set(BUILD_VERSION_CC ${CMAKE_BINARY_DIR}/build_version.cc)
-file(GENERATE OUTPUT ${BUILD_VERSION_CC} CONTENT
-  "#include \"build_version.h\"
-const char* rocksdb_build_git_sha = \"rocksdb_build_git_sha:${GIT_SHA}\";
-const char* rocksdb_build_git_datetime = \"rocksdb_build_git_datetime:${GIT_DATE_TIME}\";
-const char* rocksdb_build_compile_date = __DATE__;
-")
+configure_file(util/build_version.cc.in ${BUILD_VERSION_CC} @ONLY)
 add_library(build_version OBJECT ${BUILD_VERSION_CC})
 target_include_directories(build_version PRIVATE
  ${CMAKE_CURRENT_SOURCE_DIR}/util)
 if(WIN32)
  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /Zi /nologo  /EHsc /GS /Gd /GR /GF /fp:precise /Zc:wchar_t /Zc:forScope /errorReport:queue")

-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /FC /d2Zi+ /W3 /WX /wd4127 /wd4800 /wd4996 /wd4351")
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /FC /d2Zi+ /W3 /wd4127 /wd4800 /wd4996 /wd4351")
 else()
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -W -Wextra -Wall -Werror")
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -W -Wextra -Wall")
  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wsign-compare -Wshadow -Wno-unused-parameter -Wno-unused-variable -Woverloaded-virtual -Wnon-virtual-dtor -Wno-missing-field-initializers")
  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=native -fno-omit-frame-pointer -momit-leaf-frame-pointer")
+  if(NOT CMAKE_BUILD_TYPE STREQUAL "Debug")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fno-omit-frame-pointer")
+    include(CheckCXXCompilerFlag)
+    CHECK_CXX_COMPILER_FLAG("-momit-leaf-frame-pointer" HAVE_OMIT_LEAF_FRAME_POINTER)
+    if(HAVE_OMIT_LEAF_FRAME_POINTER)
+      set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -momit-leaf-frame-pointer")
+    endif()
+  endif()
+endif()
+
+option(FAIL_ON_WARNINGS "Treat compile warnings as errors" ON)
+if(FAIL_ON_WARNINGS)
+  if(MSVC)
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /WX")
+  else() # assume GCC
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Werror")
+  endif()
 endif()

 option(WITH_ASAN "build with ASAN" OFF)
@@ -220,19 +258,9 @@ 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)
-
 find_package(Threads REQUIRED)
-if(WIN32)
-  set(SYSTEM_LIBS ${SYSTEM_LIBS} Shlwapi.lib Rpcrt4.lib)
-else()
-  set(SYSTEM_LIBS ${CMAKE_THREAD_LIBS_INIT})
-endif()
-
-set(ROCKSDB_LIBS rocksdblib${ARTIFACT_SUFFIX})
-set(LIBS ${ROCKSDB_LIBS} ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})

 add_subdirectory(third-party/gtest-1.7.0/fused-src/gtest)
-add_subdirectory(tools)

 # Main library source code

@@ -252,11 +280,11 @@ set(SOURCES
        db/db_impl.cc
        db/db_impl_debug.cc
        db/db_impl_experimental.cc
-        db/db_impl_add_file.cc
        db/db_impl_readonly.cc
        db/db_info_dumper.cc
        db/db_iter.cc
        db/event_helpers.cc
+        db/external_sst_file_ingestion_job.cc
        db/experimental.cc
        db/filename.cc
        db/file_indexer.cc
@@ -286,7 +314,6 @@ set(SOURCES
        db/write_batch_base.cc
        db/write_controller.cc
        db/write_thread.cc
-        db/xfunc_test_points.cc
        memtable/hash_cuckoo_rep.cc
        memtable/hash_linklist_rep.cc
        memtable/hash_skiplist_rep.cc
@@ -309,10 +336,12 @@ set(SOURCES
        table/format.cc
        table/full_filter_block.cc
        table/get_context.cc
+        table/index_builder.cc
        table/iterator.cc
-        table/merger.cc
+        table/merging_iterator.cc
        table/sst_file_writer.cc
        table/meta_blocks.cc
+        table/partitioned_filter_block.cc
        table/plain_table_builder.cc
        table/plain_table_factory.cc
        table/plain_table_index.cc
@@ -349,6 +378,7 @@ set(SOURCES
        util/histogram_windowing.cc
        util/instrumented_mutex.cc
        util/iostats_context.cc
+        
        util/lru_cache.cc
        tools/ldb_cmd.cc
        tools/ldb_tool.cc
@@ -379,7 +409,6 @@ set(SOURCES
        util/thread_status_util.cc
        util/thread_status_util_debug.cc
        util/transaction_test_util.cc
-        util/xfunc.cc
        util/xxhash.cc
        utilities/backupable/backupable_db.cc
        utilities/blob_db/blob_db.cc
@@ -390,10 +419,9 @@ set(SOURCES
        utilities/document/json_document.cc
        utilities/document/json_document_builder.cc
        utilities/env_mirror.cc
-        utilities/env_registry.cc
-        utilities/flashcache/flashcache.cc
        utilities/geodb/geodb_impl.cc
        utilities/leveldb_options/leveldb_options.cc
+        utilities/lua/rocks_lua_compaction_filter.cc
        utilities/memory/memory_util.cc
        utilities/merge_operators/string_append/stringappend.cc
        utilities/merge_operators/string_append/stringappend2.cc
@@ -434,6 +462,7 @@ if(WIN32)
    port/win/env_default.cc
    port/win/port_win.cc
    port/win/win_logger.cc
+    port/win/win_thread.cc
    port/win/xpress_win.cc)
 else()
  list(APPEND SOURCES
@@ -442,6 +471,29 @@ else()
    util/io_posix.cc)
 endif()

+if(WIN32)
+  set(SYSTEM_LIBS ${SYSTEM_LIBS} Shlwapi.lib Rpcrt4.lib)
+  set(ROCKSDB_STATIC_LIB rocksdblib${ARTIFACT_SUFFIX})
+  set(ROCKSDB_IMPORT_LIB rocksdb${ARTIFACT_SUFFIX})
+  set(LIBS ${ROCKSDB_STATIC_LIB} ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})
+else()
+  set(SYSTEM_LIBS ${CMAKE_THREAD_LIBS_INIT} rt)
+  set(ROCKSDB_STATIC_LIB rocksdb${ARTIFACT_SUFFIX})
+  set(ROCKSDB_SHARED_LIB rocksdb-shared)
+  set(ROCKSDB_IMPORT_LIB ${ROCKSDB_SHARED_LIB})
+  set(LIBS ${ROCKSDB_SHARED_LIB} ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})
+
+  add_library(${ROCKSDB_SHARED_LIB} SHARED ${SOURCES})
+  target_link_libraries(${ROCKSDB_SHARED_LIB}
+    ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})
+  set_target_properties(${ROCKSDB_SHARED_LIB} PROPERTIES
+                        LINKER_LANGUAGE CXX
+                        VERSION ${ROCKSDB_VERSION}
+                        SOVERSION ${ROCKSDB_VERSION_MAJOR}
+                        CXX_STANDARD 11
+                        OUTPUT_NAME "rocksdb")
+endif()
+
 option(WITH_LIBRADOS "Build with librados" OFF)
 if(WITH_LIBRADOS)
  list(APPEND SOURCES
@@ -449,22 +501,22 @@ if(WITH_LIBRADOS)
  list(APPEND THIRDPARTY_LIBS rados)
 endif()

-add_library(rocksdblib${ARTIFACT_SUFFIX} STATIC ${SOURCES})
-target_link_libraries(rocksdblib${ARTIFACT_SUFFIX}
+add_library(${ROCKSDB_STATIC_LIB} STATIC ${SOURCES})
+target_link_libraries(${ROCKSDB_STATIC_LIB}
  ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})

 if(WIN32)
-  set_target_properties(rocksdblib${ARTIFACT_SUFFIX} PROPERTIES
-    COMPILE_FLAGS "/Fd${CMAKE_CFG_INTDIR}/rocksdblib${ARTIFACT_SUFFIX}.pdb")
+  set_target_properties(${ROCKSDB_STATIC_LIB} PROPERTIES
+    COMPILE_FLAGS "/Fd${CMAKE_CFG_INTDIR}/${ROCKSDB_STATIC_LIB}.pdb")
 endif()

-add_library(rocksdb${ARTIFACT_SUFFIX} SHARED ${SOURCES})
-target_link_libraries(rocksdb${ARTIFACT_SUFFIX}
-  ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})
-
 if(WIN32)
-  set_target_properties(rocksdb${ARTIFACT_SUFFIX} PROPERTIES
-    COMPILE_FLAGS "-DROCKSDB_DLL -DROCKSDB_LIBRARY_EXPORTS /Fd${CMAKE_CFG_INTDIR}/rocksdb${ARTIFACT_SUFFIX}.pdb")
+  add_library(${ROCKSDB_IMPORT_LIB} SHARED ${SOURCES})
+  target_link_libraries(${ROCKSDB_IMPORT_LIB}
+    ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})
+  set_target_properties(${ROCKSDB_IMPORT_LIB} PROPERTIES
+    COMPILE_FLAGS "-DROCKSDB_DLL -DROCKSDB_LIBRARY_EXPORTS /Fd${CMAKE_CFG_INTDIR}/${ROCKSDB_IMPORT_LIB}.pdb")
+else()
 endif()

 option(WITH_JNI "build with JNI" OFF)
@@ -486,12 +538,15 @@ set(TESTS
        db/comparator_db_test.cc
        db/corruption_test.cc
        db/cuckoo_table_db_test.cc
+        db/db_basic_test.cc
        db/db_compaction_filter_test.cc
        db/db_compaction_test.cc
        db/db_dynamic_level_test.cc
        db/db_inplace_update_test.cc
        db/db_iter_test.cc
        db/db_log_iter_test.cc
+        db/db_memtable_test.cc
+        db/db_merge_operator_test.cc
        db/db_options_test.cc
        db/db_properties_test.cc
        db/db_table_properties_test.cc
@@ -502,7 +557,9 @@ set(TESTS
        db/db_bloom_filter_test.cc
        db/db_flush_test.cc
        db/db_iterator_test.cc
+        db/db_range_del_test.cc
        db/db_sst_test.cc
+        db/external_sst_file_basic_test.cc
        db/external_sst_file_test.cc
        db/db_universal_compaction_test.cc
        db/db_wal_test.cc
@@ -575,7 +632,7 @@ set(TESTS
        utilities/date_tiered/date_tiered_test.cc
        utilities/document/document_db_test.cc
        utilities/document/json_document_test.cc
-        utilities/env_registry_test.cc
+        utilities/object_registry_test.cc
        utilities/geodb/geodb_test.cc
        utilities/memory/memory_test.cc
        utilities/merge_operators/string_append/stringappend_test.cc
@@ -591,6 +648,7 @@ set(TESTS
        utilities/ttl/ttl_test.cc
        utilities/write_batch_with_index/write_batch_with_index_test.cc
        utilities/column_aware_encoding_test.cc
+        utilities/lua/rocks_lua_test.cc
 )
 if(WITH_LIBRADOS)
  list(APPEND TESTS utilities/env_librados_test.cc)
@@ -608,7 +666,7 @@ foreach(sourcefile ${BENCHMARKS})
  get_filename_component(exename ${sourcefile} NAME_WE)
  add_executable(${exename}${ARTIFACT_SUFFIX} ${sourcefile}
    $<TARGET_OBJECTS:testharness>)
-  target_link_libraries(${exename}${ARTIFACT_SUFFIX} ${LIBS} gtest)
+  target_link_libraries(${exename}${ARTIFACT_SUFFIX} gtest ${LIBS})
 endforeach(sourcefile ${BENCHMARKS})

 # For test util library that is build only in DEBUG mode
@@ -646,7 +704,7 @@ foreach(sourcefile ${TEST_EXES})
      EXCLUDE_FROM_DEFAULT_BUILD_MINRELEASE 1
      EXCLUDE_FROM_DEFAULT_BUILD_RELWITHDEBINFO 1
      )
-    target_link_libraries(${exename}${ARTIFACT_SUFFIX} testutillib${ARTIFACT_SUFFIX} ${LIBS} gtest)
+    target_link_libraries(${exename}${ARTIFACT_SUFFIX} testutillib${ARTIFACT_SUFFIX} gtest ${LIBS})
    if(NOT "${exename}" MATCHES "db_sanity_test")
      add_test(NAME ${exename} COMMAND ${exename}${ARTIFACT_SUFFIX})
      add_dependencies(check ${exename}${ARTIFACT_SUFFIX})
@@ -666,7 +724,18 @@ foreach(sourcefile ${C_TEST_EXES})
      EXCLUDE_FROM_DEFAULT_BUILD_MINRELEASE 1
      EXCLUDE_FROM_DEFAULT_BUILD_RELWITHDEBINFO 1
      )
-    target_link_libraries(${exename}${ARTIFACT_SUFFIX} rocksdb${ARTIFACT_SUFFIX} testutillib${ARTIFACT_SUFFIX})
+    target_link_libraries(${exename}${ARTIFACT_SUFFIX} ${ROCKSDB_IMPORT_LIB} testutillib${ARTIFACT_SUFFIX})
    add_test(NAME ${exename} COMMAND ${exename}${ARTIFACT_SUFFIX})
    add_dependencies(check ${exename}${ARTIFACT_SUFFIX})
 endforeach(sourcefile ${C_TEST_EXES})
+add_subdirectory(tools)
+
+# Installation and packaging for Linux
+if (NOT WIN32)
+install(TARGETS ${ROCKSDB_STATIC_LIB} COMPONENT devel ARCHIVE DESTINATION lib64)
+install(TARGETS ${ROCKSDB_SHARED_LIB} COMPONENT runtime DESTINATION lib64)
+install(DIRECTORY "${PROJECT_SOURCE_DIR}/include/rocksdb/"
+        COMPONENT devel
+        DESTINATION include/rocksdb)
+set(CMAKE_INSTALL_PREFIX /usr)
+endif()
@@ -1,4 +1,11 @@
 # RocksDB default options change log
+## Unreleased
+* Change the default of delayed slowdown value to 16MB/s and further increase the L0 stop condition to 36 files.
+
+## 5.0 (11/17/2016)
+* Options::allow_concurrent_memtable_write and Options::enable_write_thread_adaptive_yield are now true by default
+* Options.level0_stop_writes_trigger default value changes from 24 to 32.
+
 ## 4.8.0 (5/2/2016)
 * options.max_open_files changes from 5000 to -1. It improves performance, but users need to set file descriptor limit to be large enough and watch memory usage for index and bloom filters.
 * options.base_background_compactions changes from max_background_compactions to 1. When users set higher max_background_compactions but the write throughput is not high, the writes are less spiky to disks.
@@ -1,4 +1,63 @@
 # Rocksdb Change Log
+## 5.3.0 (03/08/2017)
+### Public API Change
+* Remove disableDataSync option.
+* Remove timeout_hint_us option from WriteOptions. The option has been deprecated and has no effect since 3.13.0.
+* Remove option min_partial_merge_operands. Partial merge operands will always be merged in flush or compaction if there are more than one.
+* Remove option verify_checksums_in_compaction. Compaction will always verify checksum.
+
+### New Features
+* Memtable flush can be avoided during checkpoint creation if total log file size is smaller than a threshold specified by the user.
+
+### Bug Fixes
+* Fix the bug that iterator may skip keys
+
+## 5.2.0 (02/08/2017)
+### Public API Change
+* NewLRUCache() will determine number of shard bits automatically based on capacity, if the user doesn't pass one. This also impacts the default block cache when the user doesn't explict provide one.
+* Change the default of delayed slowdown value to 16MB/s and further increase the L0 stop condition to 36 files.
+* Options::use_direct_writes and Options::use_direct_reads are now ready to use.
+* (Experimental) Two-level indexing that partition the index and creates a 2nd level index on the partitions. The feature can be enabled by setting kTwoLevelIndexSearch as IndexType and configuring index_per_partition.
+
+### New Features
+* Added new overloaded function GetApproximateSizes that allows to specify if memtable stats should be computed only without computing SST files' stats approximations.
+* Added new function GetApproximateMemTableStats that approximates both number of records and size of memtables.
+* Add Direct I/O mode for SST file I/O
+
+### Bug Fixes
+* RangeSync() should work if ROCKSDB_FALLOCATE_PRESENT is not set
+* Fix wrong results in a data race case in Get()
+* Some fixes related to 2PC.
+* Fix bugs of data corruption in direct I/O
+
+## 5.1.0 (01/13/2017)
+* Support dynamically change `delete_obsolete_files_period_micros` option via SetDBOptions().
+* Added EventListener::OnExternalFileIngested which will be called when IngestExternalFile() add a file successfully.
+* BackupEngine::Open and BackupEngineReadOnly::Open now always return error statuses matching those of the backup Env.
+
+### Bug Fixes
+* Fix the bug that if 2PC is enabled, checkpoints may loss some recent transactions.
+* When file copying is needed when creating checkpoints or bulk loading files, fsync the file after the file copying.
+
+## 5.0.0 (11/17/2016)
+### Public API Change
+* Options::max_bytes_for_level_multiplier is now a double along with all getters and setters.
+* Support dynamically change `delayed_write_rate` and `max_total_wal_size` options via SetDBOptions().
+* Introduce DB::DeleteRange for optimized deletion of large ranges of contiguous keys.
+* Support dynamically change `delayed_write_rate` option via SetDBOptions().
+* Options::allow_concurrent_memtable_write and Options::enable_write_thread_adaptive_yield are now true by default.
+* Remove Tickers::SEQUENCE_NUMBER to avoid confusion if statistics object is shared among RocksDB instance. Alternatively DB::GetLatestSequenceNumber() can be used to get the same value.
+* Options.level0_stop_writes_trigger default value changes from 24 to 32.
+* New compaction filter API: CompactionFilter::FilterV2(). Allows to drop ranges of keys.
+* Removed flashcache support.
+* DB::AddFile() is deprecated and is replaced with DB::IngestExternalFile(). DB::IngestExternalFile() remove all the restrictions that existed for DB::AddFile.
+
+### New Features
+* Add avoid_flush_during_shutdown option, which speeds up DB shutdown by not flushing unpersisted data (i.e. with disableWAL = true). Unpersisted data will be lost. The options is dynamically changeable via SetDBOptions().
+* Add memtable_insert_with_hint_prefix_extractor option. The option is mean to reduce CPU usage for inserting keys into memtable, if keys can be group by prefix and insert for each prefix are sequential or almost sequential. See include/rocksdb/options.h for more details.
+* Add LuaCompactionFilter in utilities.  This allows developers to write compaction filters in Lua.  To use this feature, LUA_PATH needs to be set to the root directory of Lua.
+* No longer populate "LATEST_BACKUP" file in backup directory, which formerly contained the number of the latest backup. The latest backup can be determined by finding the highest numbered file in the "meta/" subdirectory.
+
 ## 4.13.0 (10/18/2016)
 ### Public API Change
 * DB::GetOptions() reflect dynamic changed options (i.e. through DB::SetOptions()) and return copy of options instead of reference.
@@ -29,8 +29,10 @@ your make commands, like this: `PORTABLE=1 make static_lib`
 * You can link RocksDB with following compression libraries:
  - [zlib](http://www.zlib.net/) - a library for data compression.
  - [bzip2](http://www.bzip.org/) - a library for data compression.
-  - [snappy](https://code.google.com/p/snappy/) - a library for fast
+  - [snappy](http://google.github.io/snappy/) - a library for fast
      data compression.
+  - [zstandard](http://www.zstd.net) - Fast real-time compression
+      algorithm.

 * All our tools depend on:
  - [gflags](https://gflags.github.io/gflags/) - a library that handles
@@ -40,7 +42,7 @@ your make commands, like this: `PORTABLE=1 make static_lib`
 ## Supported platforms

 * **Linux - Ubuntu**
-    * Upgrade your gcc to version at least 4.7 to get C++11 support.
+    * Upgrade your gcc to version at least 4.8 to get C++11 support.
    * Install gflags. First, try: `sudo apt-get install libgflags-dev`
      If this doesn't work and you're using Ubuntu, here's a nice tutorial:
      (http://askubuntu.com/questions/312173/installing-gflags-12-04)
@@ -48,21 +50,23 @@ your make commands, like this: `PORTABLE=1 make static_lib`
      `sudo apt-get install libsnappy-dev`.
    * Install zlib. Try: `sudo apt-get install zlib1g-dev`.
    * Install bzip2: `sudo apt-get install libbz2-dev`.
-* **Linux - CentOS**
-    * Upgrade your gcc to version at least 4.7 to get C++11 support:
-      `yum install gcc47-c++`
+    * Install zstandard: `sudo apt-get install libzstd-dev`.
+
+* **Linux - CentOS / RHEL**
+    * Upgrade your gcc to version at least 4.8 to get C++11 support:
+      `yum install gcc48-c++`
    * Install gflags:

-              wget https://gflags.googlecode.com/files/gflags-2.0-no-svn-files.tar.gz
-              tar -xzvf gflags-2.0-no-svn-files.tar.gz
-              cd gflags-2.0
+              git clone https://github.com/gflags/gflags.git
+              git checkout v2.0
+              cd gflags
              ./configure && make && sudo make install

    * Install snappy:

-              wget https://snappy.googlecode.com/files/snappy-1.1.1.tar.gz
-              tar -xzvf snappy-1.1.1.tar.gz
-              cd snappy-1.1.1
+              wget https://github.com/google/snappy/releases/download/1.1.4/snappy-1.1.4.tar.gz
+              tar -xzvf snappy-1.1.4.tar.gz
+              cd snappy-1.1.4
              ./configure && make && sudo make install

    * Install zlib:
@@ -75,12 +79,20 @@ your make commands, like this: `PORTABLE=1 make static_lib`
              sudo yum install bzip2
              sudo yum install bzip2-devel

+    * Install zstandard:
+
+             wget https://github.com/facebook/zstd/archive/v1.1.3.tar.gz
+             mv v1.1.3.tar.gz zstandard-1.1.3.tar.gz
+             tar zxvf zstd-1.1.3.tar.gz
+             cd zstd-1.1.3
+             make && sudo make install
+
 * **OS X**:
    * Install latest C++ compiler that supports C++ 11:
        * Update XCode:  run `xcode-select --install` (or install it from XCode App's settting).
        * Install via [homebrew](http://brew.sh/).
            * If you're first time developer in MacOS, you still need to run: `xcode-select --install` in your command line.
-            * run `brew tap homebrew/versions; brew install gcc47 --use-llvm` to install gcc 4.7 (or higher).
+            * run `brew tap homebrew/versions; brew install gcc48 --use-llvm` to install gcc 4.8 (or higher).
    * run `brew install rocksdb`

 * **iOS**:
@@ -10,3 +10,4 @@ This is the list of all known third-party language bindings for RocksDB. If some
 * PHP - https://github.com/Photonios/rocksdb-php
 * C# - https://github.com/warrenfalk/rocksdb-sharp
 * Rust - https://github.com/spacejam/rust-rocksdb
+* D programming language - https://github.com/b1naryth1ef/rocksdb
@@ -6,6 +6,9 @@

 #-----------------------------------------------

+BASH_EXISTS := $(shell which bash)
+SHELL := $(shell which bash)
+
 CLEAN_FILES = # deliberately empty, so we can append below.
 CFLAGS += ${EXTRA_CFLAGS}
 CXXFLAGS += ${EXTRA_CXXFLAGS}
@@ -137,7 +140,7 @@ endif

 AM_LINK = $(AM_V_CCLD)$(CXX) $^ $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
 # detect what platform we're building on
-dummy := $(shell (export ROCKSDB_ROOT="$(CURDIR)"; "$(CURDIR)/build_tools/build_detect_platform" "$(CURDIR)/make_config.mk"))
+dummy := $(shell (export ROCKSDB_ROOT="$(CURDIR)"; export PORTABLE="$(PORTABLE)"; "$(CURDIR)/build_tools/build_detect_platform" "$(CURDIR)/make_config.mk"))
 # this file is generated by the previous line to set build flags and sources
 include make_config.mk
 CLEAN_FILES += make_config.mk
@@ -166,6 +169,8 @@ ifneq ($(filter -DROCKSDB_LITE,$(OPT)),)
 	# found
 	CFLAGS += -fno-exceptions
 	CXXFLAGS += -fno-exceptions
+	# LUA is not supported under ROCKSDB_LITE
+	LUA_PATH =
 endif

 # ASAN doesn't work well with jemalloc. If we're compiling with ASAN, we should use regular malloc.
@@ -185,6 +190,8 @@ ifdef COMPILE_WITH_TSAN
        # Turn off -pg when enabling TSAN testing, because that induces
        # a link failure.  TODO: find the root cause
 	PROFILING_FLAGS =
+	# LUA is not supported under TSAN
+	LUA_PATH =
 endif

 # USAN doesn't work well with jemalloc. If we're compiling with USAN, we should use regular malloc.
@@ -220,6 +227,34 @@ ifndef DISABLE_WARNING_AS_ERROR
 	WARNING_FLAGS += -Werror
 endif

+
+ifdef LUA_PATH
+
+ifndef LUA_INCLUDE
+LUA_INCLUDE=$(LUA_PATH)/include
+endif
+
+LUA_INCLUDE_FILE=$(LUA_INCLUDE)/lualib.h
+
+ifeq ("$(wildcard $(LUA_INCLUDE_FILE))", "")
+# LUA_INCLUDE_FILE does not exist
+$(error Cannot find lualib.h under $(LUA_INCLUDE).  Try to specify both LUA_PATH and LUA_INCLUDE manually)
+endif
+LUA_FLAGS = -I$(LUA_INCLUDE) -DLUA -DLUA_COMPAT_ALL
+CFLAGS += $(LUA_FLAGS)
+CXXFLAGS += $(LUA_FLAGS)
+
+ifndef LUA_LIB
+LUA_LIB = $(LUA_PATH)/lib/liblua.a
+endif
+ifeq ("$(wildcard $(LUA_LIB))", "") # LUA_LIB does not exist
+$(error $(LUA_LIB) does not exist.  Try to specify both LUA_PATH and LUA_LIB manually)
+endif
+LDFLAGS += $(LUA_LIB)
+
+endif
+
+
 CFLAGS += $(WARNING_FLAGS) -I. -I./include $(PLATFORM_CCFLAGS) $(OPT)
 CXXFLAGS += $(WARNING_FLAGS) -I. -I./include $(PLATFORM_CXXFLAGS) $(OPT) -Woverloaded-virtual -Wnon-virtual-dtor -Wno-missing-field-initializers

@@ -231,14 +266,7 @@ ifdef FORCE_GIT_SHA
 else
 	git_sha := $(shell git rev-parse HEAD 2>/dev/null)
 endif
-gen_build_version =							\
-  printf '%s\n'								\
-    '\#include "build_version.h"'					\
-    'const char* rocksdb_build_git_sha =				\
-      "rocksdb_build_git_sha:$(git_sha)";'			\
-    'const char* rocksdb_build_git_date =				\
-      "rocksdb_build_git_date:$(date)";'				\
-    'const char* rocksdb_build_compile_date = __DATE__;'
+gen_build_version = sed -e s/@@GIT_SHA@@/$(git_sha)/ -e s/@@GIT_DATE_TIME@@/$(date)/ util/build_version.cc.in

 # Record the version of the source that we are compiling.
 # We keep a record of the git revision in this file.  It is then built
@@ -271,9 +299,26 @@ BENCHTOOLOBJECTS = $(BENCH_LIB_SOURCES:.cc=.o) $(LIBOBJECTS) $(TESTUTIL)
 EXPOBJECTS = $(EXP_LIB_SOURCES:.cc=.o) $(LIBOBJECTS) $(TESTUTIL)

 TESTS = \
-	db_test \
+	db_basic_test \
 	db_test2 \
+	external_sst_file_basic_test \
+	auto_roll_logger_test \
+	bloom_test \
+	dynamic_bloom_test \
+	c_test \
+	checkpoint_test \
+	crc32c_test \
+	coding_test \
+	inlineskiplist_test \
+	env_basic_test \
+	env_test \
+	thread_local_test \
+	rate_limiter_test \
+	perf_context_test \
+	iostats_context_test \
+	db_wal_test \
 	db_block_cache_test \
+	db_test \
 	db_bloom_filter_test \
 	db_iter_test \
 	db_log_iter_test \
@@ -283,33 +328,26 @@ TESTS = \
 	db_flush_test \
 	db_inplace_update_test \
 	db_iterator_test \
+	db_memtable_test \
+	db_merge_operator_test \
 	db_options_test \
+	db_range_del_test \
 	db_sst_test \
-	external_sst_file_test \
 	db_tailing_iter_test \
 	db_universal_compaction_test \
-	db_wal_test \
 	db_io_failure_test \
 	db_properties_test \
 	db_table_properties_test \
 	autovector_test \
+	cleanable_test \
 	column_family_test \
 	table_properties_collector_test \
 	arena_test \
-	auto_roll_logger_test \
 	block_test \
-	bloom_test \
-	dynamic_bloom_test \
-	c_test \
 	cache_test \
-	checkpoint_test \
-	coding_test \
 	corruption_test \
-	crc32c_test \
 	slice_transform_test \
 	dbformat_test \
-	env_basic_test \
-	env_test \
 	fault_injection_test \
 	filelock_test \
 	filename_test \
@@ -318,7 +356,6 @@ TESTS = \
 	full_filter_block_test \
 	hash_table_test \
 	histogram_test \
-	inlineskiplist_test \
 	log_test \
 	manual_compaction_test \
 	mock_env_test \
@@ -333,6 +370,7 @@ TESTS = \
 	reduce_levels_test \
 	plain_table_db_test \
 	comparator_db_test \
+	external_sst_file_test \
 	prefix_test \
 	skiplist_test \
 	stringappend_test \
@@ -354,9 +392,7 @@ TESTS = \
 	write_controller_test\
 	deletefile_test \
 	table_test \
-	thread_local_test \
 	geodb_test \
-	rate_limiter_test \
 	delete_scheduler_test \
 	options_test \
 	options_settable_test \
@@ -374,7 +410,6 @@ TESTS = \
 	sst_dump_test \
 	column_aware_encoding_test \
 	compact_files_test \
-	perf_context_test \
 	optimistic_transaction_test \
 	write_callback_test \
 	heap_test \
@@ -383,10 +418,13 @@ TESTS = \
 	option_change_migration_test \
 	transaction_test \
 	ldb_cmd_test \
-	iostats_context_test \
 	persistent_cache_test \
 	statistics_test \
+	lua_test \
+	range_del_aggregator_test \
 	lru_cache_test \
+	object_registry_test \
+	repair_test \

 PARALLEL_TEST = \
 	backupable_db_test \
@@ -400,7 +438,9 @@ PARALLEL_TEST = \
 	fault_injection_test \
 	inlineskiplist_test \
 	manual_compaction_test \
-	table_test
+	persistent_cache_test \
+	table_test \
+	transaction_test

 SUBSET := $(TESTS)
 ifdef ROCKSDBTESTS_START
@@ -632,6 +672,8 @@ check_0:
 	  | grep -E '$(tests-regexp)'					\
 	  | build_tools/gnu_parallel -j$(J) --plain --joblog=LOG $$eta --gnu '{} >& t/log-{/}'

+valgrind-blacklist-regexp = InlineSkipTest.ConcurrentInsert|TransactionTest.DeadlockStress|DBCompactionTest.SuggestCompactRangeNoTwoLevel0Compactions|BackupableDBTest.RateLimiting|DBTest.CloseSpeedup|DBTest.ThreadStatusFlush|DBTest.RateLimitingTest|DBTest.EncodeDecompressedBlockSizeTest|FaultInjectionTest.UninstalledCompaction|HarnessTest.Randomized|ExternalSSTFileTest.CompactDuringAddFileRandom|ExternalSSTFileTest.IngestFileWithGlobalSeqnoRandomized
+
 .PHONY: valgrind_check_0
 valgrind_check_0:
 	$(AM_V_GEN)export TEST_TMPDIR=$(TMPD);				\
@@ -645,6 +687,7 @@ valgrind_check_0:
 	}								\
 	  | $(prioritize_long_running_tests)				\
 	  | grep -E '$(tests-regexp)'					\
+	  | grep -E -v '$(valgrind-blacklist-regexp)'					\
 	  | build_tools/gnu_parallel -j$(J) --plain --joblog=LOG $$eta --gnu \
 	  '(if [[ "{}" == "./"* ]] ; then $(DRIVER) {}; else {}; fi) ' \
 	  '>& t/valgrind_log-{/}'
@@ -680,7 +723,8 @@ ifeq ($(filter -DROCKSDB_LITE,$(OPT)),)
 	sh tools/rocksdb_dump_test.sh
 endif

-check_some: $(SUBSET) ldb_tests
+# TODO add ldb_tests
+check_some: $(SUBSET)
 	for t in $(SUBSET); do echo "===== Running $$t"; ./$$t || exit 1; done

 .PHONY: ldb_tests
@@ -723,6 +767,9 @@ ubsan_crash_test:
 	COMPILE_WITH_UBSAN=1 $(MAKE) crash_test
 	$(MAKE) clean

+valgrind_test:
+	DISABLE_JEMALLOC=1 $(MAKE) valgrind_check
+
 valgrind_check: $(TESTS)
 	$(MAKE) DRIVER="$(VALGRIND_VER) $(VALGRIND_OPTS)" gen_parallel_tests
 	$(AM_V_GEN)if test "$(J)" != 1                                  \
@@ -926,6 +973,9 @@ crc32c_test: util/crc32c_test.o $(LIBOBJECTS) $(TESTHARNESS)
 slice_transform_test: util/slice_transform_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+db_basic_test: db/db_basic_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 db_test: db/db_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

@@ -959,12 +1009,24 @@ db_inplace_update_test: db/db_inplace_update_test.o db/db_test_util.o $(LIBOBJEC
 db_iterator_test: db/db_iterator_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+db_memtable_test: db/db_memtable_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
+db_merge_operator_test: db/db_merge_operator_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)

+db_range_del_test: db/db_range_del_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 db_sst_test: db/db_sst_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+external_sst_file_basic_test: db/external_sst_file_basic_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 external_sst_file_test: db/external_sst_file_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

@@ -1036,7 +1098,7 @@ env_librados_test: utilities/env_librados_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_V_CCLD)$(CXX) $^ $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
 endif

-env_registry_test: utilities/env_registry_test.o $(LIBOBJECTS) $(TESTHARNESS)
+object_registry_test: utilities/object_registry_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 ttl_test: utilities/ttl/ttl_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -1099,6 +1161,9 @@ full_filter_block_test: table/full_filter_block_test.o $(LIBOBJECTS) $(TESTHARNE
 log_test: db/log_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+cleanable_test: table/cleanable_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 table_test: table/table_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

@@ -1258,6 +1323,11 @@ statistics_test: util/statistics_test.o $(LIBOBJECTS) $(TESTHARNESS)
 lru_cache_test: util/lru_cache_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+lua_test: utilities/lua/rocks_lua_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
+range_del_aggregator_test: db/range_del_aggregator_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)

 #-------------------------------------------------
 # make install related stuff
@@ -1311,60 +1381,113 @@ ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-linu
 ROCKSDB_JAR_ALL = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH).jar
 ROCKSDB_JAVADOCS_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-javadoc.jar
 ROCKSDB_SOURCES_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-sources.jar
+SHA256_CMD = sha256sum
+
+ZLIB_VER = 1.2.11
+ZLIB_SHA256 = c3e5e9fdd5004dcb542feda5ee4f0ff0744628baf8ed2dd5d66f8ca1197cb1a1
+BZIP2_VER = 1.0.6
+BZIP2_SHA256 = a2848f34fcd5d6cf47def00461fcb528a0484d8edef8208d6d2e2909dc61d9cd
+SNAPPY_VER = 1.1.4
+SNAPPY_SHA256 = 134bfe122fd25599bb807bb8130e7ba6d9bdb851e0b16efcb83ac4f5d0b70057
+LZ4_VER = 1.7.5
+LZ4_SHA256 = 0190cacd63022ccb86f44fa5041dc6c3804407ad61550ca21c382827319e7e7e
+ZSTD_VER ?= 1.2.0
+ZSTD_SHA256 ?= 4a7e4593a3638276ca7f2a09dc4f38e674d8317bbea51626393ca73fc047cbfb
+ZSTD_DOWNLOAD_BASE ?= https://github.com/facebook/zstd/archive

 ifeq ($(PLATFORM), OS_MACOSX)
 	ROCKSDBJNILIB = librocksdbjni-osx.jnilib
 	ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-osx.jar
+	SHA256_CMD = openssl sha256 -r
 ifneq ("$(wildcard $(JAVA_HOME)/include/darwin)","")
 	JAVA_INCLUDE = -I$(JAVA_HOME)/include -I $(JAVA_HOME)/include/darwin
 else
 	JAVA_INCLUDE = -I/System/Library/Frameworks/JavaVM.framework/Headers/
 endif
 endif
+ifeq ($(PLATFORM), OS_FREEBSD)
+	JAVA_INCLUDE += -I$(JAVA_HOME)/include/freebsd
+	ROCKSDBJNILIB = librocksdbjni-freebsd$(ARCH).so
+	ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-freebsd$(ARCH).jar
+endif
 ifeq ($(PLATFORM), OS_SOLARIS)
 	ROCKSDBJNILIB = librocksdbjni-solaris$(ARCH).so
 	ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-solaris$(ARCH).jar
 	JAVA_INCLUDE = -I$(JAVA_HOME)/include/ -I$(JAVA_HOME)/include/solaris
+	SHA256_CMD = digest -a sha256
 endif

 libz.a:
-	-rm -rf zlib-1.2.8
-	curl -O -L http://zlib.net/zlib-1.2.8.tar.gz
-	tar xvzf zlib-1.2.8.tar.gz
-	cd zlib-1.2.8 && CFLAGS='-fPIC' ./configure --static && make
-	cp zlib-1.2.8/libz.a .
+	-rm -rf zlib-$(ZLIB_VER)
+	curl -O -L http://zlib.net/zlib-$(ZLIB_VER).tar.gz
+	ZLIB_SHA256_ACTUAL=`$(SHA256_CMD) zlib-$(ZLIB_VER).tar.gz | cut -d ' ' -f 1`; \
+	if [ "$(ZLIB_SHA256)" != "$$ZLIB_SHA256_ACTUAL" ]; then \
+		echo zlib-$(ZLIB_VER).tar.gz checksum mismatch, expected=\"$(ZLIB_SHA256)\" actual=\"$$ZLIB_SHA256_ACTUAL\"; \
+		exit 1; \
+	fi
+	tar xvzf zlib-$(ZLIB_VER).tar.gz
+	cd zlib-$(ZLIB_VER) && CFLAGS='-fPIC' ./configure --static && make
+	cp zlib-$(ZLIB_VER)/libz.a .

 libbz2.a:
-	-rm -rf bzip2-1.0.6
-	curl -O -L http://www.bzip.org/1.0.6/bzip2-1.0.6.tar.gz
-	tar xvzf bzip2-1.0.6.tar.gz
-	cd bzip2-1.0.6 && make CFLAGS='-fPIC -O2 -g -D_FILE_OFFSET_BITS=64'
-	cp bzip2-1.0.6/libbz2.a .
+	-rm -rf bzip2-$(BZIP2_VER)
+	curl -O -L http://www.bzip.org/$(BZIP2_VER)/bzip2-$(BZIP2_VER).tar.gz
+	BZIP2_SHA256_ACTUAL=`$(SHA256_CMD) bzip2-$(BZIP2_VER).tar.gz | cut -d ' ' -f 1`; \
+	if [ "$(BZIP2_SHA256)" != "$$BZIP2_SHA256_ACTUAL" ]; then \
+		echo bzip2-$(BZIP2_VER).tar.gz checksum mismatch, expected=\"$(BZIP2_SHA256)\" actual=\"$$BZIP2_SHA256_ACTUAL\"; \
+		exit 1; \
+	fi
+	tar xvzf bzip2-$(BZIP2_VER).tar.gz
+	cd bzip2-$(BZIP2_VER) && make CFLAGS='-fPIC -O2 -g -D_FILE_OFFSET_BITS=64'
+	cp bzip2-$(BZIP2_VER)/libbz2.a .

 libsnappy.a:
-	-rm -rf snappy-1.1.3
-	curl -O -L https://github.com/google/snappy/releases/download/1.1.3/snappy-1.1.3.tar.gz
-	tar xvzf snappy-1.1.3.tar.gz
-	cd snappy-1.1.3 && ./configure --with-pic --enable-static
-	cd snappy-1.1.3 && make
-	cp snappy-1.1.3/.libs/libsnappy.a .
+	-rm -rf snappy-$(SNAPPY_VER)
+	curl -O -L https://github.com/google/snappy/releases/download/$(SNAPPY_VER)/snappy-$(SNAPPY_VER).tar.gz
+	SNAPPY_SHA256_ACTUAL=`$(SHA256_CMD) snappy-$(SNAPPY_VER).tar.gz | cut -d ' ' -f 1`; \
+	if [ "$(SNAPPY_SHA256)" != "$$SNAPPY_SHA256_ACTUAL" ]; then \
+		echo snappy-$(SNAPPY_VER).tar.gz checksum mismatch, expected=\"$(SNAPPY_SHA256)\" actual=\"$$SNAPPY_SHA256_ACTUAL\"; \
+		exit 1; \
+	fi
+	tar xvzf snappy-$(SNAPPY_VER).tar.gz
+	cd snappy-$(SNAPPY_VER) && ./configure --with-pic --enable-static
+	cd snappy-$(SNAPPY_VER) && make
+	cp snappy-$(SNAPPY_VER)/.libs/libsnappy.a .

 liblz4.a:
-	   -rm -rf lz4-r127
-	   curl -O -L https://codeload.github.com/Cyan4973/lz4/tar.gz/r127
-	   mv r127 lz4-r127.tar.gz
-	   tar xvzf lz4-r127.tar.gz
-	   cd lz4-r127/lib && make CFLAGS='-fPIC' all
-	   cp lz4-r127/lib/liblz4.a .
+	-rm -rf lz4-$(LZ4_VER)
+	curl -O -L https://github.com/lz4/lz4/archive/v$(LZ4_VER).tar.gz
+	mv v$(LZ4_VER).tar.gz lz4-$(LZ4_VER).tar.gz
+	LZ4_SHA256_ACTUAL=`$(SHA256_CMD) lz4-$(LZ4_VER).tar.gz | cut -d ' ' -f 1`; \
+	if [ "$(LZ4_SHA256)" != "$$LZ4_SHA256_ACTUAL" ]; then \
+		echo lz4-$(LZ4_VER).tar.gz checksum mismatch, expected=\"$(LZ4_SHA256)\" actual=\"$$LZ4_SHA256_ACTUAL\"; \
+		exit 1; \
+	fi
+	tar xvzf lz4-$(LZ4_VER).tar.gz
+	cd lz4-$(LZ4_VER)/lib && make CFLAGS='-fPIC' all
+	cp lz4-$(LZ4_VER)/lib/liblz4.a .
+
+libzstd.a:
+	-rm -rf zstd-$(ZSTD_VER)
+	curl -O -L ${ZSTD_DOWNLOAD_BASE}/v$(ZSTD_VER).tar.gz
+	mv v$(ZSTD_VER).tar.gz zstd-$(ZSTD_VER).tar.gz
+	ZSTD_SHA256_ACTUAL=`$(SHA256_CMD) zstd-$(ZSTD_VER).tar.gz | cut -d ' ' -f 1`; \
+	if [ "$(ZSTD_SHA256)" != "$$ZSTD_SHA256_ACTUAL" ]; then \
+		echo zstd-$(ZSTD_VER).tar.gz checksum mismatch, expected=\"$(ZSTD_SHA256)\" actual=\"$$ZSTD_SHA256_ACTUAL\"; \
+		exit 1; \
+	fi
+	tar xvzf zstd-$(ZSTD_VER).tar.gz
+	cd zstd-$(ZSTD_VER)/lib && make CFLAGS='-fPIC -O2 ${EXTRA_CFLAGS}' all
+	cp zstd-$(ZSTD_VER)/lib/libzstd.a .

 # A version of each $(LIBOBJECTS) compiled with -fPIC and a fixed set of static compression libraries
 java_static_libobjects = $(patsubst %,jls/%,$(LIBOBJECTS))
 CLEAN_FILES += jls

-JAVA_STATIC_FLAGS = -DZLIB -DBZIP2 -DSNAPPY -DLZ4
-JAVA_STATIC_INCLUDES = -I./zlib-1.2.8 -I./bzip2-1.0.6 -I./snappy-1.1.3 -I./lz4-r127/lib
+JAVA_STATIC_FLAGS = -DZLIB -DBZIP2 -DSNAPPY -DLZ4 -DZSTD
+JAVA_STATIC_INCLUDES = -I./zlib-$(ZLIB_VER) -I./bzip2-$(BZIP2_VER) -I./snappy-$(SNAPPY_VER) -I./lz4-$(LZ4_VER)/lib -I./zstd-$(ZSTD_VER)/lib

-$(java_static_libobjects): jls/%.o: %.cc libz.a libbz2.a libsnappy.a liblz4.a
+$(java_static_libobjects): jls/%.o: %.cc libz.a libbz2.a libsnappy.a liblz4.a libzstd.a
 	$(AM_V_CC)mkdir -p $(@D) && $(CXX) $(CXXFLAGS) $(JAVA_STATIC_FLAGS) $(JAVA_STATIC_INCLUDES) -fPIC -c $< -o $@ $(COVERAGEFLAGS)

 rocksdbjavastatic: $(java_static_libobjects)
@@ -1387,8 +1510,25 @@ rocksdbjavastaticrelease: rocksdbjavastatic
 	cd java/target;jar -uf $(ROCKSDB_JAR_ALL) librocksdbjni-*.so librocksdbjni-*.jnilib
 	cd java/target/classes;jar -uf ../$(ROCKSDB_JAR_ALL) org/rocksdb/*.class org/rocksdb/util/*.class

+rocksdbjavastaticreleasedocker: rocksdbjavastatic
+	DOCKER_LINUX_X64_CONTAINER=`docker ps -aqf name=rocksdb_linux_x64-be`; \
+	if [ -z "$$DOCKER_LINUX_X64_CONTAINER" ]; then \
+		docker container create --attach stdin --attach stdout --attach stderr --volume `pwd`:/rocksdb-host --name rocksdb_linux_x64-be evolvedbinary/rocksjava:centos5_x64-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh; \
+	fi
+	docker start -a rocksdb_linux_x64-be
+	DOCKER_LINUX_X86_CONTAINER=`docker ps -aqf name=rocksdb_linux_x86-be`; \
+	if [ -z "$$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:centos5_x86-be /rocksdb-host/java/crossbuild/docker-build-linux-centos.sh; \
+	fi
+	docker start -a rocksdb_linux_x86-be
+	cd java;jar -cf target/$(ROCKSDB_JAR_ALL) HISTORY*.md
+	cd java/target;jar -uf $(ROCKSDB_JAR_ALL) librocksdbjni-*.so librocksdbjni-*.jnilib
+	cd java/target/classes;jar -uf ../$(ROCKSDB_JAR_ALL) org/rocksdb/*.class org/rocksdb/util/*.class
+
 rocksdbjavastaticpublish: rocksdbjavastaticrelease rocksdbjavastaticpublishcentral

+rocksdbjavastaticpublishdocker: rocksdbjavastaticreleasedocker rocksdbjavastaticpublishcentral
+
 rocksdbjavastaticpublishcentral:
 	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/target/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-javadoc.jar -Dclassifier=javadoc
 	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/target/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-sources.jar -Dclassifier=sources
@@ -1433,13 +1573,6 @@ commit_prereq: build_tools/rocksdb-lego-determinator \
 	J=$(J) build_tools/precommit_checker.py unit unit_481 clang_unit release release_481 clang_release tsan asan ubsan lite unit_non_shm
 	$(MAKE) clean && $(MAKE) jclean && $(MAKE) rocksdbjava;

-xfunc:
-	for xftest in $(XFUNC_TESTS); do \
-		echo "===== Running xftest $$xftest"; \
-		make check ROCKSDB_XFUNC_TEST="$$xftest" tests-regexp="DBTest" ;\
-	done
-
-
 # ---------------------------------------------------------------------------
 #  	Platform-specific compilation
 # ---------------------------------------------------------------------------
@@ -7,7 +7,7 @@ At Facebook, we use RocksDB as storage engines in multiple data management servi
 2. MongoRocks -- https://github.com/mongodb-partners/mongo-rocks
 3. ZippyDB --  Facebook's distributed key-value store with Paxos-style replication, built on top of RocksDB.[*] https://www.youtube.com/watch?v=DfiN7pG0D0khtt
 4. Laser -- Laser is a high query throughput, low (millisecond) latency, key-value storage service built on top of RocksDB.[*]
-4. Dragan -- a distributed graph query engine. https://code.facebook.com/posts/1737605303120405/dragon-a-distributed-graph-query-engine/
+4. Dragon -- a distributed graph query engine. https://code.facebook.com/posts/1737605303120405/dragon-a-distributed-graph-query-engine/
 5. Stylus -- a low-level stream processing framework writtenin C++.[*]

 [*] https://research.facebook.com/publications/realtime-data-processing-at-facebook/
@@ -64,7 +64,7 @@ Pinterest's Object Retrieval System uses RocksDB for storage: https://www.youtub
 quasardb uses a heavily tuned RocksDB as its persistence layer.

 ## Netflix
-[Netflix](http://techblog.netflix.com/2016/05/application-data-caching-using-ssds.html) uses RocksDB on spinning disks to cache application data.
+[Netflix](http://techblog.netflix.com/2016/05/application-data-caching-using-ssds.html) Netflix uses RocksDB on AWS EC2 instances with local SSD drives to cache application data.

 ## TiKV
 [TiKV](https://github.com/pingcap/tikv) is a GEO-replicated, high-performance, distributed, transactional key-value database. TiKV is powered by Rust and Raft. TiKV uses RocksDB as its persistence layer.
@@ -74,3 +74,9 @@ quasardb uses a heavily tuned RocksDB as its persistence layer.

 ## Dgraph
 [Dgraph](https://github.com/dgraph-io/dgraph) is an open-source, scalable, distributed, low latency, high throughput Graph database .They use RocksDB to store state locally on a machine.
+
+## Uber
+[Uber](http://eng.uber.com/cherami/) uses RocksDB as a durable and and scalable task queue.
+
+## 360 Pika
+[360](http://www.360.cn/) [Pika](https://github.com/Qihoo360/pika) is a nosql compatible with redis. With the huge amount of data stored, redis may suffer for a capacity bottleneck, and pika was born for solving it. It has widely been widely used in many company
@@ -1,3 +1,4 @@
+# Vagrant file
 Vagrant.configure("2") do |config|

  config.vm.provider "virtualbox" do |v|
@@ -11,7 +11,8 @@ build:
  verbosity: minimal
 test:
 test_script:
- ps: build_tools\run_ci_db_test.ps1 -EnableRerun -Run db_test2 -Concurrency 8
- ps: build_tools\run_ci_db_test.ps1 -EnableRerun -Run db_test -Exclude DBTest.GroupCommitTest -Concurrency 10
- ps: build_tools\run_ci_db_test.ps1 -Run env_test -Concurrency 1
+- ps: build_tools\run_ci_db_test.ps1 -Run db_basic_test -Concurrency 8
+- ps: build_tools\run_ci_db_test.ps1 -Run db_test2 -Concurrency 8
+- ps: build_tools\run_ci_db_test.ps1 -Run db_test -Concurrency 8
+- ps: build_tools\run_ci_db_test.ps1 -Run env_test,env_basic_test -Concurrency 2

@@ -9,6 +9,13 @@ require('RocksDBCommonHelper.php');
 define("DIFF_COMMAND", "diff");

 class FacebookArcanistConfiguration extends ArcanistConfiguration {
+  public function getCustomArgumentsForCommand($command) {
+    if ($command == "land") {
+      return array(
+          'async' => array('help' => 'Just to make tools happy'));
+    }
+    return array();
+  }

  public function didRunWorkflow($command,
                                 ArcanistWorkflow $workflow,
@@ -9,6 +9,13 @@ require('RocksDBCommonHelper.php');
 define("DIFF_COMMAND", "diff");

 class FacebookArcanistConfiguration extends ArcanistConfiguration {
+  public function getCustomArgumentsForCommand($command) {
+    if ($command == "land") {
+      return array(
+          'async' => array('help' => 'Just to make tools happy'));
+    }
+    return array();
+  }

  public function didRunWorkflow($command,
                                 ArcanistBaseWorkflow $workflow,
@@ -12,7 +12,6 @@ define("ENV_HTTPS_APP_VALUE", "HTTPS_APP_VALUE");
 define("ENV_HTTPS_TOKEN_VALUE", "HTTPS_TOKEN_VALUE");

 define("PRIMARY_TOKEN_FILE", '/home/krad/.sandcastle');
-define("SECONDARY_TOKEN_FILE", '$HOME/.sandcastle');
 define("CONT_RUN_ALIAS", "leveldb");

 //////////////////////////////////////////////////////////////////////
@@ -22,12 +21,10 @@ function postURL($diffID, $url) {
  assert(is_numeric($diffID));
  assert(strlen($url) > 0);

-  $cmd = 'echo \'{"diff_id": "' . $diffID . '", '
+  $cmd = 'echo \'{"diff_id": ' . $diffID . ', '
         . '"name":"click here for sandcastle tests for D' . $diffID . '", '
         . '"link":"' . $url . '"}\' | '
-         . 'no_proxy=facebook.com,tfbnw.net,fb.com '
-         . 'http_proxy=fwdproxy.any.facebook.com:8080 '
-         . 'https_proxy=fwdproxy.any.facebook.com:8080 arc call-conduit '
+         . 'arc call-conduit '
         . 'differential.updateunitresults';
  shell_exec($cmd);
 }
@@ -38,12 +35,10 @@ function buildUpdateTestStatusCmd($diffID, $test, $status) {
  assert(strlen($test) > 0);
  assert(strlen($status) > 0);

-  $cmd = 'echo \'{"diff_id": "' . $diffID . '", '
+  $cmd = 'echo \'{"diff_id": ' . $diffID . ', '
         . '"name":"' . $test . '", '
         . '"result":"' . $status . '"}\' | '
-         . 'no_proxy=facebook.com,tfbnw.net,fb.com '
-         . 'http_proxy=fwdproxy.any.facebook.com:8080 '
-         . 'https_proxy=fwdproxy.any.facebook.com:8080 arc call-conduit '
+         . 'arc call-conduit '
         . 'differential.updateunitresults';
  return $cmd;
 }
@@ -108,8 +103,7 @@ function getSteps($applyDiff, $diffID, $username, $test) {
    // Patch the code (keep your fingures crossed).
    $patch = array(
      "name" => "Patch " . $diffID,
-      "shell" => "no_proxy=facebook.com,tfbnw.net,fb.com "
-                  ."HTTPS_PROXY=fwdproxy:8080 arc --arcrc-file ~/.arcrc "
+      "shell" => "arc --arcrc-file ~/.arcrc "
                  . "patch --nocommit --diff " . $diffID,
      "user" => "root"
    );
@@ -178,6 +172,8 @@ function getSteps($applyDiff, $diffID, $username, $test) {
 function getSandcastleConfig() {
  $sandcastle_config = array();

+  $cwd = getcwd();
+  $cwd_token_file = "{$cwd}/.sandcastle";
  // This is a case when we're executed from a continuous run. Fetch the values
  // from the environment.
  if (getenv(ENV_POST_RECEIVE_HOOK)) {
@@ -188,20 +184,20 @@ function getSandcastleConfig() {
    // configuration files.
    for ($i = 0; $i < 50; $i++) {
      if (file_exists(PRIMARY_TOKEN_FILE) ||
-          file_exists(SECONDARY_TOKEN_FILE)) {
+          file_exists($cwd_token_file)) {
        break;
      }
      // If we failed to fetch the tokens, sleep for 0.2 second and try again
      usleep(200000);
    }
    assert(file_exists(PRIMARY_TOKEN_FILE) ||
-           file_exists(SECONDARY_TOKEN_FILE));
+           file_exists($cwd_token_file));

    // Try the primary location first, followed by a secondary.
    if (file_exists(PRIMARY_TOKEN_FILE)) {
      $cmd = 'cat ' . PRIMARY_TOKEN_FILE;
    } else {
-      $cmd = 'cat ' . SECONDARY_TOKEN_FILE;
+      $cmd = 'cat ' . $cwd_token_file;
    }

    assert(strlen($cmd) > 0);
@@ -303,11 +299,21 @@ function getSandcastleConfig() {
  );

  // Submit to Sandcastle.
-  $url = 'https://interngraph.intern.facebook.com/sandcastle/generate?'
-          .'command=SandcastleUniversalCommand'
-          .'&vcs=rocksdb-git&revision=origin%2Fmaster&type=lego'
-          .'&user=' . $username . '&alias=rocksdb-precommit'
-          .'&command-args=' . urlencode(json_encode($command));
+  $url = 'https://interngraph.intern.facebook.com/sandcastle/create';
+
+  $job = array(
+    'command' => 'SandcastleUniversalCommand',
+    'args' => $command,
+    'capabilities' => array(
+      'vcs' => 'rocksdb-int-git',
+      'type' => 'lego',
+    ),
+    'hash' => 'origin/master',
+    'user' => $username,
+    'alias' => 'rocksdb-precommit',
+    'tags' => array('rocksdb'),
+    'description' => 'Rocksdb precommit job',
+  );

  // Fetch the configuration necessary to submit a successful HTTPS request.
  $sandcastle_config = getSandcastleConfig();
@@ -315,8 +321,9 @@ function getSandcastleConfig() {
  $app = $sandcastle_config[0];
  $token = $sandcastle_config[1];

-  $cmd = 'https_proxy= HTTPS_PROXY= curl -s -k -F app=' . $app . ' '
-          . '-F token=' . $token . ' "' . $url . '"';
+  $cmd = 'curl -s -k -F app=' . $app . ' '
+          . '-F token=' . $token . ' -F job=\'' . json_encode($job)
+          .'\' "' . $url . '"';

  $output = shell_exec($cmd);
  assert(strlen($output) > 0);
@@ -466,3 +466,4 @@ echo "PROFILING_FLAGS=$PROFILING_FLAGS" >> "$OUTPUT"
 if test -n "$JEMALLOC"; then
  echo "JEMALLOC=1" >> "$OUTPUT"
 fi
+echo "LUA_PATH=$LUA_PATH" >> "$OUTPUT"
@@ -8,7 +8,7 @@
 # for a certain interval.
 #

-SRC_GIT_REPO=/data/git/rocksdb
+SRC_GIT_REPO=/data/git/rocksdb-public
 error=0

 function log {
@@ -15,3 +15,4 @@ TBB_BASE=/mnt/gvfs/third-party2/tbb/9d9a554877d0c5bef330fe818ab7178806dd316a/4.0
 KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/7c111ff27e0c466235163f00f280a9d617c3d2ec/4.0.9-36_fbk5_2933_gd092e3f/gcc-5-glibc-2.23/da39a3e
 BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/b7fd454c4b10c6a81015d4524ed06cdeab558490/2.26/centos6-native/da39a3e
 VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/d7f4d4d86674a57668e3a96f76f0e17dd0eb8765/3.10.0/gcc-4.9-glibc-2.20/e9936bf
+LUA_BASE=/mnt/gvfs/third-party2/lua/61e4abf5813bbc39bc4f548757ccfcadde175a48/5.2.3/gcc-4.9-glibc-2.20/690f0d7
@@ -15,3 +15,4 @@ TBB_BASE=/mnt/gvfs/third-party2/tbb/9d9a554877d0c5bef330fe818ab7178806dd316a/4.0
 KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/7c111ff27e0c466235163f00f280a9d617c3d2ec/4.0.9-36_fbk5_2933_gd092e3f/gcc-4.8.1-glibc-2.17/da39a3e
 BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/b7fd454c4b10c6a81015d4524ed06cdeab558490/2.26/centos6-native/da39a3e
 VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/d7f4d4d86674a57668e3a96f76f0e17dd0eb8765/3.8.1/gcc-4.8.1-glibc-2.17/c3f970a
+LUA_BASE=/mnt/gvfs/third-party2/lua/61e4abf5813bbc39bc4f548757ccfcadde175a48/5.2.3/centos6-native/730f94e
@@ -144,4 +144,12 @@ EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GF

 VALGRIND_VER="$VALGRIND_BASE/bin/"

-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_BASE"
+
+if test -z $PIC_BUILD; then
+  LUA_LIB=" $LUA_PATH/lib/liblua.a"
+else
+  LUA_LIB=" $LUA_PATH/lib/liblua_pic.a"
+fi
+
+export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD LUA_PATH LUA_LIB
@@ -110,4 +110,6 @@ EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP2_LIBS $LZ4_LIBS $ZSTD_LIBS $G

 VALGRIND_VER="$VALGRIND_BASE/bin/"

-export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE
+LUA_PATH="$LUA_BASE"
+
+export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE LUA_PATH
@@ -11,7 +11,8 @@ if ! which $CLANG_FORMAT_DIFF &> /dev/null
 then
  echo "You didn't have clang-format-diff.py available in your computer!"
  echo "You can download it by running: "
-  echo "    curl http://goo.gl/iUW1u2"
+  echo "    curl --location http://goo.gl/iUW1u2 -o ${CLANG_FORMAT_DIFF}"
+  echo "and move ${CLANG_FORMAT_DIFF} to some directory within PATH=${PATH}"
  exit 128
 fi

@@ -50,14 +51,15 @@ fi
 set -e

 uncommitted_code=`git diff HEAD`
+LAST_MASTER=`git merge-base master HEAD`

 # If there's no uncommitted changes, we assume user are doing post-commit
-# format check, in which case we'll check the modified lines from latest commit.
-# Otherwise, we'll check format of the uncommitted code only.
+# format check, in which case we'll check the modified lines since last commit
+# from master. Otherwise, we'll check format of the uncommitted code only.
 if [ -z "$uncommitted_code" ]
 then
  # Check the format of last commit
-  diffs=$(git diff -U0 HEAD^ | $CLANG_FORMAT_DIFF -p 1)
+  diffs=$(git diff -U0 $LAST_MASTER^ | $CLANG_FORMAT_DIFF -p 1)
 else
  # Check the format of uncommitted lines,
  diffs=$(git diff -U0 HEAD | $CLANG_FORMAT_DIFF -p 1)
@@ -97,7 +99,12 @@ then
 fi

 # Do in-place format adjustment.
-git diff -U0 HEAD^ | $CLANG_FORMAT_DIFF -i -p 1
+if [ -z "$uncommitted_code" ]
+then
+  git diff -U0 $LAST_MASTER^ | $CLANG_FORMAT_DIFF -i -p 1
+else
+  git diff -U0 HEAD^ | $CLANG_FORMAT_DIFF -i -p 1
+fi
 echo "Files reformatted!"

 # Amend to last commit if user do the post-commit format check
@@ -65,8 +65,8 @@ function main() {

  if [[ -d /vagrant ]]; then
    if [[ $OS = "ubuntu" ]]; then
-      package g++-4.7
-      export CXX=g++-4.7
+      package g++-4.8
+      export CXX=g++-4.8

      # the deb would depend on libgflags2, but the static lib is the only thing
      # installed by make install
@@ -1,4 +1,4 @@
-#!/usr/local/fbcode/gcc-4.8.1-glibc-2.17-fb/bin/python2.7
+#!/usr/local/fbcode/gcc-4.9-glibc-2.20-fb/bin/python2.7

 from __future__ import absolute_import
 from __future__ import division
@@ -55,7 +55,6 @@ make release
    --open_files=55000 \
    --statistics=1 \
    --histogram=1 \
-    --disable_data_sync=1 \
    --disable_wal=1 \
    --sync=0  > ${STAT_FILE}.fillseq

@@ -73,7 +72,6 @@ make release
    --open_files=55000 \
    --statistics=1 \
    --histogram=1 \
-    --disable_data_sync=1 \
    --disable_wal=1 \
    --sync=0 \
    --threads=8 > ${STAT_FILE}.overwrite
@@ -92,7 +90,6 @@ make release
    --open_files=55000 \
    --statistics=1 \
    --histogram=1 \
-    --disable_data_sync=1 \
    --disable_wal=1 \
    --sync=0 \
    --threads=1 > /dev/null
@@ -111,7 +108,6 @@ make release
    --open_files=55000 \
    --statistics=1 \
    --histogram=1 \
-    --disable_data_sync=1 \
    --disable_wal=1 \
    --sync=0 \
    --threads=16 > ${STAT_FILE}.readrandom
@@ -131,7 +127,6 @@ make release
    --use_tailing_iterator=1 \
    --statistics=1 \
    --histogram=1 \
-    --disable_data_sync=1 \
    --disable_wal=1 \
    --sync=0 \
    --threads=16 > ${STAT_FILE}.readrandomtailing
@@ -150,7 +145,6 @@ make release
    --open_files=55000 \
    --statistics=1 \
    --histogram=1 \
-    --disable_data_sync=1 \
    --disable_wal=1 \
    --sync=0 \
    --threads=16 > ${STAT_FILE}.readrandomsmallblockcache
@@ -171,7 +165,6 @@ make release
    --open_files=55000 \
    --statistics=1 \
    --histogram=1 \
-    --disable_data_sync=1 \
    --disable_wal=1 \
    --sync=0 \
    --threads=16 > ${STAT_FILE}.readrandom_mem_sst
@@ -191,7 +184,6 @@ make release
    --open_files=55000 \
    --statistics=1 \
    --histogram=1 \
-    --disable_data_sync=1 \
    --disable_wal=1 \
    --sync=0 \
    --threads=1 > /dev/null
@@ -210,7 +202,6 @@ make release
    --open_files=55000 \
    --statistics=1 \
    --histogram=1 \
-    --disable_data_sync=1 \
    --disable_wal=1 \
    --sync=0 \
    --threads=16 > /dev/null
@@ -230,7 +221,6 @@ make release
    --disable_auto_compactions=1 \
    --statistics=1 \
    --histogram=1 \
-    --disable_data_sync=1 \
    --disable_wal=1 \
    --sync=0 \
    --threads=16 > ${STAT_FILE}.readrandom_filluniquerandom
@@ -251,7 +241,6 @@ make release
    --open_files=55000 \
    --statistics=1 \
    --histogram=1 \
-    --disable_data_sync=1 \
    --disable_wal=1 \
    --sync=0 \
    --threads=16 > ${STAT_FILE}.readwhilewriting
@@ -270,7 +259,6 @@ make release
    --open_files=55000 \
    --statistics=1 \
    --histogram=1 \
-    --disable_data_sync=1 \
    --disable_wal=1 \
    --sync=0 \
    --value_size=10 \
@@ -295,7 +283,6 @@ common_in_mem_args="--db=/dev/shm/rocksdb \
    --disable_wal=0 \
    --wal_dir=/dev/shm/rocksdb \
    --sync=0 \
-    --disable_data_sync=1 \
    --verify_checksum=1 \
    --delete_obsolete_files_period_micros=314572800 \
    --max_grandparent_overlap_factor=10 \
@@ -356,7 +343,6 @@ common_in_mem_args="--db=/dev/shm/rocksdb \
    --open_files=55000 \
    --statistics=1 \
    --histogram=1 \
-    --disable_data_sync=1 \
    --disable_wal=1 \
    --sync=0  > ${STAT_FILE}.fillseq_lots_column_families

@@ -372,7 +358,6 @@ common_in_mem_args="--db=/dev/shm/rocksdb \
    --open_files=55000 \
    --statistics=1 \
    --histogram=1 \
-    --disable_data_sync=1 \
    --disable_wal=1 \
    --sync=0 \
    --threads=8 > ${STAT_FILE}.overwrite_lots_column_families
@@ -89,7 +89,7 @@ TSAN="COMPILE_WITH_TSAN=1"
 UBSAN="COMPILE_WITH_UBSAN=1"
 DISABLE_JEMALLOC="DISABLE_JEMALLOC=1"
 HTTP_PROXY="https_proxy=http://fwdproxy.29.prn1:8080 http_proxy=http://fwdproxy.29.prn1:8080 ftp_proxy=http://fwdproxy.29.prn1:8080"
-SETUP_JAVA_ENV="export $HTTP_PROXY; export JAVA_HOME=/usr/local/jdk-7u10-64/; export PATH=\$PATH:\$JAVA_HOME/bin"
+SETUP_JAVA_ENV="export $HTTP_PROXY; export JAVA_HOME=/usr/local/jdk-8u60-64/; export PATH=\$JAVA_HOME/bin:\$PATH"
 PARSER="'parser':'python build_tools/error_filter.py $1'"

 CONTRUN_NAME="ROCKSDB_CONTRUN_NAME"
@@ -100,7 +100,7 @@ CONTRUN_NAME="ROCKSDB_CONTRUN_NAME"
 # the value of $ONCALL. If it's a diff then just call `false` to make sure
 # that errors will be properly propagated to the caller.
 if [ ! -z $ONCALL ]; then
-  TASK_CREATION_TOOL="/usr/local/bin/mysql_mtr_filter --rocksdb"
+  TASK_CREATION_TOOL="/usr/local/bin/mysql_mtr_filter --rocksdb --oncall $ONCALL"
 else
  TASK_CREATION_TOOL="false"
 fi
@@ -548,7 +548,7 @@ VALGRIND_TEST_COMMANDS="[
            {
                'name':'Run RocksDB debug unit tests',
                'timeout': 86400,
-                'shell':'$DISABLE_JEMALLOC $SHM $DEBUG make $PARALLELISM valgrind_check || $CONTRUN_NAME=valgrind_check $TASK_CREATION_TOOL',
+                'shell':'$SHM $DEBUG make $PARALLELISM valgrind_test || $CONTRUN_NAME=valgrind_check $TASK_CREATION_TOOL',
                'user':'root',
                $PARSER
            },
@@ -50,7 +50,8 @@ Write-Output "Root: $RootFolder, WorkFolder: $WorkFolder"

 # Use JEMALLOC executables
 if($Run -ceq "db_test" -or
-   $Run -ceq "db_test2" ) {
+   $Run -ceq "db_test2" -or
+   $Run -ceq "db_basic_test") {

   $file_name = $Run

@@ -150,7 +151,7 @@ function MakeAndAdd([string]$token, $HashTable) {
 # Test executable name -> Log file
 function Discover-TestBinaries([string]$Pattern, $HashTable) {

-    $Exclusions = @("db_test*", "db_sanity_test*")
+    $Exclusions = @("db_test*", "db_sanity_test*", "db_basic_test*")

    $p = -join ($BinariesFolder, $pattern)

@@ -164,7 +165,8 @@ function Discover-TestBinaries([string]$Pattern, $HashTable) {
 $TestsToRun = [ordered]@{}

 if($Run -ceq "db_test" -or
-   $Run -ceq "db_test2") {
+   $Run -ceq "db_test2" -or
+   $Run -ceq "db_basic_test") {
    Normalize-DbTests -HashTable $TestsToRun
 } elseif($Run -ceq "tests") {
    if($EnableJE) {
@@ -237,7 +239,8 @@ function RunJobs($TestToLog, [int]$ConcurrencyVal, [bool]$AddForRerun)
            $log_path = ($TestToLog.$k)

            if($Run -ceq "db_test" -or
-               $Run -ceq "db_test2") {
+               $Run -ceq "db_test2" -or
+               $Run -ceq "db_basic_test") {
              $job = Start-Job -Name $k -ScriptBlock $InvokeTestCase -ArgumentList @($db_test,$k,$log_path)
            } else {
              [string]$Exe =  -Join ($BinariesFolder, $k)
@@ -87,6 +87,7 @@ get_lib_base tbb        4.0_update2 gcc-4.9-glibc-2.20
 get_lib_base kernel-headers LATEST 
 get_lib_base binutils   LATEST centos6-native 
 get_lib_base valgrind   3.10.0 gcc-4.9-glibc-2.20
+get_lib_base lua        5.2.3 gcc-4.9-glibc-2.20

 git diff $OUTPUT

@@ -125,5 +126,6 @@ get_lib_base tbb        4.0_update2 gcc-4.8.1-glibc-2.17
 get_lib_base kernel-headers LATEST gcc-4.8.1-glibc-2.17 
 get_lib_base binutils   LATEST centos6-native 
 get_lib_base valgrind   3.8.1  gcc-4.8.1-glibc-2.17
+get_lib_base lua        5.2.3 centos6-native

 git diff $OUTPUT
@@ -1,8 +1,9 @@
 #!/bin/sh
 if [ "$#" = "0" ]; then
-  echo "Usage: $0 major|minor|patch"
+  echo "Usage: $0 major|minor|patch|full"
  exit 1
 fi
+
 if [ "$1" = "major" ]; then
  cat include/rocksdb/version.h  | grep MAJOR | head -n1 | awk '{print $3}'
 fi
@@ -12,3 +13,10 @@ fi
 if [ "$1" = "patch" ]; then
  cat include/rocksdb/version.h  | grep PATCH | head -n1 | awk '{print $3}'
 fi
+if [ "$1" = "full" ]; then
+  awk '/#define ROCKSDB/ { env[$2] = $3 }
+       END { printf "%s.%s.%s\n", env["ROCKSDB_MAJOR"],
+                                  env["ROCKSDB_MINOR"],
+                                  env["ROCKSDB_PATCH"] }'  \
+      include/rocksdb/version.h
+fi
@@ -8,6 +8,7 @@

 namespace rocksdb {

+#ifndef ROCKSDB_LITE
 // -- AutoRollLogger
 Status AutoRollLogger::ResetLogger() {
  TEST_SYNC_POINT("AutoRollLogger::ResetLogger:BeforeNewLogger");
@@ -130,6 +131,7 @@ bool AutoRollLogger::LogExpired() {
  ++cached_now_access_count;
  return cached_now >= ctime_ + kLogFileTimeToRoll;
 }
+#endif  // !ROCKSDB_LITE

 Status CreateLoggerFromOptions(const std::string& dbname,
                               const DBOptions& options,
@@ -147,6 +149,7 @@ Status CreateLoggerFromOptions(const std::string& dbname,

  env->CreateDirIfMissing(dbname);  // In case it does not exist
  // Currently we only support roll by time-to-roll and log size
+#ifndef ROCKSDB_LITE
  if (options.log_file_time_to_roll > 0 || options.max_log_file_size > 0) {
    AutoRollLogger* result = new AutoRollLogger(
        env, dbname, options.db_log_dir, options.max_log_file_size,
@@ -158,17 +161,17 @@ Status CreateLoggerFromOptions(const std::string& dbname,
      logger->reset(result);
    }
    return s;
-  } else {
-    // Open a log file in the same directory as the db
-    env->RenameFile(
-        fname, OldInfoLogFileName(dbname, env->NowMicros(), db_absolute_path,
-                                  options.db_log_dir));
-    auto s = env->NewLogger(fname, logger);
-    if (logger->get() != nullptr) {
-      (*logger)->SetInfoLogLevel(options.info_log_level);
-    }
-    return s;
  }
+#endif  // !ROCKSDB_LITE
+  // Open a log file in the same directory as the db
+  env->RenameFile(fname,
+                  OldInfoLogFileName(dbname, env->NowMicros(), db_absolute_path,
+                                     options.db_log_dir));
+  auto s = env->NewLogger(fname, logger);
+  if (logger->get() != nullptr) {
+    (*logger)->SetInfoLogLevel(options.info_log_level);
+  }
+  return s;
 }

 }  // namespace rocksdb
@@ -18,6 +18,7 @@

 namespace rocksdb {

+#ifndef ROCKSDB_LITE
 // Rolls the log file by size and/or time
 class AutoRollLogger : public Logger {
 public:
@@ -120,6 +121,7 @@ class AutoRollLogger : public Logger {
  uint64_t call_NowMicros_every_N_records_;
  mutable port::Mutex mutex_;
 };
+#endif  // !ROCKSDB_LITE

 // Facade to craete logger automatically
 Status CreateLoggerFromOptions(const std::string& dbname,
@@ -3,6 +3,9 @@
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
 //
+
+#ifndef ROCKSDB_LITE
+
 #include <string>
 #include <thread>
 #include <vector>
@@ -271,7 +274,7 @@ TEST_F(AutoRollLoggerTest, LogFlushWhileRolling) {
  AutoRollLogger* auto_roll_logger =
      dynamic_cast<AutoRollLogger*>(logger.get());
  ASSERT_TRUE(auto_roll_logger);
-  std::thread flush_thread;
+  rocksdb::port::Thread flush_thread;

  // Notes:
  // (1) Need to pin the old logger before beginning the roll, as rolling grabs
@@ -293,7 +296,7 @@ TEST_F(AutoRollLoggerTest, LogFlushWhileRolling) {
       {"AutoRollLogger::Flush:PinnedLogger", "PosixLogger::Flush:Begin2"}});
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  flush_thread = std::thread([&]() { auto_roll_logger->Flush(); });
+  flush_thread = port::Thread ([&]() { auto_roll_logger->Flush(); });
  TEST_SYNC_POINT(
      "AutoRollLoggerTest::LogFlushWhileRolling:PreRollAndPostThreadInit");
  RollLogFileBySizeTest(auto_roll_logger, options.max_log_file_size,
@@ -469,3 +472,14 @@ int main(int argc, char** argv) {
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
 }
+
+#else
+#include <stdio.h>
+
+int main(int argc, char** argv) {
+  fprintf(stderr,
+          "SKIPPED as AutoRollLogger is not supported in ROCKSDB_LITE\n");
+  return 0;
+}
+
+#endif  // !ROCKSDB_LITE
@@ -61,7 +61,8 @@ TableBuilder* NewTableBuilder(
 Status BuildTable(
    const std::string& dbname, Env* env, const ImmutableCFOptions& ioptions,
    const MutableCFOptions& mutable_cf_options, const EnvOptions& env_options,
-    TableCache* table_cache, InternalIterator* iter, FileMetaData* meta,
+    TableCache* table_cache, InternalIterator* iter,
+    std::unique_ptr<InternalIterator> range_del_iter, FileMetaData* meta,
    const InternalKeyComparator& internal_comparator,
    const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
        int_tbl_prop_collector_factories,
@@ -81,6 +82,13 @@ Status BuildTable(
  Status s;
  meta->fd.file_size = 0;
  iter->SeekToFirst();
+  std::unique_ptr<RangeDelAggregator> range_del_agg(
+      new RangeDelAggregator(internal_comparator, snapshots));
+  s = range_del_agg->AddTombstones(std::move(range_del_iter));
+  if (!s.ok()) {
+    // may be non-ok if a range tombstone key is unparsable
+    return s;
+  }

  std::string fname = TableFileName(ioptions.db_paths, meta->fd.GetNumber(),
                                    meta->fd.GetPathId());
@@ -90,7 +98,7 @@ Status BuildTable(
 #endif  // !ROCKSDB_LITE
  TableProperties tp;

-  if (iter->Valid()) {
+  if (iter->Valid() || range_del_agg->ShouldAddTombstones()) {
    TableBuilder* builder;
    unique_ptr<WritableFileWriter> file_writer;
    {
@@ -104,7 +112,8 @@ Status BuildTable(
      }
      file->SetIOPriority(io_priority);

-      file_writer.reset(new WritableFileWriter(std::move(file), env_options));
+      file_writer.reset(new WritableFileWriter(std::move(file), env_options,
+                                               ioptions.statistics));

      builder = NewTableBuilder(
          ioptions, internal_comparator, int_tbl_prop_collector_factories,
@@ -112,11 +121,8 @@ Status BuildTable(
          compression_opts, level);
    }

-    std::unique_ptr<RangeDelAggregator> range_del_agg;
-    range_del_agg.reset(new RangeDelAggregator(internal_comparator, snapshots));
    MergeHelper merge(env, internal_comparator.user_comparator(),
                      ioptions.merge_operator, nullptr, ioptions.info_log,
-                      mutable_cf_options.min_partial_merge_operands,
                      true /* internal key corruption is not ok */,
                      snapshots.empty() ? 0 : snapshots.back());

@@ -138,6 +144,9 @@ Status BuildTable(
            ThreadStatus::FLUSH_BYTES_WRITTEN, IOSTATS(bytes_written));
      }
    }
+    // nullptr for table_{min,max} so all range tombstones will be flushed
+    range_del_agg->AddToBuilder(builder, nullptr /* lower_bound */,
+                                nullptr /* upper_bound */, meta);

    // Finish and check for builder errors
    bool empty = builder->NumEntries() == 0;
@@ -161,7 +170,7 @@ Status BuildTable(
    delete builder;

    // Finish and check for file errors
-    if (s.ok() && !empty && !ioptions.disable_data_sync) {
+    if (s.ok() && !empty) {
      StopWatch sw(env, ioptions.statistics, TABLE_SYNC_MICROS);
      file_writer->Sync(ioptions.use_fsync);
    }
@@ -172,7 +181,8 @@ Status BuildTable(
    if (s.ok() && !empty) {
      // Verify that the table is usable
      std::unique_ptr<InternalIterator> it(table_cache->NewIterator(
-          ReadOptions(), env_options, internal_comparator, meta->fd, nullptr,
+          ReadOptions(), env_options, internal_comparator, meta->fd,
+          nullptr /* range_del_agg */, nullptr,
          (internal_stats == nullptr) ? nullptr
                                      : internal_stats->GetFileReadHist(0),
          false /* for_compaction */, nullptr /* arena */,
@@ -17,6 +17,7 @@
 #include "rocksdb/status.h"
 #include "rocksdb/table_properties.h"
 #include "rocksdb/types.h"
+#include "table/scoped_arena_iterator.h"
 #include "util/cf_options.h"
 #include "util/event_logger.h"

@@ -63,7 +64,8 @@ TableBuilder* NewTableBuilder(
 extern Status BuildTable(
    const std::string& dbname, Env* env, const ImmutableCFOptions& options,
    const MutableCFOptions& mutable_cf_options, const EnvOptions& env_options,
-    TableCache* table_cache, InternalIterator* iter, FileMetaData* meta,
+    TableCache* table_cache, InternalIterator* iter,
+    std::unique_ptr<InternalIterator> range_del_iter, FileMetaData* meta,
    const InternalKeyComparator& internal_comparator,
    const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
        int_tbl_prop_collector_factories,
@@ -30,6 +30,7 @@
 #include "rocksdb/statistics.h"
 #include "rocksdb/slice_transform.h"
 #include "rocksdb/table.h"
+#include "rocksdb/rate_limiter.h"
 #include "rocksdb/utilities/backupable_db.h"
 #include "utilities/merge_operators.h"

@@ -47,10 +48,12 @@ using rocksdb::WALRecoveryMode;
 using rocksdb::DB;
 using rocksdb::DBOptions;
 using rocksdb::Env;
+using rocksdb::EnvOptions;
 using rocksdb::InfoLogLevel;
 using rocksdb::FileLock;
 using rocksdb::FilterPolicy;
 using rocksdb::FlushOptions;
+using rocksdb::IngestExternalFileOptions;
 using rocksdb::Iterator;
 using rocksdb::Logger;
 using rocksdb::MergeOperator;
@@ -68,6 +71,7 @@ using rocksdb::Slice;
 using rocksdb::SliceParts;
 using rocksdb::SliceTransform;
 using rocksdb::Snapshot;
+using rocksdb::SstFileWriter;
 using rocksdb::Status;
 using rocksdb::WritableFile;
 using rocksdb::WriteBatch;
@@ -78,6 +82,8 @@ using rocksdb::BackupableDBOptions;
 using rocksdb::BackupInfo;
 using rocksdb::RestoreOptions;
 using rocksdb::CompactRangeOptions;
+using rocksdb::RateLimiter;
+using rocksdb::NewGenericRateLimiter;

 using std::shared_ptr;

@@ -98,6 +104,9 @@ struct rocksdb_readoptions_t {
 };
 struct rocksdb_writeoptions_t    { WriteOptions      rep; };
 struct rocksdb_options_t         { Options           rep; };
+struct rocksdb_compactoptions_t {
+  CompactRangeOptions rep;
+};
 struct rocksdb_block_based_table_options_t  { BlockBasedTableOptions rep; };
 struct rocksdb_cuckoo_table_options_t  { CuckooTableOptions rep; };
 struct rocksdb_seqfile_t         { SequentialFile*   rep; };
@@ -108,6 +117,10 @@ struct rocksdb_logger_t          { shared_ptr<Logger>  rep; };
 struct rocksdb_cache_t           { shared_ptr<Cache>   rep; };
 struct rocksdb_livefiles_t       { std::vector<LiveFileMetaData> rep; };
 struct rocksdb_column_family_handle_t  { ColumnFamilyHandle* rep; };
+struct rocksdb_envoptions_t      { EnvOptions        rep; };
+struct rocksdb_ingestexternalfileoptions_t  { IngestExternalFileOptions rep; };
+struct rocksdb_sstfilewriter_t   { SstFileWriter*    rep; };
+struct rocksdb_ratelimiter_t     { RateLimiter*      rep; };

 struct rocksdb_compactionfiltercontext_t {
  CompactionFilter::Context rep;
@@ -435,7 +448,11 @@ rocksdb_backup_engine_t* rocksdb_backup_engine_open(
    const rocksdb_options_t* options, const char* path, char** errptr) {
  BackupEngine* be;
  if (SaveError(errptr, BackupEngine::Open(options->rep.env,
-                                           BackupableDBOptions(path), &be))) {
+                                           BackupableDBOptions(path,
+                                                               nullptr,
+                                                               true,
+                                                               options->rep.info_log.get()),
+                                           &be))) {
    return nullptr;
  }
  rocksdb_backup_engine_t* result = new rocksdb_backup_engine_t;
@@ -884,6 +901,17 @@ char* rocksdb_property_value(
  }
 }

+int rocksdb_property_int(
+    rocksdb_t* db,
+    const char* propname,
+    uint64_t *out_val) {
+  if (db->rep->GetIntProperty(Slice(propname), out_val)) {
+    return 0;
+  } else {
+    return -1;
+  }
+}
+
 char* rocksdb_property_value_cf(
    rocksdb_t* db,
    rocksdb_column_family_handle_t* column_family,
@@ -966,6 +994,30 @@ void rocksdb_compact_range_cf(
      (limit_key ? (b = Slice(limit_key, limit_key_len), &b) : nullptr));
 }

+void rocksdb_compact_range_opt(rocksdb_t* db, rocksdb_compactoptions_t* opt,
+                               const char* start_key, size_t start_key_len,
+                               const char* limit_key, size_t limit_key_len) {
+  Slice a, b;
+  db->rep->CompactRange(
+      opt->rep,
+      // Pass nullptr Slice if corresponding "const char*" is nullptr
+      (start_key ? (a = Slice(start_key, start_key_len), &a) : nullptr),
+      (limit_key ? (b = Slice(limit_key, limit_key_len), &b) : nullptr));
+}
+
+void rocksdb_compact_range_cf_opt(rocksdb_t* db,
+                                  rocksdb_column_family_handle_t* column_family,
+                                  rocksdb_compactoptions_t* opt,
+                                  const char* start_key, size_t start_key_len,
+                                  const char* limit_key, size_t limit_key_len) {
+  Slice a, b;
+  db->rep->CompactRange(
+      opt->rep, column_family->rep,
+      // Pass nullptr Slice if corresponding "const char*" is nullptr
+      (start_key ? (a = Slice(start_key, start_key_len), &a) : nullptr),
+      (limit_key ? (b = Slice(limit_key, limit_key_len), &b) : nullptr));
+}
+
 void rocksdb_flush(
    rocksdb_t* db,
    const rocksdb_flushoptions_t* options,
@@ -1021,6 +1073,11 @@ void rocksdb_iter_seek(rocksdb_iterator_t* iter, const char* k, size_t klen) {
  iter->rep->Seek(Slice(k, klen));
 }

+void rocksdb_iter_seek_for_prev(rocksdb_iterator_t* iter, const char* k,
+                                size_t klen) {
+  iter->rep->SeekForPrev(Slice(k, klen));
+}
+
 void rocksdb_iter_next(rocksdb_iterator_t* iter) {
  iter->rep->Next();
 }
@@ -1208,6 +1265,53 @@ void rocksdb_writebatch_deletev_cf(
  b->rep.Delete(column_family->rep, SliceParts(key_slices.data(), num_keys));
 }

+void rocksdb_writebatch_delete_range(rocksdb_writebatch_t* b,
+                                     const char* start_key,
+                                     size_t start_key_len, const char* end_key,
+                                     size_t end_key_len) {
+  b->rep.DeleteRange(Slice(start_key, start_key_len),
+                     Slice(end_key, end_key_len));
+}
+
+void rocksdb_writebatch_delete_range_cf(
+    rocksdb_writebatch_t* b, rocksdb_column_family_handle_t* column_family,
+    const char* start_key, size_t start_key_len, const char* end_key,
+    size_t end_key_len) {
+  b->rep.DeleteRange(column_family->rep, Slice(start_key, start_key_len),
+                     Slice(end_key, end_key_len));
+}
+
+void rocksdb_writebatch_delete_rangev(rocksdb_writebatch_t* b, int num_keys,
+                                      const char* const* start_keys_list,
+                                      const size_t* start_keys_list_sizes,
+                                      const char* const* end_keys_list,
+                                      const size_t* end_keys_list_sizes) {
+  std::vector<Slice> start_key_slices(num_keys);
+  std::vector<Slice> end_key_slices(num_keys);
+  for (int i = 0; i < num_keys; i++) {
+    start_key_slices[i] = Slice(start_keys_list[i], start_keys_list_sizes[i]);
+    end_key_slices[i] = Slice(end_keys_list[i], end_keys_list_sizes[i]);
+  }
+  b->rep.DeleteRange(SliceParts(start_key_slices.data(), num_keys),
+                     SliceParts(end_key_slices.data(), num_keys));
+}
+
+void rocksdb_writebatch_delete_rangev_cf(
+    rocksdb_writebatch_t* b, rocksdb_column_family_handle_t* column_family,
+    int num_keys, const char* const* start_keys_list,
+    const size_t* start_keys_list_sizes, const char* const* end_keys_list,
+    const size_t* end_keys_list_sizes) {
+  std::vector<Slice> start_key_slices(num_keys);
+  std::vector<Slice> end_key_slices(num_keys);
+  for (int i = 0; i < num_keys; i++) {
+    start_key_slices[i] = Slice(start_keys_list[i], start_keys_list_sizes[i]);
+    end_key_slices[i] = Slice(end_keys_list[i], end_keys_list_sizes[i]);
+  }
+  b->rep.DeleteRange(column_family->rep,
+                     SliceParts(start_key_slices.data(), num_keys),
+                     SliceParts(end_key_slices.data(), num_keys));
+}
+
 void rocksdb_writebatch_put_log_data(
    rocksdb_writebatch_t* b,
    const char* blob, size_t len) {
@@ -1243,6 +1347,15 @@ const char* rocksdb_writebatch_data(rocksdb_writebatch_t* b, size_t* size) {
  return b->rep.Data().c_str();
 }

+void rocksdb_writebatch_set_save_point(rocksdb_writebatch_t* b) {
+  b->rep.SetSavePoint();
+}
+
+void rocksdb_writebatch_rollback_to_save_point(rocksdb_writebatch_t* b,
+                                               char** errptr) {
+  SaveError(errptr, b->rep.RollbackToSavePoint());
+}
+
 rocksdb_block_based_table_options_t*
 rocksdb_block_based_options_create() {
  return new rocksdb_block_based_table_options_t;
@@ -1326,11 +1439,6 @@ void rocksdb_block_based_options_set_pin_l0_filter_and_index_blocks_in_cache(
  options->rep.pin_l0_filter_and_index_blocks_in_cache = v;
 }

-void rocksdb_block_based_options_set_skip_table_builder_flush(
-    rocksdb_block_based_table_options_t* options, unsigned char v) {
-  options->rep.skip_table_builder_flush = v;
-}
-
 void rocksdb_options_set_block_based_table_factory(
    rocksdb_options_t *opt,
    rocksdb_block_based_table_options_t* table_options) {
@@ -1385,6 +1493,14 @@ void rocksdb_options_set_cuckoo_table_factory(
  }
 }

+void rocksdb_set_options(
+    rocksdb_t* db, int count, const char* const keys[], const char* const values[], char** errptr) {
+        std::unordered_map<std::string, std::string> options_map;
+        for (int i=0; i<count; i++)
+            options_map[keys[i]] = values[i];
+        SaveError(errptr,
+            db->rep->SetOptions(options_map));
+    }

 rocksdb_options_t* rocksdb_options_create() {
  return new rocksdb_options_t;
@@ -1511,8 +1627,13 @@ void rocksdb_options_set_max_bytes_for_level_base(
  opt->rep.max_bytes_for_level_base = n;
 }

-void rocksdb_options_set_max_bytes_for_level_multiplier(
-    rocksdb_options_t* opt, int n) {
+void rocksdb_options_set_level_compaction_dynamic_level_bytes(
+    rocksdb_options_t* opt, unsigned char v) {
+  opt->rep.level_compaction_dynamic_level_bytes = v;
+}
+
+void rocksdb_options_set_max_bytes_for_level_multiplier(rocksdb_options_t* opt,
+                                                        double n) {
  opt->rep.max_bytes_for_level_multiplier = n;
 }

@@ -1587,11 +1708,6 @@ void rocksdb_options_set_prefix_extractor(
  opt->rep.prefix_extractor.reset(prefix_extractor);
 }

-void rocksdb_options_set_disable_data_sync(
-    rocksdb_options_t* opt, int disable_data_sync) {
-  opt->rep.disableDataSync = disable_data_sync;
-}
-
 void rocksdb_options_set_use_fsync(
    rocksdb_options_t* opt, int use_fsync) {
  opt->rep.use_fsync = use_fsync;
@@ -1625,9 +1741,14 @@ void rocksdb_options_set_manifest_preallocation_size(
 void rocksdb_options_set_purge_redundant_kvs_while_flush(rocksdb_options_t* opt,
                                                         unsigned char v) {}

-void rocksdb_options_set_allow_os_buffer(rocksdb_options_t* opt,
-                                         unsigned char v) {
-  opt->rep.allow_os_buffer = v;
+void rocksdb_options_set_use_direct_reads(rocksdb_options_t* opt,
+                                          unsigned char v) {
+  opt->rep.use_direct_reads = v;
+}
+
+void rocksdb_options_set_use_direct_writes(rocksdb_options_t* opt,
+                                           unsigned char v) {
+  opt->rep.use_direct_writes = v;
 }

 void rocksdb_options_set_allow_mmap_reads(
@@ -1688,9 +1809,14 @@ void rocksdb_options_set_bytes_per_sync(
  opt->rep.bytes_per_sync = v;
 }

-void rocksdb_options_set_verify_checksums_in_compaction(
+void rocksdb_options_set_allow_concurrent_memtable_write(rocksdb_options_t* opt,
+                                                         unsigned char v) {
+  opt->rep.allow_concurrent_memtable_write = v;
+}
+
+void rocksdb_options_set_enable_write_thread_adaptive_yield(
    rocksdb_options_t* opt, unsigned char v) {
-  opt->rep.verify_checksums_in_compaction = v;
+  opt->rep.enable_write_thread_adaptive_yield = v;
 }

 void rocksdb_options_set_max_sequential_skip_in_iterations(
@@ -1715,6 +1841,11 @@ void rocksdb_options_set_max_background_compactions(rocksdb_options_t* opt, int
  opt->rep.max_background_compactions = n;
 }

+void rocksdb_options_set_base_background_compactions(rocksdb_options_t* opt,
+                                                     int n) {
+  opt->rep.base_background_compactions = n;
+}
+
 void rocksdb_options_set_max_background_flushes(rocksdb_options_t* opt, int n) {
  opt->rep.max_background_flushes = n;
 }
@@ -1744,6 +1875,14 @@ void rocksdb_options_set_hard_rate_limit(rocksdb_options_t* opt, double v) {
  opt->rep.hard_rate_limit = v;
 }

+void rocksdb_options_set_soft_pending_compaction_bytes_limit(rocksdb_options_t* opt, size_t v) {
+  opt->rep.soft_pending_compaction_bytes_limit = v;
+}
+
+void rocksdb_options_set_hard_pending_compaction_bytes_limit(rocksdb_options_t* opt, size_t v) {
+  opt->rep.hard_pending_compaction_bytes_limit = v;
+}
+
 void rocksdb_options_set_rate_limit_delay_max_milliseconds(
    rocksdb_options_t* opt, unsigned int v) {
  opt->rep.rate_limit_delay_max_milliseconds = v;
@@ -1773,6 +1912,10 @@ void rocksdb_options_set_disable_auto_compactions(rocksdb_options_t* opt, int di
  opt->rep.disable_auto_compactions = disable;
 }

+void rocksdb_options_set_optimize_filters_for_hits(rocksdb_options_t* opt, int v) {
+  opt->rep.optimize_filters_for_hits = v;
+}
+
 void rocksdb_options_set_delete_obsolete_files_period_micros(
    rocksdb_options_t* opt, uint64_t v) {
  opt->rep.delete_obsolete_files_period_micros = v;
@@ -1827,11 +1970,6 @@ void rocksdb_options_set_max_successive_merges(
  opt->rep.max_successive_merges = v;
 }

-void rocksdb_options_set_min_partial_merge_operands(
-    rocksdb_options_t* opt, uint32_t v) {
-  opt->rep.min_partial_merge_operands = v;
-}
-
 void rocksdb_options_set_bloom_locality(
    rocksdb_options_t* opt, uint32_t v) {
  opt->rep.bloom_locality = v;
@@ -1874,6 +2012,28 @@ char *rocksdb_options_statistics_get_string(rocksdb_options_t *opt) {
  return nullptr;
 }

+void rocksdb_options_set_ratelimiter(rocksdb_options_t *opt, rocksdb_ratelimiter_t *limiter) {
+  opt->rep.rate_limiter.reset(limiter->rep);
+  limiter->rep = nullptr;
+}
+
+rocksdb_ratelimiter_t* rocksdb_ratelimiter_create(
+    int64_t rate_bytes_per_sec,
+    int64_t refill_period_us,
+    int32_t fairness) {
+  rocksdb_ratelimiter_t* rate_limiter = new rocksdb_ratelimiter_t;
+  rate_limiter->rep = NewGenericRateLimiter(rate_bytes_per_sec,
+                                            refill_period_us, fairness);
+  return rate_limiter;
+}
+
+void rocksdb_ratelimiter_destroy(rocksdb_ratelimiter_t *limiter) {
+  if (limiter->rep) {
+    delete limiter->rep;
+  }
+  delete limiter;
+}
+
 /*
 TODO:
 DB::OpenForReadOnly
@@ -2111,6 +2271,16 @@ void rocksdb_readoptions_set_readahead_size(
  opt->rep.readahead_size = v;
 }

+void rocksdb_readoptions_set_pin_data(rocksdb_readoptions_t* opt,
+                                      unsigned char v) {
+  opt->rep.pin_data = v;
+}
+
+void rocksdb_readoptions_set_total_order_seek(rocksdb_readoptions_t* opt,
+                                              unsigned char v) {
+  opt->rep.total_order_seek = v;
+}
+
 rocksdb_writeoptions_t* rocksdb_writeoptions_create() {
  return new rocksdb_writeoptions_t;
 }
@@ -2128,6 +2298,28 @@ void rocksdb_writeoptions_disable_WAL(rocksdb_writeoptions_t* opt, int disable)
  opt->rep.disableWAL = disable;
 }

+rocksdb_compactoptions_t* rocksdb_compactoptions_create() {
+  return new rocksdb_compactoptions_t;
+}
+
+void rocksdb_compactoptions_destroy(rocksdb_compactoptions_t* opt) {
+  delete opt;
+}
+
+void rocksdb_compactoptions_set_exclusive_manual_compaction(
+    rocksdb_compactoptions_t* opt, unsigned char v) {
+  opt->rep.exclusive_manual_compaction = v;
+}
+
+void rocksdb_compactoptions_set_change_level(rocksdb_compactoptions_t* opt,
+                                             unsigned char v) {
+  opt->rep.change_level = v;
+}
+
+void rocksdb_compactoptions_set_target_level(rocksdb_compactoptions_t* opt,
+                                             int n) {
+  opt->rep.target_level = n;
+}

 rocksdb_flushoptions_t* rocksdb_flushoptions_create() {
  return new rocksdb_flushoptions_t;
@@ -2156,6 +2348,14 @@ void rocksdb_cache_set_capacity(rocksdb_cache_t* cache, size_t capacity) {
  cache->rep->SetCapacity(capacity);
 }

+size_t rocksdb_cache_get_usage(rocksdb_cache_t* cache) {
+  return cache->rep->GetUsage();
+}
+
+size_t rocksdb_cache_get_pinned_usage(rocksdb_cache_t* cache) {
+  return cache->rep->GetPinnedUsage();
+}
+
 rocksdb_env_t* rocksdb_create_default_env() {
  rocksdb_env_t* result = new rocksdb_env_t;
  result->rep = Env::Default();
@@ -2187,6 +2387,107 @@ void rocksdb_env_destroy(rocksdb_env_t* env) {
  delete env;
 }

+rocksdb_envoptions_t* rocksdb_envoptions_create() {
+  rocksdb_envoptions_t* opt = new rocksdb_envoptions_t;
+  return opt;
+}
+
+void rocksdb_envoptions_destroy(rocksdb_envoptions_t* opt) { delete opt; }
+
+rocksdb_sstfilewriter_t* rocksdb_sstfilewriter_create(
+    const rocksdb_envoptions_t* env, const rocksdb_options_t* io_options) {
+  rocksdb_sstfilewriter_t* writer = new rocksdb_sstfilewriter_t;
+  writer->rep =
+      new SstFileWriter(env->rep, io_options->rep, io_options->rep.comparator);
+  return writer;
+}
+
+rocksdb_sstfilewriter_t* rocksdb_sstfilewriter_create_with_comparator(
+    const rocksdb_envoptions_t* env, const rocksdb_options_t* io_options,
+    const rocksdb_comparator_t* comparator) {
+  rocksdb_sstfilewriter_t* writer = new rocksdb_sstfilewriter_t;
+  writer->rep =
+      new SstFileWriter(env->rep, io_options->rep, io_options->rep.comparator);
+  return writer;
+}
+
+void rocksdb_sstfilewriter_open(rocksdb_sstfilewriter_t* writer,
+                                const char* name, char** errptr) {
+  SaveError(errptr, writer->rep->Open(std::string(name)));
+}
+
+void rocksdb_sstfilewriter_add(rocksdb_sstfilewriter_t* writer, const char* key,
+                               size_t keylen, const char* val, size_t vallen,
+                               char** errptr) {
+  SaveError(errptr, writer->rep->Add(Slice(key, keylen), Slice(val, vallen)));
+}
+
+void rocksdb_sstfilewriter_finish(rocksdb_sstfilewriter_t* writer,
+                                  char** errptr) {
+  SaveError(errptr, writer->rep->Finish(NULL));
+}
+
+void rocksdb_sstfilewriter_destroy(rocksdb_sstfilewriter_t* writer) {
+  delete writer->rep;
+  delete writer;
+}
+
+rocksdb_ingestexternalfileoptions_t*
+rocksdb_ingestexternalfileoptions_create() {
+  rocksdb_ingestexternalfileoptions_t* opt =
+      new rocksdb_ingestexternalfileoptions_t;
+  return opt;
+}
+
+void rocksdb_ingestexternalfileoptions_set_move_files(
+    rocksdb_ingestexternalfileoptions_t* opt, unsigned char move_files) {
+  opt->rep.move_files = move_files;
+}
+
+void rocksdb_ingestexternalfileoptions_set_snapshot_consistency(
+    rocksdb_ingestexternalfileoptions_t* opt,
+    unsigned char snapshot_consistency) {
+  opt->rep.snapshot_consistency = snapshot_consistency;
+}
+
+void rocksdb_ingestexternalfileoptions_set_allow_global_seqno(
+    rocksdb_ingestexternalfileoptions_t* opt,
+    unsigned char allow_global_seqno) {
+  opt->rep.allow_global_seqno = allow_global_seqno;
+}
+
+void rocksdb_ingestexternalfileoptions_set_allow_blocking_flush(
+    rocksdb_ingestexternalfileoptions_t* opt,
+    unsigned char allow_blocking_flush) {
+  opt->rep.allow_blocking_flush = allow_blocking_flush;
+}
+
+void rocksdb_ingestexternalfileoptions_destroy(
+    rocksdb_ingestexternalfileoptions_t* opt) {
+  delete opt;
+}
+
+void rocksdb_ingest_external_file(
+    rocksdb_t* db, const char* const* file_list, const size_t list_len,
+    const rocksdb_ingestexternalfileoptions_t* opt, char** errptr) {
+  std::vector<std::string> files(list_len);
+  for (size_t i = 0; i < list_len; ++i) {
+    files[i] = std::string(file_list[i]);
+  }
+  SaveError(errptr, db->rep->IngestExternalFile(files, opt->rep));
+}
+
+void rocksdb_ingest_external_file_cf(
+    rocksdb_t* db, rocksdb_column_family_handle_t* handle,
+    const char* const* file_list, const size_t list_len,
+    const rocksdb_ingestexternalfileoptions_t* opt, char** errptr) {
+  std::vector<std::string> files(list_len);
+  for (size_t i = 0; i < list_len; ++i) {
+    files[i] = std::string(file_list[i]);
+  }
+  SaveError(errptr, db->rep->IngestExternalFile(handle->rep, files, opt->rep));
+}
+
 rocksdb_slicetransform_t* rocksdb_slicetransform_create(
    void* state,
    void (*destructor)(void*),
@@ -2,17 +2,18 @@
   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 <stdio.h>
+
 #ifndef ROCKSDB_LITE  // Lite does not support C API

 #include "rocksdb/c.h"

 #include <stddef.h>
-#include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
 #include <sys/types.h>
 #ifndef OS_WIN
-#  include <unistd.h>
+#include <unistd.h>
 #endif
 #include <inttypes.h>

@@ -21,11 +22,10 @@

 #include <Windows.h>

-# define snprintf _snprintf
+#define snprintf _snprintf

 // Ok for uniqueness
 int geteuid() {
-
  int result = 0;

  result = ((int)GetCurrentProcessId() << 16);
@@ -38,13 +38,13 @@ int geteuid() {

 const char* phase = "";
 static char dbname[200];
+static char sstfilename[200];
 static char dbbackupname[200];

 static void StartPhase(const char* name) {
  fprintf(stderr, "=== Test %s\n", name);
  phase = name;
 }
-
 static const char* GetTempDir(void) {
    const char* ret = getenv("TEST_TMPDIR");
    if (ret == NULL || ret[0] == '\0')
@@ -288,9 +288,11 @@ int main(int argc, char** argv) {
  rocksdb_cache_t* cache;
  rocksdb_env_t* env;
  rocksdb_options_t* options;
+  rocksdb_compactoptions_t* coptions;
  rocksdb_block_based_table_options_t* table_options;
  rocksdb_readoptions_t* roptions;
  rocksdb_writeoptions_t* woptions;
+  rocksdb_ratelimiter_t* rate_limiter;
  char* err = NULL;
  int run = -1;

@@ -304,6 +306,11 @@ int main(int argc, char** argv) {
           GetTempDir(),
           ((int) geteuid()));

+  snprintf(sstfilename, sizeof(sstfilename),
+           "%s/rocksdb_c_test-%d-sst",
+           GetTempDir(),
+           ((int)geteuid()));
+
  StartPhase("create_objects");
  cmp = rocksdb_comparator_create(NULL, CmpDestroy, CmpCompare, CmpName);
  env = rocksdb_create_default_env();
@@ -317,6 +324,7 @@ int main(int argc, char** argv) {
  rocksdb_options_set_write_buffer_size(options, 100000);
  rocksdb_options_set_paranoid_checks(options, 1);
  rocksdb_options_set_max_open_files(options, 10);
+  rocksdb_options_set_base_background_compactions(options, 1);
  table_options = rocksdb_block_based_options_create();
  rocksdb_block_based_options_set_block_cache(table_options, cache);
  rocksdb_options_set_block_based_table_factory(options, table_options);
@@ -326,14 +334,20 @@ int main(int argc, char** argv) {
  int compression_levels[] = {rocksdb_no_compression, rocksdb_no_compression,
                              rocksdb_no_compression, rocksdb_no_compression};
  rocksdb_options_set_compression_per_level(options, compression_levels, 4);
+  rate_limiter = rocksdb_ratelimiter_create(1000 * 1024 * 1024, 100 * 1000, 10);
+  rocksdb_options_set_ratelimiter(options, rate_limiter);
+  rocksdb_ratelimiter_destroy(rate_limiter);

  roptions = rocksdb_readoptions_create();
  rocksdb_readoptions_set_verify_checksums(roptions, 1);
-  rocksdb_readoptions_set_fill_cache(roptions, 0);
+  rocksdb_readoptions_set_fill_cache(roptions, 1);

  woptions = rocksdb_writeoptions_create();
  rocksdb_writeoptions_set_sync(woptions, 1);

+  coptions = rocksdb_compactoptions_create();
+  rocksdb_compactoptions_set_exclusive_manual_compaction(coptions, 1);
+
  StartPhase("destroy");
  rocksdb_destroy_db(options, dbname, &err);
  Free(&err);
@@ -415,6 +429,95 @@ int main(int argc, char** argv) {
  rocksdb_compact_range(db, "a", 1, "z", 1);
  CheckGet(db, roptions, "foo", "hello");

+  StartPhase("compactallopt");
+  rocksdb_compact_range_opt(db, coptions, NULL, 0, NULL, 0);
+  CheckGet(db, roptions, "foo", "hello");
+
+  StartPhase("compactrangeopt");
+  rocksdb_compact_range_opt(db, coptions, "a", 1, "z", 1);
+  CheckGet(db, roptions, "foo", "hello");
+
+  // Simple check cache usage
+  StartPhase("cache_usage");
+  {
+    rocksdb_readoptions_set_pin_data(roptions, 1);
+    rocksdb_iterator_t* iter = rocksdb_create_iterator(db, roptions);
+    rocksdb_iter_seek(iter, "foo", 3);
+
+    size_t usage = rocksdb_cache_get_usage(cache);
+    CheckCondition(usage > 0);
+
+    size_t pin_usage = rocksdb_cache_get_pinned_usage(cache);
+    CheckCondition(pin_usage > 0);
+
+    rocksdb_iter_next(iter);
+    rocksdb_iter_destroy(iter);
+    rocksdb_readoptions_set_pin_data(roptions, 0);
+  }
+
+  StartPhase("addfile");
+  {
+    rocksdb_envoptions_t* env_opt = rocksdb_envoptions_create();
+    rocksdb_options_t* io_options = rocksdb_options_create();
+    rocksdb_sstfilewriter_t* writer =
+        rocksdb_sstfilewriter_create(env_opt, io_options);
+
+    unlink(sstfilename);
+    rocksdb_sstfilewriter_open(writer, sstfilename, &err);
+    CheckNoError(err);
+    rocksdb_sstfilewriter_add(writer, "sstk1", 5, "v1", 2, &err);
+    CheckNoError(err);
+    rocksdb_sstfilewriter_add(writer, "sstk2", 5, "v2", 2, &err);
+    CheckNoError(err);
+    rocksdb_sstfilewriter_add(writer, "sstk3", 5, "v3", 2, &err);
+    CheckNoError(err);
+    rocksdb_sstfilewriter_finish(writer, &err);
+    CheckNoError(err);
+
+    rocksdb_ingestexternalfileoptions_t* ing_opt =
+        rocksdb_ingestexternalfileoptions_create();
+    const char* file_list[1] = {sstfilename};
+    rocksdb_ingest_external_file(db, file_list, 1, ing_opt, &err);
+    CheckNoError(err);
+    CheckGet(db, roptions, "sstk1", "v1");
+    CheckGet(db, roptions, "sstk2", "v2");
+    CheckGet(db, roptions, "sstk3", "v3");
+
+    unlink(sstfilename);
+    rocksdb_sstfilewriter_open(writer, sstfilename, &err);
+    CheckNoError(err);
+    rocksdb_sstfilewriter_add(writer, "sstk2", 5, "v4", 2, &err);
+    CheckNoError(err);
+    rocksdb_sstfilewriter_add(writer, "sstk22", 6, "v5", 2, &err);
+    CheckNoError(err);
+    rocksdb_sstfilewriter_add(writer, "sstk3", 5, "v6", 2, &err);
+    CheckNoError(err);
+    rocksdb_sstfilewriter_finish(writer, &err);
+    CheckNoError(err);
+
+    rocksdb_ingest_external_file(db, file_list, 1, ing_opt, &err);
+    CheckNoError(err);
+    CheckGet(db, roptions, "sstk1", "v1");
+    CheckGet(db, roptions, "sstk2", "v4");
+    CheckGet(db, roptions, "sstk22", "v5");
+    CheckGet(db, roptions, "sstk3", "v6");
+
+    rocksdb_ingestexternalfileoptions_destroy(ing_opt);
+    rocksdb_sstfilewriter_destroy(writer);
+    rocksdb_options_destroy(io_options);
+    rocksdb_envoptions_destroy(env_opt);
+
+    // Delete all keys we just ingested
+    rocksdb_delete(db, woptions, "sstk1", 5, &err);
+    CheckNoError(err);
+    rocksdb_delete(db, woptions, "sstk2", 5, &err);
+    CheckNoError(err);
+    rocksdb_delete(db, woptions, "sstk22", 6, &err);
+    CheckNoError(err);
+    rocksdb_delete(db, woptions, "sstk3", 5, &err);
+    CheckNoError(err);
+  }
+
  StartPhase("writebatch");
  {
    rocksdb_writebatch_t* wb = rocksdb_writebatch_create();
@@ -431,6 +534,24 @@ int main(int argc, char** argv) {
    int pos = 0;
    rocksdb_writebatch_iterate(wb, &pos, CheckPut, CheckDel);
    CheckCondition(pos == 3);
+    rocksdb_writebatch_clear(wb);
+    rocksdb_writebatch_put(wb, "bar", 3, "b", 1);
+    rocksdb_writebatch_put(wb, "bay", 3, "d", 1);
+    rocksdb_writebatch_delete_range(wb, "bar", 3, "bay", 3);
+    rocksdb_write(db, woptions, wb, &err);
+    CheckNoError(err);
+    CheckGet(db, roptions, "bar", NULL);
+    CheckGet(db, roptions, "bay", "d");
+    rocksdb_writebatch_clear(wb);
+    const char* start_list[1] = {"bay"};
+    const size_t start_sizes[1] = {3};
+    const char* end_list[1] = {"baz"};
+    const size_t end_sizes[1] = {3};
+    rocksdb_writebatch_delete_rangev(wb, 1, start_list, start_sizes, end_list,
+                                     end_sizes);
+    rocksdb_write(db, woptions, wb, &err);
+    CheckNoError(err);
+    CheckGet(db, roptions, "bay", NULL);
    rocksdb_writebatch_destroy(wb);
  }

@@ -452,6 +573,23 @@ int main(int argc, char** argv) {
    rocksdb_writebatch_destroy(wb);
  }

+  StartPhase("writebatch_savepoint");
+  {
+    rocksdb_writebatch_t* wb = rocksdb_writebatch_create();
+    rocksdb_writebatch_set_save_point(wb);
+    const char* k_list[2] = {"z", "ap"};
+    const size_t k_sizes[2] = {1, 2};
+    const char* v_list[3] = {"x", "y", "z"};
+    const size_t v_sizes[3] = {1, 1, 1};
+    rocksdb_writebatch_putv(wb, 2, k_list, k_sizes, 3, v_list, v_sizes);
+    rocksdb_writebatch_rollback_to_save_point(wb, &err);
+    CheckNoError(err);
+    rocksdb_write(db, woptions, wb, &err);
+    CheckNoError(err);
+    CheckGet(db, roptions, "zap", NULL);
+    rocksdb_writebatch_destroy(wb);
+  }
+
  StartPhase("writebatch_rep");
  {
    rocksdb_writebatch_t* wb1 = rocksdb_writebatch_create();
@@ -487,6 +625,10 @@ int main(int argc, char** argv) {
    CheckIter(iter, "foo", "hello");
    rocksdb_iter_seek(iter, "b", 1);
    CheckIter(iter, "box", "c");
+    rocksdb_iter_seek_for_prev(iter, "g", 1);
+    CheckIter(iter, "foo", "hello");
+    rocksdb_iter_seek_for_prev(iter, "box", 3);
+    CheckIter(iter, "box", "c");
    rocksdb_iter_get_error(iter, &err);
    CheckNoError(err);
    rocksdb_iter_destroy(iter);
@@ -698,7 +840,7 @@ int main(int argc, char** argv) {
  {
    rocksdb_close(db);
    rocksdb_destroy_db(options, dbname, &err);
-    CheckNoError(err)
+    CheckNoError(err);

    rocksdb_options_t* db_options = rocksdb_options_create();
    rocksdb_options_set_create_if_missing(db_options, 1);
@@ -829,6 +971,7 @@ int main(int argc, char** argv) {
    rocksdb_options_set_prefix_extractor(options, rocksdb_slicetransform_create_fixed_prefix(3));
    rocksdb_options_set_hash_skip_list_rep(options, 5000, 4, 4);
    rocksdb_options_set_plain_table_factory(options, 4, 10, 0.75, 16);
+    rocksdb_options_set_allow_concurrent_memtable_write(options, 0);

    db = rocksdb_open(options, dbname, &err);
    CheckNoError(err);
@@ -863,6 +1006,19 @@ int main(int argc, char** argv) {
    CheckNoError(err);
    rocksdb_iter_destroy(iter);

+    rocksdb_readoptions_set_total_order_seek(roptions, 1);
+    iter = rocksdb_create_iterator(db, roptions);
+    CheckCondition(!rocksdb_iter_valid(iter));
+
+    rocksdb_iter_seek(iter, "ba", 2);
+    rocksdb_iter_get_error(iter, &err);
+    CheckNoError(err);
+    CheckCondition(rocksdb_iter_valid(iter));
+    CheckIter(iter, "bar1", "bar");
+
+    rocksdb_iter_destroy(iter);
+    rocksdb_readoptions_set_total_order_seek(roptions, 0);
+
    rocksdb_close(db);
    rocksdb_destroy_db(options, dbname, &err);
  }
@@ -972,6 +1128,7 @@ int main(int argc, char** argv) {
  rocksdb_block_based_options_destroy(table_options);
  rocksdb_readoptions_destroy(roptions);
  rocksdb_writeoptions_destroy(woptions);
+  rocksdb_compactoptions_destroy(coptions);
  rocksdb_cache_destroy(cache);
  rocksdb_comparator_destroy(cmp);
  rocksdb_env_destroy(env);
@@ -981,7 +1138,6 @@ int main(int argc, char** argv) {
 }

 #else
-#include <stdio.h>

 int main() {
  fprintf(stderr, "SKIPPED\n");
@@ -27,11 +27,11 @@
 #include "db/version_set.h"
 #include "db/write_controller.h"
 #include "memtable/hash_skiplist_rep.h"
+#include "table/block_based_table_factory.h"
 #include "util/autovector.h"
 #include "util/compression.h"
 #include "util/options_helper.h"
 #include "util/thread_status_util.h"
-#include "util/xfunc.h"

 namespace rocksdb {

@@ -139,10 +139,8 @@ Status CheckConcurrentWritesSupported(const ColumnFamilyOptions& cf_options) {
 }

 ColumnFamilyOptions SanitizeOptions(const ImmutableDBOptions& db_options,
-                                    const InternalKeyComparator* icmp,
                                    const ColumnFamilyOptions& src) {
  ColumnFamilyOptions result = src;
-  result.comparator = icmp;
  size_t clamp_max = std::conditional<
      sizeof(size_t) == 4, std::integral_constant<size_t, 0xffffffff>,
      std::integral_constant<uint64_t, 64ull << 30>>::type::value;
@@ -160,6 +158,10 @@ ColumnFamilyOptions SanitizeOptions(const ImmutableDBOptions& db_options,
  result.min_write_buffer_number_to_merge =
      std::min(result.min_write_buffer_number_to_merge,
               result.max_write_buffer_number - 1);
+  if (result.min_write_buffer_number_to_merge < 1) {
+    result.min_write_buffer_number_to_merge = 1;
+  }
+
  if (result.num_levels < 1) {
    result.num_levels = 1;
  }
@@ -179,14 +181,6 @@ ColumnFamilyOptions SanitizeOptions(const ImmutableDBOptions& db_options,
  } else if (result.memtable_prefix_bloom_size_ratio < 0) {
    result.memtable_prefix_bloom_size_ratio = 0;
  }
-  XFUNC_TEST("memtablelist_history", "transaction_xftest_SanitizeOptions",
-             xf_transaction_set_memtable_history1,
-             xf_transaction_set_memtable_history,
-             &result.max_write_buffer_number_to_maintain);
-  XFUNC_TEST("memtablelist_history_clear", "transaction_xftest_SanitizeOptions",
-             xf_transaction_clear_memtable_history1,
-             xf_transaction_clear_memtable_history,
-             &result.max_write_buffer_number_to_maintain);

  if (!result.prefix_extractor) {
    assert(result.memtable_factory);
@@ -349,10 +343,11 @@ ColumnFamilyData::ColumnFamilyData(
      refs_(0),
      dropped_(false),
      internal_comparator_(cf_options.comparator),
-      initial_cf_options_(
-          SanitizeOptions(db_options, &internal_comparator_, cf_options)),
+      initial_cf_options_(SanitizeOptions(db_options, cf_options)),
      ioptions_(db_options, initial_cf_options_),
      mutable_cf_options_(initial_cf_options_),
+      is_delete_range_supported_(strcmp(cf_options.table_factory->Name(),
+                                        BlockBasedTableFactory().Name()) == 0),
      write_buffer_manager_(write_buffer_manager),
      mem_(nullptr),
      imm_(ioptions_.min_write_buffer_number_to_merge,
@@ -366,7 +361,8 @@ ColumnFamilyData::ColumnFamilyData(
      column_family_set_(column_family_set),
      pending_flush_(false),
      pending_compaction_(false),
-      prev_compaction_needed_bytes_(0) {
+      prev_compaction_needed_bytes_(0),
+      allow_2pc_(db_options.allow_2pc) {
  Ref();

  // Convert user defined table properties collector factories to internal ones.
@@ -488,15 +484,39 @@ ColumnFamilyOptions ColumnFamilyData::GetLatestCFOptions() const {
  return BuildColumnFamilyOptions(initial_cf_options_, mutable_cf_options_);
 }

-const double kSlowdownRatio = 1.2;
+uint64_t ColumnFamilyData::OldestLogToKeep() {
+  auto current_log = GetLogNumber();
+
+  if (allow_2pc_) {
+    auto imm_prep_log = imm()->GetMinLogContainingPrepSection();
+    auto mem_prep_log = mem()->GetMinLogContainingPrepSection();
+
+    if (imm_prep_log > 0 && imm_prep_log < current_log) {
+      current_log = imm_prep_log;
+    }
+
+    if (mem_prep_log > 0 && mem_prep_log < current_log) {
+      current_log = mem_prep_log;
+    }
+  }
+
+  return current_log;
+}
+
+const double kIncSlowdownRatio = 0.8;
+const double kDecSlowdownRatio = 1 / kIncSlowdownRatio;
+const double kNearStopSlowdownRatio = 0.6;
+const double kDelayRecoverSlowdownRatio = 1.4;

 namespace {
+// If penalize_stop is true, we further reduce slowdown rate.
 std::unique_ptr<WriteControllerToken> SetupDelay(
-    uint64_t max_write_rate, WriteController* write_controller,
-    uint64_t compaction_needed_bytes, uint64_t prev_compaction_neeed_bytes,
+    WriteController* write_controller, uint64_t compaction_needed_bytes,
+    uint64_t prev_compaction_need_bytes, bool penalize_stop,
    bool auto_comapctions_disabled) {
-  const uint64_t kMinWriteRate = 1024u;  // Minimum write rate 1KB/s.
+  const uint64_t kMinWriteRate = 16 * 1024u;  // Minimum write rate 16KB/s.

+  uint64_t max_write_rate = write_controller->max_delayed_write_rate();
  uint64_t write_rate = write_controller->delayed_write_rate();

  if (auto_comapctions_disabled) {
@@ -519,19 +539,32 @@ std::unique_ptr<WriteControllerToken> SetupDelay(
    // insert to mem tables, so we need to actively slow down before we get
    // feedback signal from compaction and flushes to avoid the full stop
    // because of hitting the max write buffer number.
-    if (prev_compaction_neeed_bytes > 0 &&
-        prev_compaction_neeed_bytes <= compaction_needed_bytes) {
-      write_rate = static_cast<uint64_t>(static_cast<double>(write_rate) /
-                                         kSlowdownRatio);
+    //
+    // If DB just falled into the stop condition, we need to further reduce
+    // the write rate to avoid the stop condition.
+    if (penalize_stop) {
+      // Penalize the near stop or stop condition by more agressive slowdown.
+      // This is to provide the long term slowdown increase signal.
+      // The penalty is more than the reward of recovering to the normal
+      // condition.
+      write_rate = static_cast<uint64_t>(static_cast<double>(write_rate) *
+                                         kNearStopSlowdownRatio);
      if (write_rate < kMinWriteRate) {
        write_rate = kMinWriteRate;
      }
-    } else if (prev_compaction_neeed_bytes > compaction_needed_bytes) {
+    } else if (prev_compaction_need_bytes > 0 &&
+               prev_compaction_need_bytes <= compaction_needed_bytes) {
+      write_rate = static_cast<uint64_t>(static_cast<double>(write_rate) *
+                                         kIncSlowdownRatio);
+      if (write_rate < kMinWriteRate) {
+        write_rate = kMinWriteRate;
+      }
+    } else if (prev_compaction_need_bytes > compaction_needed_bytes) {
      // We are speeding up by ratio of kSlowdownRatio when we have paid
      // compaction debt. But we'll never speed up to faster than the write rate
      // given by users.
      write_rate = static_cast<uint64_t>(static_cast<double>(write_rate) *
-                                         kSlowdownRatio);
+                                         kDecSlowdownRatio);
      if (write_rate > max_write_rate) {
        write_rate = max_write_rate;
      }
@@ -545,14 +578,31 @@ int GetL0ThresholdSpeedupCompaction(int level0_file_num_compaction_trigger,
  // SanitizeOptions() ensures it.
  assert(level0_file_num_compaction_trigger <= level0_slowdown_writes_trigger);

+  if (level0_file_num_compaction_trigger < 0) {
+    return std::numeric_limits<int>::max();
+  }
+
+  const int64_t twice_level0_trigger =
+      static_cast<int64_t>(level0_file_num_compaction_trigger) * 2;
+
+  const int64_t one_fourth_trigger_slowdown =
+      static_cast<int64_t>(level0_file_num_compaction_trigger) +
+      ((level0_slowdown_writes_trigger - level0_file_num_compaction_trigger) /
+       4);
+
+  assert(twice_level0_trigger >= 0);
+  assert(one_fourth_trigger_slowdown >= 0);
+
  // 1/4 of the way between L0 compaction trigger threshold and slowdown
  // condition.
  // Or twice as compaction trigger, if it is smaller.
-  return std::min(level0_file_num_compaction_trigger * 2,
-                  level0_file_num_compaction_trigger +
-                      (level0_slowdown_writes_trigger -
-                       level0_file_num_compaction_trigger) /
-                          4);
+  int64_t res = std::min(twice_level0_trigger, one_fourth_trigger_slowdown);
+  if (res >= port::kMaxInt32) {
+    return port::kMaxInt32;
+  } else {
+    // res fits in int
+    return static_cast<int>(res);
+  }
 }
 }  // namespace

@@ -564,6 +614,9 @@ void ColumnFamilyData::RecalculateWriteStallConditions(
    uint64_t compaction_needed_bytes =
        vstorage->estimated_compaction_needed_bytes();

+    bool was_stopped = write_controller->IsStopped();
+    bool needed_delay = write_controller->NeedsDelay();
+
    if (imm()->NumNotFlushed() >= mutable_cf_options.max_write_buffer_number) {
      write_controller_token_ = write_controller->GetStopToken();
      internal_stats_->AddCFStats(InternalStats::MEMTABLE_COMPACTION, 1);
@@ -599,8 +652,8 @@ void ColumnFamilyData::RecalculateWriteStallConditions(
               imm()->NumNotFlushed() >=
                   mutable_cf_options.max_write_buffer_number - 1) {
      write_controller_token_ =
-          SetupDelay(ioptions_.delayed_write_rate, write_controller,
-                     compaction_needed_bytes, prev_compaction_needed_bytes_,
+          SetupDelay(write_controller, compaction_needed_bytes,
+                     prev_compaction_needed_bytes_, was_stopped,
                     mutable_cf_options.disable_auto_compactions);
      internal_stats_->AddCFStats(InternalStats::MEMTABLE_SLOWDOWN, 1);
      Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
@@ -614,9 +667,12 @@ void ColumnFamilyData::RecalculateWriteStallConditions(
               mutable_cf_options.level0_slowdown_writes_trigger >= 0 &&
               vstorage->l0_delay_trigger_count() >=
                   mutable_cf_options.level0_slowdown_writes_trigger) {
+      // L0 is the last two files from stopping.
+      bool near_stop = vstorage->l0_delay_trigger_count() >=
+                       mutable_cf_options.level0_stop_writes_trigger - 2;
      write_controller_token_ =
-          SetupDelay(ioptions_.delayed_write_rate, write_controller,
-                     compaction_needed_bytes, prev_compaction_needed_bytes_,
+          SetupDelay(write_controller, compaction_needed_bytes,
+                     prev_compaction_needed_bytes_, was_stopped || near_stop,
                     mutable_cf_options.disable_auto_compactions);
      internal_stats_->AddCFStats(InternalStats::LEVEL0_SLOWDOWN_TOTAL, 1);
      if (compaction_picker_->IsLevel0CompactionInProgress()) {
@@ -632,9 +688,20 @@ void ColumnFamilyData::RecalculateWriteStallConditions(
               mutable_cf_options.soft_pending_compaction_bytes_limit > 0 &&
               vstorage->estimated_compaction_needed_bytes() >=
                   mutable_cf_options.soft_pending_compaction_bytes_limit) {
+      // If the distance to hard limit is less than 1/4 of the gap between soft
+      // and
+      // hard bytes limit, we think it is near stop and speed up the slowdown.
+      bool near_stop =
+          mutable_cf_options.hard_pending_compaction_bytes_limit > 0 &&
+          (compaction_needed_bytes -
+           mutable_cf_options.soft_pending_compaction_bytes_limit) >
+              3 * (mutable_cf_options.hard_pending_compaction_bytes_limit -
+                   mutable_cf_options.soft_pending_compaction_bytes_limit) /
+                  4;
+
      write_controller_token_ =
-          SetupDelay(ioptions_.delayed_write_rate, write_controller,
-                     compaction_needed_bytes, prev_compaction_needed_bytes_,
+          SetupDelay(write_controller, compaction_needed_bytes,
+                     prev_compaction_needed_bytes_, was_stopped || near_stop,
                     mutable_cf_options.disable_auto_compactions);
      internal_stats_->AddCFStats(
          InternalStats::SOFT_PENDING_COMPACTION_BYTES_LIMIT, 1);
@@ -643,31 +710,43 @@ void ColumnFamilyData::RecalculateWriteStallConditions(
          "bytes %" PRIu64 " rate %" PRIu64,
          name_.c_str(), vstorage->estimated_compaction_needed_bytes(),
          write_controller->delayed_write_rate());
-    } else if (vstorage->l0_delay_trigger_count() >=
-               GetL0ThresholdSpeedupCompaction(
-                   mutable_cf_options.level0_file_num_compaction_trigger,
-                   mutable_cf_options.level0_slowdown_writes_trigger)) {
-      write_controller_token_ = write_controller->GetCompactionPressureToken();
-      Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
-          "[%s] Increasing compaction threads because we have %d level-0 "
-          "files ",
-          name_.c_str(), vstorage->l0_delay_trigger_count());
-    } else if (vstorage->estimated_compaction_needed_bytes() >=
-               mutable_cf_options.soft_pending_compaction_bytes_limit / 4) {
-      // Increase compaction threads if bytes needed for compaction exceeds
-      // 1/4 of threshold for slowing down.
-      // If soft pending compaction byte limit is not set, always speed up
-      // compaction.
-      write_controller_token_ = write_controller->GetCompactionPressureToken();
-      if (mutable_cf_options.soft_pending_compaction_bytes_limit > 0) {
-        Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
-            "[%s] Increasing compaction threads because of estimated pending "
-            "compaction "
-            "bytes %" PRIu64,
-            name_.c_str(), vstorage->estimated_compaction_needed_bytes());
-      }
    } else {
-      write_controller_token_.reset();
+      if (vstorage->l0_delay_trigger_count() >=
+          GetL0ThresholdSpeedupCompaction(
+              mutable_cf_options.level0_file_num_compaction_trigger,
+              mutable_cf_options.level0_slowdown_writes_trigger)) {
+        write_controller_token_ =
+            write_controller->GetCompactionPressureToken();
+        Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
+            "[%s] Increasing compaction threads because we have %d level-0 "
+            "files ",
+            name_.c_str(), vstorage->l0_delay_trigger_count());
+      } else if (vstorage->estimated_compaction_needed_bytes() >=
+                 mutable_cf_options.soft_pending_compaction_bytes_limit / 4) {
+        // Increase compaction threads if bytes needed for compaction exceeds
+        // 1/4 of threshold for slowing down.
+        // If soft pending compaction byte limit is not set, always speed up
+        // compaction.
+        write_controller_token_ =
+            write_controller->GetCompactionPressureToken();
+        if (mutable_cf_options.soft_pending_compaction_bytes_limit > 0) {
+          Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
+              "[%s] Increasing compaction threads because of estimated pending "
+              "compaction "
+              "bytes %" PRIu64,
+              name_.c_str(), vstorage->estimated_compaction_needed_bytes());
+        }
+      } else {
+        write_controller_token_.reset();
+      }
+      // If the DB recovers from delay conditions, we reward with reducing
+      // double the slowdown ratio. This is to balance the long term slowdown
+      // increase signal.
+      if (needed_delay) {
+        uint64_t write_rate = write_controller->delayed_write_rate();
+        write_controller->set_delayed_write_rate(static_cast<uint64_t>(
+            static_cast<double>(write_rate) * kDelayRecoverSlowdownRatio));
+      }
    }
    prev_compaction_needed_bytes_ = compaction_needed_bytes;
  }
@@ -42,7 +42,7 @@ class LogBuffer;
 class InstrumentedMutex;
 class InstrumentedMutexLock;

-extern const double kSlowdownRatio;
+extern const double kIncSlowdownRatio;

 // ColumnFamilyHandleImpl is the class that clients use to access different
 // column families. It has non-trivial destructor, which gets called when client
@@ -137,7 +137,6 @@ extern Status CheckConcurrentWritesSupported(
    const ColumnFamilyOptions& cf_options);

 extern ColumnFamilyOptions SanitizeOptions(const ImmutableDBOptions& db_options,
-                                           const InternalKeyComparator* icmp,
                                           const ColumnFamilyOptions& src);
 // Wrap user defined table proproties collector factories `from cf_options`
 // into internal ones in int_tbl_prop_collector_factories. Add a system internal
@@ -221,6 +220,8 @@ class ColumnFamilyData {
  // options.
  ColumnFamilyOptions GetLatestCFOptions() const;

+  bool is_delete_range_supported() { return is_delete_range_supported_; }
+
 #ifndef ROCKSDB_LITE
  // REQUIRES: DB mutex held
  Status SetOptions(
@@ -238,6 +239,9 @@ class ColumnFamilyData {
  uint64_t GetTotalSstFilesSize() const;  // REQUIRE: DB mutex held
  void SetMemtable(MemTable* new_mem) { mem_ = new_mem; }

+  // calculate the oldest log needed for the durability of this column family
+  uint64_t OldestLogToKeep();
+
  // See Memtable constructor for explanation of earliest_seq param.
  MemTable* ConstructNewMemtable(const MutableCFOptions& mutable_cf_options,
                                 SequenceNumber earliest_seq);
@@ -355,6 +359,8 @@ class ColumnFamilyData {
  const ImmutableCFOptions ioptions_;
  MutableCFOptions mutable_cf_options_;

+  const bool is_delete_range_supported_;
+
  std::unique_ptr<TableCache> table_cache_;

  std::unique_ptr<InternalStats> internal_stats_;
@@ -401,6 +407,9 @@ class ColumnFamilyData {
  bool pending_compaction_;

  uint64_t prev_compaction_needed_bytes_;
+
+  // if the database was opened with 2pc enabled
+  bool allow_2pc_;
 };

 // ColumnFamilySet has interesting thread-safety requirements
@@ -15,6 +15,7 @@

 #include "db/db_impl.h"
 #include "db/db_test_util.h"
+#include "port/port.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
 #include "rocksdb/iterator.h"
@@ -212,6 +213,30 @@ class ColumnFamilyTest : public testing::Test {
 #endif
  }

+  bool IsDbWriteStopped() {
+#ifndef ROCKSDB_LITE
+    uint64_t v;
+    EXPECT_TRUE(dbfull()->GetIntProperty("rocksdb.is-write-stopped", &v));
+    return (v == 1);
+#else
+    return dbfull()->TEST_write_controler().IsStopped();
+#endif  // !ROCKSDB_LITE
+  }
+
+  uint64_t GetDbDelayedWriteRate() {
+#ifndef ROCKSDB_LITE
+    uint64_t v;
+    EXPECT_TRUE(
+        dbfull()->GetIntProperty("rocksdb.actual-delayed-write-rate", &v));
+    return v;
+#else
+    if (!dbfull()->TEST_write_controler().NeedsDelay()) {
+      return 0;
+    }
+    return dbfull()->TEST_write_controler().delayed_write_rate();
+#endif  // !ROCKSDB_LITE
+  }
+
  void Destroy() {
    Close();
    ASSERT_OK(DestroyDB(dbname_, Options(db_options_, column_family_options_)));
@@ -941,6 +966,12 @@ TEST_F(ColumnFamilyTest, CrashAfterFlush) {
  db_options_.env = env_;
 }

+TEST_F(ColumnFamilyTest, OpenNonexistentColumnFamily) {
+  ASSERT_OK(TryOpen({"default"}));
+  Close();
+  ASSERT_TRUE(TryOpen({"default", "dne"}).IsInvalidArgument());
+}
+
 #ifndef ROCKSDB_LITE  // WaitForFlush() is not supported
 // Makes sure that obsolete log files get deleted
 TEST_F(ColumnFamilyTest, DifferentWriteBufferSizes) {
@@ -1050,6 +1081,7 @@ TEST_F(ColumnFamilyTest, DifferentWriteBufferSizes) {

 #ifndef ROCKSDB_LITE  // Cuckoo is not supported in lite
 TEST_F(ColumnFamilyTest, MemtableNotSupportSnapshot) {
+  db_options_.allow_concurrent_memtable_write = false;
  Open();
  auto* s1 = dbfull()->GetSnapshot();
  ASSERT_TRUE(s1 != nullptr);
@@ -1140,7 +1172,6 @@ TEST_F(ColumnFamilyTest, DifferentCompactionStyles) {
  CreateColumnFamilies({"one", "two"});
  ColumnFamilyOptions default_cf, one, two;
  db_options_.max_open_files = 20;  // only 10 files in file cache
-  db_options_.disableDataSync = true;

  default_cf.compaction_style = kCompactionStyleLevel;
  default_cf.num_levels = 3;
@@ -1213,7 +1244,6 @@ TEST_F(ColumnFamilyTest, MultipleManualCompactions) {
  CreateColumnFamilies({"one", "two"});
  ColumnFamilyOptions default_cf, one, two;
  db_options_.max_open_files = 20;  // only 10 files in file cache
-  db_options_.disableDataSync = true;
  db_options_.max_background_compactions = 3;

  default_cf.compaction_style = kCompactionStyleLevel;
@@ -1261,7 +1291,7 @@ TEST_F(ColumnFamilyTest, MultipleManualCompactions) {
      });

  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-  std::vector<std::thread> threads;
+  std::vector<port::Thread> threads;
  threads.emplace_back([&] {
    CompactRangeOptions compact_options;
    compact_options.exclusive_manual_compaction = false;
@@ -1311,7 +1341,6 @@ TEST_F(ColumnFamilyTest, AutomaticAndManualCompactions) {
  CreateColumnFamilies({"one", "two"});
  ColumnFamilyOptions default_cf, one, two;
  db_options_.max_open_files = 20;  // only 10 files in file cache
-  db_options_.disableDataSync = true;
  db_options_.max_background_compactions = 3;
  db_options_.base_background_compactions = 3;

@@ -1369,7 +1398,7 @@ TEST_F(ColumnFamilyTest, AutomaticAndManualCompactions) {
    WaitForFlush(2);
    AssertFilesPerLevel(ToString(i + 1), 2);
  }
-  std::thread threads([&] {
+  rocksdb::port::Thread threads([&] {
    CompactRangeOptions compact_options;
    compact_options.exclusive_manual_compaction = false;
    ASSERT_OK(
@@ -1404,7 +1433,6 @@ TEST_F(ColumnFamilyTest, ManualAndAutomaticCompactions) {
  CreateColumnFamilies({"one", "two"});
  ColumnFamilyOptions default_cf, one, two;
  db_options_.max_open_files = 20;  // only 10 files in file cache
-  db_options_.disableDataSync = true;
  db_options_.max_background_compactions = 3;
  db_options_.base_background_compactions = 3;

@@ -1457,7 +1485,7 @@ TEST_F(ColumnFamilyTest, ManualAndAutomaticCompactions) {
      });

  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-  std::thread threads([&] {
+  rocksdb::port::Thread threads([&] {
    CompactRangeOptions compact_options;
    compact_options.exclusive_manual_compaction = false;
    ASSERT_OK(
@@ -1500,7 +1528,6 @@ TEST_F(ColumnFamilyTest, SameCFManualManualCompactions) {
  CreateColumnFamilies({"one"});
  ColumnFamilyOptions default_cf, one;
  db_options_.max_open_files = 20;  // only 10 files in file cache
-  db_options_.disableDataSync = true;
  db_options_.max_background_compactions = 3;
  db_options_.base_background_compactions = 3;

@@ -1550,7 +1577,7 @@ TEST_F(ColumnFamilyTest, SameCFManualManualCompactions) {
      });

  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-  std::thread threads([&] {
+  rocksdb::port::Thread threads([&] {
    CompactRangeOptions compact_options;
    compact_options.exclusive_manual_compaction = true;
    ASSERT_OK(
@@ -1570,7 +1597,7 @@ TEST_F(ColumnFamilyTest, SameCFManualManualCompactions) {
                        1);
  }

-  std::thread threads1([&] {
+  rocksdb::port::Thread threads1([&] {
    CompactRangeOptions compact_options;
    compact_options.exclusive_manual_compaction = false;
    ASSERT_OK(
@@ -1599,7 +1626,6 @@ TEST_F(ColumnFamilyTest, SameCFManualAutomaticCompactions) {
  CreateColumnFamilies({"one"});
  ColumnFamilyOptions default_cf, one;
  db_options_.max_open_files = 20;  // only 10 files in file cache
-  db_options_.disableDataSync = true;
  db_options_.max_background_compactions = 3;
  db_options_.base_background_compactions = 3;

@@ -1648,7 +1674,7 @@ TEST_F(ColumnFamilyTest, SameCFManualAutomaticCompactions) {
      });

  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-  std::thread threads([&] {
+  rocksdb::port::Thread threads([&] {
    CompactRangeOptions compact_options;
    compact_options.exclusive_manual_compaction = false;
    ASSERT_OK(
@@ -1689,7 +1715,6 @@ TEST_F(ColumnFamilyTest, SameCFManualAutomaticCompactionsLevel) {
  CreateColumnFamilies({"one"});
  ColumnFamilyOptions default_cf, one;
  db_options_.max_open_files = 20;  // only 10 files in file cache
-  db_options_.disableDataSync = true;
  db_options_.max_background_compactions = 3;
  db_options_.base_background_compactions = 3;

@@ -1740,7 +1765,7 @@ TEST_F(ColumnFamilyTest, SameCFManualAutomaticCompactionsLevel) {
      });

  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-  std::thread threads([&] {
+  rocksdb::port::Thread threads([&] {
    CompactRangeOptions compact_options;
    compact_options.exclusive_manual_compaction = false;
    ASSERT_OK(
@@ -1786,7 +1811,6 @@ TEST_F(ColumnFamilyTest, SameCFManualAutomaticConflict) {
  CreateColumnFamilies({"one"});
  ColumnFamilyOptions default_cf, one;
  db_options_.max_open_files = 20;  // only 10 files in file cache
-  db_options_.disableDataSync = true;
  db_options_.max_background_compactions = 3;
  db_options_.base_background_compactions = 3;

@@ -1851,7 +1875,7 @@ TEST_F(ColumnFamilyTest, SameCFManualAutomaticConflict) {
      });

  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-  std::thread threads([&] {
+  rocksdb::port::Thread threads([&] {
    CompactRangeOptions compact_options;
    compact_options.exclusive_manual_compaction = false;
    ASSERT_OK(
@@ -1906,7 +1930,6 @@ TEST_F(ColumnFamilyTest, SameCFAutomaticManualCompactions) {
  CreateColumnFamilies({"one"});
  ColumnFamilyOptions default_cf, one;
  db_options_.max_open_files = 20;  // only 10 files in file cache
-  db_options_.disableDataSync = true;
  db_options_.max_background_compactions = 3;
  db_options_.base_background_compactions = 3;

@@ -2182,8 +2205,8 @@ TEST_F(ColumnFamilyTest, SanitizeOptions) {
            original.write_buffer_size =
                l * 4 * 1024 * 1024 + i * 1024 * 1024 + j * 1024 + k;

-            ColumnFamilyOptions result = SanitizeOptions(
-                ImmutableDBOptions(db_options), nullptr, original);
+            ColumnFamilyOptions result =
+                SanitizeOptions(ImmutableDBOptions(db_options), original);
            ASSERT_TRUE(result.level0_stop_writes_trigger >=
                        result.level0_slowdown_writes_trigger);
            ASSERT_TRUE(result.level0_slowdown_writes_trigger >=
@@ -2316,7 +2339,7 @@ TEST_F(ColumnFamilyTest, FlushAndDropRaceCondition) {
  int kKeysNum = 10000;
  PutRandomData(1, kKeysNum, 100);

-  std::vector<std::thread> threads;
+  std::vector<port::Thread> threads;
  threads.emplace_back([&] { ASSERT_OK(db_->DropColumnFamily(handles_[1])); });

  sleeping_task.WakeUp();
@@ -2417,7 +2440,7 @@ TEST_F(ColumnFamilyTest, CreateAndDropRace) {

  // Start a thread that will drop the first column family
  // and its comparator
-  std::thread drop_cf_thread(DropSingleColumnFamily, this, 1, &comparators);
+  rocksdb::port::Thread drop_cf_thread(DropSingleColumnFamily, this, 1, &comparators);

  DropColumnFamilies({2});

@@ -2433,7 +2456,7 @@ TEST_F(ColumnFamilyTest, CreateAndDropRace) {
 #endif  // !ROCKSDB_LITE

 TEST_F(ColumnFamilyTest, WriteStallSingleColumnFamily) {
-  const uint64_t kBaseRate = 810000u;
+  const uint64_t kBaseRate = 800000u;
  db_options_.delayed_write_rate = kBaseRate;
  db_options_.base_background_compactions = 2;
  db_options_.max_background_compactions = 6;
@@ -2454,174 +2477,165 @@ TEST_F(ColumnFamilyTest, WriteStallSingleColumnFamily) {

  vstorage->TEST_set_estimated_compaction_needed_bytes(50);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());

  vstorage->TEST_set_estimated_compaction_needed_bytes(201);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate, GetDbDelayedWriteRate());
  ASSERT_EQ(6, dbfull()->TEST_BGCompactionsAllowed());

  vstorage->TEST_set_estimated_compaction_needed_bytes(400);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate / 1.25, GetDbDelayedWriteRate());
  ASSERT_EQ(6, dbfull()->TEST_BGCompactionsAllowed());

  vstorage->TEST_set_estimated_compaction_needed_bytes(500);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2 / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate / 1.25 / 1.25, GetDbDelayedWriteRate());

  vstorage->TEST_set_estimated_compaction_needed_bytes(450);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate / 1.25, GetDbDelayedWriteRate());

  vstorage->TEST_set_estimated_compaction_needed_bytes(205);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate, GetDbDelayedWriteRate());

  vstorage->TEST_set_estimated_compaction_needed_bytes(202);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate, GetDbDelayedWriteRate());

  vstorage->TEST_set_estimated_compaction_needed_bytes(201);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate, GetDbDelayedWriteRate());

  vstorage->TEST_set_estimated_compaction_needed_bytes(198);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());

  vstorage->TEST_set_estimated_compaction_needed_bytes(399);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate, GetDbDelayedWriteRate());

  vstorage->TEST_set_estimated_compaction_needed_bytes(599);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate / 1.25, GetDbDelayedWriteRate());

  vstorage->TEST_set_estimated_compaction_needed_bytes(2001);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(IsDbWriteStopped());
  ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
  ASSERT_EQ(6, dbfull()->TEST_BGCompactionsAllowed());

  vstorage->TEST_set_estimated_compaction_needed_bytes(3001);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(IsDbWriteStopped());
  ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());

  vstorage->TEST_set_estimated_compaction_needed_bytes(390);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate / 1.25, GetDbDelayedWriteRate());

  vstorage->TEST_set_estimated_compaction_needed_bytes(100);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());

  vstorage->set_l0_delay_trigger_count(100);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate, GetDbDelayedWriteRate());
  ASSERT_EQ(6, dbfull()->TEST_BGCompactionsAllowed());

  vstorage->set_l0_delay_trigger_count(101);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2 / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate / 1.25, GetDbDelayedWriteRate());

  vstorage->set_l0_delay_trigger_count(0);
  vstorage->TEST_set_estimated_compaction_needed_bytes(300);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2 / 1.2 / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate / 1.25 / 1.25, GetDbDelayedWriteRate());

  vstorage->set_l0_delay_trigger_count(101);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2 / 1.2 / 1.2 / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate / 1.25 / 1.25 / 1.25, GetDbDelayedWriteRate());

  vstorage->TEST_set_estimated_compaction_needed_bytes(200);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2 / 1.2 / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate / 1.25 / 1.25, GetDbDelayedWriteRate());

  vstorage->set_l0_delay_trigger_count(0);
  vstorage->TEST_set_estimated_compaction_needed_bytes(0);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());

  mutable_cf_options.disable_auto_compactions = true;
  dbfull()->TEST_write_controler().set_delayed_write_rate(kBaseRate);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());

  vstorage->set_l0_delay_trigger_count(50);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
+  ASSERT_EQ(0, GetDbDelayedWriteRate());
  ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());

  vstorage->set_l0_delay_trigger_count(60);
  vstorage->TEST_set_estimated_compaction_needed_bytes(300);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
+  ASSERT_EQ(0, GetDbDelayedWriteRate());
  ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());

  mutable_cf_options.disable_auto_compactions = false;
  vstorage->set_l0_delay_trigger_count(70);
  vstorage->TEST_set_estimated_compaction_needed_bytes(500);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate, GetDbDelayedWriteRate());

  vstorage->set_l0_delay_trigger_count(71);
  vstorage->TEST_set_estimated_compaction_needed_bytes(501);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate / 1.25, GetDbDelayedWriteRate());
 }

 TEST_F(ColumnFamilyTest, CompactionSpeedupSingleColumnFamily) {
@@ -2713,60 +2727,55 @@ TEST_F(ColumnFamilyTest, WriteStallTwoColumnFamilies) {

  vstorage->TEST_set_estimated_compaction_needed_bytes(50);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());

  vstorage1->TEST_set_estimated_compaction_needed_bytes(201);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());

  vstorage1->TEST_set_estimated_compaction_needed_bytes(600);
  cfd1->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate, GetDbDelayedWriteRate());

  vstorage->TEST_set_estimated_compaction_needed_bytes(70);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate, GetDbDelayedWriteRate());

  vstorage1->TEST_set_estimated_compaction_needed_bytes(800);
  cfd1->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate / 1.25, GetDbDelayedWriteRate());

  vstorage->TEST_set_estimated_compaction_needed_bytes(300);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2 / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate / 1.25 / 1.25, GetDbDelayedWriteRate());

  vstorage1->TEST_set_estimated_compaction_needed_bytes(700);
  cfd1->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate / 1.25, GetDbDelayedWriteRate());

  vstorage->TEST_set_estimated_compaction_needed_bytes(500);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2 / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate / 1.25 / 1.25, GetDbDelayedWriteRate());

  vstorage1->TEST_set_estimated_compaction_needed_bytes(600);
  cfd1->RecalculateWriteStallConditions(mutable_cf_options);
-  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
+  ASSERT_TRUE(!IsDbWriteStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
-  ASSERT_EQ(kBaseRate / 1.2,
-            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(kBaseRate / 1.25, GetDbDelayedWriteRate());
 }

 TEST_F(ColumnFamilyTest, CompactionSpeedupTwoColumnFamilies) {
@@ -3072,7 +3081,7 @@ TEST_F(ColumnFamilyTest, LogSyncConflictFlush) {

  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  std::thread thread([&] { db_->SyncWAL(); });
+  rocksdb::port::Thread thread([&] { db_->SyncWAL(); });

  TEST_SYNC_POINT("ColumnFamilyTest::LogSyncConflictFlush:1");
  Flush(1);
@@ -11,6 +11,7 @@
 #include <vector>

 #include "db/db_impl.h"
+#include "port/port.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
 #include "util/string_util.h"
@@ -231,7 +232,7 @@ TEST_F(CompactFilesTest, CapturingPendingFiles) {
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

  // Start compacting files.
-  std::thread compaction_thread(
+  rocksdb::port::Thread compaction_thread(
      [&] { EXPECT_OK(db->CompactFiles(CompactionOptions(), l0_files, 1)); });

  // In the meantime flush another file.
@@ -253,6 +254,61 @@ TEST_F(CompactFilesTest, CapturingPendingFiles) {
  delete db;
 }

+TEST_F(CompactFilesTest, CompactionFilterWithGetSv) {
+  class FilterWithGet : public CompactionFilter {
+   public:
+    virtual bool Filter(int level, const Slice& key, const Slice& value,
+                        std::string* new_value,
+                        bool* value_changed) const override {
+      if (db_ == nullptr) {
+        return true;
+      }
+      std::string res;
+      db_->Get(ReadOptions(), "", &res);
+      return true;
+    }
+
+    void SetDB(DB* db) {
+      db_ = db;
+    }
+
+    virtual const char* Name() const override { return "FilterWithGet"; }
+
+   private:
+    DB* db_;
+  };
+
+
+  std::shared_ptr<FilterWithGet> cf(new FilterWithGet());
+
+  Options options;
+  options.create_if_missing = true;
+  options.compaction_filter = cf.get();
+
+  DB* db = nullptr;
+  DestroyDB(db_name_, options);
+  Status s = DB::Open(options, db_name_, &db);
+  ASSERT_OK(s);
+
+  cf->SetDB(db);
+
+  // Write one L0 file
+  db->Put(WriteOptions(), "K1", "V1");
+  db->Flush(FlushOptions());
+
+  // Compact all L0 files using CompactFiles
+  rocksdb::ColumnFamilyMetaData meta;
+  db->GetColumnFamilyMetaData(&meta);
+  for (auto& file : meta.levels[0].files) {
+    std::string fname = file.db_path + "/" + file.name;
+    ASSERT_OK(
+        db->CompactFiles(rocksdb::CompactionOptions(), {fname}, 0));
+  }
+
+
+  delete db;
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -46,7 +46,7 @@ Status CompactedDBImpl::Get(const ReadOptions& options,
     ColumnFamilyHandle*, const Slice& key, std::string* value) {
  GetContext get_context(user_comparator_, nullptr, nullptr, nullptr,
                         GetContext::kNotFound, key, value, nullptr, nullptr,
-                         nullptr);
+                         nullptr, nullptr);
  LookupKey lkey(key, kMaxSequenceNumber);
  files_.files[FindFile(key)].fd.table_reader->Get(
      options, lkey.internal_key(), &get_context);
@@ -77,7 +77,7 @@ std::vector<Status> CompactedDBImpl::MultiGet(const ReadOptions& options,
    if (r != nullptr) {
      GetContext get_context(user_comparator_, nullptr, nullptr, nullptr,
                             GetContext::kNotFound, keys[idx], &(*values)[idx],
-                             nullptr, nullptr, nullptr);
+                             nullptr, nullptr, nullptr, nullptr);
      LookupKey lkey(keys[idx], kMaxSequenceNumber);
      r->Get(options, lkey.internal_key(), &get_context);
      if (get_context.State() == GetContext::kFound) {
@@ -76,6 +76,13 @@ class CompactedDBImpl : public DBImpl {
                       ColumnFamilyHandle* column_family) override {
    return Status::NotSupported("Not supported in compacted db mode.");
  }
+  using DB::IngestExternalFile;
+  virtual Status IngestExternalFile(
+      ColumnFamilyHandle* column_family,
+      const std::vector<std::string>& external_files,
+      const IngestExternalFileOptions& ingestion_options) override {
+    return Status::NotSupported("Not supported in compacted db mode.");
+  }

 private:
  friend class DB;
@@ -150,7 +150,8 @@ Compaction::Compaction(VersionStorageInfo* vstorage,
                       bool _manual_compaction, double _score,
                       bool _deletion_compaction,
                       CompactionReason _compaction_reason)
-    : start_level_(_inputs[0].level),
+    : input_vstorage_(vstorage),
+      start_level_(_inputs[0].level),
      output_level_(_output_level),
      max_output_file_size_(_target_file_size),
      max_compaction_bytes_(_max_compaction_bytes),
@@ -204,11 +205,10 @@ Compaction::~Compaction() {
 }

 bool Compaction::InputCompressionMatchesOutput() const {
-  VersionStorageInfo* vstorage = input_version_->storage_info();
-  int base_level = vstorage->base_level();
-  bool matches =
-      (GetCompressionType(*cfd_->ioptions(), vstorage, mutable_cf_options_,
-                          start_level_, base_level) == output_compression_);
+  int base_level = input_vstorage_->base_level();
+  bool matches = (GetCompressionType(immutable_cf_options_, input_vstorage_,
+                                     mutable_cf_options_, start_level_,
+                                     base_level) == output_compression_);
  if (matches) {
    TEST_SYNC_POINT("Compaction::InputCompressionMatchesOutput:Matches");
    return true;
@@ -225,15 +225,14 @@ bool Compaction::IsTrivialMove() const {
  // filter to be applied to that level, and thus cannot be a trivial move.

  // Check if start level have files with overlapping ranges
-  if (start_level_ == 0 &&
-      input_version_->storage_info()->level0_non_overlapping() == false) {
+  if (start_level_ == 0 && input_vstorage_->level0_non_overlapping() == false) {
    // We cannot move files from L0 to L1 if the files are overlapping
    return false;
  }

  if (is_manual_compaction_ &&
-      (cfd_->ioptions()->compaction_filter != nullptr ||
-       cfd_->ioptions()->compaction_filter_factory != nullptr)) {
+      (immutable_cf_options_.compaction_filter != nullptr ||
+       immutable_cf_options_.compaction_filter_factory != nullptr)) {
    // This is a manual compaction and we have a compaction filter that should
    // be executed, we cannot do a trivial move
    return false;
@@ -241,15 +240,34 @@ bool Compaction::IsTrivialMove() const {

  // Used in universal compaction, where trivial move can be done if the
  // input files are non overlapping
-  if ((cfd_->ioptions()->compaction_options_universal.allow_trivial_move) &&
+  if ((immutable_cf_options_.compaction_options_universal.allow_trivial_move) &&
      (output_level_ != 0)) {
    return is_trivial_move_;
  }

-  return (start_level_ != output_level_ && num_input_levels() == 1 &&
+  if (!(start_level_ != output_level_ && num_input_levels() == 1 &&
          input(0, 0)->fd.GetPathId() == output_path_id() &&
-          InputCompressionMatchesOutput() &&
-          TotalFileSize(grandparents_) <= max_compaction_bytes_);
+          InputCompressionMatchesOutput())) {
+    return false;
+  }
+
+  // assert inputs_.size() == 1
+
+  for (const auto& file : inputs_.front().files) {
+    std::vector<FileMetaData*> file_grand_parents;
+    if (output_level_ + 1 >= number_levels_) {
+      continue;
+    }
+    input_vstorage_->GetOverlappingInputs(output_level_ + 1, &file->smallest,
+                                          &file->largest, &file_grand_parents);
+    const auto compaction_size =
+        file->fd.GetFileSize() + TotalFileSize(file_grand_parents);
+    if (compaction_size > max_compaction_bytes_) {
+      return false;
+    }
+  }
+
+  return true;
 }

 void Compaction::AddInputDeletions(VersionEdit* out_edit) {
@@ -272,8 +290,7 @@ bool Compaction::KeyNotExistsBeyondOutputLevel(
  // Maybe use binary search to find right entry instead of linear search?
  const Comparator* user_cmp = cfd_->user_comparator();
  for (int lvl = output_level_ + 1; lvl < number_levels_; lvl++) {
-    const std::vector<FileMetaData*>& files =
-        input_version_->storage_info()->LevelFiles(lvl);
+    const std::vector<FileMetaData*>& files = input_vstorage_->LevelFiles(lvl);
    for (; level_ptrs->at(lvl) < files.size(); level_ptrs->at(lvl)++) {
      auto* f = files[level_ptrs->at(lvl)];
      if (user_cmp->Compare(user_key, f->largest.user_key()) <= 0) {
@@ -345,7 +362,7 @@ void Compaction::ReleaseCompactionFiles(Status status) {

 void Compaction::ResetNextCompactionIndex() {
  assert(input_version_ != nullptr);
-  input_version_->storage_info()->ResetNextCompactionIndex(start_level_);
+  input_vstorage_->ResetNextCompactionIndex(start_level_);
 }

 namespace {
@@ -258,6 +258,8 @@ class Compaction {
  static bool IsFullCompaction(VersionStorageInfo* vstorage,
                               const std::vector<CompactionInputFiles>& inputs);

+  VersionStorageInfo* input_vstorage_;
+
  const int start_level_;    // the lowest level to be compacted
  const int output_level_;  // levels to which output files are stored
  uint64_t max_output_file_size_;
@@ -0,0 +1,33 @@
+//  Copyright (c) 2016-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#pragma once
+
+struct CompactionIterationStats {
+  // Compaction statistics
+
+  // Doesn't include records skipped because of
+  // CompactionFilter::Decision::kRemoveAndSkipUntil.
+  int64_t num_record_drop_user = 0;
+
+  int64_t num_record_drop_hidden = 0;
+  int64_t num_record_drop_obsolete = 0;
+  int64_t num_record_drop_range_del = 0;
+  int64_t num_range_del_drop_obsolete = 0;
+  uint64_t total_filter_time = 0;
+
+  // Input statistics
+  // TODO(noetzli): The stats are incomplete. They are lacking everything
+  // consumed by MergeHelper.
+  uint64_t num_input_records = 0;
+  uint64_t num_input_deletion_records = 0;
+  uint64_t num_input_corrupt_records = 0;
+  uint64_t total_input_raw_key_bytes = 0;
+  uint64_t total_input_raw_value_bytes = 0;
+
+  // Single-Delete diagnostics for exceptional situations
+  uint64_t num_single_del_fallthru = 0;
+  uint64_t num_single_del_mismatch = 0;
+};
@@ -16,7 +16,23 @@ CompactionIterator::CompactionIterator(
    SequenceNumber earliest_write_conflict_snapshot, Env* env,
    bool expect_valid_internal_key, RangeDelAggregator* range_del_agg,
    const Compaction* compaction, const CompactionFilter* compaction_filter,
-    LogBuffer* log_buffer)
+    const std::atomic<bool>* shutting_down)
+    : CompactionIterator(
+          input, cmp, merge_helper, last_sequence, snapshots,
+          earliest_write_conflict_snapshot, env, expect_valid_internal_key,
+          range_del_agg,
+          std::unique_ptr<CompactionProxy>(
+              compaction ? new CompactionProxy(compaction) : nullptr),
+          compaction_filter, shutting_down) {}
+
+CompactionIterator::CompactionIterator(
+    InternalIterator* input, const Comparator* cmp, MergeHelper* merge_helper,
+    SequenceNumber last_sequence, std::vector<SequenceNumber>* snapshots,
+    SequenceNumber earliest_write_conflict_snapshot, Env* env,
+    bool expect_valid_internal_key, RangeDelAggregator* range_del_agg,
+    std::unique_ptr<CompactionProxy> compaction,
+    const CompactionFilter* compaction_filter,
+    const std::atomic<bool>* shutting_down)
    : input_(input),
      cmp_(cmp),
      merge_helper_(merge_helper),
@@ -25,9 +41,9 @@ CompactionIterator::CompactionIterator(
      env_(env),
      expect_valid_internal_key_(expect_valid_internal_key),
      range_del_agg_(range_del_agg),
-      compaction_(compaction),
+      compaction_(std::move(compaction)),
      compaction_filter_(compaction_filter),
-      log_buffer_(log_buffer),
+      shutting_down_(shutting_down),
      merge_out_iter_(merge_helper_) {
  assert(compaction_filter_ == nullptr || compaction_ != nullptr);
  bottommost_level_ =
@@ -63,6 +79,8 @@ void CompactionIterator::ResetRecordCounts() {
  iter_stats_.num_record_drop_user = 0;
  iter_stats_.num_record_drop_hidden = 0;
  iter_stats_.num_record_drop_obsolete = 0;
+  iter_stats_.num_record_drop_range_del = 0;
+  iter_stats_.num_range_del_drop_obsolete = 0;
 }

 void CompactionIterator::SeekToFirst() {
@@ -108,7 +126,7 @@ void CompactionIterator::Next() {
  }

  if (valid_) {
-    // Record that we've ouputted a record for the current key.
+    // Record that we've outputted a record for the current key.
    has_outputted_key_ = true;
  }

@@ -119,7 +137,7 @@ void CompactionIterator::NextFromInput() {
  at_next_ = false;
  valid_ = false;

-  while (!valid_ && input_->Valid()) {
+  while (!valid_ && input_->Valid() && !IsShuttingDown()) {
    key_ = input_->key();
    value_ = input_->value();
    iter_stats_.num_input_records++;
@@ -149,6 +167,13 @@ void CompactionIterator::NextFromInput() {
    iter_stats_.total_input_raw_key_bytes += key_.size();
    iter_stats_.total_input_raw_value_bytes += value_.size();

+    // If need_skip is true, we should seek the input iterator
+    // to internal key skip_until and continue from there.
+    bool need_skip = false;
+    // Points either into compaction_filter_skip_until_ or into
+    // merge_helper_->compaction_filter_skip_until_.
+    Slice skip_until;
+
    // Check whether the user key changed. After this if statement current_key_
    // is a copy of the current input key (maybe converted to a delete by the
    // compaction filter). ikey_.user_key is pointing to the copy.
@@ -171,26 +196,42 @@ void CompactionIterator::NextFromInput() {
        // number is greater than any external snapshot, then invoke the
        // filter. If the return value of the compaction filter is true,
        // replace the entry with a deletion marker.
-        bool value_changed = false;
-        bool to_delete = false;
+        CompactionFilter::Decision filter;
        compaction_filter_value_.clear();
+        compaction_filter_skip_until_.Clear();
        {
          StopWatchNano timer(env_, true);
-          to_delete = compaction_filter_->Filter(
-              compaction_->level(), ikey_.user_key, value_,
-              &compaction_filter_value_, &value_changed);
+          filter = compaction_filter_->FilterV2(
+              compaction_->level(), ikey_.user_key,
+              CompactionFilter::ValueType::kValue, value_,
+              &compaction_filter_value_, compaction_filter_skip_until_.rep());
          iter_stats_.total_filter_time +=
              env_ != nullptr ? timer.ElapsedNanos() : 0;
        }
-        if (to_delete) {
-          // convert the current key to a delete
+
+        if (filter == CompactionFilter::Decision::kRemoveAndSkipUntil &&
+            cmp_->Compare(*compaction_filter_skip_until_.rep(),
+                          ikey_.user_key) <= 0) {
+          // Can't skip to a key smaller than the current one.
+          // Keep the key as per FilterV2 documentation.
+          filter = CompactionFilter::Decision::kKeep;
+        }
+
+        if (filter == CompactionFilter::Decision::kRemove) {
+          // convert the current key to a delete; key_ is pointing into
+          // current_key_ at this point, so updating current_key_ updates key()
          ikey_.type = kTypeDeletion;
          current_key_.UpdateInternalKey(ikey_.sequence, kTypeDeletion);
          // no value associated with delete
          value_.clear();
          iter_stats_.num_record_drop_user++;
-        } else if (value_changed) {
+        } else if (filter == CompactionFilter::Decision::kChangeValue) {
          value_ = compaction_filter_value_;
+        } else if (filter == CompactionFilter::Decision::kRemoveAndSkipUntil) {
+          need_skip = true;
+          compaction_filter_skip_until_.ConvertFromUserKey(kMaxSequenceNumber,
+                                                           kValueTypeForSeek);
+          skip_until = compaction_filter_skip_until_.Encode();
        }
      }
    } else {
@@ -217,7 +258,9 @@ void CompactionIterator::NextFromInput() {
            ? earliest_snapshot_
            : findEarliestVisibleSnapshot(ikey_.sequence, &prev_snapshot);

-    if (clear_and_output_next_key_) {
+    if (need_skip) {
+      // This case is handled below.
+    } else if (clear_and_output_next_key_) {
      // In the previous iteration we encountered a single delete that we could
      // not compact out.  We will keep this Put, but can drop it's data.
      // (See Optimization 3, below.)
@@ -308,7 +351,6 @@ void CompactionIterator::NextFromInput() {
            // to know that a write happened in this snapshot (Rule 2 above).
            // Clear the value and output the SingleDelete. (The value will be
            // outputted on the next iteration.)
-            ++iter_stats_.num_record_drop_hidden;

            // Setting valid_ to true will output the current SingleDelete
            valid_ = true;
@@ -382,7 +424,6 @@ void CompactionIterator::NextFromInput() {
      input_->Next();
    } else if (ikey_.type == kTypeMerge) {
      if (!merge_helper_->HasOperator()) {
-        LogToBuffer(log_buffer_, "Options::merge_operator is null.");
        status_ = Status::InvalidArgument(
            "merge_operator is not properly initialized.");
        return;
@@ -393,11 +434,14 @@ void CompactionIterator::NextFromInput() {
      // have hit (A)
      // We encapsulate the merge related state machine in a different
      // object to minimize change to the existing flow.
-      merge_helper_->MergeUntil(input_, range_del_agg_, prev_snapshot,
-                                bottommost_level_);
+      Status s = merge_helper_->MergeUntil(input_, range_del_agg_,
+                                           prev_snapshot, bottommost_level_);
      merge_out_iter_.SeekToFirst();

-      if (merge_out_iter_.Valid()) {
+      if (!s.ok() && !s.IsMergeInProgress()) {
+        status_ = s;
+        return;
+      } else if (merge_out_iter_.Valid()) {
        // NOTE: key, value, and ikey_ refer to old entries.
        //       These will be correctly set below.
        key_ = merge_out_iter_.key();
@@ -418,18 +462,32 @@ void CompactionIterator::NextFromInput() {
        // coming after the merges
        has_current_user_key_ = false;
        pinned_iters_mgr_.ReleasePinnedData();
+
+        if (merge_helper_->FilteredUntil(&skip_until)) {
+          need_skip = true;
+        }
      }
    } else {
      // 1. new user key -OR-
      // 2. different snapshot stripe
-      bool should_delete =
-          range_del_agg_->ShouldDelete(key_, true /* for_compaction */);
+      bool should_delete = range_del_agg_->ShouldDelete(
+          key_, RangeDelAggregator::RangePositioningMode::kForwardTraversal);
      if (should_delete) {
+        ++iter_stats_.num_record_drop_hidden;
+        ++iter_stats_.num_record_drop_range_del;
        input_->Next();
      } else {
        valid_ = true;
      }
    }
+
+    if (need_skip) {
+      input_->Seek(skip_until);
+    }
+  }
+
+  if (!valid_ && IsShuttingDown()) {
+    status_ = Status::ShutdownInProgress();
  }
 }

@@ -12,37 +12,46 @@
 #include <vector>

 #include "db/compaction.h"
+#include "db/compaction_iteration_stats.h"
 #include "db/merge_helper.h"
 #include "db/pinned_iterators_manager.h"
 #include "db/range_del_aggregator.h"
 #include "rocksdb/compaction_filter.h"
-#include "util/log_buffer.h"

 namespace rocksdb {

-struct CompactionIteratorStats {
-  // Compaction statistics
-  int64_t num_record_drop_user = 0;
-  int64_t num_record_drop_hidden = 0;
-  int64_t num_record_drop_obsolete = 0;
-  uint64_t total_filter_time = 0;
-
-  // Input statistics
-  // TODO(noetzli): The stats are incomplete. They are lacking everything
-  // consumed by MergeHelper.
-  uint64_t num_input_records = 0;
-  uint64_t num_input_deletion_records = 0;
-  uint64_t num_input_corrupt_records = 0;
-  uint64_t total_input_raw_key_bytes = 0;
-  uint64_t total_input_raw_value_bytes = 0;
-
-  // Single-Delete diagnostics for exceptional situations
-  uint64_t num_single_del_fallthru = 0;
-  uint64_t num_single_del_mismatch = 0;
-};
-
 class CompactionIterator {
 public:
+  // A wrapper around Compaction. Has a much smaller interface, only what
+  // CompactionIterator uses. Tests can override it.
+  class CompactionProxy {
+   public:
+    explicit CompactionProxy(const Compaction* compaction)
+        : compaction_(compaction) {}
+
+    virtual ~CompactionProxy() = default;
+    virtual int level(size_t compaction_input_level = 0) const {
+      return compaction_->level();
+    }
+    virtual bool KeyNotExistsBeyondOutputLevel(
+        const Slice& user_key, std::vector<size_t>* level_ptrs) const {
+      return compaction_->KeyNotExistsBeyondOutputLevel(user_key, level_ptrs);
+    }
+    virtual bool bottommost_level() const {
+      return compaction_->bottommost_level();
+    }
+    virtual int number_levels() const { return compaction_->number_levels(); }
+    virtual Slice GetLargestUserKey() const {
+      return compaction_->GetLargestUserKey();
+    }
+
+   protected:
+    CompactionProxy() = default;
+
+   private:
+    const Compaction* compaction_;
+  };
+
  CompactionIterator(InternalIterator* input, const Comparator* cmp,
                     MergeHelper* merge_helper, SequenceNumber last_sequence,
                     std::vector<SequenceNumber>* snapshots,
@@ -51,7 +60,18 @@ class CompactionIterator {
                     RangeDelAggregator* range_del_agg,
                     const Compaction* compaction = nullptr,
                     const CompactionFilter* compaction_filter = nullptr,
-                     LogBuffer* log_buffer = nullptr);
+                     const std::atomic<bool>* shutting_down = nullptr);
+
+  // Constructor with custom CompactionProxy, used for tests.
+  CompactionIterator(InternalIterator* input, const Comparator* cmp,
+                     MergeHelper* merge_helper, SequenceNumber last_sequence,
+                     std::vector<SequenceNumber>* snapshots,
+                     SequenceNumber earliest_write_conflict_snapshot, Env* env,
+                     bool expect_valid_internal_key,
+                     RangeDelAggregator* range_del_agg,
+                     std::unique_ptr<CompactionProxy> compaction,
+                     const CompactionFilter* compaction_filter = nullptr,
+                     const std::atomic<bool>* shutting_down = nullptr);

  ~CompactionIterator();

@@ -74,7 +94,7 @@ class CompactionIterator {
  const ParsedInternalKey& ikey() const { return ikey_; }
  bool Valid() const { return valid_; }
  const Slice& user_key() const { return current_user_key_; }
-  const CompactionIteratorStats& iter_stats() const { return iter_stats_; }
+  const CompactionIterationStats& iter_stats() const { return iter_stats_; }

 private:
  // Processes the input stream to find the next output
@@ -102,9 +122,9 @@ class CompactionIterator {
  Env* env_;
  bool expect_valid_internal_key_;
  RangeDelAggregator* range_del_agg_;
-  const Compaction* compaction_;
+  std::unique_ptr<CompactionProxy> compaction_;
  const CompactionFilter* compaction_filter_;
-  LogBuffer* log_buffer_;
+  const std::atomic<bool>* shutting_down_;
  bool bottommost_level_;
  bool valid_ = false;
  bool visible_at_tip_;
@@ -150,6 +170,7 @@ class CompactionIterator {
  // merge operands and then releasing them after consuming them.
  PinnedIteratorsManager pinned_iters_mgr_;
  std::string compaction_filter_value_;
+  InternalKey compaction_filter_skip_until_;
  // "level_ptrs" holds indices that remember which file of an associated
  // level we were last checking during the last call to compaction->
  // KeyNotExistsBeyondOutputLevel(). This allows future calls to the function
@@ -157,6 +178,11 @@ class CompactionIterator {
  // increasing so a later call to the function must be looking for a key that
  // is in or beyond the last file checked during the previous call
  std::vector<size_t> level_ptrs_;
-  CompactionIteratorStats iter_stats_;
+  CompactionIterationStats iter_stats_;
+
+  bool IsShuttingDown() {
+    // This is a best-effort facility, so memory_order_relaxed is sufficient.
+    return shutting_down_ && shutting_down_->load(std::memory_order_relaxed);
+  }
 };
 }  // namespace rocksdb
@@ -8,11 +8,156 @@
 #include <string>
 #include <vector>

+#include "port/port.h"
 #include "util/testharness.h"
 #include "util/testutil.h"

 namespace rocksdb {

+// Expects no merging attempts.
+class NoMergingMergeOp : public MergeOperator {
+ public:
+  bool FullMergeV2(const MergeOperationInput& merge_in,
+                   MergeOperationOutput* merge_out) const override {
+    ADD_FAILURE();
+    return false;
+  }
+  bool PartialMergeMulti(const Slice& key,
+                         const std::deque<Slice>& operand_list,
+                         std::string* new_value,
+                         Logger* logger) const override {
+    ADD_FAILURE();
+    return false;
+  }
+  const char* Name() const override {
+    return "CompactionIteratorTest NoMergingMergeOp";
+  }
+};
+
+// Compaction filter that gets stuck when it sees a particular key,
+// then gets unstuck when told to.
+// Always returns Decition::kRemove.
+class StallingFilter : public CompactionFilter {
+ public:
+  virtual Decision FilterV2(int level, const Slice& key, ValueType t,
+                            const Slice& existing_value, std::string* new_value,
+                            std::string* skip_until) const override {
+    int k = std::atoi(key.ToString().c_str());
+    last_seen.store(k);
+    while (k >= stall_at.load()) {
+      std::this_thread::yield();
+    }
+    return Decision::kRemove;
+  }
+
+  const char* Name() const override {
+    return "CompactionIteratorTest StallingFilter";
+  }
+
+  // Wait until the filter sees a key >= k and stalls at that key.
+  // If `exact`, asserts that the seen key is equal to k.
+  void WaitForStall(int k, bool exact = true) {
+    stall_at.store(k);
+    while (last_seen.load() < k) {
+      std::this_thread::yield();
+    }
+    if (exact) {
+      EXPECT_EQ(k, last_seen.load());
+    }
+  }
+
+  // Filter will stall on key >= stall_at. Advance stall_at to unstall.
+  mutable std::atomic<int> stall_at{0};
+  // Last key the filter was called with.
+  mutable std::atomic<int> last_seen{0};
+};
+
+class LoggingForwardVectorIterator : public InternalIterator {
+ public:
+  struct Action {
+    enum class Type {
+      SEEK_TO_FIRST,
+      SEEK,
+      NEXT,
+    };
+
+    Type type;
+    std::string arg;
+
+    explicit Action(Type _type, std::string _arg = "")
+        : type(_type), arg(_arg) {}
+
+    bool operator==(const Action& rhs) const {
+      return std::tie(type, arg) == std::tie(rhs.type, rhs.arg);
+    }
+  };
+
+  LoggingForwardVectorIterator(const std::vector<std::string>& keys,
+                               const std::vector<std::string>& values)
+      : keys_(keys), values_(values), current_(keys.size()) {
+    assert(keys_.size() == values_.size());
+  }
+
+  virtual bool Valid() const override { return current_ < keys_.size(); }
+
+  virtual void SeekToFirst() override {
+    log.emplace_back(Action::Type::SEEK_TO_FIRST);
+    current_ = 0;
+  }
+  virtual void SeekToLast() override { assert(false); }
+
+  virtual void Seek(const Slice& target) override {
+    log.emplace_back(Action::Type::SEEK, target.ToString());
+    current_ = std::lower_bound(keys_.begin(), keys_.end(), target.ToString()) -
+               keys_.begin();
+  }
+
+  virtual void SeekForPrev(const Slice& target) override { assert(false); }
+
+  virtual void Next() override {
+    assert(Valid());
+    log.emplace_back(Action::Type::NEXT);
+    current_++;
+  }
+  virtual void Prev() override { assert(false); }
+
+  virtual Slice key() const override {
+    assert(Valid());
+    return Slice(keys_[current_]);
+  }
+  virtual Slice value() const override {
+    assert(Valid());
+    return Slice(values_[current_]);
+  }
+
+  virtual Status status() const override { return Status::OK(); }
+
+  std::vector<Action> log;
+
+ private:
+  std::vector<std::string> keys_;
+  std::vector<std::string> values_;
+  size_t current_;
+};
+
+class FakeCompaction : public CompactionIterator::CompactionProxy {
+ public:
+  FakeCompaction() = default;
+
+  virtual int level(size_t compaction_input_level) const { return 0; }
+  virtual bool KeyNotExistsBeyondOutputLevel(
+      const Slice& user_key, std::vector<size_t>* level_ptrs) const {
+    return key_not_exists_beyond_output_level;
+  }
+  virtual bool bottommost_level() const { return false; }
+  virtual int number_levels() const { return 1; }
+  virtual Slice GetLargestUserKey() const {
+    return "\xff\xff\xff\xff\xff\xff\xff\xff\xff";
+  }
+
+  bool key_not_exists_beyond_output_level = false;
+};
+
 class CompactionIteratorTest : public testing::Test {
 public:
  CompactionIteratorTest()
@@ -22,19 +167,29 @@ class CompactionIteratorTest : public testing::Test {
                     const std::vector<std::string>& vs,
                     const std::vector<std::string>& range_del_ks,
                     const std::vector<std::string>& range_del_vs,
-                     SequenceNumber last_sequence) {
+                     SequenceNumber last_sequence,
+                     MergeOperator* merge_op = nullptr,
+                     CompactionFilter* filter = nullptr) {
    std::unique_ptr<InternalIterator> range_del_iter(
        new test::VectorIterator(range_del_ks, range_del_vs));
    range_del_agg_.reset(new RangeDelAggregator(icmp_, snapshots_));
    ASSERT_OK(range_del_agg_->AddTombstones(std::move(range_del_iter)));

-    merge_helper_.reset(new MergeHelper(Env::Default(), cmp_, nullptr, nullptr,
-                                        nullptr, 0U, false, 0));
-    iter_.reset(new test::VectorIterator(ks, vs));
+    std::unique_ptr<CompactionIterator::CompactionProxy> compaction;
+    if (filter) {
+      compaction_proxy_ = new FakeCompaction();
+      compaction.reset(compaction_proxy_);
+    }
+
+    merge_helper_.reset(new MergeHelper(Env::Default(), cmp_, merge_op, filter,
+                                        nullptr, false, 0, 0, nullptr,
+                                        &shutting_down_));
+    iter_.reset(new LoggingForwardVectorIterator(ks, vs));
    iter_->SeekToFirst();
    c_iter_.reset(new CompactionIterator(
        iter_.get(), cmp_, merge_helper_.get(), last_sequence, &snapshots_,
-        kMaxSequenceNumber, Env::Default(), false, range_del_agg_.get()));
+        kMaxSequenceNumber, Env::Default(), false, range_del_agg_.get(),
+        std::move(compaction), filter, &shutting_down_));
  }

  void AddSnapshot(SequenceNumber snapshot) { snapshots_.push_back(snapshot); }
@@ -43,9 +198,11 @@ class CompactionIteratorTest : public testing::Test {
  const InternalKeyComparator icmp_;
  std::vector<SequenceNumber> snapshots_;
  std::unique_ptr<MergeHelper> merge_helper_;
-  std::unique_ptr<test::VectorIterator> iter_;
+  std::unique_ptr<LoggingForwardVectorIterator> iter_;
  std::unique_ptr<CompactionIterator> c_iter_;
  std::unique_ptr<RangeDelAggregator> range_del_agg_;
+  std::atomic<bool> shutting_down_{false};
+  FakeCompaction* compaction_proxy_;
 };

 // It is possible that the output of the compaction iterator is empty even if
@@ -116,6 +273,187 @@ TEST_F(CompactionIteratorTest, RangeDeletionWithSnapshots) {
  ASSERT_FALSE(c_iter_->Valid());
 }

+TEST_F(CompactionIteratorTest, CompactionFilterSkipUntil) {
+  class Filter : public CompactionFilter {
+    virtual Decision FilterV2(int level, const Slice& key, ValueType t,
+                              const Slice& existing_value,
+                              std::string* new_value,
+                              std::string* skip_until) const override {
+      std::string k = key.ToString();
+      std::string v = existing_value.ToString();
+      // See InitIterators() call below for the sequence of keys and their
+      // filtering decisions. Here we closely assert that compaction filter is
+      // called with the expected keys and only them, and with the right values.
+      if (k == "a") {
+        EXPECT_EQ(ValueType::kValue, t);
+        EXPECT_EQ("av50", v);
+        return Decision::kKeep;
+      }
+      if (k == "b") {
+        EXPECT_EQ(ValueType::kValue, t);
+        EXPECT_EQ("bv60", v);
+        *skip_until = "d+";
+        return Decision::kRemoveAndSkipUntil;
+      }
+      if (k == "e") {
+        EXPECT_EQ(ValueType::kMergeOperand, t);
+        EXPECT_EQ("em71", v);
+        return Decision::kKeep;
+      }
+      if (k == "f") {
+        if (v == "fm65") {
+          EXPECT_EQ(ValueType::kMergeOperand, t);
+          *skip_until = "f";
+        } else {
+          EXPECT_EQ("fm30", v);
+          EXPECT_EQ(ValueType::kMergeOperand, t);
+          *skip_until = "g+";
+        }
+        return Decision::kRemoveAndSkipUntil;
+      }
+      if (k == "h") {
+        EXPECT_EQ(ValueType::kValue, t);
+        EXPECT_EQ("hv91", v);
+        return Decision::kKeep;
+      }
+      if (k == "i") {
+        EXPECT_EQ(ValueType::kMergeOperand, t);
+        EXPECT_EQ("im95", v);
+        *skip_until = "z";
+        return Decision::kRemoveAndSkipUntil;
+      }
+      ADD_FAILURE();
+      return Decision::kKeep;
+    }
+
+    const char* Name() const override {
+      return "CompactionIteratorTest.CompactionFilterSkipUntil::Filter";
+    }
+  };
+
+  NoMergingMergeOp merge_op;
+  Filter filter;
+  InitIterators(
+      {test::KeyStr("a", 50, kTypeValue),  // keep
+       test::KeyStr("a", 45, kTypeMerge),
+       test::KeyStr("b", 60, kTypeValue),  // skip to "d+"
+       test::KeyStr("b", 40, kTypeValue), test::KeyStr("c", 35, kTypeValue),
+       test::KeyStr("d", 70, kTypeMerge),
+       test::KeyStr("e", 71, kTypeMerge),  // keep
+       test::KeyStr("f", 65, kTypeMerge),  // skip to "f", aka keep
+       test::KeyStr("f", 30, kTypeMerge),  // skip to "g+"
+       test::KeyStr("f", 25, kTypeValue), test::KeyStr("g", 90, kTypeValue),
+       test::KeyStr("h", 91, kTypeValue),  // keep
+       test::KeyStr("i", 95, kTypeMerge),  // skip to "z"
+       test::KeyStr("j", 99, kTypeValue)},
+      {"av50", "am45", "bv60", "bv40", "cv35", "dm70", "em71", "fm65", "fm30",
+       "fv25", "gv90", "hv91", "im95", "jv99"},
+      {}, {}, kMaxSequenceNumber, &merge_op, &filter);
+
+  // Compaction should output just "a", "e" and "h" keys.
+  c_iter_->SeekToFirst();
+  ASSERT_TRUE(c_iter_->Valid());
+  ASSERT_EQ(test::KeyStr("a", 50, kTypeValue), c_iter_->key().ToString());
+  ASSERT_EQ("av50", c_iter_->value().ToString());
+  c_iter_->Next();
+  ASSERT_TRUE(c_iter_->Valid());
+  ASSERT_EQ(test::KeyStr("e", 71, kTypeMerge), c_iter_->key().ToString());
+  ASSERT_EQ("em71", c_iter_->value().ToString());
+  c_iter_->Next();
+  ASSERT_TRUE(c_iter_->Valid());
+  ASSERT_EQ(test::KeyStr("h", 91, kTypeValue), c_iter_->key().ToString());
+  ASSERT_EQ("hv91", c_iter_->value().ToString());
+  c_iter_->Next();
+  ASSERT_FALSE(c_iter_->Valid());
+
+  // Check that the compaction iterator did the correct sequence of calls on
+  // the underlying iterator.
+  using A = LoggingForwardVectorIterator::Action;
+  using T = A::Type;
+  std::vector<A> expected_actions = {
+      A(T::SEEK_TO_FIRST),
+      A(T::NEXT),
+      A(T::NEXT),
+      A(T::SEEK, test::KeyStr("d+", kMaxSequenceNumber, kValueTypeForSeek)),
+      A(T::NEXT),
+      A(T::NEXT),
+      A(T::SEEK, test::KeyStr("g+", kMaxSequenceNumber, kValueTypeForSeek)),
+      A(T::NEXT),
+      A(T::SEEK, test::KeyStr("z", kMaxSequenceNumber, kValueTypeForSeek))};
+  ASSERT_EQ(expected_actions, iter_->log);
+}
+
+TEST_F(CompactionIteratorTest, ShuttingDownInFilter) {
+  NoMergingMergeOp merge_op;
+  StallingFilter filter;
+  InitIterators(
+      {test::KeyStr("1", 1, kTypeValue), test::KeyStr("2", 2, kTypeValue),
+       test::KeyStr("3", 3, kTypeValue), test::KeyStr("4", 4, kTypeValue)},
+      {"v1", "v2", "v3", "v4"}, {}, {}, kMaxSequenceNumber, &merge_op, &filter);
+  // Don't leave tombstones (kTypeDeletion) for filtered keys.
+  compaction_proxy_->key_not_exists_beyond_output_level = true;
+
+  std::atomic<bool> seek_done{false};
+  rocksdb::port::Thread compaction_thread([&] {
+    c_iter_->SeekToFirst();
+    EXPECT_FALSE(c_iter_->Valid());
+    EXPECT_TRUE(c_iter_->status().IsShutdownInProgress());
+    seek_done.store(true);
+  });
+
+  // Let key 1 through.
+  filter.WaitForStall(1);
+
+  // Shutdown during compaction filter call for key 2.
+  filter.WaitForStall(2);
+  shutting_down_.store(true);
+  EXPECT_FALSE(seek_done.load());
+
+  // Unstall filter and wait for SeekToFirst() to return.
+  filter.stall_at.store(3);
+  compaction_thread.join();
+  assert(seek_done.load());
+
+  // Check that filter was never called again.
+  EXPECT_EQ(2, filter.last_seen.load());
+}
+
+// Same as ShuttingDownInFilter, but shutdown happens during filter call for
+// a merge operand, not for a value.
+TEST_F(CompactionIteratorTest, ShuttingDownInMerge) {
+  NoMergingMergeOp merge_op;
+  StallingFilter filter;
+  InitIterators(
+      {test::KeyStr("1", 1, kTypeValue), test::KeyStr("2", 2, kTypeMerge),
+       test::KeyStr("3", 3, kTypeMerge), test::KeyStr("4", 4, kTypeValue)},
+      {"v1", "v2", "v3", "v4"}, {}, {}, kMaxSequenceNumber, &merge_op, &filter);
+  compaction_proxy_->key_not_exists_beyond_output_level = true;
+
+  std::atomic<bool> seek_done{false};
+  rocksdb::port::Thread compaction_thread([&] {
+    c_iter_->SeekToFirst();
+    ASSERT_FALSE(c_iter_->Valid());
+    ASSERT_TRUE(c_iter_->status().IsShutdownInProgress());
+    seek_done.store(true);
+  });
+
+  // Let key 1 through.
+  filter.WaitForStall(1);
+
+  // Shutdown during compaction filter call for key 2.
+  filter.WaitForStall(2);
+  shutting_down_.store(true);
+  EXPECT_FALSE(seek_done.load());
+
+  // Unstall filter and wait for SeekToFirst() to return.
+  filter.stall_at.store(3);
+  compaction_thread.join();
+  assert(seek_done.load());
+
+  // Check that filter was never called again.
+  EXPECT_EQ(2, filter.last_seen.load());
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -45,7 +45,7 @@
 #include "rocksdb/table.h"
 #include "table/block.h"
 #include "table/block_based_table_factory.h"
-#include "table/merger.h"
+#include "table/merging_iterator.h"
 #include "table/table_builder.h"
 #include "util/coding.h"
 #include "util/file_reader_writer.h"
@@ -264,7 +264,7 @@ void CompactionJob::AggregateStatistics() {
 CompactionJob::CompactionJob(
    int job_id, Compaction* compaction, const ImmutableDBOptions& db_options,
    const EnvOptions& env_options, VersionSet* versions,
-    std::atomic<bool>* shutting_down, LogBuffer* log_buffer,
+    const std::atomic<bool>* shutting_down, LogBuffer* log_buffer,
    Directory* db_directory, Directory* output_directory, Statistics* stats,
    InstrumentedMutex* db_mutex, Status* db_bg_error,
    std::vector<SequenceNumber> existing_snapshots,
@@ -484,9 +484,8 @@ void CompactionJob::GenSubcompactionBoundaries() {
                static_cast<uint64_t>(db_options_.max_subcompactions),
                max_output_files});

-  double mean = sum * 1.0 / subcompactions;
-
  if (subcompactions > 1) {
+    double mean = sum * 1.0 / subcompactions;
    // Greedily add ranges to the subcompaction until the sum of the ranges'
    // sizes becomes >= the expected mean size of a subcompaction
    sum = 0;
@@ -523,7 +522,7 @@ Status CompactionJob::Run() {
  const uint64_t start_micros = env_->NowMicros();

  // Launch a thread for each of subcompactions 1...num_threads-1
-  std::vector<std::thread> thread_pool;
+  std::vector<port::Thread> thread_pool;
  thread_pool.reserve(num_threads - 1);
  for (size_t i = 1; i < compact_->sub_compact_states.size(); i++) {
    thread_pool.emplace_back(&CompactionJob::ProcessKeyValueCompaction, this,
@@ -539,7 +538,7 @@ Status CompactionJob::Run() {
    thread.join();
  }

-  if (output_directory_ && !db_options_.disable_data_sync) {
+  if (output_directory_) {
    output_directory_->Fsync();
  }

@@ -591,6 +590,16 @@ Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options) {
  VersionStorageInfo::LevelSummaryStorage tmp;
  auto vstorage = cfd->current()->storage_info();
  const auto& stats = compaction_stats_;
+
+  double read_write_amp = 0.0;
+  double write_amp = 0.0;
+  if (stats.bytes_read_non_output_levels > 0) {
+    read_write_amp = (stats.bytes_written + stats.bytes_read_output_level +
+                      stats.bytes_read_non_output_levels) /
+                     static_cast<double>(stats.bytes_read_non_output_levels);
+    write_amp = stats.bytes_written /
+                static_cast<double>(stats.bytes_read_non_output_levels);
+  }
  LogToBuffer(
      log_buffer_,
      "[%s] compacted to: %s, MB/sec: %.1f rd, %.1f wr, level %d, "
@@ -603,16 +612,10 @@ Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options) {
      stats.bytes_written / static_cast<double>(stats.micros),
      compact_->compaction->output_level(),
      stats.num_input_files_in_non_output_levels,
-      stats.num_input_files_in_output_level,
-      stats.num_output_files,
+      stats.num_input_files_in_output_level, stats.num_output_files,
      stats.bytes_read_non_output_levels / 1048576.0,
      stats.bytes_read_output_level / 1048576.0,
-      stats.bytes_written / 1048576.0,
-      (stats.bytes_written + stats.bytes_read_output_level +
-       stats.bytes_read_non_output_levels) /
-          static_cast<double>(stats.bytes_read_non_output_levels),
-      stats.bytes_written /
-          static_cast<double>(stats.bytes_read_non_output_levels),
+      stats.bytes_written / 1048576.0, read_write_amp, write_amp,
      status.ToString().c_str(), stats.num_input_records,
      stats.num_dropped_records);

@@ -718,10 +721,10 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
  MergeHelper merge(
      env_, cfd->user_comparator(), cfd->ioptions()->merge_operator,
      compaction_filter, db_options_.info_log.get(),
-      mutable_cf_options->min_partial_merge_operands,
      false /* internal key corruption is expected */,
      existing_snapshots_.empty() ? 0 : existing_snapshots_.back(),
-      compact_->compaction->level(), db_options_.statistics.get());
+      compact_->compaction->level(), db_options_.statistics.get(),
+      shutting_down_);

  TEST_SYNC_POINT("CompactionJob::Run():Inprogress");

@@ -739,9 +742,18 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
  sub_compact->c_iter.reset(new CompactionIterator(
      input.get(), cfd->user_comparator(), &merge, versions_->LastSequence(),
      &existing_snapshots_, earliest_write_conflict_snapshot_, env_, false,
-      range_del_agg.get(), sub_compact->compaction, compaction_filter));
+      range_del_agg.get(), sub_compact->compaction, compaction_filter,
+      shutting_down_));
  auto c_iter = sub_compact->c_iter.get();
  c_iter->SeekToFirst();
+  if (c_iter->Valid() &&
+      sub_compact->compaction->output_level() != 0) {
+    // ShouldStopBefore() maintains state based on keys processed so far. The
+    // compaction loop always calls it on the "next" key, thus won't tell it the
+    // first key. So we do that here.
+    sub_compact->ShouldStopBefore(
+      c_iter->key(), sub_compact->current_output_file_size);
+  }
  const auto& c_iter_stats = c_iter->iter_stats();
  auto sample_begin_offset_iter = sample_begin_offsets.cbegin();
  // data_begin_offset and compression_dict are only valid while generating
@@ -750,10 +762,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
  std::string compression_dict;
  compression_dict.reserve(cfd->ioptions()->compression_opts.max_dict_bytes);

-  // TODO(noetzli): check whether we could check !shutting_down_->... only
-  // only occasionally (see diff D42687)
-  while (status.ok() && !shutting_down_->load(std::memory_order_acquire) &&
-         !cfd->IsDropped() && c_iter->Valid()) {
+  while (status.ok() && !cfd->IsDropped() && c_iter->Valid()) {
    // Invariant: c_iter.status() is guaranteed to be OK if c_iter->Valid()
    // returns true.
    const Slice& key = c_iter->key();
@@ -764,17 +773,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
    if (end != nullptr &&
        cfd->user_comparator()->Compare(c_iter->user_key(), *end) >= 0) {
      break;
-    } else if (sub_compact->compaction->output_level() != 0 &&
-               sub_compact->ShouldStopBefore(
-                   key, sub_compact->current_output_file_size) &&
-               sub_compact->builder != nullptr) {
-      status = FinishCompactionOutputFile(input->status(), sub_compact,
-                                          range_del_agg.get());
-      if (!status.ok()) {
-        break;
-      }
    }
-
    if (c_iter_stats.num_input_records % kRecordStatsEvery ==
        kRecordStatsEvery - 1) {
      RecordDroppedKeys(c_iter_stats, &sub_compact->compaction_job_stats);
@@ -846,23 +845,47 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
      }
    }

-    Status input_status = input->status();
-    c_iter->Next();
-
-    // Close output file if it is big enough
+    // Close output file if it is big enough. Two possibilities determine it's
+    // time to close it: (1) the current key should be this file's last key, (2)
+    // the next key should not be in this file.
+    //
    // TODO(aekmekji): determine if file should be closed earlier than this
    // during subcompactions (i.e. if output size, estimated by input size, is
    // going to be 1.2MB and max_output_file_size = 1MB, prefer to have 0.6MB
    // and 0.6MB instead of 1MB and 0.2MB)
+    bool output_file_ended = false;
+    Status input_status;
    if (sub_compact->compaction->output_level() != 0 &&
        sub_compact->current_output_file_size >=
            sub_compact->compaction->max_output_file_size()) {
+      // (1) this key terminates the file. For historical reasons, the iterator
+      // status before advancing will be given to FinishCompactionOutputFile().
+      input_status = input->status();
+      output_file_ended = true;
+    }
+    c_iter->Next();
+    if (!output_file_ended && c_iter->Valid() &&
+        sub_compact->compaction->output_level() != 0 &&
+        sub_compact->ShouldStopBefore(
+          c_iter->key(), sub_compact->current_output_file_size) &&
+        sub_compact->builder != nullptr) {
+      // (2) this key belongs to the next file. For historical reasons, the
+      // iterator status after advancing will be given to
+      // FinishCompactionOutputFile().
+      input_status = input->status();
+      output_file_ended = true;
+    }
+    if (output_file_ended) {
      const Slice* next_key = nullptr;
      if (c_iter->Valid()) {
        next_key = &c_iter->key();
      }
+      CompactionIterationStats range_del_out_stats;
      status = FinishCompactionOutputFile(input_status, sub_compact,
-                                          range_del_agg.get(), next_key);
+                                          range_del_agg.get(),
+                                          &range_del_out_stats, next_key);
+      RecordDroppedKeys(range_del_out_stats,
+                        &sub_compact->compaction_job_stats);
      if (sub_compact->outputs.size() == 1) {
        // Use dictionary from first output file for compression of subsequent
        // files.
@@ -890,22 +913,36 @@ 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_acquire) || cfd->IsDropped())) {
+  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() && sub_compact->builder == nullptr &&
-      range_del_agg->ShouldAddTombstones(bottommost_level_)) {
-    status = OpenCompactionOutputFile(sub_compact);
-  }
-  if (status.ok() && sub_compact->builder != nullptr) {
-    status = FinishCompactionOutputFile(input->status(), sub_compact,
-                                        range_del_agg.get());
-  }
  if (status.ok()) {
    status = input->status();
  }
+  if (status.ok()) {
+    status = c_iter->status();
+  }
+
+  if (status.ok() && sub_compact->builder == nullptr &&
+      sub_compact->outputs.size() == 0 &&
+      range_del_agg->ShouldAddTombstones(bottommost_level_)) {
+    // handle subcompaction containing only range deletions
+    status = OpenCompactionOutputFile(sub_compact);
+  }
+
+  // Call FinishCompactionOutputFile() even if status is not ok: it needs to
+  // close the output file.
+  if (sub_compact->builder != nullptr) {
+    CompactionIterationStats range_del_out_stats;
+    Status s = FinishCompactionOutputFile(
+        status, sub_compact, range_del_agg.get(), &range_del_out_stats);
+    if (status.ok()) {
+      status = s;
+    }
+    RecordDroppedKeys(range_del_out_stats, &sub_compact->compaction_job_stats);
+  }

  if (measure_io_stats_) {
    sub_compact->compaction_job_stats.file_write_nanos +=
@@ -927,7 +964,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
 }

 void CompactionJob::RecordDroppedKeys(
-    const CompactionIteratorStats& c_iter_stats,
+    const CompactionIterationStats& c_iter_stats,
    CompactionJobStats* compaction_job_stats) {
  if (c_iter_stats.num_record_drop_user > 0) {
    RecordTick(stats_, COMPACTION_KEY_DROP_USER,
@@ -949,11 +986,20 @@ void CompactionJob::RecordDroppedKeys(
          c_iter_stats.num_record_drop_obsolete;
    }
  }
+  if (c_iter_stats.num_record_drop_range_del > 0) {
+    RecordTick(stats_, COMPACTION_KEY_DROP_RANGE_DEL,
+               c_iter_stats.num_record_drop_range_del);
+  }
+  if (c_iter_stats.num_range_del_drop_obsolete > 0) {
+    RecordTick(stats_, COMPACTION_RANGE_DEL_DROP_OBSOLETE,
+               c_iter_stats.num_range_del_drop_obsolete);
+  }
 }

 Status CompactionJob::FinishCompactionOutputFile(
    const Status& input_status, SubcompactionState* sub_compact,
    RangeDelAggregator* range_del_agg,
+    CompactionIterationStats* range_del_out_stats,
    const Slice* next_table_min_key /* = nullptr */) {
  AutoThreadOperationStageUpdater stage_updater(
      ThreadStatus::STAGE_COMPACTION_SYNC_FILE);
@@ -970,14 +1016,34 @@ Status CompactionJob::FinishCompactionOutputFile(
  Status s = input_status;
  auto meta = &sub_compact->current_output()->meta;
  if (s.ok()) {
-    // For the first output table, include range tombstones before the min key
-    // boundary. For subsequent output tables, this is unnecessary because we
-    // extend each file's max key boundary up until the next file's min key when
-    // range tombstones fall in the gap.
-    range_del_agg->AddToBuilder(
-        sub_compact->builder.get(),
-        sub_compact->outputs.size() == 1 /* extend_before_min_key */,
-        next_table_min_key, meta, bottommost_level_);
+    Slice lower_bound_guard, upper_bound_guard;
+    const Slice *lower_bound, *upper_bound;
+    if (sub_compact->outputs.size() == 1) {
+      // For the first output table, include range tombstones before the min key
+      // but after the subcompaction boundary.
+      lower_bound = sub_compact->start;
+    } else if (meta->smallest.size() > 0) {
+      // For subsequent output tables, only include range tombstones from min
+      // key onwards since the previous file was extended to contain range
+      // tombstones falling before min key.
+      lower_bound_guard = meta->smallest.user_key();
+      lower_bound = &lower_bound_guard;
+    } else {
+      lower_bound = nullptr;
+    }
+    if (next_table_min_key != nullptr) {
+      // This isn't the last file in the subcompaction, so extend until the next
+      // file starts.
+      upper_bound_guard = ExtractUserKey(*next_table_min_key);
+      upper_bound = &upper_bound_guard;
+    } else {
+      // This is the last file in the subcompaction, so extend until the
+      // subcompaction ends.
+      upper_bound = sub_compact->end;
+    }
+    range_del_agg->AddToBuilder(sub_compact->builder.get(), lower_bound,
+                                upper_bound, meta, range_del_out_stats,
+                                bottommost_level_);
  }
  const uint64_t current_entries = sub_compact->builder->NumEntries();
  meta->marked_for_compaction = sub_compact->builder->NeedCompact();
@@ -992,7 +1058,7 @@ Status CompactionJob::FinishCompactionOutputFile(
  sub_compact->total_bytes += current_bytes;

  // Finish and check for file errors
-  if (s.ok() && !db_options_.disable_data_sync) {
+  if (s.ok()) {
    StopWatch sw(env_, stats_, COMPACTION_OUTFILE_SYNC_MICROS);
    s = sub_compact->outfile->Sync(db_options_.use_fsync);
  }
@@ -1007,8 +1073,9 @@ Status CompactionJob::FinishCompactionOutputFile(
    // Verify that the table is usable
    InternalIterator* iter = cfd->table_cache()->NewIterator(
        ReadOptions(), env_options_, cfd->internal_comparator(), meta->fd,
-        nullptr, cfd->internal_stats()->GetFileReadHist(
-                     compact_->compaction->output_level()),
+        nullptr /* range_del_agg */, nullptr,
+        cfd->internal_stats()->GetFileReadHist(
+            compact_->compaction->output_level()),
        false);
    s = iter->status();

@@ -1038,6 +1105,7 @@ Status CompactionJob::FinishCompactionOutputFile(
      event_logger_, cfd->ioptions()->listeners, dbname_, cfd->GetName(), fname,
      job_id_, meta->fd, tp, TableFileCreationReason::kCompaction, s);

+#ifndef ROCKSDB_LITE
  // Report new file to SstFileManagerImpl
  auto sfm =
      static_cast<SstFileManagerImpl*>(db_options_.sst_file_manager.get());
@@ -1055,6 +1123,7 @@ Status CompactionJob::FinishCompactionOutputFile(
      }
    }
  }
+#endif

  sub_compact->builder.reset();
  sub_compact->current_output_file_size = 0;
@@ -1153,8 +1222,8 @@ Status CompactionJob::OpenCompactionOutputFile(
  writable_file->SetIOPriority(Env::IO_LOW);
  writable_file->SetPreallocationBlockSize(static_cast<size_t>(
      sub_compact->compaction->OutputFilePreallocationSize()));
-  sub_compact->outfile.reset(
-      new WritableFileWriter(std::move(writable_file), env_options_));
+  sub_compact->outfile.reset(new WritableFileWriter(
+      std::move(writable_file), env_options_, db_options_.statistics.get()));

  // If the Column family flag is to only optimize filters for hits,
  // we can skip creating filters if this is the bottommost_level where
@@ -1215,8 +1284,7 @@ void CompactionJob::UpdateCompactionStats() {
  for (int input_level = 0;
       input_level < static_cast<int>(compaction->num_input_levels());
       ++input_level) {
-    if (compaction->start_level() + input_level
-        != compaction->output_level()) {
+    if (compaction->level(input_level) != compaction->output_level()) {
      UpdateCompactionInputStatsHelper(
          &compaction_stats_.num_input_files_in_non_output_levels,
          &compaction_stats_.bytes_read_non_output_levels,
@@ -57,7 +57,7 @@ class CompactionJob {
  CompactionJob(int job_id, Compaction* compaction,
                const ImmutableDBOptions& db_options,
                const EnvOptions& env_options, VersionSet* versions,
-                std::atomic<bool>* shutting_down, LogBuffer* log_buffer,
+                const std::atomic<bool>* shutting_down, LogBuffer* log_buffer,
                Directory* db_directory, Directory* output_directory,
                Statistics* stats, InstrumentedMutex* db_mutex,
                Status* db_bg_error,
@@ -96,17 +96,18 @@ class CompactionJob {
  // kv-pairs
  void ProcessKeyValueCompaction(SubcompactionState* sub_compact);

-  Status FinishCompactionOutputFile(const Status& input_status,
-                                    SubcompactionState* sub_compact,
-                                    RangeDelAggregator* range_del_agg = nullptr,
-                                    const Slice* next_table_min_key = nullptr);
+  Status FinishCompactionOutputFile(
+      const Status& input_status, SubcompactionState* sub_compact,
+      RangeDelAggregator* range_del_agg,
+      CompactionIterationStats* range_del_out_stats,
+      const Slice* next_table_min_key = nullptr);
  Status InstallCompactionResults(const MutableCFOptions& mutable_cf_options);
  void RecordCompactionIOStats();
  Status OpenCompactionOutputFile(SubcompactionState* sub_compact);
  void CleanupCompaction();
  void UpdateCompactionJobStats(
    const InternalStats::CompactionStats& stats) const;
-  void RecordDroppedKeys(const CompactionIteratorStats& c_iter_stats,
+  void RecordDroppedKeys(const CompactionIterationStats& c_iter_stats,
                         CompactionJobStats* compaction_job_stats = nullptr);

  void UpdateCompactionStats();
@@ -130,7 +131,7 @@ class CompactionJob {

  Env* env_;
  VersionSet* versions_;
-  std::atomic<bool>* shutting_down_;
+  const std::atomic<bool>* shutting_down_;
  LogBuffer* log_buffer_;
  Directory* db_directory_;
  Directory* output_directory_;
@@ -61,7 +61,6 @@
 #include "util/testharness.h"
 #include "util/testutil.h"
 #include "util/thread_status_util.h"
-#include "util/xfunc.h"
 #include "utilities/merge_operators.h"

 #if !defined(IOS_CROSS_COMPILE)
@@ -18,7 +18,6 @@
 #include <queue>
 #include <string>
 #include <utility>
-
 #include "db/column_family.h"
 #include "db/filename.h"
 #include "util/log_buffer.h"
@@ -451,13 +450,14 @@ bool CompactionPicker::SetupOtherInputs(
  vstorage->GetOverlappingInputs(output_level, &smallest, &largest,
                                 &output_level_inputs->files, *parent_index,
                                 parent_index);
-  if (!output_level_inputs->empty()) {
-    ExpandWhileOverlapping(cf_name, vstorage, output_level_inputs);
-  }
-
  if (FilesInCompaction(output_level_inputs->files)) {
    return false;
  }
+  if (!output_level_inputs->empty()) {
+    if (!ExpandWhileOverlapping(cf_name, vstorage, output_level_inputs)) {
+      return false;
+    }
+  }

  // See if we can further grow the number of inputs in "level" without
  // changing the number of "level+1" files we pick up. We also choose NOT
@@ -465,45 +465,66 @@ bool CompactionPicker::SetupOtherInputs(
  // user key, while excluding other entries for the same user key. This
  // can happen when one user key spans multiple files.
  if (!output_level_inputs->empty()) {
-    CompactionInputFiles expanded0;
-    expanded0.level = input_level;
-    // Get entire range covered by compaction
+    const uint64_t limit = mutable_cf_options.max_compaction_bytes;
+    const uint64_t output_level_inputs_size =
+        TotalCompensatedFileSize(output_level_inputs->files);
+    const uint64_t inputs_size = TotalCompensatedFileSize(inputs->files);
+    bool expand_inputs = false;
+
+    CompactionInputFiles expanded_inputs;
+    expanded_inputs.level = input_level;
+    // Get closed interval of output level
    InternalKey all_start, all_limit;
    GetRange(*inputs, *output_level_inputs, &all_start, &all_limit);
-
+    bool try_overlapping_inputs = true;
    vstorage->GetOverlappingInputs(input_level, &all_start, &all_limit,
-                                   &expanded0.files, base_index, nullptr);
-    const uint64_t inputs0_size = TotalCompensatedFileSize(inputs->files);
-    const uint64_t inputs1_size =
-        TotalCompensatedFileSize(output_level_inputs->files);
-    const uint64_t expanded0_size = TotalCompensatedFileSize(expanded0.files);
-    uint64_t limit = mutable_cf_options.max_compaction_bytes;
-    if (expanded0.size() > inputs->size() &&
-        inputs1_size + expanded0_size < limit &&
-        !FilesInCompaction(expanded0.files) &&
-        !vstorage->HasOverlappingUserKey(&expanded0.files, input_level)) {
+                                   &expanded_inputs.files, base_index, nullptr);
+    uint64_t expanded_inputs_size =
+        TotalCompensatedFileSize(expanded_inputs.files);
+    if (!ExpandWhileOverlapping(cf_name, vstorage, &expanded_inputs)) {
+      try_overlapping_inputs = false;
+    }
+    if (try_overlapping_inputs && expanded_inputs.size() > inputs->size() &&
+        output_level_inputs_size + expanded_inputs_size < limit &&
+        !FilesInCompaction(expanded_inputs.files)) {
      InternalKey new_start, new_limit;
-      GetRange(expanded0, &new_start, &new_limit);
-      std::vector<FileMetaData*> expanded1;
+      GetRange(expanded_inputs, &new_start, &new_limit);
+      CompactionInputFiles expanded_output_level_inputs;
+      expanded_output_level_inputs.level = output_level;
      vstorage->GetOverlappingInputs(output_level, &new_start, &new_limit,
-                                     &expanded1, *parent_index, parent_index);
-      if (expanded1.size() == output_level_inputs->size() &&
-          !FilesInCompaction(expanded1)) {
-        Log(InfoLogLevel::INFO_LEVEL, ioptions_.info_log,
-            "[%s] Expanding@%d %" ROCKSDB_PRIszt "+%" ROCKSDB_PRIszt "(%" PRIu64
-            "+%" PRIu64 " bytes) to %" ROCKSDB_PRIszt "+%" ROCKSDB_PRIszt
-            " (%" PRIu64 "+%" PRIu64 "bytes)\n",
-            cf_name.c_str(), input_level, inputs->size(),
-            output_level_inputs->size(), inputs0_size, inputs1_size,
-            expanded0.size(), expanded1.size(), expanded0_size, inputs1_size);
-        smallest = new_start;
-        largest = new_limit;
-        inputs->files = expanded0.files;
-        output_level_inputs->files = expanded1;
+                                     &expanded_output_level_inputs.files,
+                                     *parent_index, parent_index);
+      assert(!expanded_output_level_inputs.empty());
+      if (!FilesInCompaction(expanded_output_level_inputs.files) &&
+          ExpandWhileOverlapping(cf_name, vstorage,
+                                 &expanded_output_level_inputs) &&
+          expanded_output_level_inputs.size() == output_level_inputs->size()) {
+        expand_inputs = true;
      }
    }
+    if (!expand_inputs) {
+      vstorage->GetCleanInputsWithinInterval(input_level, &all_start,
+                                             &all_limit, &expanded_inputs.files,
+                                             base_index, nullptr);
+      expanded_inputs_size = TotalCompensatedFileSize(expanded_inputs.files);
+      if (expanded_inputs.size() > inputs->size() &&
+          output_level_inputs_size + expanded_inputs_size < limit &&
+          !FilesInCompaction(expanded_inputs.files)) {
+        expand_inputs = true;
+      }
+    }
+    if (expand_inputs) {
+      Log(InfoLogLevel::INFO_LEVEL, ioptions_.info_log,
+          "[%s] Expanding@%d %" ROCKSDB_PRIszt "+%" ROCKSDB_PRIszt "(%" PRIu64
+          "+%" PRIu64 " bytes) to %" ROCKSDB_PRIszt "+%" ROCKSDB_PRIszt
+          " (%" PRIu64 "+%" PRIu64 "bytes)\n",
+          cf_name.c_str(), input_level, inputs->size(),
+          output_level_inputs->size(), inputs_size, output_level_inputs_size,
+          expanded_inputs.size(), output_level_inputs->size(),
+          expanded_inputs_size, output_level_inputs_size);
+      inputs->files = expanded_inputs.files;
+    }
  }
-
  return true;
 }

@@ -1179,7 +1200,8 @@ uint32_t LevelCompactionPicker::GetPathId(
        return p;
      } else {
        current_path_size -= level_size;
-        level_size *= mutable_cf_options.max_bytes_for_level_multiplier;
+        level_size = static_cast<uint64_t>(
+            level_size * mutable_cf_options.max_bytes_for_level_multiplier);
        cur_level++;
        continue;
      }
@@ -119,9 +119,6 @@ class CompactionPicker {
                                  const Slice& largest_user_key,
                                  int level) const;

- protected:
-  int NumberLevels() const { return ioptions_.num_levels; }
-
  // Stores the minimal range that covers all entries in inputs in
  // *smallest, *largest.
  // REQUIRES: inputs is not empty
@@ -141,6 +138,9 @@ class CompactionPicker {
  void GetRange(const std::vector<CompactionInputFiles>& inputs,
                InternalKey* smallest, InternalKey* largest) const;

+ protected:
+  int NumberLevels() const { return ioptions_.num_levels; }
+
  // Add more files to the inputs on "level" to make sure that
  // no newer version of a key is compacted to "level+1" while leaving an older
  // version in a "level". Otherwise, any Get() will search "level" first,
@@ -121,6 +121,7 @@ class CompactionPickerTest : public testing::Test {
    vstorage_->GenerateLevelFilesBrief();
    vstorage_->ComputeCompactionScore(ioptions_, mutable_cf_options_);
    vstorage_->GenerateLevel0NonOverlapping();
+    vstorage_->ComputeFilesMarkedForCompaction();
    vstorage_->SetFinalized();
  }
 };
@@ -693,6 +694,134 @@ TEST_F(CompactionPickerTest, OverlappingUserKeys4) {
  ASSERT_EQ(7U, compaction->input(1, 0)->fd.GetNumber());
 }

+TEST_F(CompactionPickerTest, OverlappingUserKeys5) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  // Overlapping user keys on same level and output level
+  Add(1, 1U, "200", "400", 1000000000U);
+  Add(1, 2U, "400", "500", 1U, 0, 0);
+  Add(2, 3U, "000", "100", 1U);
+  Add(2, 4U, "100", "600", 1U, 0, 0);
+  Add(2, 5U, "600", "700", 1U, 0, 0);
+
+  vstorage_->LevelFiles(2)[2]->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);
+}
+
+TEST_F(CompactionPickerTest, OverlappingUserKeys6) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  // Overlapping user keys on same level and output level
+  Add(1, 1U, "200", "400", 1U, 0, 0);
+  Add(1, 2U, "401", "500", 1U, 0, 0);
+  Add(2, 3U, "000", "100", 1U);
+  Add(2, 4U, "100", "300", 1U, 0, 0);
+  Add(2, 5U, "305", "450", 1U, 0, 0);
+  Add(2, 6U, "460", "600", 1U, 0, 0);
+  Add(2, 7U, "600", "700", 1U, 0, 0);
+
+  vstorage_->LevelFiles(1)[0]->marked_for_compaction = true;
+  vstorage_->LevelFiles(1)[1]->marked_for_compaction = 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(2U, compaction->num_input_levels());
+  ASSERT_EQ(1U, compaction->num_input_files(0));
+  ASSERT_EQ(3U, compaction->num_input_files(1));
+}
+
+TEST_F(CompactionPickerTest, OverlappingUserKeys7) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  mutable_cf_options_.max_compaction_bytes = 100000000000u;
+  // Overlapping user keys on same level and output level
+  Add(1, 1U, "200", "400", 1U, 0, 0);
+  Add(1, 2U, "401", "500", 1000000000U, 0, 0);
+  Add(2, 3U, "100", "250", 1U);
+  Add(2, 4U, "300", "600", 1U, 0, 0);
+  Add(2, 5U, "600", "800", 1U, 0, 0);
+
+  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(2U, compaction->num_input_levels());
+  ASSERT_GE(1U, compaction->num_input_files(0));
+  ASSERT_GE(2U, compaction->num_input_files(1));
+  // File 5 has to be included in the compaction
+  ASSERT_EQ(5U, compaction->inputs(1)->back()->fd.GetNumber());
+}
+
+TEST_F(CompactionPickerTest, OverlappingUserKeys8) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  mutable_cf_options_.max_compaction_bytes = 100000000000u;
+  // grow the number of inputs in "level" without
+  // changing the number of "level+1" files we pick up
+  // Expand input level as much as possible
+  // no overlapping case
+  Add(1, 1U, "101", "150", 1U);
+  Add(1, 2U, "151", "200", 1U);
+  Add(1, 3U, "201", "300", 1000000000U);
+  Add(1, 4U, "301", "400", 1U);
+  Add(1, 5U, "401", "500", 1U);
+  Add(2, 6U, "150", "200", 1U);
+  Add(2, 7U, "200", "450", 1U, 0, 0);
+  Add(2, 8U, "500", "600", 1U);
+
+  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(2U, compaction->num_input_levels());
+  ASSERT_EQ(3U, compaction->num_input_files(0));
+  ASSERT_EQ(2U, compaction->num_input_files(1));
+  ASSERT_EQ(2U, compaction->input(0, 0)->fd.GetNumber());
+  ASSERT_EQ(3U, compaction->input(0, 1)->fd.GetNumber());
+  ASSERT_EQ(4U, compaction->input(0, 2)->fd.GetNumber());
+  ASSERT_EQ(6U, compaction->input(1, 0)->fd.GetNumber());
+  ASSERT_EQ(7U, compaction->input(1, 1)->fd.GetNumber());
+}
+
+TEST_F(CompactionPickerTest, OverlappingUserKeys9) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  mutable_cf_options_.max_compaction_bytes = 100000000000u;
+  // grow the number of inputs in "level" without
+  // changing the number of "level+1" files we pick up
+  // Expand input level as much as possible
+  // overlapping case
+  Add(1, 1U, "121", "150", 1U);
+  Add(1, 2U, "151", "200", 1U);
+  Add(1, 3U, "201", "300", 1000000000U);
+  Add(1, 4U, "301", "400", 1U);
+  Add(1, 5U, "401", "500", 1U);
+  Add(2, 6U, "100", "120", 1U);
+  Add(2, 7U, "150", "200", 1U);
+  Add(2, 8U, "200", "450", 1U, 0, 0);
+  Add(2, 9U, "501", "600", 1U);
+
+  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(2U, compaction->num_input_levels());
+  ASSERT_EQ(5U, compaction->num_input_files(0));
+  ASSERT_EQ(2U, compaction->num_input_files(1));
+  ASSERT_EQ(1U, compaction->input(0, 0)->fd.GetNumber());
+  ASSERT_EQ(2U, compaction->input(0, 1)->fd.GetNumber());
+  ASSERT_EQ(3U, compaction->input(0, 2)->fd.GetNumber());
+  ASSERT_EQ(4U, compaction->input(0, 3)->fd.GetNumber());
+  ASSERT_EQ(7U, compaction->input(1, 0)->fd.GetNumber());
+  ASSERT_EQ(8U, compaction->input(1, 1)->fd.GetNumber());
+}
+
 TEST_F(CompactionPickerTest, NotScheduleL1IfL0WithHigherPri1) {
  NewVersionStorage(6, kCompactionStyleLevel);
  mutable_cf_options_.level0_file_num_compaction_trigger = 2;
@@ -1067,13 +1196,13 @@ TEST_F(CompactionPickerTest, MaxCompactionBytesHit) {
  ioptions_.level_compaction_dynamic_level_bytes = false;
  NewVersionStorage(6, kCompactionStyleLevel);
  // A compaction should be triggered and pick file 2 and 5.
-  // It cannot expand because adding file 1 and 3, the compaction size will
+  // It can expand because adding file 1 and 3, the compaction size will
  // exceed mutable_cf_options_.max_bytes_for_level_base.
  Add(1, 1U, "100", "150", 300000U);
  Add(1, 2U, "151", "200", 300001U, 0, 0);
  Add(1, 3U, "201", "250", 300000U, 0, 0);
  Add(1, 4U, "251", "300", 300000U, 0, 0);
-  Add(2, 5U, "160", "256", 1U);
+  Add(2, 5U, "100", "256", 1U);
  UpdateVersionStorageInfo();

  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
@@ -1087,7 +1216,7 @@ TEST_F(CompactionPickerTest, MaxCompactionBytesHit) {
 }

 TEST_F(CompactionPickerTest, MaxCompactionBytesNotHit) {
-  mutable_cf_options_.max_bytes_for_level_base = 1000000u;
+  mutable_cf_options_.max_bytes_for_level_base = 800000u;
  mutable_cf_options_.max_compaction_bytes = 1000000u;
  ioptions_.level_compaction_dynamic_level_bytes = false;
  NewVersionStorage(6, kCompactionStyleLevel);
@@ -1097,7 +1226,7 @@ TEST_F(CompactionPickerTest, MaxCompactionBytesNotHit) {
  Add(1, 2U, "151", "200", 300001U, 0, 0);
  Add(1, 3U, "201", "250", 300000U, 0, 0);
  Add(1, 4U, "251", "300", 300000U, 0, 0);
-  Add(2, 5U, "000", "233", 1U);
+  Add(2, 5U, "000", "251", 1U);
  UpdateVersionStorageInfo();

  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
@@ -1112,6 +1241,50 @@ TEST_F(CompactionPickerTest, MaxCompactionBytesNotHit) {
  ASSERT_EQ(5U, compaction->input(1, 0)->fd.GetNumber());
 }

+TEST_F(CompactionPickerTest, IsTrivialMoveOn) {
+  mutable_cf_options_.max_bytes_for_level_base = 10000u;
+  mutable_cf_options_.max_compaction_bytes = 10001u;
+  ioptions_.level_compaction_dynamic_level_bytes = false;
+  NewVersionStorage(6, kCompactionStyleLevel);
+  // A compaction should be triggered and pick file 2
+  Add(1, 1U, "100", "150", 3000U);
+  Add(1, 2U, "151", "200", 3001U);
+  Add(1, 3U, "201", "250", 3000U);
+  Add(1, 4U, "251", "300", 3000U);
+
+  Add(3, 5U, "120", "130", 7000U);
+  Add(3, 6U, "170", "180", 7000U);
+  Add(3, 5U, "220", "230", 7000U);
+  Add(3, 5U, "270", "280", 7000U);
+  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_TRUE(compaction->IsTrivialMove());
+}
+
+TEST_F(CompactionPickerTest, IsTrivialMoveOff) {
+  mutable_cf_options_.max_bytes_for_level_base = 1000000u;
+  mutable_cf_options_.max_compaction_bytes = 10000u;
+  ioptions_.level_compaction_dynamic_level_bytes = false;
+  NewVersionStorage(6, kCompactionStyleLevel);
+  // A compaction should be triggered and pick all files from level 1
+  Add(1, 1U, "100", "150", 300000U, 0, 0);
+  Add(1, 2U, "150", "200", 300000U, 0, 0);
+  Add(1, 3U, "200", "250", 300000U, 0, 0);
+  Add(1, 4U, "250", "300", 300000U, 0, 0);
+
+  Add(3, 5U, "120", "130", 6000U);
+  Add(3, 6U, "140", "150", 6000U);
+  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_FALSE(compaction->IsTrivialMove());
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -41,7 +41,10 @@ class CorruptionTest : public testing::Test {
  DB* db_;

  CorruptionTest() {
-    tiny_cache_ = NewLRUCache(100);
+    // If LRU cache shard bit is smaller than 2 (or -1 which will automatically
+    // set it to 0), test SequenceNumberRecovery will fail, likely because of a
+    // bug in recovery code. Keep it 4 for now to make the test passes.
+    tiny_cache_ = NewLRUCache(100, 4);
    options_.wal_recovery_mode = WALRecoveryMode::kTolerateCorruptedTailRecords;
    options_.env = &env_;
    dbname_ = test::TmpDir() + "/corruption_test";
@@ -41,6 +41,7 @@ class CuckooTableDBTest : public testing::Test {
    options.memtable_factory.reset(NewHashLinkListRepFactory(4, 0, 3, true));
    options.allow_mmap_reads = true;
    options.create_if_missing = true;
+    options.allow_concurrent_memtable_write = false;
    return options;
  }

@@ -0,0 +1,852 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same 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 "db/db_test_util.h"
+#include "port/stack_trace.h"
+#include "rocksdb/perf_context.h"
+#if !defined(ROCKSDB_LITE)
+#include "util/sync_point.h"
+#endif
+
+namespace rocksdb {
+
+class DBBasicTest : public DBTestBase {
+ public:
+  DBBasicTest() : DBTestBase("/db_basic_test") {}
+};
+
+TEST_F(DBBasicTest, OpenWhenOpen) {
+  Options options = CurrentOptions();
+  options.env = env_;
+  rocksdb::DB* db2 = nullptr;
+  rocksdb::Status s = DB::Open(options, dbname_, &db2);
+
+  ASSERT_EQ(Status::Code::kIOError, s.code());
+  ASSERT_EQ(Status::SubCode::kNone, s.subcode());
+  ASSERT_TRUE(strstr(s.getState(), "lock ") != nullptr);
+
+  delete db2;
+}
+
+#ifndef ROCKSDB_LITE
+TEST_F(DBBasicTest, ReadOnlyDB) {
+  ASSERT_OK(Put("foo", "v1"));
+  ASSERT_OK(Put("bar", "v2"));
+  ASSERT_OK(Put("foo", "v3"));
+  Close();
+
+  auto options = CurrentOptions();
+  assert(options.env = env_);
+  ASSERT_OK(ReadOnlyReopen(options));
+  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());
+}
+
+TEST_F(DBBasicTest, CompactedDB) {
+  const uint64_t kFileSize = 1 << 20;
+  Options options = CurrentOptions();
+  options.disable_auto_compactions = true;
+  options.write_buffer_size = kFileSize;
+  options.target_file_size_base = kFileSize;
+  options.max_bytes_for_level_base = 1 << 30;
+  options.compression = kNoCompression;
+  Reopen(options);
+  // 1 L0 file, use CompactedDB if max_open_files = -1
+  ASSERT_OK(Put("aaa", DummyString(kFileSize / 2, '1')));
+  Flush();
+  Close();
+  ASSERT_OK(ReadOnlyReopen(options));
+  Status s = Put("new", "value");
+  ASSERT_EQ(s.ToString(),
+            "Not implemented: Not supported operation in read only mode.");
+  ASSERT_EQ(DummyString(kFileSize / 2, '1'), Get("aaa"));
+  Close();
+  options.max_open_files = -1;
+  ASSERT_OK(ReadOnlyReopen(options));
+  s = Put("new", "value");
+  ASSERT_EQ(s.ToString(),
+            "Not implemented: Not supported in compacted db mode.");
+  ASSERT_EQ(DummyString(kFileSize / 2, '1'), Get("aaa"));
+  Close();
+  Reopen(options);
+  // Add more L0 files
+  ASSERT_OK(Put("bbb", DummyString(kFileSize / 2, '2')));
+  Flush();
+  ASSERT_OK(Put("aaa", DummyString(kFileSize / 2, 'a')));
+  Flush();
+  ASSERT_OK(Put("bbb", DummyString(kFileSize / 2, 'b')));
+  ASSERT_OK(Put("eee", DummyString(kFileSize / 2, 'e')));
+  Flush();
+  Close();
+
+  ASSERT_OK(ReadOnlyReopen(options));
+  // Fallback to read-only DB
+  s = Put("new", "value");
+  ASSERT_EQ(s.ToString(),
+            "Not implemented: Not supported operation in read only mode.");
+  Close();
+
+  // Full compaction
+  Reopen(options);
+  // Add more keys
+  ASSERT_OK(Put("fff", DummyString(kFileSize / 2, 'f')));
+  ASSERT_OK(Put("hhh", DummyString(kFileSize / 2, 'h')));
+  ASSERT_OK(Put("iii", DummyString(kFileSize / 2, 'i')));
+  ASSERT_OK(Put("jjj", DummyString(kFileSize / 2, 'j')));
+  db_->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+  ASSERT_EQ(3, NumTableFilesAtLevel(1));
+  Close();
+
+  // CompactedDB
+  ASSERT_OK(ReadOnlyReopen(options));
+  s = Put("new", "value");
+  ASSERT_EQ(s.ToString(),
+            "Not implemented: Not supported in compacted db mode.");
+  ASSERT_EQ("NOT_FOUND", Get("abc"));
+  ASSERT_EQ(DummyString(kFileSize / 2, 'a'), Get("aaa"));
+  ASSERT_EQ(DummyString(kFileSize / 2, 'b'), Get("bbb"));
+  ASSERT_EQ("NOT_FOUND", Get("ccc"));
+  ASSERT_EQ(DummyString(kFileSize / 2, 'e'), Get("eee"));
+  ASSERT_EQ(DummyString(kFileSize / 2, 'f'), Get("fff"));
+  ASSERT_EQ("NOT_FOUND", Get("ggg"));
+  ASSERT_EQ(DummyString(kFileSize / 2, 'h'), Get("hhh"));
+  ASSERT_EQ(DummyString(kFileSize / 2, 'i'), Get("iii"));
+  ASSERT_EQ(DummyString(kFileSize / 2, 'j'), Get("jjj"));
+  ASSERT_EQ("NOT_FOUND", Get("kkk"));
+
+  // MultiGet
+  std::vector<std::string> values;
+  std::vector<Status> status_list = dbfull()->MultiGet(
+      ReadOptions(),
+      std::vector<Slice>({Slice("aaa"), Slice("ccc"), Slice("eee"),
+                          Slice("ggg"), Slice("iii"), Slice("kkk")}),
+      &values);
+  ASSERT_EQ(status_list.size(), static_cast<uint64_t>(6));
+  ASSERT_EQ(values.size(), static_cast<uint64_t>(6));
+  ASSERT_OK(status_list[0]);
+  ASSERT_EQ(DummyString(kFileSize / 2, 'a'), values[0]);
+  ASSERT_TRUE(status_list[1].IsNotFound());
+  ASSERT_OK(status_list[2]);
+  ASSERT_EQ(DummyString(kFileSize / 2, 'e'), values[2]);
+  ASSERT_TRUE(status_list[3].IsNotFound());
+  ASSERT_OK(status_list[4]);
+  ASSERT_EQ(DummyString(kFileSize / 2, 'i'), values[4]);
+  ASSERT_TRUE(status_list[5].IsNotFound());
+
+  Reopen(options);
+  // Add a key
+  ASSERT_OK(Put("fff", DummyString(kFileSize / 2, 'f')));
+  Close();
+  ASSERT_OK(ReadOnlyReopen(options));
+  s = Put("new", "value");
+  ASSERT_EQ(s.ToString(),
+            "Not implemented: Not supported operation in read only mode.");
+}
+
+TEST_F(DBBasicTest, LevelLimitReopen) {
+  Options options = CurrentOptions();
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  const std::string value(1024 * 1024, ' ');
+  int i = 0;
+  while (NumTableFilesAtLevel(2, 1) == 0) {
+    ASSERT_OK(Put(1, Key(i++), value));
+  }
+
+  options.num_levels = 1;
+  options.max_bytes_for_level_multiplier_additional.resize(1, 1);
+  Status s = TryReopenWithColumnFamilies({"default", "pikachu"}, options);
+  ASSERT_EQ(s.IsInvalidArgument(), true);
+  ASSERT_EQ(s.ToString(),
+            "Invalid argument: db has more levels than options.num_levels");
+
+  options.num_levels = 10;
+  options.max_bytes_for_level_multiplier_additional.resize(10, 1);
+  ASSERT_OK(TryReopenWithColumnFamilies({"default", "pikachu"}, options));
+}
+#endif  // ROCKSDB_LITE
+
+TEST_F(DBBasicTest, PutDeleteGet) {
+  do {
+    CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
+    ASSERT_OK(Put(1, "foo", "v1"));
+    ASSERT_EQ("v1", Get(1, "foo"));
+    ASSERT_OK(Put(1, "foo", "v2"));
+    ASSERT_EQ("v2", Get(1, "foo"));
+    ASSERT_OK(Delete(1, "foo"));
+    ASSERT_EQ("NOT_FOUND", Get(1, "foo"));
+  } while (ChangeOptions());
+}
+
+TEST_F(DBBasicTest, PutSingleDeleteGet) {
+  do {
+    CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
+    ASSERT_OK(Put(1, "foo", "v1"));
+    ASSERT_EQ("v1", Get(1, "foo"));
+    ASSERT_OK(Put(1, "foo2", "v2"));
+    ASSERT_EQ("v2", Get(1, "foo2"));
+    ASSERT_OK(SingleDelete(1, "foo"));
+    ASSERT_EQ("NOT_FOUND", Get(1, "foo"));
+    // Skip HashCuckooRep as it does not support single delete. FIFO and
+    // universal compaction do not apply to the test case. Skip MergePut
+    // because single delete does not get removed when it encounters a merge.
+  } while (ChangeOptions(kSkipHashCuckoo | kSkipFIFOCompaction |
+                         kSkipUniversalCompaction | kSkipMergePut));
+}
+
+TEST_F(DBBasicTest, EmptyFlush) {
+  // It is possible to produce empty flushes when using single deletes. Tests
+  // whether empty flushes cause issues.
+  do {
+    Random rnd(301);
+
+    Options options = CurrentOptions();
+    options.disable_auto_compactions = true;
+    CreateAndReopenWithCF({"pikachu"}, options);
+
+    Put(1, "a", Slice());
+    SingleDelete(1, "a");
+    ASSERT_OK(Flush(1));
+
+    ASSERT_EQ("[ ]", AllEntriesFor("a", 1));
+    // Skip HashCuckooRep as it does not support single delete. FIFO and
+    // universal compaction do not apply to the test case. Skip MergePut
+    // because merges cannot be combined with single deletions.
+  } while (ChangeOptions(kSkipHashCuckoo | kSkipFIFOCompaction |
+                         kSkipUniversalCompaction | kSkipMergePut));
+}
+
+TEST_F(DBBasicTest, GetFromVersions) {
+  do {
+    CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
+    ASSERT_OK(Put(1, "foo", "v1"));
+    ASSERT_OK(Flush(1));
+    ASSERT_EQ("v1", Get(1, "foo"));
+    ASSERT_EQ("NOT_FOUND", Get(0, "foo"));
+  } while (ChangeOptions());
+}
+
+#ifndef ROCKSDB_LITE
+TEST_F(DBBasicTest, GetSnapshot) {
+  anon::OptionsOverride options_override;
+  options_override.skip_policy = kSkipNoSnapshot;
+  do {
+    CreateAndReopenWithCF({"pikachu"}, CurrentOptions(options_override));
+    // Try with both a short key and a long key
+    for (int i = 0; i < 2; i++) {
+      std::string key = (i == 0) ? std::string("foo") : std::string(200, 'x');
+      ASSERT_OK(Put(1, key, "v1"));
+      const Snapshot* s1 = db_->GetSnapshot();
+      if (option_config_ == kHashCuckoo) {
+        // Unsupported case.
+        ASSERT_TRUE(s1 == nullptr);
+        break;
+      }
+      ASSERT_OK(Put(1, key, "v2"));
+      ASSERT_EQ("v2", Get(1, key));
+      ASSERT_EQ("v1", Get(1, key, s1));
+      ASSERT_OK(Flush(1));
+      ASSERT_EQ("v2", Get(1, key));
+      ASSERT_EQ("v1", Get(1, key, s1));
+      db_->ReleaseSnapshot(s1);
+    }
+  } while (ChangeOptions());
+}
+#endif  // ROCKSDB_LITE
+
+TEST_F(DBBasicTest, CheckLock) {
+  do {
+    DB* localdb;
+    Options options = CurrentOptions();
+    ASSERT_OK(TryReopen(options));
+
+    // second open should fail
+    ASSERT_TRUE(!(DB::Open(options, dbname_, &localdb)).ok());
+  } while (ChangeCompactOptions());
+}
+
+TEST_F(DBBasicTest, FlushMultipleMemtable) {
+  do {
+    Options options = CurrentOptions();
+    WriteOptions writeOpt = WriteOptions();
+    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;
+    CreateAndReopenWithCF({"pikachu"}, options);
+    ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v1"));
+    ASSERT_OK(Flush(1));
+    ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v1"));
+
+    ASSERT_EQ("v1", Get(1, "foo"));
+    ASSERT_EQ("v1", Get(1, "bar"));
+    ASSERT_OK(Flush(1));
+  } while (ChangeCompactOptions());
+}
+
+TEST_F(DBBasicTest, FlushEmptyColumnFamily) {
+  // Block flush thread and disable compaction thread
+  env_->SetBackgroundThreads(1, Env::HIGH);
+  env_->SetBackgroundThreads(1, Env::LOW);
+  test::SleepingBackgroundTask sleeping_task_low;
+  env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task_low,
+                 Env::Priority::LOW);
+  test::SleepingBackgroundTask sleeping_task_high;
+  env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask,
+                 &sleeping_task_high, Env::Priority::HIGH);
+
+  Options options = CurrentOptions();
+  // disable compaction
+  options.disable_auto_compactions = true;
+  WriteOptions writeOpt = WriteOptions();
+  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;
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  // Compaction can still go through even if no thread can flush the
+  // mem table.
+  ASSERT_OK(Flush(0));
+  ASSERT_OK(Flush(1));
+
+  // Insert can go through
+  ASSERT_OK(dbfull()->Put(writeOpt, handles_[0], "foo", "v1"));
+  ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v1"));
+
+  ASSERT_EQ("v1", Get(0, "foo"));
+  ASSERT_EQ("v1", Get(1, "bar"));
+
+  sleeping_task_high.WakeUp();
+  sleeping_task_high.WaitUntilDone();
+
+  // Flush can still go through.
+  ASSERT_OK(Flush(0));
+  ASSERT_OK(Flush(1));
+
+  sleeping_task_low.WakeUp();
+  sleeping_task_low.WaitUntilDone();
+}
+
+TEST_F(DBBasicTest, FLUSH) {
+  do {
+    CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
+    WriteOptions writeOpt = WriteOptions();
+    writeOpt.disableWAL = true;
+    SetPerfLevel(kEnableTime);
+    ;
+    ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v1"));
+    // this will now also flush the last 2 writes
+    ASSERT_OK(Flush(1));
+    ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v1"));
+
+    perf_context.Reset();
+    Get(1, "foo");
+    ASSERT_TRUE((int)perf_context.get_from_output_files_time > 0);
+
+    ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
+    ASSERT_EQ("v1", Get(1, "foo"));
+    ASSERT_EQ("v1", Get(1, "bar"));
+
+    writeOpt.disableWAL = true;
+    ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v2"));
+    ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v2"));
+    ASSERT_OK(Flush(1));
+
+    ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
+    ASSERT_EQ("v2", Get(1, "bar"));
+    perf_context.Reset();
+    ASSERT_EQ("v2", Get(1, "foo"));
+    ASSERT_TRUE((int)perf_context.get_from_output_files_time > 0);
+
+    writeOpt.disableWAL = false;
+    ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v3"));
+    ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v3"));
+    ASSERT_OK(Flush(1));
+
+    ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
+    // 'foo' should be there because its put
+    // has WAL enabled.
+    ASSERT_EQ("v3", Get(1, "foo"));
+    ASSERT_EQ("v3", Get(1, "bar"));
+
+    SetPerfLevel(kDisable);
+  } while (ChangeCompactOptions());
+}
+
+TEST_F(DBBasicTest, ManifestRollOver) {
+  do {
+    Options options;
+    options.max_manifest_file_size = 10;  // 10 bytes
+    options = CurrentOptions(options);
+    CreateAndReopenWithCF({"pikachu"}, options);
+    {
+      ASSERT_OK(Put(1, "manifest_key1", std::string(1000, '1')));
+      ASSERT_OK(Put(1, "manifest_key2", std::string(1000, '2')));
+      ASSERT_OK(Put(1, "manifest_key3", std::string(1000, '3')));
+      uint64_t manifest_before_flush = dbfull()->TEST_Current_Manifest_FileNo();
+      ASSERT_OK(Flush(1));  // This should trigger LogAndApply.
+      uint64_t manifest_after_flush = dbfull()->TEST_Current_Manifest_FileNo();
+      ASSERT_GT(manifest_after_flush, manifest_before_flush);
+      ReopenWithColumnFamilies({"default", "pikachu"}, options);
+      ASSERT_GT(dbfull()->TEST_Current_Manifest_FileNo(), manifest_after_flush);
+      // check if a new manifest file got inserted or not.
+      ASSERT_EQ(std::string(1000, '1'), Get(1, "manifest_key1"));
+      ASSERT_EQ(std::string(1000, '2'), Get(1, "manifest_key2"));
+      ASSERT_EQ(std::string(1000, '3'), Get(1, "manifest_key3"));
+    }
+  } while (ChangeCompactOptions());
+}
+
+TEST_F(DBBasicTest, IdentityAcrossRestarts) {
+  do {
+    std::string id1;
+    ASSERT_OK(db_->GetDbIdentity(id1));
+
+    Options options = CurrentOptions();
+    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);
+  } while (ChangeCompactOptions());
+}
+
+#ifndef ROCKSDB_LITE
+TEST_F(DBBasicTest, Snapshot) {
+  anon::OptionsOverride options_override;
+  options_override.skip_policy = kSkipNoSnapshot;
+  do {
+    CreateAndReopenWithCF({"pikachu"}, CurrentOptions(options_override));
+    Put(0, "foo", "0v1");
+    Put(1, "foo", "1v1");
+
+    const Snapshot* s1 = db_->GetSnapshot();
+    ASSERT_EQ(1U, GetNumSnapshots());
+    uint64_t time_snap1 = GetTimeOldestSnapshots();
+    ASSERT_GT(time_snap1, 0U);
+    Put(0, "foo", "0v2");
+    Put(1, "foo", "1v2");
+
+    env_->addon_time_.fetch_add(1);
+
+    const Snapshot* s2 = db_->GetSnapshot();
+    ASSERT_EQ(2U, GetNumSnapshots());
+    ASSERT_EQ(time_snap1, GetTimeOldestSnapshots());
+    Put(0, "foo", "0v3");
+    Put(1, "foo", "1v3");
+
+    {
+      ManagedSnapshot s3(db_);
+      ASSERT_EQ(3U, GetNumSnapshots());
+      ASSERT_EQ(time_snap1, GetTimeOldestSnapshots());
+
+      Put(0, "foo", "0v4");
+      Put(1, "foo", "1v4");
+      ASSERT_EQ("0v1", Get(0, "foo", s1));
+      ASSERT_EQ("1v1", Get(1, "foo", s1));
+      ASSERT_EQ("0v2", Get(0, "foo", s2));
+      ASSERT_EQ("1v2", Get(1, "foo", s2));
+      ASSERT_EQ("0v3", Get(0, "foo", s3.snapshot()));
+      ASSERT_EQ("1v3", Get(1, "foo", s3.snapshot()));
+      ASSERT_EQ("0v4", Get(0, "foo"));
+      ASSERT_EQ("1v4", Get(1, "foo"));
+    }
+
+    ASSERT_EQ(2U, GetNumSnapshots());
+    ASSERT_EQ(time_snap1, GetTimeOldestSnapshots());
+    ASSERT_EQ("0v1", Get(0, "foo", s1));
+    ASSERT_EQ("1v1", Get(1, "foo", s1));
+    ASSERT_EQ("0v2", Get(0, "foo", s2));
+    ASSERT_EQ("1v2", Get(1, "foo", s2));
+    ASSERT_EQ("0v4", Get(0, "foo"));
+    ASSERT_EQ("1v4", Get(1, "foo"));
+
+    db_->ReleaseSnapshot(s1);
+    ASSERT_EQ("0v2", Get(0, "foo", s2));
+    ASSERT_EQ("1v2", Get(1, "foo", s2));
+    ASSERT_EQ("0v4", Get(0, "foo"));
+    ASSERT_EQ("1v4", Get(1, "foo"));
+    ASSERT_EQ(1U, GetNumSnapshots());
+    ASSERT_LT(time_snap1, GetTimeOldestSnapshots());
+
+    db_->ReleaseSnapshot(s2);
+    ASSERT_EQ(0U, GetNumSnapshots());
+    ASSERT_EQ("0v4", Get(0, "foo"));
+    ASSERT_EQ("1v4", Get(1, "foo"));
+  } while (ChangeOptions(kSkipHashCuckoo));
+}
+
+#endif  // ROCKSDB_LITE
+
+TEST_F(DBBasicTest, CompactBetweenSnapshots) {
+  anon::OptionsOverride options_override;
+  options_override.skip_policy = kSkipNoSnapshot;
+  do {
+    Options options = CurrentOptions(options_override);
+    options.disable_auto_compactions = true;
+    CreateAndReopenWithCF({"pikachu"}, options);
+    Random rnd(301);
+    FillLevels("a", "z", 1);
+
+    Put(1, "foo", "first");
+    const Snapshot* snapshot1 = db_->GetSnapshot();
+    Put(1, "foo", "second");
+    Put(1, "foo", "third");
+    Put(1, "foo", "fourth");
+    const Snapshot* snapshot2 = db_->GetSnapshot();
+    Put(1, "foo", "fifth");
+    Put(1, "foo", "sixth");
+
+    // All entries (including duplicates) exist
+    // before any compaction or flush is triggered.
+    ASSERT_EQ(AllEntriesFor("foo", 1),
+              "[ sixth, fifth, fourth, third, second, first ]");
+    ASSERT_EQ("sixth", Get(1, "foo"));
+    ASSERT_EQ("fourth", Get(1, "foo", snapshot2));
+    ASSERT_EQ("first", Get(1, "foo", snapshot1));
+
+    // After a flush, "second", "third" and "fifth" should
+    // be removed
+    ASSERT_OK(Flush(1));
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ sixth, fourth, first ]");
+
+    // after we release the snapshot1, only two values left
+    db_->ReleaseSnapshot(snapshot1);
+    FillLevels("a", "z", 1);
+    dbfull()->CompactRange(CompactRangeOptions(), handles_[1], nullptr,
+                           nullptr);
+
+    // We have only one valid snapshot snapshot2. Since snapshot1 is
+    // not valid anymore, "first" should be removed by a compaction.
+    ASSERT_EQ("sixth", Get(1, "foo"));
+    ASSERT_EQ("fourth", Get(1, "foo", snapshot2));
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ sixth, fourth ]");
+
+    // after we release the snapshot2, only one value should be left
+    db_->ReleaseSnapshot(snapshot2);
+    FillLevels("a", "z", 1);
+    dbfull()->CompactRange(CompactRangeOptions(), handles_[1], nullptr,
+                           nullptr);
+    ASSERT_EQ("sixth", Get(1, "foo"));
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ sixth ]");
+    // skip HashCuckooRep as it does not support snapshot
+  } while (ChangeOptions(kSkipHashCuckoo | kSkipFIFOCompaction));
+}
+
+TEST_F(DBBasicTest, DBOpen_Options) {
+  Options options = CurrentOptions();
+  std::string dbname = test::TmpDir(env_) + "/db_options_test";
+  ASSERT_OK(DestroyDB(dbname, options));
+
+  // Does not exist, and create_if_missing == false: error
+  DB* db = nullptr;
+  options.create_if_missing = false;
+  Status s = DB::Open(options, dbname, &db);
+  ASSERT_TRUE(strstr(s.ToString().c_str(), "does not exist") != nullptr);
+  ASSERT_TRUE(db == nullptr);
+
+  // Does not exist, and create_if_missing == true: OK
+  options.create_if_missing = true;
+  s = DB::Open(options, dbname, &db);
+  ASSERT_OK(s);
+  ASSERT_TRUE(db != nullptr);
+
+  delete db;
+  db = nullptr;
+
+  // Does exist, and error_if_exists == true: error
+  options.create_if_missing = false;
+  options.error_if_exists = true;
+  s = DB::Open(options, dbname, &db);
+  ASSERT_TRUE(strstr(s.ToString().c_str(), "exists") != nullptr);
+  ASSERT_TRUE(db == nullptr);
+
+  // Does exist, and error_if_exists == false: OK
+  options.create_if_missing = true;
+  options.error_if_exists = false;
+  s = DB::Open(options, dbname, &db);
+  ASSERT_OK(s);
+  ASSERT_TRUE(db != nullptr);
+
+  delete db;
+  db = nullptr;
+}
+
+TEST_F(DBBasicTest, CompactOnFlush) {
+  anon::OptionsOverride options_override;
+  options_override.skip_policy = kSkipNoSnapshot;
+  do {
+    Options options = CurrentOptions(options_override);
+    options.disable_auto_compactions = true;
+    CreateAndReopenWithCF({"pikachu"}, options);
+
+    Put(1, "foo", "v1");
+    ASSERT_OK(Flush(1));
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ v1 ]");
+
+    // Write two new keys
+    Put(1, "a", "begin");
+    Put(1, "z", "end");
+    Flush(1);
+
+    // Case1: Delete followed by a put
+    Delete(1, "foo");
+    Put(1, "foo", "v2");
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ v2, DEL, v1 ]");
+
+    // After the current memtable is flushed, the DEL should
+    // have been removed
+    ASSERT_OK(Flush(1));
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ v2, v1 ]");
+
+    dbfull()->CompactRange(CompactRangeOptions(), handles_[1], nullptr,
+                           nullptr);
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ v2 ]");
+
+    // Case 2: Delete followed by another delete
+    Delete(1, "foo");
+    Delete(1, "foo");
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ DEL, DEL, v2 ]");
+    ASSERT_OK(Flush(1));
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ DEL, v2 ]");
+    dbfull()->CompactRange(CompactRangeOptions(), handles_[1], nullptr,
+                           nullptr);
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ ]");
+
+    // Case 3: Put followed by a delete
+    Put(1, "foo", "v3");
+    Delete(1, "foo");
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ DEL, v3 ]");
+    ASSERT_OK(Flush(1));
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ DEL ]");
+    dbfull()->CompactRange(CompactRangeOptions(), handles_[1], nullptr,
+                           nullptr);
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ ]");
+
+    // Case 4: Put followed by another Put
+    Put(1, "foo", "v4");
+    Put(1, "foo", "v5");
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ v5, v4 ]");
+    ASSERT_OK(Flush(1));
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ v5 ]");
+    dbfull()->CompactRange(CompactRangeOptions(), handles_[1], nullptr,
+                           nullptr);
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ v5 ]");
+
+    // clear database
+    Delete(1, "foo");
+    dbfull()->CompactRange(CompactRangeOptions(), handles_[1], nullptr,
+                           nullptr);
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ ]");
+
+    // Case 5: Put followed by snapshot followed by another Put
+    // Both puts should remain.
+    Put(1, "foo", "v6");
+    const Snapshot* snapshot = db_->GetSnapshot();
+    Put(1, "foo", "v7");
+    ASSERT_OK(Flush(1));
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ v7, v6 ]");
+    db_->ReleaseSnapshot(snapshot);
+
+    // clear database
+    Delete(1, "foo");
+    dbfull()->CompactRange(CompactRangeOptions(), handles_[1], nullptr,
+                           nullptr);
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ ]");
+
+    // Case 5: snapshot followed by a put followed by another Put
+    // Only the last put should remain.
+    const Snapshot* snapshot1 = db_->GetSnapshot();
+    Put(1, "foo", "v8");
+    Put(1, "foo", "v9");
+    ASSERT_OK(Flush(1));
+    ASSERT_EQ(AllEntriesFor("foo", 1), "[ v9 ]");
+    db_->ReleaseSnapshot(snapshot1);
+  } while (ChangeCompactOptions());
+}
+
+TEST_F(DBBasicTest, FlushOneColumnFamily) {
+  Options options = CurrentOptions();
+  CreateAndReopenWithCF({"pikachu", "ilya", "muromec", "dobrynia", "nikitich",
+                         "alyosha", "popovich"},
+                        options);
+
+  ASSERT_OK(Put(0, "Default", "Default"));
+  ASSERT_OK(Put(1, "pikachu", "pikachu"));
+  ASSERT_OK(Put(2, "ilya", "ilya"));
+  ASSERT_OK(Put(3, "muromec", "muromec"));
+  ASSERT_OK(Put(4, "dobrynia", "dobrynia"));
+  ASSERT_OK(Put(5, "nikitich", "nikitich"));
+  ASSERT_OK(Put(6, "alyosha", "alyosha"));
+  ASSERT_OK(Put(7, "popovich", "popovich"));
+
+  for (int i = 0; i < 8; ++i) {
+    Flush(i);
+    auto tables = ListTableFiles(env_, dbname_);
+    ASSERT_EQ(tables.size(), i + 1U);
+  }
+}
+
+TEST_F(DBBasicTest, MultiGetSimple) {
+  do {
+    CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
+    ASSERT_OK(Put(1, "k1", "v1"));
+    ASSERT_OK(Put(1, "k2", "v2"));
+    ASSERT_OK(Put(1, "k3", "v3"));
+    ASSERT_OK(Put(1, "k4", "v4"));
+    ASSERT_OK(Delete(1, "k4"));
+    ASSERT_OK(Put(1, "k5", "v5"));
+    ASSERT_OK(Delete(1, "no_key"));
+
+    std::vector<Slice> keys({"k1", "k2", "k3", "k4", "k5", "no_key"});
+
+    std::vector<std::string> values(20, "Temporary data to be overwritten");
+    std::vector<ColumnFamilyHandle*> cfs(keys.size(), handles_[1]);
+
+    std::vector<Status> s = db_->MultiGet(ReadOptions(), cfs, keys, &values);
+    ASSERT_EQ(values.size(), keys.size());
+    ASSERT_EQ(values[0], "v1");
+    ASSERT_EQ(values[1], "v2");
+    ASSERT_EQ(values[2], "v3");
+    ASSERT_EQ(values[4], "v5");
+
+    ASSERT_OK(s[0]);
+    ASSERT_OK(s[1]);
+    ASSERT_OK(s[2]);
+    ASSERT_TRUE(s[3].IsNotFound());
+    ASSERT_OK(s[4]);
+    ASSERT_TRUE(s[5].IsNotFound());
+  } while (ChangeCompactOptions());
+}
+
+TEST_F(DBBasicTest, MultiGetEmpty) {
+  do {
+    CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
+    // Empty Key Set
+    std::vector<Slice> keys;
+    std::vector<std::string> values;
+    std::vector<ColumnFamilyHandle*> cfs;
+    std::vector<Status> s = db_->MultiGet(ReadOptions(), cfs, keys, &values);
+    ASSERT_EQ(s.size(), 0U);
+
+    // Empty Database, Empty Key Set
+    Options options = CurrentOptions();
+    options.create_if_missing = true;
+    DestroyAndReopen(options);
+    CreateAndReopenWithCF({"pikachu"}, options);
+    s = db_->MultiGet(ReadOptions(), cfs, keys, &values);
+    ASSERT_EQ(s.size(), 0U);
+
+    // Empty Database, Search for Keys
+    keys.resize(2);
+    keys[0] = "a";
+    keys[1] = "b";
+    cfs.push_back(handles_[0]);
+    cfs.push_back(handles_[1]);
+    s = db_->MultiGet(ReadOptions(), cfs, keys, &values);
+    ASSERT_EQ(static_cast<int>(s.size()), 2);
+    ASSERT_TRUE(s[0].IsNotFound() && s[1].IsNotFound());
+  } while (ChangeCompactOptions());
+}
+
+TEST_F(DBBasicTest, ChecksumTest) {
+  BlockBasedTableOptions table_options;
+  Options options = CurrentOptions();
+
+  table_options.checksum = kCRC32c;
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  Reopen(options);
+  ASSERT_OK(Put("a", "b"));
+  ASSERT_OK(Put("c", "d"));
+  ASSERT_OK(Flush());  // table with crc checksum
+
+  table_options.checksum = kxxHash;
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  Reopen(options);
+  ASSERT_OK(Put("e", "f"));
+  ASSERT_OK(Put("g", "h"));
+  ASSERT_OK(Flush());  // table with xxhash checksum
+
+  table_options.checksum = kCRC32c;
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  Reopen(options);
+  ASSERT_EQ("b", Get("a"));
+  ASSERT_EQ("d", Get("c"));
+  ASSERT_EQ("f", Get("e"));
+  ASSERT_EQ("h", Get("g"));
+
+  table_options.checksum = kCRC32c;
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  Reopen(options);
+  ASSERT_EQ("b", Get("a"));
+  ASSERT_EQ("d", Get("c"));
+  ASSERT_EQ("f", Get("e"));
+  ASSERT_EQ("h", Get("g"));
+}
+
+// On Windows you can have either memory mapped file or a file
+// with unbuffered access. So this asserts and does not make
+// sense to run
+#ifndef OS_WIN
+TEST_F(DBBasicTest, MmapAndBufferOptions) {
+  Options options = CurrentOptions();
+
+  options.use_direct_reads = true;
+  options.allow_mmap_reads = true;
+  ASSERT_NOK(TryReopen(options));
+
+  // All other combinations are acceptable
+  options.use_direct_reads = false;
+  ASSERT_OK(TryReopen(options));
+
+  if (IsDirectIOSupported()) {
+    options.use_direct_reads = true;
+    options.allow_mmap_reads = false;
+    ASSERT_OK(TryReopen(options));
+  }
+
+  options.use_direct_reads = false;
+  ASSERT_OK(TryReopen(options));
+}
+#endif
+
+}  // namespace rocksdb
+
+int main(int argc, char** argv) {
+  rocksdb::port::InstallStackTraceHandler();
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}
@@ -788,8 +788,6 @@ void PrefixScanInit(DBBloomFilterTest* dbtest) {
 }  // namespace

 TEST_F(DBBloomFilterTest, PrefixScan) {
-  XFUNC_TEST("", "dbtest_prefix", prefix_skip1, XFuncPoint::SetSkip,
-             kSkipNoPrefix);
  while (ChangeFilterOptions()) {
    int count;
    Slice prefix;
@@ -810,6 +808,7 @@ TEST_F(DBBloomFilterTest, PrefixScan) {
    options.max_background_compactions = 2;
    options.create_if_missing = true;
    options.memtable_factory.reset(NewHashSkipListRepFactory(16));
+    options.allow_concurrent_memtable_write = false;

    BlockBasedTableOptions table_options;
    table_options.no_block_cache = true;
@@ -835,7 +834,6 @@ TEST_F(DBBloomFilterTest, PrefixScan) {
    ASSERT_EQ(env_->random_read_counter_.Read(), 2);
    Close();
  }  // end of while
-  XFUNC_TEST("", "dbtest_prefix", prefix_skip1, XFuncPoint::SetSkip, 0);
 }

 TEST_F(DBBloomFilterTest, OptimizeFiltersForHits) {
@@ -13,6 +13,7 @@
 namespace rocksdb {

 static int cfilter_count = 0;
+static int cfilter_skips = 0;

 // This is a static filter used for filtering
 // kvs during the compaction process.
@@ -65,6 +66,30 @@ class DeleteISFilter : public CompactionFilter {
  virtual const char* Name() const override { return "DeleteFilter"; }
 };

+// Skip x if floor(x/10) is even, use range skips. Requires that keys are
+// zero-padded to length 10.
+class SkipEvenFilter : public CompactionFilter {
+ public:
+  virtual Decision FilterV2(int level, const Slice& key, ValueType value_type,
+                            const Slice& existing_value, std::string* new_value,
+                            std::string* skip_until) const override {
+    cfilter_count++;
+    int i = std::stoi(key.ToString());
+    if (i / 10 % 2 == 0) {
+      char key_str[100];
+      snprintf(key_str, sizeof(key), "%010d", i / 10 * 10 + 10);
+      *skip_until = key_str;
+      ++cfilter_skips;
+      return Decision::kRemoveAndSkipUntil;
+    }
+    return Decision::kKeep;
+  }
+
+  virtual bool IgnoreSnapshots() const override { return true; }
+
+  virtual const char* Name() const override { return "DeleteFilter"; }
+};
+
 class DelayFilter : public CompactionFilter {
 public:
  explicit DelayFilter(DBTestBase* d) : db_test(d) {}
@@ -174,6 +199,20 @@ class DeleteISFilterFactory : public CompactionFilterFactory {
  virtual const char* Name() const override { return "DeleteFilterFactory"; }
 };

+class SkipEvenFilterFactory : public CompactionFilterFactory {
+ public:
+  virtual std::unique_ptr<CompactionFilter> CreateCompactionFilter(
+      const CompactionFilter::Context& context) override {
+    if (context.is_manual_compaction) {
+      return std::unique_ptr<CompactionFilter>(new SkipEvenFilter());
+    } else {
+      return std::unique_ptr<CompactionFilter>(nullptr);
+    }
+  }
+
+  virtual const char* Name() const override { return "SkipEvenFilterFactory"; }
+};
+
 class DelayFilterFactory : public CompactionFilterFactory {
 public:
  explicit DelayFilterFactory(DBTestBase* d) : db_test(d) {}
@@ -261,8 +300,10 @@ TEST_F(DBTestCompactionFilter, CompactionFilter) {
  int total = 0;
  Arena arena;
  {
+    InternalKeyComparator icmp(options.comparator);
+    RangeDelAggregator range_del_agg(icmp, {} /* snapshots */);
    ScopedArenaIterator iter(
-        dbfull()->NewInternalIterator(&arena, handles_[1]));
+        dbfull()->NewInternalIterator(&arena, &range_del_agg, handles_[1]));
    iter->SeekToFirst();
    ASSERT_OK(iter->status());
    while (iter->Valid()) {
@@ -349,8 +390,10 @@ TEST_F(DBTestCompactionFilter, CompactionFilter) {
  // level Lmax because this record is at the tip
  count = 0;
  {
+    InternalKeyComparator icmp(options.comparator);
+    RangeDelAggregator range_del_agg(icmp, {} /* snapshots */);
    ScopedArenaIterator iter(
-        dbfull()->NewInternalIterator(&arena, handles_[1]));
+        dbfull()->NewInternalIterator(&arena, &range_del_agg, handles_[1]));
    iter->SeekToFirst();
    ASSERT_OK(iter->status());
    while (iter->Valid()) {
@@ -566,7 +609,10 @@ TEST_F(DBTestCompactionFilter, CompactionFilterContextManual) {
    int count = 0;
    int total = 0;
    Arena arena;
-    ScopedArenaIterator iter(dbfull()->NewInternalIterator(&arena));
+    InternalKeyComparator icmp(options.comparator);
+    RangeDelAggregator range_del_agg(icmp, {} /* snapshots */);
+    ScopedArenaIterator iter(
+        dbfull()->NewInternalIterator(&arena, &range_del_agg));
    iter->SeekToFirst();
    ASSERT_OK(iter->status());
    while (iter->Valid()) {
@@ -712,6 +758,47 @@ TEST_F(DBTestCompactionFilter, CompactionFilterIgnoreSnapshot) {
 }
 #endif  // ROCKSDB_LITE

+TEST_F(DBTestCompactionFilter, SkipUntil) {
+  Options options = CurrentOptions();
+  options.compaction_filter_factory = std::make_shared<SkipEvenFilterFactory>();
+  options.disable_auto_compactions = true;
+  options.create_if_missing = true;
+  DestroyAndReopen(options);
+
+  // Write 100K keys, these are written to a few files in L0.
+  for (int table = 0; table < 4; ++table) {
+    // Key ranges in tables are [0, 38], [106, 149], [212, 260], [318, 371].
+    for (int i = table * 6; i < 39 + table * 11; ++i) {
+      char key[100];
+      snprintf(key, sizeof(key), "%010d", table * 100 + i);
+      Put(key, std::to_string(table * 1000 + i));
+    }
+    Flush();
+  }
+
+  cfilter_skips = 0;
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  // Numberof skips in tables: 2, 3, 3, 3.
+  ASSERT_EQ(11, cfilter_skips);
+
+  for (int table = 0; table < 4; ++table) {
+    for (int i = table * 6; i < 39 + table * 11; ++i) {
+      int k = table * 100 + i;
+      char key[100];
+      snprintf(key, sizeof(key), "%010d", table * 100 + i);
+      auto expected = std::to_string(table * 1000 + i);
+      std::string val;
+      Status s = db_->Get(ReadOptions(), key, &val);
+      if (k / 10 % 2 == 0) {
+        ASSERT_TRUE(s.IsNotFound());
+      } else {
+        ASSERT_OK(s);
+        ASSERT_EQ(expected, val);
+      }
+    }
+  }
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -9,6 +9,7 @@

 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
+#include "port/port.h"
 #include "rocksdb/experimental.h"
 #include "rocksdb/utilities/convenience.h"
 #include "util/sync_point.h"
@@ -305,8 +306,8 @@ TEST_F(DBCompactionTest, TestTableReaderForCompaction) {
  dbfull()->TEST_WaitForCompact();
  // Preloading iterator issues one table cache lookup and creates
  // a new table reader. One file is created for flush and one for compaction.
-  // Compaction inputs make no table cache look-up for data iterators or
-  // range tombstone iterators since they're already cached.
+  // Compaction inputs make no table cache look-up for data/range deletion
+  // iterators
  ASSERT_EQ(num_table_cache_lookup, 2);
  // Create new iterator for:
  // (1) 1 for verifying flush results
@@ -327,7 +328,8 @@ TEST_F(DBCompactionTest, TestTableReaderForCompaction) {
  cro.target_level = 2;
  cro.bottommost_level_compaction = BottommostLevelCompaction::kForce;
  db_->CompactRange(cro, nullptr, nullptr);
-  // Only verifying compaction outputs issues one table cache lookup.
+  // Only verifying compaction outputs issues one table cache lookup
+  // for both data block and range deletion block).
  ASSERT_EQ(num_table_cache_lookup, 1);
  // One for compaction input, one for verifying compaction results.
  ASSERT_EQ(num_new_table_reader, 2);
@@ -1100,7 +1102,7 @@ TEST_P(DBCompactionTestWithParam, ManualCompactionPartial) {
  ASSERT_EQ(trivial_move, 6);
  ASSERT_EQ(non_trivial_move, 0);

-  std::thread threads([&] {
+  rocksdb::port::Thread threads([&] {
    compact_options.change_level = false;
    compact_options.exclusive_manual_compaction = false;
    std::string begin_string = Key(0);
@@ -1232,7 +1234,7 @@ TEST_F(DBCompactionTest, ManualPartialFill) {
  ASSERT_EQ(trivial_move, 2);
  ASSERT_EQ(non_trivial_move, 0);

-  std::thread threads([&] {
+  rocksdb::port::Thread threads([&] {
    compact_options.change_level = false;
    compact_options.exclusive_manual_compaction = false;
    std::string begin_string = Key(0);
@@ -1259,7 +1261,8 @@ TEST_F(DBCompactionTest, ManualPartialFill) {
  uint64_t target_size = 4 * options.max_bytes_for_level_base;
  for (int32_t i = 1; i < options.num_levels; i++) {
    ASSERT_LE(SizeAtLevel(i), target_size);
-    target_size *= options.max_bytes_for_level_multiplier;
+    target_size = static_cast<uint64_t>(target_size *
+                                        options.max_bytes_for_level_multiplier);
  }

  TEST_SYNC_POINT("DBCompaction::PartialFill:2");
@@ -1336,7 +1339,8 @@ TEST_F(DBCompactionTest, DeleteFileRange) {
  uint64_t target_size = 4 * options.max_bytes_for_level_base;
  for (int32_t i = 1; i < options.num_levels; i++) {
    ASSERT_LE(SizeAtLevel(i), target_size);
-    target_size *= options.max_bytes_for_level_multiplier;
+    target_size = static_cast<uint64_t>(target_size *
+                                        options.max_bytes_for_level_multiplier);
  }

  size_t old_num_files = CountFiles();
@@ -13,6 +13,7 @@
 #if !defined(ROCKSDB_LITE)

 #include "db/db_test_util.h"
+#include "port/port.h"
 #include "port/stack_trace.h"

 namespace rocksdb {
@@ -251,7 +252,7 @@ TEST_F(DBTestDynamicLevel, DynamicLevelMaxBytesBase2) {
  });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  std::thread thread([this] {
+  rocksdb::port::Thread thread([this] {
    TEST_SYNC_POINT("DynamicLevelMaxBytesBase2:compact_range_start");
    ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
    TEST_SYNC_POINT("DynamicLevelMaxBytesBase2:compact_range_finish");
@@ -406,7 +407,7 @@ TEST_F(DBTestDynamicLevel, DynamicLevelMaxBytesBaseInc) {
  env_->SetBackgroundThreads(1, Env::HIGH);
 }

-TEST_F(DBTestDynamicLevel, MigrateToDynamicLevelMaxBytesBase) {
+TEST_F(DBTestDynamicLevel, DISABLED_MigrateToDynamicLevelMaxBytesBase) {
  Random rnd(301);
  const int kMaxKey = 2000;

@@ -462,7 +463,7 @@ TEST_F(DBTestDynamicLevel, MigrateToDynamicLevelMaxBytesBase) {
  compaction_finished = false;
  // Issue manual compaction in one thread and still verify DB state
  // in main thread.
-  std::thread t([&]() {
+  rocksdb::port::Thread t([&]() {
    CompactRangeOptions compact_options;
    compact_options.change_level = true;
    compact_options.target_level = options.num_levels - 1;
@@ -27,7 +27,6 @@
 #include "util/file_util.h"
 #include "util/mutexlock.h"
 #include "util/sync_point.h"
-#include "util/testharness.h"

 namespace rocksdb {

@@ -9,6 +9,7 @@

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

 namespace rocksdb {
@@ -47,6 +48,40 @@ TEST_F(DBFlushTest, FlushWhileWritingManifest) {
 #endif  // ROCKSDB_LITE
 }

+TEST_F(DBFlushTest, SyncFail) {
+  std::unique_ptr<FaultInjectionTestEnv> fault_injection_env(
+      new FaultInjectionTestEnv(Env::Default()));
+  Options options;
+  options.disable_auto_compactions = true;
+  options.env = fault_injection_env.get();
+
+  SyncPoint::GetInstance()->LoadDependency(
+      {{"DBFlushTest::SyncFail:1", "DBImpl::SyncClosedLogs:Start"},
+       {"DBImpl::SyncClosedLogs:Failed", "DBFlushTest::SyncFail:2"}});
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  Reopen(options);
+  Put("key", "value");
+  auto* cfd =
+      reinterpret_cast<ColumnFamilyHandleImpl*>(db_->DefaultColumnFamily())
+          ->cfd();
+  int refs_before = cfd->current()->TEST_refs();
+  FlushOptions flush_options;
+  flush_options.wait = false;
+  ASSERT_OK(dbfull()->Flush(flush_options));
+  fault_injection_env->SetFilesystemActive(false);
+  TEST_SYNC_POINT("DBFlushTest::SyncFail:1");
+  TEST_SYNC_POINT("DBFlushTest::SyncFail:2");
+  fault_injection_env->SetFilesystemActive(true);
+  dbfull()->TEST_WaitForFlushMemTable();
+#ifndef ROCKSDB_LITE
+  ASSERT_EQ("", FilesPerLevel());  // flush failed.
+#endif                             // ROCKSDB_LITE
+  // Flush job should release ref count to current version.
+  ASSERT_EQ(refs_before, cfd->current()->TEST_refs());
+  Destroy(options);
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -13,6 +13,7 @@
 #include <functional>
 #include <limits>
 #include <list>
+#include <map>
 #include <queue>
 #include <set>
 #include <string>
@@ -22,6 +23,7 @@
 #include "db/column_family.h"
 #include "db/compaction_job.h"
 #include "db/dbformat.h"
+#include "db/external_sst_file_ingestion_job.h"
 #include "db/flush_job.h"
 #include "db/flush_scheduler.h"
 #include "db/internal_stats.h"
@@ -123,6 +125,10 @@ class DBImpl : public DB {
  using DB::GetProperty;
  virtual bool GetProperty(ColumnFamilyHandle* column_family,
                           const Slice& property, std::string* value) override;
+  using DB::GetMapProperty;
+  virtual bool GetMapProperty(ColumnFamilyHandle* column_family,
+                              const Slice& property,
+                              std::map<std::string, double>* value) override;
  using DB::GetIntProperty;
  virtual bool GetIntProperty(ColumnFamilyHandle* column_family,
                              const Slice& property, uint64_t* value) override;
@@ -132,7 +138,13 @@ class DBImpl : public DB {
  using DB::GetApproximateSizes;
  virtual void GetApproximateSizes(ColumnFamilyHandle* column_family,
                                   const Range* range, int n, uint64_t* sizes,
-                                   bool include_memtable = false) override;
+                                   uint8_t include_flags
+                                   = INCLUDE_FILES) override;
+  using DB::GetApproximateMemTableStats;
+  virtual void GetApproximateMemTableStats(ColumnFamilyHandle* column_family,
+                                           const Range& range,
+                                           uint64_t* const count,
+                                           uint64_t* const size) override;
  using DB::CompactRange;
  virtual Status CompactRange(const CompactRangeOptions& options,
                              ColumnFamilyHandle* column_family,
@@ -260,13 +272,11 @@ class DBImpl : public DB {
                                 bool cache_only, SequenceNumber* seq,
                                 bool* found_record_for_key);

-  using DB::AddFile;
-  virtual Status AddFile(ColumnFamilyHandle* column_family,
-                         const std::vector<ExternalSstFileInfo>& file_info_list,
-                         bool move_file, bool skip_snapshot_check) override;
-  virtual Status AddFile(ColumnFamilyHandle* column_family,
-                         const std::vector<std::string>& file_path_list,
-                         bool move_file, bool skip_snapshot_check) override;
+  using DB::IngestExternalFile;
+  virtual Status IngestExternalFile(
+      ColumnFamilyHandle* column_family,
+      const std::vector<std::string>& external_files,
+      const IngestExternalFileOptions& ingestion_options) override;

 #endif  // ROCKSDB_LITE

@@ -292,7 +302,8 @@ class DBImpl : public DB {
  // The keys of this iterator are internal keys (see format.h).
  // The returned iterator should be deleted when no longer needed.
  InternalIterator* NewInternalIterator(
-      Arena* arena, ColumnFamilyHandle* column_family = nullptr);
+      Arena* arena, RangeDelAggregator* range_del_agg,
+      ColumnFamilyHandle* column_family = nullptr);

 #ifndef NDEBUG
  // Extra methods (for testing) that are not in the public DB interface
@@ -303,6 +314,16 @@ class DBImpl : public DB {
                           ColumnFamilyHandle* column_family = nullptr,
                           bool disallow_trivial_move = false);

+  void TEST_MaybeFlushColumnFamilies();
+
+  bool TEST_UnableToFlushOldestLog() {
+    return unable_to_flush_oldest_log_;
+  }
+
+  bool TEST_IsLogGettingFlushed() {
+    return alive_log_files_.begin()->getting_flushed;
+  }
+
  // Force current memtable contents to be flushed.
  Status TEST_FlushMemTable(bool wait = true,
                            ColumnFamilyHandle* cfh = nullptr);
@@ -375,6 +396,8 @@ class DBImpl : public DB {
  // schedule a purge
  void ScheduleBgLogWriterClose(JobContext* job_context);

+  uint64_t MinLogNumberToKeep();
+
  // Returns the list of live files in 'live' and the list
  // of all files in the filesystem in 'candidate_files'.
  // If force == false and the last call was less than
@@ -413,9 +436,6 @@ class DBImpl : public DB {
  // mutex is held.
  SuperVersion* GetAndRefSuperVersion(uint32_t column_family_id);

-  // Same as above, should called without mutex held and not on write thread.
-  SuperVersion* GetAndRefSuperVersionUnlocked(uint32_t column_family_id);
-
  // Un-reference the super version and return it to thread local cache if
  // needed. If it is the last reference of the super version. Clean it up
  // after un-referencing it.
@@ -426,18 +446,11 @@ class DBImpl : public DB {
  // REQUIRED: this function should only be called on the write thread.
  void ReturnAndCleanupSuperVersion(uint32_t colun_family_id, SuperVersion* sv);

-  // Same as above, should called without mutex held and not on write thread.
-  void ReturnAndCleanupSuperVersionUnlocked(uint32_t colun_family_id,
-                                            SuperVersion* sv);
-
  // REQUIRED: this function should only be called on the write thread or if the
  // mutex is held.  Return value only valid until next call to this function or
  // mutex is released.
  ColumnFamilyHandle* GetColumnFamilyHandle(uint32_t column_family_id);

-  // Same as above, should called without mutex held and not on write thread.
-  ColumnFamilyHandle* GetColumnFamilyHandleUnlocked(uint32_t column_family_id);
-
  // Returns the number of currently running flushes.
  // REQUIREMENT: mutex_ must be held when calling this function.
  int num_running_flushes() {
@@ -452,6 +465,8 @@ class DBImpl : public DB {
    return num_running_compactions_;
  }

+  const WriteController& write_controller() { return write_controller_; }
+
  // hollow transactions shell used for recovery.
  // these will then be passed to TransactionDB so that
  // locks can be reacquired before writing can resume.
@@ -527,7 +542,8 @@ class DBImpl : public DB {
  InternalIterator* NewInternalIterator(const ReadOptions&,
                                        ColumnFamilyData* cfd,
                                        SuperVersion* super_version,
-                                        Arena* arena);
+                                        Arena* arena,
+                                        RangeDelAggregator* range_del_agg);

  // Except in DB::Open(), WriteOptionsFile can only be called when:
  // 1. WriteThread::Writer::EnterUnbatched() is used.
@@ -551,6 +567,11 @@ class DBImpl : public DB {
  void NotifyOnMemTableSealed(ColumnFamilyData* cfd,
                              const MemTableInfo& mem_table_info);

+#ifndef ROCKSDB_LITE
+  void NotifyOnExternalFileIngested(
+      ColumnFamilyData* cfd, const ExternalSstFileIngestionJob& ingestion_job);
+#endif  // !ROCKSDB_LITE
+
  void NewThreadStatusCfInfo(ColumnFamilyData* cfd) const;

  void EraseThreadStatusCfInfo(ColumnFamilyData* cfd) const;
@@ -643,27 +664,20 @@ class DBImpl : public DB {

  // num_bytes: for slowdown case, delay time is calculated based on
  //            `num_bytes` going through.
-  Status DelayWrite(uint64_t num_bytes);
+  Status DelayWrite(uint64_t num_bytes, const WriteOptions& write_options);

  Status ScheduleFlushes(WriteContext* context);

  Status SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context);

  // Force current memtable contents to be flushed.
-  Status FlushMemTable(ColumnFamilyData* cfd, const FlushOptions& options);
+  Status FlushMemTable(ColumnFamilyData* cfd, const FlushOptions& options,
+                       bool writes_stopped = false);

  // Wait for memtable flushed
  Status WaitForFlushMemTable(ColumnFamilyData* cfd);

 #ifndef ROCKSDB_LITE
-  // Finds the lowest level in the DB that the ingested file can be added to
-  // REQUIRES: mutex_ held
-  int PickLevelForIngestedFile(ColumnFamilyData* cfd,
-                               const ExternalSstFileInfo& file_info);
-
-  // Wait for current AddFile() calls to finish.
-  // REQUIRES: mutex_ held
-  void WaitForAddFile();

  Status CompactFilesImpl(const CompactionOptions& compact_options,
                          ColumnFamilyData* cfd, Version* version,
@@ -671,14 +685,14 @@ class DBImpl : public DB {
                          const int output_level, int output_path_id,
                          JobContext* job_context, LogBuffer* log_buffer);

-  Status ReadExternalSstFileInfo(ColumnFamilyHandle* column_family,
-                                 const std::string& file_path,
-                                 ExternalSstFileInfo* file_info);
+  // Wait for current IngestExternalFile() calls to finish.
+  // REQUIRES: mutex_ held
+  void WaitForIngestFile();

 #else
-  // AddFile is not supported in ROCKSDB_LITE so this function
+  // IngestExternalFile is not supported in ROCKSDB_LITE so this function
  // will be no-op
-  void WaitForAddFile() {}
+  void WaitForIngestFile() {}
 #endif  // ROCKSDB_LITE

  ColumnFamilyData* GetColumnFamilyDataByName(const std::string& cf_name);
@@ -730,6 +744,10 @@ class DBImpl : public DB {
  // REQUIRES: mutex locked
  Status PersistOptions();

+  void MaybeFlushColumnFamilies();
+
+  uint64_t GetMaxTotalWalSize() const;
+
  // table_cache_ provides its own synchronization
  std::shared_ptr<Cache> table_cache_;

@@ -752,7 +770,7 @@ class DBImpl : public DB {
  // * whenever bg_flush_scheduled_ or bg_purge_scheduled_ value decreases
  // (i.e. whenever a flush is done, even if it didn't make any progress)
  // * whenever there is an error in background purge, flush or compaction
-  // * whenever num_running_addfile_ goes to 0.
+  // * whenever num_running_ingest_file_ goes to 0.
  InstrumentedCondVar bg_cv_;
  uint64_t logfile_number_;
  std::deque<uint64_t>
@@ -970,8 +988,9 @@ class DBImpl : public DB {
  // without any synchronization
  int disable_delete_obsolete_files_;

-  // next time when we should run DeleteObsoleteFiles with full scan
-  uint64_t delete_obsolete_files_next_run_;
+  // last time when DeleteObsoleteFiles with full scan was executed. Originaly
+  // initialized with startup time.
+  uint64_t delete_obsolete_files_last_run_;

  // last time stats were dumped to LOG
  std::atomic<uint64_t> last_stats_dump_time_microsec_;
@@ -983,7 +1002,15 @@ class DBImpl : public DB {
  // A flag indicating whether the current rocksdb database has any
  // data that is not yet persisted into either WAL or SST file.
  // Used when disableWAL is true.
-  bool has_unpersisted_data_;
+  std::atomic<bool> has_unpersisted_data_;
+
+  // if an attempt was made to flush all column families that
+  // the oldest log depends on but uncommited data in the oldest
+  // log prevents the log from being released.
+  // We must attempt to free the dependent memtables again
+  // at a later time after the transaction in the oldest
+  // log is fully commited.
+  bool unable_to_flush_oldest_log_;

  static const int KEEP_LOG_FILE_NUM = 1000;
  // MSVC version 1800 still does not have constexpr for ::max()
@@ -994,9 +1021,9 @@ class DBImpl : public DB {
  // The options to access storage files
  const EnvOptions env_options_;

-  // Number of running AddFile() calls.
+  // Number of running IngestExternalFile() calls.
  // REQUIRES: mutex held
-  int num_running_addfile_;
+  int num_running_ingest_file_;

 #ifndef ROCKSDB_LITE
  WalManager wal_manager_;
@@ -1098,7 +1125,6 @@ class DBImpl : public DB {
 };

 extern Options SanitizeOptions(const std::string& db,
-                               const InternalKeyComparator* icmp,
                               const Options& src);

 extern DBOptions SanitizeOptions(const std::string& db, const DBOptions& src);
@@ -1,430 +0,0 @@
-//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
-//  This source code is licensed under the BSD-style license found in the
-//  LICENSE file in the root directory of this source tree. An additional grant
-//  of patent rights can be found in the PATENTS file in the same directory.
-
-#include "db/db_impl.h"
-
-#ifndef __STDC_FORMAT_MACROS
-#define __STDC_FORMAT_MACROS
-#endif
-
-#include <inttypes.h>
-
-#include "rocksdb/db.h"
-#include "rocksdb/env.h"
-#include "rocksdb/sst_file_writer.h"
-
-#include "db/builder.h"
-#include "table/sst_file_writer_collectors.h"
-#include "table/table_builder.h"
-#include "util/file_reader_writer.h"
-#include "util/file_util.h"
-#include "util/sync_point.h"
-
-namespace rocksdb {
-
-#ifndef ROCKSDB_LITE
-
-Status DBImpl::ReadExternalSstFileInfo(ColumnFamilyHandle* column_family,
-                                       const std::string& file_path,
-                                       ExternalSstFileInfo* file_info) {
-  Status status;
-  auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
-  auto cfd = cfh->cfd();
-
-  file_info->file_path = file_path;
-  status = env_->GetFileSize(file_path, &file_info->file_size);
-  if (!status.ok()) {
-    return status;
-  }
-
-  // Access the file using TableReader to extract
-  // version, number of entries, smallest user key, largest user key
-  std::unique_ptr<RandomAccessFile> sst_file;
-  status = env_->NewRandomAccessFile(file_path, &sst_file, env_options_);
-  if (!status.ok()) {
-    return status;
-  }
-  std::unique_ptr<RandomAccessFileReader> sst_file_reader;
-  sst_file_reader.reset(new RandomAccessFileReader(std::move(sst_file)));
-
-  std::unique_ptr<TableReader> table_reader;
-  status = cfd->ioptions()->table_factory->NewTableReader(
-      TableReaderOptions(*cfd->ioptions(), env_options_,
-                         cfd->internal_comparator()),
-      std::move(sst_file_reader), file_info->file_size, &table_reader);
-  if (!status.ok()) {
-    return status;
-  }
-
-  // Get the external sst file version from table properties
-  const UserCollectedProperties& user_collected_properties =
-      table_reader->GetTableProperties()->user_collected_properties;
-  UserCollectedProperties::const_iterator external_sst_file_version_iter =
-      user_collected_properties.find(ExternalSstFilePropertyNames::kVersion);
-  if (external_sst_file_version_iter == user_collected_properties.end()) {
-    return Status::InvalidArgument("Generated table version not found");
-  }
-
-  file_info->version =
-      DecodeFixed32(external_sst_file_version_iter->second.c_str());
-  if (file_info->version == 2) {
-    // version 2 imply that we have global sequence number
-
-    // TODO(tec): Implement version 2 ingestion
-    file_info->sequence_number = 0;
-  } else if (file_info->version == 1) {
-    // version 1 imply that all sequence numbers in table equal 0
-    file_info->sequence_number = 0;
-  } else {
-    return Status::InvalidArgument("Generated table version is not supported");
-  }
-  // Get number of entries in table
-  file_info->num_entries = table_reader->GetTableProperties()->num_entries;
-
-  ParsedInternalKey key;
-  std::unique_ptr<InternalIterator> iter(
-      table_reader->NewIterator(ReadOptions()));
-
-  // Get first (smallest) key from file
-  iter->SeekToFirst();
-  if (!ParseInternalKey(iter->key(), &key)) {
-    return Status::Corruption("Generated table have corrupted keys");
-  }
-  if (key.sequence != 0) {
-    return Status::Corruption("Generated table have non zero sequence number");
-  }
-  file_info->smallest_key = key.user_key.ToString();
-
-  // Get last (largest) key from file
-  iter->SeekToLast();
-  if (!ParseInternalKey(iter->key(), &key)) {
-    return Status::Corruption("Generated table have corrupted keys");
-  }
-  if (key.sequence != 0) {
-    return Status::Corruption("Generated table have non zero sequence number");
-  }
-  file_info->largest_key = key.user_key.ToString();
-
-  return Status::OK();
-}
-
-Status DBImpl::AddFile(ColumnFamilyHandle* column_family,
-                       const std::vector<std::string>& file_path_list,
-                       bool move_file, bool skip_snapshot_check) {
-  Status status;
-  auto num_files = file_path_list.size();
-  if (num_files == 0) {
-    return Status::InvalidArgument("The list of files is empty");
-  }
-
-  std::vector<ExternalSstFileInfo> file_info_list(num_files);
-  for (size_t i = 0; i < num_files; i++) {
-    status = ReadExternalSstFileInfo(column_family, file_path_list[i],
-                                     &file_info_list[i]);
-    if (!status.ok()) {
-      return status;
-    }
-  }
-  return AddFile(column_family, file_info_list, move_file, skip_snapshot_check);
-}
-
-Status DBImpl::AddFile(ColumnFamilyHandle* column_family,
-                       const std::vector<ExternalSstFileInfo>& file_info_list,
-                       bool move_file, bool skip_snapshot_check) {
-  Status status;
-  auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
-  ColumnFamilyData* cfd = cfh->cfd();
-  const Comparator* user_cmp = cfd->internal_comparator().user_comparator();
-
-  auto num_files = file_info_list.size();
-  if (num_files == 0) {
-    return Status::InvalidArgument("The list of files is empty");
-  }
-
-  // Verify that passed files dont have overlapping ranges
-  if (num_files > 1) {
-    std::vector<const ExternalSstFileInfo*> sorted_file_info_list(num_files);
-    for (size_t i = 0; i < num_files; i++) {
-      sorted_file_info_list[i] = &file_info_list[i];
-    }
-
-    std::sort(sorted_file_info_list.begin(), sorted_file_info_list.end(),
-              [&user_cmp, &file_info_list](const ExternalSstFileInfo* info1,
-                                           const ExternalSstFileInfo* info2) {
-                return user_cmp->Compare(info1->smallest_key,
-                                         info2->smallest_key) < 0;
-              });
-
-    for (size_t i = 0; i < num_files - 1; i++) {
-      if (user_cmp->Compare(sorted_file_info_list[i]->largest_key,
-                            sorted_file_info_list[i + 1]->smallest_key) >= 0) {
-        return Status::NotSupported("Files have overlapping ranges");
-      }
-    }
-  }
-
-  std::vector<uint64_t> micro_list(num_files, 0);
-  std::vector<FileMetaData> meta_list(num_files);
-  for (size_t i = 0; i < num_files; i++) {
-    StopWatch sw(env_, nullptr, 0, &micro_list[i], false);
-    if (file_info_list[i].num_entries == 0) {
-      return Status::InvalidArgument("File contain no entries");
-    }
-
-    if (file_info_list[i].version == 2) {
-      // version 2 imply that file have only Put Operations
-      // with global Sequence Number
-
-      // TODO(tec): Implement changing file global sequence number
-    } else if (file_info_list[i].version == 1) {
-      // version 1 imply that file have only Put Operations
-      // with Sequence Number = 0
-    } else {
-      // Unknown version !
-      return Status::InvalidArgument(
-          "Generated table version is not supported");
-    }
-
-    meta_list[i].smallest =
-        InternalKey(file_info_list[i].smallest_key,
-                    file_info_list[i].sequence_number, ValueType::kTypeValue);
-    meta_list[i].largest =
-        InternalKey(file_info_list[i].largest_key,
-                    file_info_list[i].sequence_number, ValueType::kTypeValue);
-    if (!meta_list[i].smallest.Valid() || !meta_list[i].largest.Valid()) {
-      return Status::Corruption("Generated table have corrupted keys");
-    }
-    meta_list[i].smallest_seqno = file_info_list[i].sequence_number;
-    meta_list[i].largest_seqno = file_info_list[i].sequence_number;
-    if (meta_list[i].smallest_seqno != 0 || meta_list[i].largest_seqno != 0) {
-      return Status::InvalidArgument(
-          "Non zero sequence numbers are not supported");
-    }
-  }
-
-  std::vector<std::list<uint64_t>::iterator> pending_outputs_inserted_elem_list(
-      num_files);
-  // Generate locations for the new tables
-  {
-    InstrumentedMutexLock l(&mutex_);
-    for (size_t i = 0; i < num_files; i++) {
-      StopWatch sw(env_, nullptr, 0, &micro_list[i], false);
-      pending_outputs_inserted_elem_list[i] =
-          CaptureCurrentFileNumberInPendingOutputs();
-      meta_list[i].fd = FileDescriptor(versions_->NewFileNumber(), 0,
-                                       file_info_list[i].file_size);
-    }
-  }
-
-  // Copy/Move external files into DB
-  std::vector<std::string> db_fname_list(num_files);
-  size_t j = 0;
-  for (; j < num_files; j++) {
-    StopWatch sw(env_, nullptr, 0, &micro_list[j], false);
-    db_fname_list[j] =
-        TableFileName(immutable_db_options_.db_paths,
-                      meta_list[j].fd.GetNumber(), meta_list[j].fd.GetPathId());
-    if (move_file) {
-      status = env_->LinkFile(file_info_list[j].file_path, db_fname_list[j]);
-      if (status.IsNotSupported()) {
-        // Original file is on a different FS, use copy instead of hard linking
-        status =
-            CopyFile(env_, file_info_list[j].file_path, db_fname_list[j], 0);
-      }
-    } else {
-      status = CopyFile(env_, file_info_list[j].file_path, db_fname_list[j], 0);
-    }
-    TEST_SYNC_POINT("DBImpl::AddFile:FileCopied");
-    if (!status.ok()) {
-      for (size_t i = 0; i < j; i++) {
-        Status s = env_->DeleteFile(db_fname_list[i]);
-        if (!s.ok()) {
-          Log(InfoLogLevel::WARN_LEVEL, immutable_db_options_.info_log,
-              "AddFile() clean up for file %s failed : %s",
-              db_fname_list[i].c_str(), s.ToString().c_str());
-        }
-      }
-      return status;
-    }
-  }
-
-  {
-    InstrumentedMutexLock l(&mutex_);
-    TEST_SYNC_POINT("DBImpl::AddFile:MutexLock");
-
-    const MutableCFOptions mutable_cf_options =
-        *cfd->GetLatestMutableCFOptions();
-
-    WriteThread::Writer w;
-    write_thread_.EnterUnbatched(&w, &mutex_);
-
-    num_running_addfile_++;
-
-    if (!skip_snapshot_check && !snapshots_.empty()) {
-      // Check that no snapshots are being held
-      status =
-          Status::NotSupported("Cannot add a file while holding snapshots");
-    }
-
-    if (status.ok()) {
-      // Verify that added file key range dont overlap with any keys in DB
-      SuperVersion* sv = cfd->GetSuperVersion()->Ref();
-      Arena arena;
-      ReadOptions ro;
-      ro.total_order_seek = true;
-      ScopedArenaIterator iter(NewInternalIterator(ro, cfd, sv, &arena));
-
-      for (size_t i = 0; i < num_files; i++) {
-        StopWatch sw(env_, nullptr, 0, &micro_list[i], false);
-        InternalKey range_start(file_info_list[i].smallest_key,
-                                kMaxSequenceNumber, kValueTypeForSeek);
-        iter->Seek(range_start.Encode());
-        status = iter->status();
-
-        if (status.ok() && iter->Valid()) {
-          ParsedInternalKey seek_result;
-          if (ParseInternalKey(iter->key(), &seek_result)) {
-            if (user_cmp->Compare(seek_result.user_key,
-                                  file_info_list[i].largest_key) <= 0) {
-              status = Status::NotSupported("Cannot add overlapping range");
-              break;
-            }
-          } else {
-            status = Status::Corruption("DB have corrupted keys");
-            break;
-          }
-        }
-      }
-    }
-
-    // The levels that the files will be ingested into
-    std::vector<int> target_level_list(num_files, 0);
-    if (status.ok()) {
-      VersionEdit edit;
-      edit.SetColumnFamily(cfd->GetID());
-      for (size_t i = 0; i < num_files; i++) {
-        StopWatch sw(env_, nullptr, 0, &micro_list[i], false);
-        // Add file to the lowest possible level
-        target_level_list[i] = PickLevelForIngestedFile(cfd, file_info_list[i]);
-        edit.AddFile(target_level_list[i], meta_list[i].fd.GetNumber(),
-                     meta_list[i].fd.GetPathId(), meta_list[i].fd.GetFileSize(),
-                     meta_list[i].smallest, meta_list[i].largest,
-                     meta_list[i].smallest_seqno, meta_list[i].largest_seqno,
-                     meta_list[i].marked_for_compaction);
-      }
-      status = versions_->LogAndApply(cfd, mutable_cf_options, &edit, &mutex_,
-                                      directories_.GetDbDir());
-    }
-    write_thread_.ExitUnbatched(&w);
-
-    if (status.ok()) {
-      delete InstallSuperVersionAndScheduleWork(cfd, nullptr,
-                                                mutable_cf_options);
-
-      // Update internal stats for new ingested files
-      uint64_t total_keys = 0;
-      uint64_t total_l0_files = 0;
-      for (size_t i = 0; i < num_files; i++) {
-        InternalStats::CompactionStats stats(1);
-        stats.micros = micro_list[i];
-        stats.bytes_written = meta_list[i].fd.GetFileSize();
-        stats.num_output_files = 1;
-        cfd->internal_stats()->AddCompactionStats(target_level_list[i], stats);
-        cfd->internal_stats()->AddCFStats(
-            InternalStats::BYTES_INGESTED_ADD_FILE,
-            meta_list[i].fd.GetFileSize());
-        total_keys += file_info_list[i].num_entries;
-        if (target_level_list[i] == 0) {
-          total_l0_files += 1;
-        }
-        Log(InfoLogLevel::INFO_LEVEL, immutable_db_options_.info_log,
-            "[AddFile] External SST file %s was ingested in L%d with path %s\n",
-            file_info_list[i].file_path.c_str(), target_level_list[i],
-            db_fname_list[i].c_str());
-      }
-      cfd->internal_stats()->AddCFStats(InternalStats::INGESTED_NUM_KEYS_TOTAL,
-                                        total_keys);
-      cfd->internal_stats()->AddCFStats(InternalStats::INGESTED_NUM_FILES_TOTAL,
-                                        num_files);
-      cfd->internal_stats()->AddCFStats(
-          InternalStats::INGESTED_LEVEL0_NUM_FILES_TOTAL, total_l0_files);
-    }
-
-    for (size_t i = 0; i < num_files; i++) {
-      ReleaseFileNumberFromPendingOutputs(
-          pending_outputs_inserted_elem_list[i]);
-    }
-
-    num_running_addfile_--;
-    if (num_running_addfile_ == 0) {
-      bg_cv_.SignalAll();
-    }
-    TEST_SYNC_POINT("DBImpl::AddFile:MutexUnlock");
-  }  // mutex_ is unlocked here;
-
-  if (!status.ok()) {
-    // We failed to add the files to the database
-    for (size_t i = 0; i < num_files; i++) {
-      Status s = env_->DeleteFile(db_fname_list[i]);
-      if (!s.ok()) {
-        Log(InfoLogLevel::WARN_LEVEL, immutable_db_options_.info_log,
-            "AddFile() clean up for file %s failed : %s",
-            db_fname_list[i].c_str(), s.ToString().c_str());
-      }
-    }
-  } else if (status.ok() && move_file) {
-    // The files were moved and added successfully, remove original file links
-    for (size_t i = 0; i < num_files; i++) {
-      Status s = env_->DeleteFile(file_info_list[i].file_path);
-      if (!s.ok()) {
-        Log(InfoLogLevel::WARN_LEVEL, immutable_db_options_.info_log,
-            "%s was added to DB successfully but failed to remove original "
-            "file "
-            "link : %s",
-            file_info_list[i].file_path.c_str(), s.ToString().c_str());
-      }
-    }
-  }
-  return status;
-}
-
-// Finds the lowest level in the DB that the ingested file can be added to
-int DBImpl::PickLevelForIngestedFile(ColumnFamilyData* cfd,
-                                     const ExternalSstFileInfo& file_info) {
-  mutex_.AssertHeld();
-
-  int target_level = 0;
-  auto* vstorage = cfd->current()->storage_info();
-  Slice file_smallest_user_key(file_info.smallest_key);
-  Slice file_largest_user_key(file_info.largest_key);
-
-  for (int lvl = cfd->NumberLevels() - 1; lvl >= vstorage->base_level();
-       lvl--) {
-    // Make sure that the file fits in Level `lvl` and dont overlap with
-    // the output of any compaction running right now.
-    if (vstorage->OverlapInLevel(lvl, &file_smallest_user_key,
-                                 &file_largest_user_key) == false &&
-        cfd->RangeOverlapWithCompaction(file_smallest_user_key,
-                                        file_largest_user_key, lvl) == false) {
-      // Level lvl is the lowest level that dont have any files with key
-      // range overlapping with our file key range and no compactions
-      // planning to add overlapping files in it.
-      target_level = lvl;
-      break;
-    }
-  }
-
-  return target_level;
-}
-
-void DBImpl::WaitForAddFile() {
-  mutex_.AssertHeld();
-  while (num_running_addfile_ > 0) {
-    bg_cv_.Wait();
-  }
-}
-#endif  // ROCKSDB_LITE
-
-}  // namespace rocksdb
@@ -19,6 +19,11 @@ uint64_t DBImpl::TEST_GetLevel0TotalSize() {
  return default_cf_handle_->cfd()->current()->storage_info()->NumLevelBytes(0);
 }

+void DBImpl::TEST_MaybeFlushColumnFamilies() {
+  InstrumentedMutexLock l(&mutex_);
+  MaybeFlushColumnFamilies();
+}
+
 int64_t DBImpl::TEST_MaxNextLevelOverlappingBytes(
    ColumnFamilyHandle* column_family) {
  ColumnFamilyData* cfd;
@@ -8,8 +8,9 @@

 #include "db/compacted_db_impl.h"
 #include "db/db_impl.h"
-#include "db/merge_context.h"
 #include "db/db_iter.h"
+#include "db/merge_context.h"
+#include "db/range_del_aggregator.h"
 #include "util/perf_context_imp.h"

 namespace rocksdb {
@@ -37,11 +38,14 @@ Status DBImplReadOnly::Get(const ReadOptions& read_options,
  auto cfd = cfh->cfd();
  SuperVersion* super_version = cfd->GetSuperVersion();
  MergeContext merge_context;
+  RangeDelAggregator range_del_agg(cfd->internal_comparator(), snapshot);
  LookupKey lkey(key, snapshot);
-  if (super_version->mem->Get(lkey, value, &s, &merge_context)) {
+  if (super_version->mem->Get(lkey, value, &s, &merge_context, &range_del_agg,
+                              read_options)) {
  } else {
    PERF_TIMER_GUARD(get_from_output_files_time);
-    super_version->current->Get(read_options, lkey, value, &s, &merge_context);
+    super_version->current->Get(read_options, lkey, value, &s, &merge_context,
+                                &range_del_agg);
  }
  return s;
 }
@@ -60,8 +64,9 @@ Iterator* DBImplReadOnly::NewIterator(const ReadOptions& read_options,
           : latest_snapshot),
      super_version->mutable_cf_options.max_sequential_skip_in_iterations,
      super_version->version_number);
-  auto internal_iter = NewInternalIterator(
-      read_options, cfd, super_version, db_iter->GetArena());
+  auto internal_iter =
+      NewInternalIterator(read_options, cfd, super_version, db_iter->GetArena(),
+                          db_iter->GetRangeDelAggregator());
  db_iter->SetIterUnderDBIter(internal_iter);
  return db_iter;
 }
@@ -88,8 +93,9 @@ Status DBImplReadOnly::NewIterators(
             : latest_snapshot),
        sv->mutable_cf_options.max_sequential_skip_in_iterations,
        sv->version_number);
-    auto* internal_iter = NewInternalIterator(
-        read_options, cfd, sv, db_iter->GetArena());
+    auto* internal_iter =
+        NewInternalIterator(read_options, cfd, sv, db_iter->GetArena(),
+                            db_iter->GetRangeDelAggregator());
    db_iter->SetIterUnderDBIter(internal_iter);
    iterators->push_back(db_iter);
  }
@@ -79,7 +79,6 @@ class DBImplReadOnly : public DBImpl {
    return Status::NotSupported("Not supported operation in read only mode.");
  }

-#ifndef ROCKSDB_LITE
  virtual Status DisableFileDeletions() override {
    return Status::NotSupported("Not supported operation in read only mode.");
  }
@@ -92,7 +91,6 @@ class DBImplReadOnly : public DBImpl {
                              bool flush_memtable = true) override {
    return Status::NotSupported("Not supported operation in read only mode.");
  }
-#endif  // ROCKSDB_LITE

  using DBImpl::Flush;
  virtual Status Flush(const FlushOptions& options,
@@ -105,6 +103,14 @@ class DBImplReadOnly : public DBImpl {
    return Status::NotSupported("Not supported operation in read only mode.");
  }

+  using DB::IngestExternalFile;
+  virtual Status IngestExternalFile(
+      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.");
+  }
+
 private:
  friend class DB;

@@ -23,6 +23,8 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdate) {
    options.inplace_update_support = true;
    options.env = env_;
    options.write_buffer_size = 100000;
+    options.allow_concurrent_memtable_write = false;
+    Reopen(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Update key with values of smaller size
@@ -45,6 +47,8 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateLargeNewValue) {
    options.inplace_update_support = true;
    options.env = env_;
    options.write_buffer_size = 100000;
+    options.allow_concurrent_memtable_write = false;
+    Reopen(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Update key with values of larger size
@@ -70,6 +74,8 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackSmallerSize) {
    options.write_buffer_size = 100000;
    options.inplace_callback =
      rocksdb::DBTestInPlaceUpdate::updateInPlaceSmallerSize;
+    options.allow_concurrent_memtable_write = false;
+    Reopen(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Update key with values of smaller size
@@ -97,6 +103,8 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackSmallerVarintSize) {
    options.write_buffer_size = 100000;
    options.inplace_callback =
      rocksdb::DBTestInPlaceUpdate::updateInPlaceSmallerVarintSize;
+    options.allow_concurrent_memtable_write = false;
+    Reopen(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Update key with values of smaller varint size
@@ -124,6 +132,8 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackLargeNewValue) {
    options.write_buffer_size = 100000;
    options.inplace_callback =
      rocksdb::DBTestInPlaceUpdate::updateInPlaceLargerSize;
+    options.allow_concurrent_memtable_write = false;
+    Reopen(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Update key with values of larger size
@@ -148,7 +158,9 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackNoAction) {
    options.env = env_;
    options.write_buffer_size = 100000;
    options.inplace_callback =
-      rocksdb::DBTestInPlaceUpdate::updateInPlaceNoAction;
+        rocksdb::DBTestInPlaceUpdate::updateInPlaceNoAction;
+    options.allow_concurrent_memtable_write = false;
+    Reopen(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Callback function requests no actions from db
@@ -33,7 +33,7 @@ TEST_F(DBIOFailureTest, DropWrites) {
    // Force out-of-space errors
    env_->drop_writes_.store(true, std::memory_order_release);
    env_->sleep_counter_.Reset();
-    env_->no_sleep_ = true;
+    env_->no_slowdown_ = true;
    for (int i = 0; i < 5; i++) {
      if (option_config_ != kUniversalCompactionMultiLevel &&
          option_config_ != kUniversalSubcompactions) {
@@ -122,7 +122,9 @@ class DBIter: public Iterator {
        iterate_upper_bound_(iterate_upper_bound),
        prefix_same_as_start_(prefix_same_as_start),
        pin_thru_lifetime_(pin_data),
-        total_order_seek_(total_order_seek) {
+        total_order_seek_(total_order_seek),
+        range_del_agg_(ioptions.internal_comparator, s,
+                       true /* collapse_deletions */) {
    RecordTick(statistics_, NO_ITERATORS);
    prefix_extractor_ = ioptions.prefix_extractor;
    max_skip_ = max_sequential_skip_in_iterations;
@@ -151,6 +153,10 @@ class DBIter: public Iterator {
    iter_ = iter;
    iter_->SetPinnedItersMgr(&pinned_iters_mgr_);
  }
+  virtual RangeDelAggregator* GetRangeDelAggregator() {
+    return &range_del_agg_;
+  }
+
  virtual bool Valid() const override { return valid_; }
  virtual Slice key() const override {
    assert(valid_);
@@ -273,6 +279,7 @@ class DBIter: public Iterator {
  const bool total_order_seek_;
  // List of operands for merge operator.
  MergeContext merge_context_;
+  RangeDelAggregator range_del_agg_;
  LocalStatistics local_stats_;
  PinnedIteratorsManager pinned_iters_mgr_;

@@ -333,7 +340,8 @@ void DBIter::Next() {
 //           saved_value_             => the merged value
 //
 // NOTE: In between, saved_key_ can point to a user key that has
-//       a delete marker
+//       a delete marker or a sequence number higher than sequence_
+//       saved_key_ MUST have a proper user_key before calling this function
 //
 // The prefix_check parameter controls whether we check the iterated
 // keys against the prefix of the seeked key. Set to false when
@@ -350,71 +358,140 @@ void DBIter::FindNextUserEntryInternal(bool skipping, bool prefix_check) {
  assert(iter_->Valid());
  assert(direction_ == kForward);
  current_entry_is_merged_ = false;
+
+  // How many times in a row we have skipped an entry with user key less than
+  // or equal to saved_key_. We could skip these entries either because
+  // sequence numbers were too high or because skipping = true.
+  // What saved_key_ contains throughout this method:
+  //  - if skipping        : saved_key_ contains the key that we need to skip,
+  //                         and we haven't seen any keys greater than that,
+  //  - if num_skipped > 0 : saved_key_ contains the key that we have skipped
+  //                         num_skipped times, and we haven't seen any keys
+  //                         greater than that,
+  //  - none of the above  : saved_key_ can contain anything, it doesn't matter.
  uint64_t num_skipped = 0;
+
  do {
    ParsedInternalKey ikey;

-    if (ParseKey(&ikey)) {
-      if (iterate_upper_bound_ != nullptr &&
-          user_comparator_->Compare(ikey.user_key, *iterate_upper_bound_) >= 0) {
-        break;
-      }
+    if (!ParseKey(&ikey)) {
+      // Skip corrupted keys.
+      iter_->Next();
+      continue;
+    }

-      if (prefix_extractor_ && prefix_check &&
-          prefix_extractor_->Transform(ikey.user_key).compare(prefix_start_key_) != 0) {
-        break;
-      }
+    if (iterate_upper_bound_ != nullptr &&
+        user_comparator_->Compare(ikey.user_key, *iterate_upper_bound_) >= 0) {
+      break;
+    }

-      if (ikey.sequence <= sequence_) {
-        if (skipping &&
-           user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) <= 0) {
-          num_skipped++;  // skip this entry
-          PERF_COUNTER_ADD(internal_key_skipped_count, 1);
-        } else {
-          switch (ikey.type) {
-            case kTypeDeletion:
-            case kTypeSingleDeletion:
+    if (prefix_extractor_ && prefix_check &&
+        prefix_extractor_->Transform(ikey.user_key)
+          .compare(prefix_start_key_) != 0) {
+      break;
+    }
+
+    if (ikey.sequence <= sequence_) {
+      if (skipping &&
+          user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) <= 0) {
+        num_skipped++;  // skip this entry
+        PERF_COUNTER_ADD(internal_key_skipped_count, 1);
+      } else {
+        num_skipped = 0;
+        switch (ikey.type) {
+          case kTypeDeletion:
+          case kTypeSingleDeletion:
+            // Arrange to skip all upcoming entries for this key since
+            // they are hidden by this deletion.
+            saved_key_.SetKey(
+                ikey.user_key,
+                !iter_->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
+            skipping = true;
+            PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
+            break;
+          case kTypeValue:
+            saved_key_.SetKey(
+                ikey.user_key,
+                !iter_->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
+            if (range_del_agg_.ShouldDelete(
+                    ikey, RangeDelAggregator::RangePositioningMode::
+                              kForwardTraversal)) {
              // Arrange to skip all upcoming entries for this key since
              // they are hidden by this deletion.
-              saved_key_.SetKey(
-                  ikey.user_key,
-                  !iter_->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
              skipping = true;
              num_skipped = 0;
              PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
-              break;
-            case kTypeValue:
+            } else {
              valid_ = true;
-              saved_key_.SetKey(
-                  ikey.user_key,
-                  !iter_->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
              return;
-            case kTypeMerge:
-              // By now, we are sure the current ikey is going to yield a value
-              saved_key_.SetKey(
-                  ikey.user_key,
-                  !iter_->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
+            }
+            break;
+          case kTypeMerge:
+            saved_key_.SetKey(
+                ikey.user_key,
+                !iter_->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
+            if (range_del_agg_.ShouldDelete(
+                    ikey, RangeDelAggregator::RangePositioningMode::
+                              kForwardTraversal)) {
+              // Arrange to skip all upcoming entries for this key since
+              // they are hidden by this deletion.
+              skipping = true;
+              num_skipped = 0;
+              PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
+            } else {
+              // By now, we are sure the current ikey is going to yield a
+              // value
              current_entry_is_merged_ = true;
              valid_ = true;
              MergeValuesNewToOld();  // Go to a different state machine
              return;
-            default:
-              assert(false);
-              break;
-          }
+            }
+            break;
+          default:
+            assert(false);
+            break;
        }
      }
+    } else {
+      // This key was inserted after our snapshot was taken.
+      PERF_COUNTER_ADD(internal_recent_skipped_count, 1);
+
+      // Here saved_key_ may contain some old key, or the default empty key, or
+      // key assigned by some random other method. We don't care.
+      if (user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) <= 0) {
+        num_skipped++;
+      } else {
+        saved_key_.SetKey(
+                ikey.user_key,
+                !iter_->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
+        skipping = false;
+        num_skipped = 0;
+      }
    }
-    // If we have sequentially iterated via numerous keys and still not
-    // found the next user-key, then it is better to seek so that we can
-    // avoid too many key comparisons. We seek to the last occurrence of
-    // our current key by looking for sequence number 0 and type deletion
-    // (the smallest type).
-    if (skipping && num_skipped > max_skip_) {
+
+    // 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_) {
      num_skipped = 0;
      std::string last_key;
-      AppendInternalKey(&last_key, ParsedInternalKey(saved_key_.GetKey(), 0,
-                                                     kTypeDeletion));
+      if (skipping) {
+        // We're looking for the next user-key but all we see are the same
+        // user-key with decreasing sequence numbers. Fast forward to
+        // sequence number 0 and type deletion (the smallest type).
+        AppendInternalKey(&last_key, ParsedInternalKey(saved_key_.GetKey(), 0,
+                                                       kTypeDeletion));
+        // Don't set skipping = false because we may still see more user-keys
+        // equal to saved_key_.
+      } else {
+        // We saw multiple entries with this user key and sequence numbers
+        // higher than sequence_. Fast forward to sequence_.
+        // Note that this only covers a case when a higher key was overwritten
+        // many times since our snapshot was taken, not the case when a lot of
+        // different keys were inserted after our snapshot was taken.
+        AppendInternalKey(&last_key,
+                          ParsedInternalKey(saved_key_.GetKey(), sequence_,
+                                            kValueTypeForSeek));
+      }
      iter_->Seek(last_key);
      RecordTick(statistics_, NUMBER_OF_RESEEKS_IN_ITERATION);
    } else {
@@ -447,6 +524,7 @@ void DBIter::MergeValuesNewToOld() {
                             iter_->IsValuePinned() /* operand_pinned */);

  ParsedInternalKey ikey;
+  Status s;
  for (iter_->Next(); iter_->Valid(); iter_->Next()) {
    if (!ParseKey(&ikey)) {
      // skip corrupted key
@@ -456,7 +534,10 @@ void DBIter::MergeValuesNewToOld() {
    if (!user_comparator_->Equal(ikey.user_key, saved_key_.GetKey())) {
      // hit the next user key, stop right here
      break;
-    } else if (kTypeDeletion == ikey.type || kTypeSingleDeletion == ikey.type) {
+    } else if (kTypeDeletion == ikey.type || kTypeSingleDeletion == ikey.type ||
+               range_del_agg_.ShouldDelete(
+                   ikey, RangeDelAggregator::RangePositioningMode::
+                             kForwardTraversal)) {
      // hit a delete with the same user key, stop right here
      // iter_ is positioned after delete
      iter_->Next();
@@ -466,9 +547,12 @@ void DBIter::MergeValuesNewToOld() {
      // final result in saved_value_. We are done!
      // ignore corruption if there is any.
      const Slice val = iter_->value();
-      MergeHelper::TimedFullMerge(merge_operator_, ikey.user_key, &val,
-                                  merge_context_.GetOperands(), &saved_value_,
-                                  logger_, statistics_, env_, &pinned_value_);
+      s = MergeHelper::TimedFullMerge(
+          merge_operator_, ikey.user_key, &val, merge_context_.GetOperands(),
+          &saved_value_, logger_, statistics_, env_, &pinned_value_);
+      if (!s.ok()) {
+        status_ = s;
+      }
      // iter_ is positioned after put
      iter_->Next();
      return;
@@ -477,6 +561,7 @@ void DBIter::MergeValuesNewToOld() {
      // when complete, add result to operands and continue.
      merge_context_.PushOperand(iter_->value(),
                                 iter_->IsValuePinned() /* operand_pinned */);
+      PERF_COUNTER_ADD(internal_merge_count, 1);
    } else {
      assert(false);
    }
@@ -486,9 +571,12 @@ void DBIter::MergeValuesNewToOld() {
  // a deletion marker.
  // feed null as the existing value to the merge operator, such that
  // client can differentiate this scenario and do things accordingly.
-  MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetKey(), nullptr,
-                              merge_context_.GetOperands(), &saved_value_,
-                              logger_, statistics_, env_, &pinned_value_);
+  s = MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetKey(), nullptr,
+                                  merge_context_.GetOperands(), &saved_value_,
+                                  logger_, statistics_, env_, &pinned_value_);
+  if (!s.ok()) {
+    status_ = s;
+  }
 }

 void DBIter::Prev() {
@@ -518,6 +606,7 @@ void DBIter::ReverseToForward() {
  direction_ = kForward;
  if (!iter_->Valid()) {
    iter_->SeekToFirst();
+    range_del_agg_.InvalidateTombstoneMapPositions();
  }
 }

@@ -533,11 +622,17 @@ void DBIter::ReverseToBackward() {
    // previous key.
    if (!iter_->Valid()) {
      iter_->SeekToLast();
+      range_del_agg_.InvalidateTombstoneMapPositions();
    }
    ParsedInternalKey ikey;
    FindParseableKey(&ikey, kReverse);
    while (iter_->Valid() &&
           user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) > 0) {
+      if (ikey.sequence > sequence_) {
+        PERF_COUNTER_ADD(internal_recent_skipped_count, 1);
+      } else {
+        PERF_COUNTER_ADD(internal_key_skipped_count, 1);
+      }
      iter_->Prev();
      FindParseableKey(&ikey, kReverse);
    }
@@ -624,10 +719,17 @@ bool DBIter::FindValueForCurrentKey() {
    last_key_entry_type = ikey.type;
    switch (last_key_entry_type) {
      case kTypeValue:
+        if (range_del_agg_.ShouldDelete(
+                ikey,
+                RangeDelAggregator::RangePositioningMode::kBackwardTraversal)) {
+          last_key_entry_type = kTypeRangeDeletion;
+          PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
+        } else {
+          assert(iter_->IsValuePinned());
+          pinned_value_ = iter_->value();
+        }
        merge_context_.Clear();
-        assert(iter_->IsValuePinned());
-        pinned_value_ = iter_->value();
-        last_not_merge_type = kTypeValue;
+        last_not_merge_type = last_key_entry_type;
        break;
      case kTypeDeletion:
      case kTypeSingleDeletion:
@@ -636,9 +738,19 @@ bool DBIter::FindValueForCurrentKey() {
        PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
        break;
      case kTypeMerge:
-        assert(merge_operator_ != nullptr);
-        merge_context_.PushOperandBack(
-            iter_->value(), iter_->IsValuePinned() /* operand_pinned */);
+        if (range_del_agg_.ShouldDelete(
+                ikey,
+                RangeDelAggregator::RangePositioningMode::kBackwardTraversal)) {
+          merge_context_.Clear();
+          last_key_entry_type = kTypeRangeDeletion;
+          last_not_merge_type = last_key_entry_type;
+          PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
+        } else {
+          assert(merge_operator_ != nullptr);
+          merge_context_.PushOperandBack(
+              iter_->value(), iter_->IsValuePinned() /* operand_pinned */);
+          PERF_COUNTER_ADD(internal_merge_count, 1);
+        }
        break;
      default:
        assert(false);
@@ -651,25 +763,28 @@ bool DBIter::FindValueForCurrentKey() {
    FindParseableKey(&ikey, kReverse);
  }

+  Status s;
  switch (last_key_entry_type) {
    case kTypeDeletion:
    case kTypeSingleDeletion:
+    case kTypeRangeDeletion:
      valid_ = false;
      return false;
    case kTypeMerge:
      current_entry_is_merged_ = true;
      if (last_not_merge_type == kTypeDeletion ||
-          last_not_merge_type == kTypeSingleDeletion) {
-        MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetKey(),
-                                    nullptr, merge_context_.GetOperands(),
-                                    &saved_value_, logger_, statistics_, env_,
-                                    &pinned_value_);
+          last_not_merge_type == kTypeSingleDeletion ||
+          last_not_merge_type == kTypeRangeDeletion) {
+        s = MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetKey(),
+                                        nullptr, merge_context_.GetOperands(),
+                                        &saved_value_, logger_, statistics_,
+                                        env_, &pinned_value_);
      } else {
        assert(last_not_merge_type == kTypeValue);
-        MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetKey(),
-                                    &pinned_value_,
-                                    merge_context_.GetOperands(), &saved_value_,
-                                    logger_, statistics_, env_, &pinned_value_);
+        s = MergeHelper::TimedFullMerge(
+            merge_operator_, saved_key_.GetKey(), &pinned_value_,
+            merge_context_.GetOperands(), &saved_value_, logger_, statistics_,
+            env_, &pinned_value_);
      }
      break;
    case kTypeValue:
@@ -680,6 +795,9 @@ bool DBIter::FindValueForCurrentKey() {
      break;
  }
  valid_ = true;
+  if (!s.ok()) {
+    status_ = s;
+  }
  return true;
 }

@@ -699,37 +817,46 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
  ParsedInternalKey ikey;
  FindParseableKey(&ikey, kForward);

-  if (ikey.type == kTypeValue || ikey.type == kTypeDeletion ||
-      ikey.type == kTypeSingleDeletion) {
-    if (ikey.type == kTypeValue) {
-      assert(iter_->IsValuePinned());
-      pinned_value_ = iter_->value();
-      valid_ = true;
-      return true;
-    }
+  if (ikey.type == kTypeDeletion || ikey.type == kTypeSingleDeletion ||
+      range_del_agg_.ShouldDelete(
+          ikey, RangeDelAggregator::RangePositioningMode::kBackwardTraversal)) {
    valid_ = false;
    return false;
  }
+  if (ikey.type == kTypeValue) {
+    assert(iter_->IsValuePinned());
+    pinned_value_ = iter_->value();
+    valid_ = true;
+    return true;
+  }

  // kTypeMerge. We need to collect all kTypeMerge values and save them
  // in operands
  current_entry_is_merged_ = true;
  merge_context_.Clear();
-  while (iter_->Valid() &&
-         user_comparator_->Equal(ikey.user_key, saved_key_.GetKey()) &&
-         ikey.type == kTypeMerge) {
+  while (
+      iter_->Valid() &&
+      user_comparator_->Equal(ikey.user_key, saved_key_.GetKey()) &&
+      ikey.type == kTypeMerge &&
+      !range_del_agg_.ShouldDelete(
+          ikey, RangeDelAggregator::RangePositioningMode::kBackwardTraversal)) {
    merge_context_.PushOperand(iter_->value(),
                               iter_->IsValuePinned() /* operand_pinned */);
+    PERF_COUNTER_ADD(internal_merge_count, 1);
    iter_->Next();
    FindParseableKey(&ikey, kForward);
  }

+  Status s;
  if (!iter_->Valid() ||
      !user_comparator_->Equal(ikey.user_key, saved_key_.GetKey()) ||
-      ikey.type == kTypeDeletion || ikey.type == kTypeSingleDeletion) {
-    MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetKey(), nullptr,
-                                merge_context_.GetOperands(), &saved_value_,
-                                logger_, statistics_, env_, &pinned_value_);
+      ikey.type == kTypeDeletion || ikey.type == kTypeSingleDeletion ||
+      range_del_agg_.ShouldDelete(
+          ikey, RangeDelAggregator::RangePositioningMode::kBackwardTraversal)) {
+    s = MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetKey(),
+                                    nullptr, merge_context_.GetOperands(),
+                                    &saved_value_, logger_, statistics_, env_,
+                                    &pinned_value_);
    // Make iter_ valid and point to saved_key_
    if (!iter_->Valid() ||
        !user_comparator_->Equal(ikey.user_key, saved_key_.GetKey())) {
@@ -737,14 +864,20 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
      RecordTick(statistics_, NUMBER_OF_RESEEKS_IN_ITERATION);
    }
    valid_ = true;
+    if (!s.ok()) {
+      status_ = s;
+    }
    return true;
  }

  const Slice& val = iter_->value();
-  MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetKey(), &val,
-                              merge_context_.GetOperands(), &saved_value_,
-                              logger_, statistics_, env_, &pinned_value_);
+  s = MergeHelper::TimedFullMerge(merge_operator_, saved_key_.GetKey(), &val,
+                                  merge_context_.GetOperands(), &saved_value_,
+                                  logger_, statistics_, env_, &pinned_value_);
  valid_ = true;
+  if (!s.ok()) {
+    status_ = s;
+  }
  return true;
 }

@@ -789,6 +922,11 @@ void DBIter::FindPrevUserKey() {
        ++num_skipped;
      }
    }
+    if (ikey.sequence > sequence_) {
+      PERF_COUNTER_ADD(internal_recent_skipped_count, 1);
+    } else {
+      PERF_COUNTER_ADD(internal_key_skipped_count, 1);
+    }
    iter_->Prev();
    FindParseableKey(&ikey, kReverse);
  }
@@ -809,12 +947,12 @@ void DBIter::Seek(const Slice& target) {
  StopWatch sw(env_, statistics_, DB_SEEK);
  ReleaseTempPinnedData();
  saved_key_.Clear();
-  // now savved_key is used to store internal key.
  saved_key_.SetInternalKey(target, sequence_);

  {
    PERF_TIMER_GUARD(seek_internal_seek_time);
    iter_->Seek(saved_key_.GetKey());
+    range_del_agg_.InvalidateTombstoneMapPositions();
  }
  RecordTick(statistics_, NUMBER_DB_SEEK);
  if (iter_->Valid()) {
@@ -823,6 +961,9 @@ void DBIter::Seek(const Slice& target) {
    }
    direction_ = kForward;
    ClearSavedValue();
+    // convert the InternalKey to UserKey in saved_key_ before
+    // passed to FindNextUserEntry
+    saved_key_.Reserve(saved_key_.Size() - 8);
    FindNextUserEntry(false /* not skipping */, prefix_same_as_start_);
    if (!valid_) {
      prefix_start_key_.clear();
@@ -854,6 +995,7 @@ void DBIter::SeekForPrev(const Slice& target) {
  {
    PERF_TIMER_GUARD(seek_internal_seek_time);
    iter_->SeekForPrev(saved_key_.GetKey());
+    range_del_agg_.InvalidateTombstoneMapPositions();
  }

  RecordTick(statistics_, NUMBER_DB_SEEK);
@@ -895,10 +1037,13 @@ void DBIter::SeekToFirst() {
  {
    PERF_TIMER_GUARD(seek_internal_seek_time);
    iter_->SeekToFirst();
+    range_del_agg_.InvalidateTombstoneMapPositions();
  }

  RecordTick(statistics_, NUMBER_DB_SEEK);
  if (iter_->Valid()) {
+    saved_key_.SetKey(ExtractUserKey(iter_->key()),
+                      !iter_->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
    FindNextUserEntry(false /* not skipping */, false /* no prefix check */);
    if (statistics_ != nullptr) {
      if (valid_) {
@@ -928,12 +1073,14 @@ void DBIter::SeekToLast() {
  {
    PERF_TIMER_GUARD(seek_internal_seek_time);
    iter_->SeekToLast();
+    range_del_agg_.InvalidateTombstoneMapPositions();
  }
  // When the iterate_upper_bound is set to a value,
  // it will seek to the last key before the
  // ReadOptions.iterate_upper_bound
  if (iter_->Valid() && iterate_upper_bound_ != nullptr) {
    SeekForPrev(*iterate_upper_bound_);
+    range_del_agg_.InvalidateTombstoneMapPositions();
    if (!Valid()) {
      return;
    } else if (user_comparator_->Equal(*iterate_upper_bound_, key())) {
@@ -972,6 +1119,10 @@ ArenaWrappedDBIter::~ArenaWrappedDBIter() { db_iter_->~DBIter(); }

 void ArenaWrappedDBIter::SetDBIter(DBIter* iter) { db_iter_ = iter; }

+RangeDelAggregator* ArenaWrappedDBIter::GetRangeDelAggregator() {
+  return db_iter_->GetRangeDelAggregator();
+}
+
 void ArenaWrappedDBIter::SetIterUnderDBIter(InternalIterator* iter) {
  static_cast<DBIter*>(db_iter_)->SetIter(iter);
 }
@@ -11,6 +11,7 @@
 #include <stdint.h>
 #include <string>
 #include "db/dbformat.h"
+#include "db/range_del_aggregator.h"
 #include "rocksdb/db.h"
 #include "rocksdb/iterator.h"
 #include "util/arena.h"
@@ -46,6 +47,7 @@ class ArenaWrappedDBIter : public Iterator {
  // Get the arena to be used to allocate memory for DBIter to be wrapped,
  // as well as child iterators in it.
  virtual Arena* GetArena() { return &arena_; }
+  virtual RangeDelAggregator* GetRangeDelAggregator();

  // Set the DB Iterator to be wrapped

@@ -16,7 +16,7 @@
 #include "rocksdb/slice.h"
 #include "rocksdb/statistics.h"
 #include "table/iterator_wrapper.h"
-#include "table/merger.h"
+#include "table/merging_iterator.h"
 #include "util/string_util.h"
 #include "util/sync_point.h"
 #include "util/testharness.h"
@@ -1897,6 +1897,66 @@ TEST_F(DBIteratorTest, DBIterator12) {
  ASSERT_FALSE(db_iter->Valid());
 }

+TEST_F(DBIteratorTest, DBIterator13) {
+  Options options;
+  options.merge_operator = nullptr;
+
+  std::string key;
+  key.resize(9);
+  key.assign(9, static_cast<char>(0));
+  key[0] = 'b';
+
+  TestIterator* internal_iter = new TestIterator(BytewiseComparator());
+  internal_iter->AddPut(key, "0");
+  internal_iter->AddPut(key, "1");
+  internal_iter->AddPut(key, "2");
+  internal_iter->AddPut(key, "3");
+  internal_iter->AddPut(key, "4");
+  internal_iter->AddPut(key, "5");
+  internal_iter->AddPut(key, "6");
+  internal_iter->AddPut(key, "7");
+  internal_iter->AddPut(key, "8");
+  internal_iter->Finish();
+
+  std::unique_ptr<Iterator> db_iter(
+      NewDBIterator(env_, ImmutableCFOptions(options), BytewiseComparator(),
+                    internal_iter, 2, 3, 0));
+  db_iter->Seek("b");
+  ASSERT_TRUE(db_iter->Valid());
+  ASSERT_EQ(db_iter->key().ToString(), key);
+  ASSERT_EQ(db_iter->value().ToString(), "2");
+}
+
+TEST_F(DBIteratorTest, DBIterator14) {
+  Options options;
+  options.merge_operator = nullptr;
+
+  std::string key("b");
+  TestIterator* internal_iter = new TestIterator(BytewiseComparator());
+  internal_iter->AddPut("b", "0");
+  internal_iter->AddPut("b", "1");
+  internal_iter->AddPut("b", "2");
+  internal_iter->AddPut("b", "3");
+  internal_iter->AddPut("a", "4");
+  internal_iter->AddPut("a", "5");
+  internal_iter->AddPut("a", "6");
+  internal_iter->AddPut("c", "7");
+  internal_iter->AddPut("c", "8");
+  internal_iter->AddPut("c", "9");
+  internal_iter->Finish();
+
+  std::unique_ptr<Iterator> db_iter(
+      NewDBIterator(env_, ImmutableCFOptions(options), BytewiseComparator(),
+                    internal_iter, 4, 1, 0));
+  db_iter->Seek("b");
+  ASSERT_TRUE(db_iter->Valid());
+  ASSERT_EQ(db_iter->key().ToString(), "b");
+  ASSERT_EQ(db_iter->value().ToString(), "3");
+  db_iter->SeekToFirst();
+  ASSERT_EQ(db_iter->key().ToString(), "a");
+  ASSERT_EQ(db_iter->value().ToString(), "4");
+}
+
 class DBIterWithMergeIterTest : public testing::Test {
 public:
  DBIterWithMergeIterTest()
@@ -7,11 +7,13 @@
 // 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 <functional>
+
 #include "db/db_test_util.h"
+#include "port/port.h"
 #include "port/stack_trace.h"
 #include "rocksdb/iostats_context.h"
 #include "rocksdb/perf_context.h"
-#include "port/port.h"

 namespace rocksdb {

@@ -1642,7 +1644,7 @@ TEST_F(DBIteratorTest, IteratorWithLocalStatistics) {
  std::atomic<uint64_t> total_prev_found(0);
  std::atomic<uint64_t> total_bytes(0);

-  std::vector<std::thread> threads;
+  std::vector<port::Thread> threads;
  std::function<void()> reader_func_next = [&]() {
    Iterator* iter = db_->NewIterator(ReadOptions());

@@ -1698,13 +1700,14 @@ TEST_F(DBIteratorTest, IteratorWithLocalStatistics) {
    t.join();
  }

-  ASSERT_EQ(TestGetTickerCount(options, NUMBER_DB_NEXT), total_next);
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_DB_NEXT), (uint64_t)total_next);
  ASSERT_EQ(TestGetTickerCount(options, NUMBER_DB_NEXT_FOUND),
-            total_next_found);
-  ASSERT_EQ(TestGetTickerCount(options, NUMBER_DB_PREV), total_prev);
+            (uint64_t)total_next_found);
+  ASSERT_EQ(TestGetTickerCount(options, NUMBER_DB_PREV), (uint64_t)total_prev);
  ASSERT_EQ(TestGetTickerCount(options, NUMBER_DB_PREV_FOUND),
-            total_prev_found);
-  ASSERT_EQ(TestGetTickerCount(options, ITER_BYTES_READ), total_bytes);
+            (uint64_t)total_prev_found);
+  ASSERT_EQ(TestGetTickerCount(options, ITER_BYTES_READ), (uint64_t)total_bytes);
+
 }

 TEST_F(DBIteratorTest, ReadAhead) {
@@ -1777,6 +1780,62 @@ TEST_F(DBIteratorTest, ReadAhead) {
  delete iter;
 }

+// Insert a key, create a snapshot iterator, overwrite key lots of times,
+// seek to a smaller key. Expect DBIter to fall back to a seek instead of
+// going through all the overwrites linearly.
+TEST_F(DBIteratorTest, DBIteratorSkipRecentDuplicatesTest) {
+  Options options = CurrentOptions();
+  options.env = env_;
+  options.create_if_missing = true;
+  options.max_sequential_skip_in_iterations = 3;
+  options.prefix_extractor = nullptr;
+  options.write_buffer_size = 1 << 27;  // big enough to avoid flush
+  options.statistics = rocksdb::CreateDBStatistics();
+  DestroyAndReopen(options);
+
+  // Insert.
+  ASSERT_OK(Put("b", "0"));
+
+  // Create iterator.
+  ReadOptions ro;
+  std::unique_ptr<Iterator> iter(db_->NewIterator(ro));
+
+  // Insert a lot.
+  for (int i = 0; i < 100; ++i) {
+    ASSERT_OK(Put("b", std::to_string(i + 1).c_str()));
+  }
+
+#ifndef ROCKSDB_LITE
+  // Check that memtable wasn't flushed.
+  std::string val;
+  ASSERT_TRUE(db_->GetProperty("rocksdb.num-files-at-level0", &val));
+  EXPECT_EQ("0", val);
+#endif
+
+  // Seek iterator to a smaller key.
+  perf_context.Reset();
+  iter->Seek("a");
+  ASSERT_TRUE(iter->Valid());
+  EXPECT_EQ("b", iter->key().ToString());
+  EXPECT_EQ("0", iter->value().ToString());
+
+  // Check that the seek didn't do too much work.
+  // Checks are not tight, just make sure that everything is well below 100.
+  EXPECT_LT(perf_context.internal_key_skipped_count, 4);
+  EXPECT_LT(perf_context.internal_recent_skipped_count, 8);
+  EXPECT_LT(perf_context.seek_on_memtable_count, 10);
+  EXPECT_LT(perf_context.next_on_memtable_count, 10);
+  EXPECT_LT(perf_context.prev_on_memtable_count, 10);
+
+  // Check that iterator did something like what we expect.
+  EXPECT_EQ(perf_context.internal_delete_skipped_count, 0);
+  EXPECT_EQ(perf_context.internal_merge_count, 0);
+  EXPECT_GE(perf_context.internal_recent_skipped_count, 2);
+  EXPECT_GE(perf_context.seek_on_memtable_count, 2);
+  EXPECT_EQ(1, options.statistics->getTickerCount(
+                 NUMBER_OF_RESEEKS_IN_ITERATION));
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -0,0 +1,163 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#include <memory>
+#include <string>
+
+#include "db/db_test_util.h"
+#include "db/memtable.h"
+#include "port/stack_trace.h"
+#include "rocksdb/memtablerep.h"
+#include "rocksdb/slice_transform.h"
+
+namespace rocksdb {
+
+class DBMemTableTest : public DBTestBase {
+ public:
+  DBMemTableTest() : DBTestBase("/db_memtable_test") {}
+};
+
+class MockMemTableRep : public MemTableRep {
+ public:
+  explicit MockMemTableRep(MemTableAllocator* allocator, MemTableRep* rep)
+      : MemTableRep(allocator), rep_(rep), num_insert_with_hint_(0) {}
+
+  virtual KeyHandle Allocate(const size_t len, char** buf) override {
+    return rep_->Allocate(len, buf);
+  }
+
+  virtual void Insert(KeyHandle handle) override {
+    return rep_->Insert(handle);
+  }
+
+  virtual void InsertWithHint(KeyHandle handle, void** hint) override {
+    num_insert_with_hint_++;
+    ASSERT_NE(nullptr, hint);
+    last_hint_in_ = *hint;
+    rep_->InsertWithHint(handle, hint);
+    last_hint_out_ = *hint;
+  }
+
+  virtual bool Contains(const char* key) const override {
+    return rep_->Contains(key);
+  }
+
+  virtual void Get(const LookupKey& k, void* callback_args,
+                   bool (*callback_func)(void* arg,
+                                         const char* entry)) override {
+    rep_->Get(k, callback_args, callback_func);
+  }
+
+  virtual size_t ApproximateMemoryUsage() override {
+    return rep_->ApproximateMemoryUsage();
+  }
+
+  virtual Iterator* GetIterator(Arena* arena) override {
+    return rep_->GetIterator(arena);
+  }
+
+  void* last_hint_in() { return last_hint_in_; }
+  void* last_hint_out() { return last_hint_out_; }
+  int num_insert_with_hint() { return num_insert_with_hint_; }
+
+ private:
+  std::unique_ptr<MemTableRep> rep_;
+  void* last_hint_in_;
+  void* last_hint_out_;
+  int num_insert_with_hint_;
+};
+
+class MockMemTableRepFactory : public MemTableRepFactory {
+ public:
+  virtual MemTableRep* CreateMemTableRep(const MemTableRep::KeyComparator& cmp,
+                                         MemTableAllocator* allocator,
+                                         const SliceTransform* transform,
+                                         Logger* logger) override {
+    SkipListFactory factory;
+    MemTableRep* skiplist_rep =
+        factory.CreateMemTableRep(cmp, allocator, transform, logger);
+    mock_rep_ = new MockMemTableRep(allocator, skiplist_rep);
+    return mock_rep_;
+  }
+
+  virtual const char* Name() const override { return "MockMemTableRepFactory"; }
+
+  MockMemTableRep* rep() { return mock_rep_; }
+
+  bool IsInsertConcurrentlySupported() const override { return false; }
+
+ private:
+  MockMemTableRep* mock_rep_;
+};
+
+class TestPrefixExtractor : public SliceTransform {
+ public:
+  virtual const char* Name() const override { return "TestPrefixExtractor"; }
+
+  virtual Slice Transform(const Slice& key) const override {
+    const char* p = separator(key);
+    if (p == nullptr) {
+      return Slice();
+    }
+    return Slice(key.data(), p - key.data() + 1);
+  }
+
+  virtual bool InDomain(const Slice& key) const override {
+    return separator(key) != nullptr;
+  }
+
+  virtual bool InRange(const Slice& key) const override { return false; }
+
+ private:
+  const char* separator(const Slice& key) const {
+    return reinterpret_cast<const char*>(memchr(key.data(), '_', key.size()));
+  }
+};
+
+TEST_F(DBMemTableTest, InsertWithHint) {
+  Options options;
+  options.allow_concurrent_memtable_write = false;
+  options.create_if_missing = true;
+  options.memtable_factory.reset(new MockMemTableRepFactory());
+  options.memtable_insert_with_hint_prefix_extractor.reset(
+      new TestPrefixExtractor());
+  Reopen(options);
+  MockMemTableRep* rep =
+      reinterpret_cast<MockMemTableRepFactory*>(options.memtable_factory.get())
+          ->rep();
+  ASSERT_OK(Put("foo_k1", "foo_v1"));
+  ASSERT_EQ(nullptr, rep->last_hint_in());
+  void* hint_foo = rep->last_hint_out();
+  ASSERT_OK(Put("foo_k2", "foo_v2"));
+  ASSERT_EQ(hint_foo, rep->last_hint_in());
+  ASSERT_EQ(hint_foo, rep->last_hint_out());
+  ASSERT_OK(Put("foo_k3", "foo_v3"));
+  ASSERT_EQ(hint_foo, rep->last_hint_in());
+  ASSERT_EQ(hint_foo, rep->last_hint_out());
+  ASSERT_OK(Put("bar_k1", "bar_v1"));
+  ASSERT_EQ(nullptr, rep->last_hint_in());
+  void* hint_bar = rep->last_hint_out();
+  ASSERT_NE(hint_foo, hint_bar);
+  ASSERT_OK(Put("bar_k2", "bar_v2"));
+  ASSERT_EQ(hint_bar, rep->last_hint_in());
+  ASSERT_EQ(hint_bar, rep->last_hint_out());
+  ASSERT_EQ(5, rep->num_insert_with_hint());
+  ASSERT_OK(Put("whitelisted", "vvv"));
+  ASSERT_EQ(5, rep->num_insert_with_hint());
+  ASSERT_EQ("foo_v1", Get("foo_k1"));
+  ASSERT_EQ("foo_v2", Get("foo_k2"));
+  ASSERT_EQ("foo_v3", Get("foo_k3"));
+  ASSERT_EQ("bar_v1", Get("bar_k1"));
+  ASSERT_EQ("bar_v2", Get("bar_k2"));
+  ASSERT_EQ("vvv", Get("whitelisted"));
+}
+
+}  // namespace rocksdb
+
+int main(int argc, char** argv) {
+  rocksdb::port::InstallStackTraceHandler();
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}
@@ -0,0 +1,98 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+#include <string>
+#include <vector>
+
+#include "db/db_test_util.h"
+#include "db/forward_iterator.h"
+#include "port/stack_trace.h"
+#include "utilities/merge_operators.h"
+
+namespace rocksdb {
+
+// Test merge operator functionality.
+class DBMergeOperatorTest : public DBTestBase {
+ public:
+  DBMergeOperatorTest() : DBTestBase("/db_merge_operator_test") {}
+};
+
+// A test merge operator mimics put but also fails if one of merge operands is
+// "corrupted".
+class TestPutOperator : public MergeOperator {
+ public:
+  virtual bool FullMergeV2(const MergeOperationInput& merge_in,
+                           MergeOperationOutput* merge_out) const override {
+    if (merge_in.existing_value != nullptr &&
+        *(merge_in.existing_value) == "corrupted") {
+      return false;
+    }
+    for (auto value : merge_in.operand_list) {
+      if (value == "corrupted") {
+        return false;
+      }
+    }
+    merge_out->existing_operand = merge_in.operand_list.back();
+    return true;
+  }
+
+  virtual const char* Name() const override { return "TestPutOperator"; }
+};
+
+TEST_F(DBMergeOperatorTest, MergeErrorOnRead) {
+  Options options;
+  options.create_if_missing = true;
+  options.merge_operator.reset(new TestPutOperator());
+  Reopen(options);
+  ASSERT_OK(Merge("k1", "v1"));
+  ASSERT_OK(Merge("k1", "corrupted"));
+  std::string value;
+  ASSERT_TRUE(db_->Get(ReadOptions(), "k1", &value).IsCorruption());
+  VerifyDBInternal({{"k1", "corrupted"}, {"k1", "v1"}});
+}
+
+TEST_F(DBMergeOperatorTest, MergeErrorOnWrite) {
+  Options options;
+  options.create_if_missing = true;
+  options.merge_operator.reset(new TestPutOperator());
+  options.max_successive_merges = 3;
+  Reopen(options);
+  ASSERT_OK(Merge("k1", "v1"));
+  ASSERT_OK(Merge("k1", "v2"));
+  // Will trigger a merge when hitting max_successive_merges and the merge
+  // will fail. The delta will be inserted nevertheless.
+  ASSERT_OK(Merge("k1", "corrupted"));
+  // Data should stay unmerged after the error.
+  VerifyDBInternal({{"k1", "corrupted"}, {"k1", "v2"}, {"k1", "v1"}});
+}
+
+TEST_F(DBMergeOperatorTest, MergeErrorOnIteration) {
+  Options options;
+  options.create_if_missing = true;
+  options.merge_operator.reset(new TestPutOperator());
+
+  DestroyAndReopen(options);
+  ASSERT_OK(Merge("k1", "v1"));
+  ASSERT_OK(Merge("k1", "corrupted"));
+  ASSERT_OK(Put("k2", "v2"));
+  VerifyDBFromMap({{"k1", ""}, {"k2", "v2"}}, nullptr, false,
+                  {{"k1", Status::Corruption()}});
+  VerifyDBInternal({{"k1", "corrupted"}, {"k1", "v1"}, {"k2", "v2"}});
+
+  DestroyAndReopen(options);
+  ASSERT_OK(Merge("k1", "v1"));
+  ASSERT_OK(Put("k2", "v2"));
+  ASSERT_OK(Merge("k2", "corrupted"));
+  VerifyDBFromMap({{"k1", "v1"}, {"k2", ""}}, nullptr, false,
+                  {{"k2", Status::Corruption()}});
+  VerifyDBInternal({{"k1", "v1"}, {"k2", "corrupted"}, {"k2", "v2"}});
+}
+
+}  // namespace rocksdb
+
+int main(int argc, char** argv) {
+  rocksdb::port::InstallStackTraceHandler();
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}
@@ -64,7 +64,7 @@ class DBOptionsTest : public DBTestBase {
    Options options;
    ImmutableDBOptions db_options(options);
    test::RandomInitCFOptions(&options, rnd);
-    auto sanitized_options = SanitizeOptions(db_options, nullptr, options);
+    auto sanitized_options = SanitizeOptions(db_options, options);
    auto opt_map = GetMutableCFOptionsMap(sanitized_options);
    delete options.compaction_filter;
    return opt_map;
@@ -256,6 +256,107 @@ TEST_F(DBOptionsTest, SetBackgroundCompactionThreads) {
  ASSERT_EQ(3, dbfull()->TEST_BGCompactionsAllowed());
 }

+TEST_F(DBOptionsTest, AvoidFlushDuringShutdown) {
+  Options options;
+  options.create_if_missing = true;
+  options.disable_auto_compactions = true;
+  WriteOptions write_without_wal;
+  write_without_wal.disableWAL = true;
+
+  ASSERT_FALSE(options.avoid_flush_during_shutdown);
+  DestroyAndReopen(options);
+  ASSERT_OK(Put("foo", "v1", write_without_wal));
+  Reopen(options);
+  ASSERT_EQ("v1", Get("foo"));
+  ASSERT_EQ("1", FilesPerLevel());
+
+  DestroyAndReopen(options);
+  ASSERT_OK(Put("foo", "v2", write_without_wal));
+  ASSERT_OK(dbfull()->SetDBOptions({{"avoid_flush_during_shutdown", "true"}}));
+  Reopen(options);
+  ASSERT_EQ("NOT_FOUND", Get("foo"));
+  ASSERT_EQ("", FilesPerLevel());
+}
+
+TEST_F(DBOptionsTest, SetDelayedWriteRateOption) {
+  Options options;
+  options.create_if_missing = true;
+  options.delayed_write_rate = 2 * 1024U * 1024U;
+  Reopen(options);
+  ASSERT_EQ(2 * 1024U * 1024U, dbfull()->TEST_write_controler().max_delayed_write_rate());
+
+  ASSERT_OK(dbfull()->SetDBOptions({{"delayed_write_rate", "20000"}}));
+  ASSERT_EQ(20000, dbfull()->TEST_write_controler().max_delayed_write_rate());
+}
+
+TEST_F(DBOptionsTest, MaxTotalWalSizeChange) {
+  Random rnd(1044);
+  const auto value_size = size_t(1024);
+  std::string value;
+  test::RandomString(&rnd, value_size, &value);
+
+  Options options;
+  options.create_if_missing = true;
+  CreateColumnFamilies({"1", "2", "3"}, options);
+  ReopenWithColumnFamilies({"default", "1", "2", "3"}, options);
+
+  WriteOptions write_options;
+
+  const int key_count = 100;
+  for (int i = 0; i < key_count; ++i) {
+    for (size_t cf = 0; cf < handles_.size(); ++cf) {
+      ASSERT_OK(Put(static_cast<int>(cf), Key(i), value));
+    }
+  }
+  ASSERT_OK(dbfull()->SetDBOptions({{"max_total_wal_size", "10"}}));
+
+  for (size_t cf = 0; cf < handles_.size(); ++cf) {
+    dbfull()->TEST_WaitForFlushMemTable(handles_[cf]);
+    ASSERT_EQ("1", FilesPerLevel(static_cast<int>(cf)));
+  }
+}
+
+static void assert_candidate_files_empty(DBImpl* dbfull, const bool empty) {
+  dbfull->TEST_LockMutex();
+  JobContext job_context(0);
+  dbfull->FindObsoleteFiles(&job_context, false);
+  ASSERT_EQ(empty, job_context.full_scan_candidate_files.empty());
+  job_context.Clean();
+  dbfull->TEST_UnlockMutex();
+}
+
+TEST_F(DBOptionsTest, DeleteObsoleteFilesPeriodChange) {
+  SpecialEnv env(Env::Default());
+  env.time_elapse_only_sleep_ = true;
+  Options options;
+  options.env = &env;
+  options.create_if_missing = true;
+  ASSERT_OK(TryReopen(options));
+
+  // Verify that candidate files set is empty when no full scan requested.
+  assert_candidate_files_empty(dbfull(), true);
+
+  ASSERT_OK(
+      dbfull()->SetDBOptions({{"delete_obsolete_files_period_micros", "0"}}));
+
+  // After delete_obsolete_files_period_micros updated to 0, the next call
+  // to FindObsoleteFiles should make a full scan
+  assert_candidate_files_empty(dbfull(), false);
+
+  ASSERT_OK(
+      dbfull()->SetDBOptions({{"delete_obsolete_files_period_micros", "20"}}));
+
+  assert_candidate_files_empty(dbfull(), true);
+
+  env.addon_time_.store(20);
+  assert_candidate_files_empty(dbfull(), true);
+
+  env.addon_time_.store(21);
+  assert_candidate_files_empty(dbfull(), false);
+
+  Close();
+}
+
 #endif  // ROCKSDB_LITE

 }  // namespace rocksdb
@@ -34,6 +34,7 @@ TEST_F(DBPropertiesTest, Empty) {
    Options options;
    options.env = env_;
    options.write_buffer_size = 100000;  // Small write buffer
+    options.allow_concurrent_memtable_write = false;
    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

@@ -491,6 +492,7 @@ TEST_F(DBPropertiesTest, NumImmutableMemTable) {

    std::string big_value(1000000 * 2, 'x');
    std::string num;
+    uint64_t value;
    SetPerfLevel(kEnableTime);
    ASSERT_TRUE(GetPerfLevel() == kEnableTime);

@@ -555,11 +557,11 @@ TEST_F(DBPropertiesTest, NumImmutableMemTable) {
    ASSERT_TRUE(dbfull()->GetProperty(
        handles_[1], DB::Properties::kNumImmutableMemTableFlushed, &num));
    ASSERT_EQ(num, "3");
-    ASSERT_TRUE(dbfull()->GetProperty(
-        handles_[1], "rocksdb.cur-size-active-mem-table", &num));
-    // "384" is the size of the metadata of two empty skiplists, this would
-    // break if we change the default vectorrep/skiplist implementation
-    ASSERT_EQ(num, "384");
+    ASSERT_TRUE(dbfull()->GetIntProperty(
+        handles_[1], "rocksdb.cur-size-active-mem-table", &value));
+    // "192" is the size of the metadata of two empty skiplists, this would
+    // break if we change the default skiplist implementation
+    ASSERT_GE(value, 192);

    uint64_t int_num;
    uint64_t base_total_size;
@@ -0,0 +1,818 @@
+//  Copyright (c) 2016-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#include "db/db_test_util.h"
+#include "port/stack_trace.h"
+#include "util/testutil.h"
+#include "utilities/merge_operators.h"
+
+namespace rocksdb {
+
+class DBRangeDelTest : public DBTestBase {
+ public:
+  DBRangeDelTest() : DBTestBase("/db_range_del_test") {}
+
+  std::string GetNumericStr(int key) {
+    uint64_t uint64_key = static_cast<uint64_t>(key);
+    std::string str;
+    str.resize(8);
+    memcpy(&str[0], static_cast<void*>(&uint64_key), 8);
+    return str;
+  }
+};
+
+// PlainTableFactory and NumTableFilesAtLevel() are not supported in
+// ROCKSDB_LITE
+#ifndef ROCKSDB_LITE
+TEST_F(DBRangeDelTest, NonBlockBasedTableNotSupported) {
+  Options opts = CurrentOptions();
+  opts.table_factory.reset(new PlainTableFactory());
+  opts.prefix_extractor.reset(NewNoopTransform());
+  opts.allow_mmap_reads = true;
+  opts.max_sequential_skip_in_iterations = 999999;
+  Reopen(opts);
+
+  ASSERT_TRUE(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "dr1", "dr1")
+          .IsNotSupported());
+}
+
+TEST_F(DBRangeDelTest, FlushOutputHasOnlyRangeTombstones) {
+  ASSERT_OK(db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "dr1",
+                             "dr2"));
+  ASSERT_OK(db_->Flush(FlushOptions()));
+  ASSERT_EQ(1, NumTableFilesAtLevel(0));
+}
+
+TEST_F(DBRangeDelTest, CompactionOutputHasOnlyRangeTombstone) {
+  Options opts = CurrentOptions();
+  opts.disable_auto_compactions = true;
+  opts.statistics = CreateDBStatistics();
+  Reopen(opts);
+
+  // snapshot protects range tombstone from dropping due to becoming obsolete.
+  const Snapshot* snapshot = db_->GetSnapshot();
+  db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "z");
+  db_->Flush(FlushOptions());
+
+  ASSERT_EQ(1, NumTableFilesAtLevel(0));
+  ASSERT_EQ(0, NumTableFilesAtLevel(1));
+  dbfull()->TEST_CompactRange(0, nullptr, nullptr, nullptr,
+                              true /* disallow_trivial_move */);
+  ASSERT_EQ(0, NumTableFilesAtLevel(0));
+  ASSERT_EQ(1, NumTableFilesAtLevel(1));
+  ASSERT_EQ(0, TestGetTickerCount(opts, COMPACTION_RANGE_DEL_DROP_OBSOLETE));
+  db_->ReleaseSnapshot(snapshot);
+}
+
+TEST_F(DBRangeDelTest, CompactionOutputFilesExactlyFilled) {
+  // regression test for exactly filled compaction output files. Previously
+  // another file would be generated containing all range deletions, which
+  // could invalidate the non-overlapping file boundary invariant.
+  const int kNumPerFile = 4, kNumFiles = 2, kFileBytes = 9 << 10;
+  Options options = CurrentOptions();
+  options.disable_auto_compactions = true;
+  options.level0_file_num_compaction_trigger = kNumFiles;
+  options.memtable_factory.reset(new SpecialSkipListFactory(kNumPerFile));
+  options.num_levels = 2;
+  options.target_file_size_base = kFileBytes;
+  BlockBasedTableOptions table_options;
+  table_options.block_size_deviation = 50;  // each block holds two keys
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  Reopen(options);
+
+  // snapshot protects range tombstone from dropping due to becoming obsolete.
+  const Snapshot* snapshot = db_->GetSnapshot();
+  db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), Key(0), Key(1));
+
+  Random rnd(301);
+  for (int i = 0; i < kNumFiles; ++i) {
+    std::vector<std::string> values;
+    // Write 12K (4 values, each 3K)
+    for (int j = 0; j < kNumPerFile; j++) {
+      values.push_back(RandomString(&rnd, 3 << 10));
+      ASSERT_OK(Put(Key(i * kNumPerFile + j), values[j]));
+      if (j == 0 && i > 0) {
+        dbfull()->TEST_WaitForFlushMemTable();
+      }
+    }
+  }
+  // put extra key to trigger final flush
+  ASSERT_OK(Put("", ""));
+  dbfull()->TEST_WaitForFlushMemTable();
+  ASSERT_EQ(kNumFiles, NumTableFilesAtLevel(0));
+  ASSERT_EQ(0, NumTableFilesAtLevel(1));
+
+  dbfull()->TEST_CompactRange(0, nullptr, nullptr, nullptr,
+                              true /* disallow_trivial_move */);
+  ASSERT_EQ(0, NumTableFilesAtLevel(0));
+  ASSERT_EQ(2, NumTableFilesAtLevel(1));
+  db_->ReleaseSnapshot(snapshot);
+}
+
+TEST_F(DBRangeDelTest, MaxCompactionBytesCutsOutputFiles) {
+  // Ensures range deletion spanning multiple compaction output files that are
+  // cut by max_compaction_bytes will have non-overlapping key-ranges.
+  // https://github.com/facebook/rocksdb/issues/1778
+  const int kNumFiles = 2, kNumPerFile = 1 << 8, kBytesPerVal = 1 << 12;
+  Options opts = CurrentOptions();
+  opts.comparator = test::Uint64Comparator();
+  opts.disable_auto_compactions = true;
+  opts.level0_file_num_compaction_trigger = kNumFiles;
+  opts.max_compaction_bytes = kNumPerFile * kBytesPerVal;
+  opts.memtable_factory.reset(new SpecialSkipListFactory(kNumPerFile));
+  // Want max_compaction_bytes to trigger the end of compaction output file, not
+  // target_file_size_base, so make the latter much bigger
+  opts.target_file_size_base = 100 * opts.max_compaction_bytes;
+  Reopen(opts);
+
+  // snapshot protects range tombstone from dropping due to becoming obsolete.
+  const Snapshot* snapshot = db_->GetSnapshot();
+
+  // It spans the whole key-range, thus will be included in all output files
+  ASSERT_OK(db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(),
+                             GetNumericStr(0),
+                             GetNumericStr(kNumFiles * kNumPerFile - 1)));
+  Random rnd(301);
+  for (int i = 0; i < kNumFiles; ++i) {
+    std::vector<std::string> values;
+    // Write 1MB (256 values, each 4K)
+    for (int j = 0; j < kNumPerFile; j++) {
+      values.push_back(RandomString(&rnd, kBytesPerVal));
+      ASSERT_OK(Put(GetNumericStr(kNumPerFile * i + j), values[j]));
+    }
+    // extra entry to trigger SpecialSkipListFactory's flush
+    ASSERT_OK(Put(GetNumericStr(kNumPerFile), ""));
+    dbfull()->TEST_WaitForFlushMemTable();
+    ASSERT_EQ(i + 1, NumTableFilesAtLevel(0));
+  }
+
+  dbfull()->TEST_CompactRange(0, nullptr, nullptr, nullptr,
+                              true /* disallow_trivial_move */);
+  ASSERT_EQ(0, NumTableFilesAtLevel(0));
+  ASSERT_GE(NumTableFilesAtLevel(1), 2);
+
+  std::vector<std::vector<FileMetaData>> files;
+  dbfull()->TEST_GetFilesMetaData(db_->DefaultColumnFamily(), &files);
+
+  for (size_t i = 0; i < files[1].size() - 1; ++i) {
+    ASSERT_TRUE(InternalKeyComparator(opts.comparator)
+                    .Compare(files[1][i].largest, files[1][i + 1].smallest) <
+                0);
+  }
+  db_->ReleaseSnapshot(snapshot);
+}
+
+TEST_F(DBRangeDelTest, SentinelsOmittedFromOutputFile) {
+  // Regression test for bug where sentinel range deletions (i.e., ones with
+  // sequence number of zero) were included in output files.
+  // snapshot protects range tombstone from dropping due to becoming obsolete.
+  const Snapshot* snapshot = db_->GetSnapshot();
+
+  // gaps between ranges creates sentinels in our internal representation
+  std::vector<std::pair<std::string, std::string>> range_dels = {{"a", "b"}, {"c", "d"}, {"e", "f"}};
+  for (const auto& range_del : range_dels) {
+    ASSERT_OK(db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(),
+                               range_del.first, range_del.second));
+  }
+  ASSERT_OK(db_->Flush(FlushOptions()));
+  ASSERT_EQ(1, NumTableFilesAtLevel(0));
+
+  std::vector<std::vector<FileMetaData>> files;
+  dbfull()->TEST_GetFilesMetaData(db_->DefaultColumnFamily(), &files);
+  ASSERT_GT(files[0][0].smallest_seqno, 0);
+
+  db_->ReleaseSnapshot(snapshot);
+}
+
+TEST_F(DBRangeDelTest, FlushRangeDelsSameStartKey) {
+  db_->Put(WriteOptions(), "b1", "val");
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "c"));
+  db_->Put(WriteOptions(), "b2", "val");
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "b"));
+  // first iteration verifies query correctness in memtable, second verifies
+  // query correctness for a single SST file
+  for (int i = 0; i < 2; ++i) {
+    if (i > 0) {
+      ASSERT_OK(db_->Flush(FlushOptions()));
+      ASSERT_EQ(1, NumTableFilesAtLevel(0));
+    }
+    std::string value;
+    ASSERT_TRUE(db_->Get(ReadOptions(), "b1", &value).IsNotFound());
+    ASSERT_OK(db_->Get(ReadOptions(), "b2", &value));
+  }
+}
+
+TEST_F(DBRangeDelTest, CompactRangeDelsSameStartKey) {
+  db_->Put(WriteOptions(), "unused", "val");  // prevents empty after compaction
+  db_->Put(WriteOptions(), "b1", "val");
+  ASSERT_OK(db_->Flush(FlushOptions()));
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "c"));
+  ASSERT_OK(db_->Flush(FlushOptions()));
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "b"));
+  ASSERT_OK(db_->Flush(FlushOptions()));
+  ASSERT_EQ(3, NumTableFilesAtLevel(0));
+
+  for (int i = 0; i < 2; ++i) {
+    if (i > 0) {
+      dbfull()->TEST_CompactRange(0, nullptr, nullptr, nullptr,
+                                  true /* disallow_trivial_move */);
+      ASSERT_EQ(0, NumTableFilesAtLevel(0));
+      ASSERT_EQ(1, NumTableFilesAtLevel(1));
+    }
+    std::string value;
+    ASSERT_TRUE(db_->Get(ReadOptions(), "b1", &value).IsNotFound());
+  }
+}
+#endif  // ROCKSDB_LITE
+
+TEST_F(DBRangeDelTest, FlushRemovesCoveredKeys) {
+  const int kNum = 300, kRangeBegin = 50, kRangeEnd = 250;
+  Options opts = CurrentOptions();
+  opts.comparator = test::Uint64Comparator();
+  Reopen(opts);
+
+  // Write a third before snapshot, a third between snapshot and tombstone, and
+  // a third after the tombstone. Keys older than snapshot or newer than the
+  // tombstone should be preserved.
+  const Snapshot* snapshot = nullptr;
+  for (int i = 0; i < kNum; ++i) {
+    if (i == kNum / 3) {
+      snapshot = db_->GetSnapshot();
+    } else if (i == 2 * kNum / 3) {
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(),
+                       GetNumericStr(kRangeBegin), GetNumericStr(kRangeEnd));
+    }
+    db_->Put(WriteOptions(), GetNumericStr(i), "val");
+  }
+  db_->Flush(FlushOptions());
+
+  for (int i = 0; i < kNum; ++i) {
+    ReadOptions read_opts;
+    read_opts.ignore_range_deletions = true;
+    std::string value;
+    if (i < kRangeBegin || i > kRangeEnd || i < kNum / 3 || i >= 2 * kNum / 3) {
+      ASSERT_OK(db_->Get(read_opts, GetNumericStr(i), &value));
+    } else {
+      ASSERT_TRUE(db_->Get(read_opts, GetNumericStr(i), &value).IsNotFound());
+    }
+  }
+  db_->ReleaseSnapshot(snapshot);
+}
+
+// NumTableFilesAtLevel() is not supported in ROCKSDB_LITE
+#ifndef ROCKSDB_LITE
+TEST_F(DBRangeDelTest, CompactionRemovesCoveredKeys) {
+  const int kNumPerFile = 100, kNumFiles = 4;
+  Options opts = CurrentOptions();
+  opts.comparator = test::Uint64Comparator();
+  opts.disable_auto_compactions = true;
+  opts.memtable_factory.reset(new SpecialSkipListFactory(kNumPerFile));
+  opts.num_levels = 2;
+  opts.statistics = CreateDBStatistics();
+  Reopen(opts);
+
+  for (int i = 0; i < kNumFiles; ++i) {
+    if (i > 0) {
+      // range tombstone covers first half of the previous file
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(),
+                       GetNumericStr((i - 1) * kNumPerFile),
+                       GetNumericStr((i - 1) * kNumPerFile + kNumPerFile / 2));
+    }
+    // Make sure a given key appears in each file so compaction won't be able to
+    // use trivial move, which would happen if the ranges were non-overlapping.
+    // Also, we need an extra element since flush is only triggered when the
+    // number of keys is one greater than SpecialSkipListFactory's limit.
+    // We choose a key outside the key-range used by the test to avoid conflict.
+    db_->Put(WriteOptions(), GetNumericStr(kNumPerFile * kNumFiles), "val");
+
+    for (int j = 0; j < kNumPerFile; ++j) {
+      db_->Put(WriteOptions(), GetNumericStr(i * kNumPerFile + j), "val");
+    }
+    dbfull()->TEST_WaitForFlushMemTable();
+    ASSERT_EQ(i + 1, NumTableFilesAtLevel(0));
+  }
+  db_->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+  ASSERT_EQ(0, NumTableFilesAtLevel(0));
+  ASSERT_GT(NumTableFilesAtLevel(1), 0);
+  ASSERT_EQ((kNumFiles - 1) * kNumPerFile / 2,
+            TestGetTickerCount(opts, COMPACTION_KEY_DROP_RANGE_DEL));
+
+  for (int i = 0; i < kNumFiles; ++i) {
+    for (int j = 0; j < kNumPerFile; ++j) {
+      ReadOptions read_opts;
+      read_opts.ignore_range_deletions = true;
+      std::string value;
+      if (i == kNumFiles - 1 || j >= kNumPerFile / 2) {
+        ASSERT_OK(
+            db_->Get(read_opts, GetNumericStr(i * kNumPerFile + j), &value));
+      } else {
+        ASSERT_TRUE(
+            db_->Get(read_opts, GetNumericStr(i * kNumPerFile + j), &value)
+                .IsNotFound());
+      }
+    }
+  }
+}
+
+TEST_F(DBRangeDelTest, ValidLevelSubcompactionBoundaries) {
+  const int kNumPerFile = 100, kNumFiles = 4, kFileBytes = 100 << 10;
+  Options options = CurrentOptions();
+  options.level0_file_num_compaction_trigger = kNumFiles;
+  options.max_bytes_for_level_base = 2 * kFileBytes;
+  options.max_subcompactions = 4;
+  options.memtable_factory.reset(new SpecialSkipListFactory(kNumPerFile));
+  options.num_levels = 3;
+  options.target_file_size_base = kFileBytes;
+  options.target_file_size_multiplier = 1;
+  Reopen(options);
+
+  Random rnd(301);
+  for (int i = 0; i < 2; ++i) {
+    for (int j = 0; j < kNumFiles; ++j) {
+      if (i > 0) {
+        // delete [95,105) in two files, [295,305) in next two
+        int mid = (j + (1 - j % 2)) * kNumPerFile;
+        db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(),
+                         Key(mid - 5), Key(mid + 5));
+      }
+      std::vector<std::string> values;
+      // Write 100KB (100 values, each 1K)
+      for (int k = 0; k < kNumPerFile; k++) {
+        values.push_back(RandomString(&rnd, 990));
+        ASSERT_OK(Put(Key(j * kNumPerFile + k), values[k]));
+      }
+      // put extra key to trigger flush
+      ASSERT_OK(Put("", ""));
+      dbfull()->TEST_WaitForFlushMemTable();
+      if (j < kNumFiles - 1) {
+        // background compaction may happen early for kNumFiles'th file
+        ASSERT_EQ(NumTableFilesAtLevel(0), j + 1);
+      }
+      if (j == options.level0_file_num_compaction_trigger - 1) {
+        // When i == 1, compaction will output some files to L1, at which point
+        // L1 is not bottommost so range deletions cannot be compacted away. The
+        // new L1 files must be generated with non-overlapping key ranges even
+        // though multiple subcompactions see the same ranges deleted, else an
+        // assertion will fail.
+        dbfull()->TEST_WaitForCompact();
+        ASSERT_EQ(NumTableFilesAtLevel(0), 0);
+        ASSERT_GT(NumTableFilesAtLevel(1), 0);
+        ASSERT_GT(NumTableFilesAtLevel(2), 0);
+      }
+    }
+  }
+}
+
+TEST_F(DBRangeDelTest, ValidUniversalSubcompactionBoundaries) {
+  const int kNumPerFile = 100, kFilesPerLevel = 4, kNumLevels = 4;
+  Options options = CurrentOptions();
+  options.compaction_options_universal.min_merge_width = kFilesPerLevel;
+  options.compaction_options_universal.max_merge_width = kFilesPerLevel;
+  options.compaction_options_universal.size_ratio = 10;
+  options.compaction_style = kCompactionStyleUniversal;
+  options.level0_file_num_compaction_trigger = kFilesPerLevel;
+  options.max_subcompactions = 4;
+  options.memtable_factory.reset(new SpecialSkipListFactory(kNumPerFile));
+  options.num_levels = kNumLevels;
+  options.target_file_size_base = kNumPerFile << 10;
+  options.target_file_size_multiplier = 1;
+  Reopen(options);
+
+  Random rnd(301);
+  for (int i = 0; i < kNumLevels - 1; ++i) {
+    for (int j = 0; j < kFilesPerLevel; ++j) {
+      if (i == kNumLevels - 2) {
+        // insert range deletions [95,105) in two files, [295,305) in next two
+        // to prepare L1 for later manual compaction.
+        int mid = (j + (1 - j % 2)) * kNumPerFile;
+        db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(),
+                         Key(mid - 5), Key(mid + 5));
+      }
+      std::vector<std::string> values;
+      // Write 100KB (100 values, each 1K)
+      for (int k = 0; k < kNumPerFile; k++) {
+        values.push_back(RandomString(&rnd, 990));
+        ASSERT_OK(Put(Key(j * kNumPerFile + k), values[k]));
+      }
+      // put extra key to trigger flush
+      ASSERT_OK(Put("", ""));
+      dbfull()->TEST_WaitForFlushMemTable();
+      if (j < kFilesPerLevel - 1) {
+        // background compaction may happen early for kFilesPerLevel'th file
+        ASSERT_EQ(NumTableFilesAtLevel(0), j + 1);
+      }
+    }
+    dbfull()->TEST_WaitForCompact();
+    ASSERT_EQ(NumTableFilesAtLevel(0), 0);
+    ASSERT_GT(NumTableFilesAtLevel(kNumLevels - 1 - i), kFilesPerLevel - 1);
+  }
+  // Now L1-L3 are full, when we compact L1->L2 we should see (1) subcompactions
+  // happen since input level > 0; (2) range deletions are not dropped since
+  // output level is not bottommost. If no file boundary assertion fails, that
+  // probably means universal compaction + subcompaction + range deletion are
+  // compatible.
+  ASSERT_OK(dbfull()->RunManualCompaction(
+      reinterpret_cast<ColumnFamilyHandleImpl*>(db_->DefaultColumnFamily())
+          ->cfd(),
+      1 /* input_level */, 2 /* output_level */, 0 /* output_path_id */,
+      nullptr /* begin */, nullptr /* end */, true /* exclusive */,
+      true /* disallow_trivial_move */));
+}
+#endif  // ROCKSDB_LITE
+
+TEST_F(DBRangeDelTest, CompactionRemovesCoveredMergeOperands) {
+  const int kNumPerFile = 3, kNumFiles = 3;
+  Options opts = CurrentOptions();
+  opts.disable_auto_compactions = true;
+  opts.memtable_factory.reset(new SpecialSkipListFactory(2 * kNumPerFile));
+  opts.merge_operator = MergeOperators::CreateUInt64AddOperator();
+  opts.num_levels = 2;
+  Reopen(opts);
+
+  // Iterates kNumFiles * kNumPerFile + 1 times since flushing the last file
+  // requires an extra entry.
+  for (int i = 0; i <= kNumFiles * kNumPerFile; ++i) {
+    if (i % kNumPerFile == 0 && i / kNumPerFile == kNumFiles - 1) {
+      // Delete merge operands from all but the last file
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "key",
+                       "key_");
+    }
+    std::string val;
+    PutFixed64(&val, i);
+    db_->Merge(WriteOptions(), "key", val);
+    // we need to prevent trivial move using Puts so compaction will actually
+    // process the merge operands.
+    db_->Put(WriteOptions(), "prevent_trivial_move", "");
+    if (i > 0 && i % kNumPerFile == 0) {
+      dbfull()->TEST_WaitForFlushMemTable();
+    }
+  }
+
+  ReadOptions read_opts;
+  read_opts.ignore_range_deletions = true;
+  std::string expected, actual;
+  ASSERT_OK(db_->Get(read_opts, "key", &actual));
+  PutFixed64(&expected, 45);  // 1+2+...+9
+  ASSERT_EQ(expected, actual);
+
+  db_->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+
+  expected.clear();
+  ASSERT_OK(db_->Get(read_opts, "key", &actual));
+  uint64_t tmp;
+  Slice tmp2(actual);
+  GetFixed64(&tmp2, &tmp);
+  PutFixed64(&expected, 30);  // 6+7+8+9 (earlier operands covered by tombstone)
+  ASSERT_EQ(expected, actual);
+}
+
+// NumTableFilesAtLevel() is not supported in ROCKSDB_LITE
+#ifndef ROCKSDB_LITE
+TEST_F(DBRangeDelTest, ObsoleteTombstoneCleanup) {
+  // During compaction to bottommost level, verify range tombstones older than
+  // the oldest snapshot are removed, while others are preserved.
+  Options opts = CurrentOptions();
+  opts.disable_auto_compactions = true;
+  opts.num_levels = 2;
+  opts.statistics = CreateDBStatistics();
+  Reopen(opts);
+
+  db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "dr1",
+                   "dr1");  // obsolete after compaction
+  db_->Put(WriteOptions(), "key", "val");
+  db_->Flush(FlushOptions());
+  const Snapshot* snapshot = db_->GetSnapshot();
+  db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "dr2",
+                   "dr2");  // protected by snapshot
+  db_->Put(WriteOptions(), "key", "val");
+  db_->Flush(FlushOptions());
+
+  ASSERT_EQ(2, NumTableFilesAtLevel(0));
+  ASSERT_EQ(0, NumTableFilesAtLevel(1));
+  db_->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+  ASSERT_EQ(0, NumTableFilesAtLevel(0));
+  ASSERT_EQ(1, NumTableFilesAtLevel(1));
+  ASSERT_EQ(1, TestGetTickerCount(opts, COMPACTION_RANGE_DEL_DROP_OBSOLETE));
+
+  db_->ReleaseSnapshot(snapshot);
+}
+
+TEST_F(DBRangeDelTest, TableEvictedDuringScan) {
+  // The RangeDelAggregator holds pointers into range deletion blocks created by
+  // table readers. This test ensures the aggregator can still access those
+  // blocks even if it outlives the table readers that created them.
+  //
+  // DBIter always keeps readers open for L0 files. So, in order to test
+  // aggregator outliving reader, we need to have deletions in L1 files, which
+  // are opened/closed on-demand during the scan. This is accomplished by
+  // setting kNumRanges > level0_stop_writes_trigger, which prevents deletions
+  // from all lingering in L0 (there is at most one range deletion per L0 file).
+  //
+  // The first L1 file will contain a range deletion since its begin key is 0.
+  // SeekToFirst() references that table's reader and adds its range tombstone
+  // to the aggregator. Upon advancing beyond that table's key-range via Next(),
+  // the table reader will be unreferenced by the iterator. Since we manually
+  // call Evict() on all readers before the full scan, this unreference causes
+  // the reader's refcount to drop to zero and thus be destroyed.
+  //
+  // When it is destroyed, we do not remove its range deletions from the
+  // aggregator. So, subsequent calls to Next() must be able to use these
+  // deletions to decide whether a key is covered. This will work as long as
+  // the aggregator properly references the range deletion block.
+  const int kNum = 25, kRangeBegin = 0, kRangeEnd = 7, kNumRanges = 5;
+  Options opts = CurrentOptions();
+  opts.comparator = test::Uint64Comparator();
+  opts.level0_file_num_compaction_trigger = 4;
+  opts.level0_stop_writes_trigger = 4;
+  opts.memtable_factory.reset(new SpecialSkipListFactory(1));
+  opts.num_levels = 2;
+  BlockBasedTableOptions bbto;
+  bbto.cache_index_and_filter_blocks = true;
+  bbto.block_cache = NewLRUCache(8 << 20);
+  opts.table_factory.reset(NewBlockBasedTableFactory(bbto));
+  Reopen(opts);
+
+  // Hold a snapshot so range deletions can't become obsolete during compaction
+  // to bottommost level (i.e., L1).
+  const Snapshot* snapshot = db_->GetSnapshot();
+  for (int i = 0; i < kNum; ++i) {
+    db_->Put(WriteOptions(), GetNumericStr(i), "val");
+    if (i > 0) {
+      dbfull()->TEST_WaitForFlushMemTable();
+    }
+    if (i >= kNum / 2 && i < kNum / 2 + kNumRanges) {
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(),
+                       GetNumericStr(kRangeBegin), GetNumericStr(kRangeEnd));
+    }
+  }
+  // Must be > 1 so the first L1 file can be closed before scan finishes
+  dbfull()->TEST_WaitForCompact();
+  ASSERT_GT(NumTableFilesAtLevel(1), 1);
+  std::vector<uint64_t> file_numbers = ListTableFiles(env_, dbname_);
+
+  ReadOptions read_opts;
+  auto* iter = db_->NewIterator(read_opts);
+  int expected = kRangeEnd;
+  iter->SeekToFirst();
+  for (auto file_number : file_numbers) {
+    // This puts table caches in the state of being externally referenced only
+    // so they are destroyed immediately upon iterator unreferencing.
+    TableCache::Evict(dbfull()->TEST_table_cache(), file_number);
+  }
+  for (; iter->Valid(); iter->Next()) {
+    ASSERT_EQ(GetNumericStr(expected), iter->key());
+    ++expected;
+    // Keep clearing block cache's LRU so range deletion block can be freed as
+    // soon as its refcount drops to zero.
+    bbto.block_cache->EraseUnRefEntries();
+  }
+  ASSERT_EQ(kNum, expected);
+  delete iter;
+  db_->ReleaseSnapshot(snapshot);
+}
+
+TEST_F(DBRangeDelTest, GetCoveredKeyFromMutableMemtable) {
+  db_->Put(WriteOptions(), "key", "val");
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "z"));
+
+  ReadOptions read_opts;
+  std::string value;
+  ASSERT_TRUE(db_->Get(read_opts, "key", &value).IsNotFound());
+}
+
+TEST_F(DBRangeDelTest, GetCoveredKeyFromImmutableMemtable) {
+  Options opts = CurrentOptions();
+  opts.max_write_buffer_number = 3;
+  opts.min_write_buffer_number_to_merge = 2;
+  // SpecialSkipListFactory lets us specify maximum number of elements the
+  // memtable can hold. It switches the active memtable to immutable (flush is
+  // prevented by the above options) upon inserting an element that would
+  // overflow the memtable.
+  opts.memtable_factory.reset(new SpecialSkipListFactory(1));
+  Reopen(opts);
+
+  db_->Put(WriteOptions(), "key", "val");
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "z"));
+  db_->Put(WriteOptions(), "blah", "val");
+
+  ReadOptions read_opts;
+  std::string value;
+  ASSERT_TRUE(db_->Get(read_opts, "key", &value).IsNotFound());
+}
+
+TEST_F(DBRangeDelTest, GetCoveredKeyFromSst) {
+  db_->Put(WriteOptions(), "key", "val");
+  // snapshot prevents key from being deleted during flush
+  const Snapshot* snapshot = db_->GetSnapshot();
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "z"));
+  ASSERT_OK(db_->Flush(FlushOptions()));
+
+  ReadOptions read_opts;
+  std::string value;
+  ASSERT_TRUE(db_->Get(read_opts, "key", &value).IsNotFound());
+  db_->ReleaseSnapshot(snapshot);
+}
+
+TEST_F(DBRangeDelTest, GetCoveredMergeOperandFromMemtable) {
+  const int kNumMergeOps = 10;
+  Options opts = CurrentOptions();
+  opts.merge_operator = MergeOperators::CreateUInt64AddOperator();
+  Reopen(opts);
+
+  for (int i = 0; i < kNumMergeOps; ++i) {
+    std::string val;
+    PutFixed64(&val, i);
+    db_->Merge(WriteOptions(), "key", val);
+    if (i == kNumMergeOps / 2) {
+      // deletes [0, 5]
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "key",
+                       "key_");
+    }
+  }
+
+  ReadOptions read_opts;
+  std::string expected, actual;
+  ASSERT_OK(db_->Get(read_opts, "key", &actual));
+  PutFixed64(&expected, 30);  // 6+7+8+9
+  ASSERT_EQ(expected, actual);
+
+  expected.clear();
+  read_opts.ignore_range_deletions = true;
+  ASSERT_OK(db_->Get(read_opts, "key", &actual));
+  PutFixed64(&expected, 45);  // 0+1+2+...+9
+  ASSERT_EQ(expected, actual);
+}
+
+TEST_F(DBRangeDelTest, GetIgnoresRangeDeletions) {
+  Options opts = CurrentOptions();
+  opts.max_write_buffer_number = 4;
+  opts.min_write_buffer_number_to_merge = 3;
+  opts.memtable_factory.reset(new SpecialSkipListFactory(1));
+  Reopen(opts);
+
+  db_->Put(WriteOptions(), "sst_key", "val");
+  // snapshot prevents key from being deleted during flush
+  const Snapshot* snapshot = db_->GetSnapshot();
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "z"));
+  ASSERT_OK(db_->Flush(FlushOptions()));
+  db_->Put(WriteOptions(), "imm_key", "val");
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "z"));
+  db_->Put(WriteOptions(), "mem_key", "val");
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "z"));
+
+  ReadOptions read_opts;
+  read_opts.ignore_range_deletions = true;
+  for (std::string key : {"sst_key", "imm_key", "mem_key"}) {
+    std::string value;
+    ASSERT_OK(db_->Get(read_opts, key, &value));
+  }
+  db_->ReleaseSnapshot(snapshot);
+}
+
+TEST_F(DBRangeDelTest, IteratorRemovesCoveredKeys) {
+  const int kNum = 200, kRangeBegin = 50, kRangeEnd = 150, kNumPerFile = 25;
+  Options opts = CurrentOptions();
+  opts.comparator = test::Uint64Comparator();
+  opts.memtable_factory.reset(new SpecialSkipListFactory(kNumPerFile));
+  Reopen(opts);
+
+  // Write half of the keys before the tombstone and half after the tombstone.
+  // Only covered keys (i.e., within the range and older than the tombstone)
+  // should be deleted.
+  for (int i = 0; i < kNum; ++i) {
+    if (i == kNum / 2) {
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(),
+                       GetNumericStr(kRangeBegin), GetNumericStr(kRangeEnd));
+    }
+    db_->Put(WriteOptions(), GetNumericStr(i), "val");
+  }
+  ReadOptions read_opts;
+  auto* iter = db_->NewIterator(read_opts);
+
+  int expected = 0;
+  for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
+    ASSERT_EQ(GetNumericStr(expected), iter->key());
+    if (expected == kRangeBegin - 1) {
+      expected = kNum / 2;
+    } else {
+      ++expected;
+    }
+  }
+  ASSERT_EQ(kNum, expected);
+  delete iter;
+}
+
+TEST_F(DBRangeDelTest, IteratorOverUserSnapshot) {
+  const int kNum = 200, kRangeBegin = 50, kRangeEnd = 150, kNumPerFile = 25;
+  Options opts = CurrentOptions();
+  opts.comparator = test::Uint64Comparator();
+  opts.memtable_factory.reset(new SpecialSkipListFactory(kNumPerFile));
+  Reopen(opts);
+
+  const Snapshot* snapshot = nullptr;
+  // Put a snapshot before the range tombstone, verify an iterator using that
+  // snapshot sees all inserted keys.
+  for (int i = 0; i < kNum; ++i) {
+    if (i == kNum / 2) {
+      snapshot = db_->GetSnapshot();
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(),
+                       GetNumericStr(kRangeBegin), GetNumericStr(kRangeEnd));
+    }
+    db_->Put(WriteOptions(), GetNumericStr(i), "val");
+  }
+  ReadOptions read_opts;
+  read_opts.snapshot = snapshot;
+  auto* iter = db_->NewIterator(read_opts);
+
+  int expected = 0;
+  for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
+    ASSERT_EQ(GetNumericStr(expected), iter->key());
+    ++expected;
+  }
+  ASSERT_EQ(kNum / 2, expected);
+  delete iter;
+  db_->ReleaseSnapshot(snapshot);
+}
+
+TEST_F(DBRangeDelTest, IteratorIgnoresRangeDeletions) {
+  Options opts = CurrentOptions();
+  opts.max_write_buffer_number = 4;
+  opts.min_write_buffer_number_to_merge = 3;
+  opts.memtable_factory.reset(new SpecialSkipListFactory(1));
+  Reopen(opts);
+
+  db_->Put(WriteOptions(), "sst_key", "val");
+  // snapshot prevents key from being deleted during flush
+  const Snapshot* snapshot = db_->GetSnapshot();
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "z"));
+  ASSERT_OK(db_->Flush(FlushOptions()));
+  db_->Put(WriteOptions(), "imm_key", "val");
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "z"));
+  db_->Put(WriteOptions(), "mem_key", "val");
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "z"));
+
+  ReadOptions read_opts;
+  read_opts.ignore_range_deletions = true;
+  auto* iter = db_->NewIterator(read_opts);
+  int i = 0;
+  std::string expected[] = {"imm_key", "mem_key", "sst_key"};
+  for (iter->SeekToFirst(); iter->Valid(); iter->Next(), ++i) {
+    std::string key;
+    ASSERT_EQ(expected[i], iter->key());
+  }
+  ASSERT_EQ(3, i);
+  delete iter;
+  db_->ReleaseSnapshot(snapshot);
+}
+
+TEST_F(DBRangeDelTest, TailingIteratorRangeTombstoneUnsupported) {
+  db_->Put(WriteOptions(), "key", "val");
+  // snapshot prevents key from being deleted during flush
+  const Snapshot* snapshot = db_->GetSnapshot();
+  ASSERT_OK(
+      db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), "a", "z"));
+
+  // iterations check unsupported in memtable, l0, and then l1
+  for (int i = 0; i < 3; ++i) {
+    ReadOptions read_opts;
+    read_opts.tailing = true;
+    auto* iter = db_->NewIterator(read_opts);
+    if (i == 2) {
+      // For L1+, iterators over files are created on-demand, so need seek
+      iter->SeekToFirst();
+    }
+    ASSERT_TRUE(iter->status().IsNotSupported());
+    delete iter;
+    if (i == 0) {
+      ASSERT_OK(db_->Flush(FlushOptions()));
+    } else if (i == 1) {
+      MoveFilesToLevel(1);
+    }
+  }
+  db_->ReleaseSnapshot(snapshot);
+}
+#endif  // ROCKSDB_LITE
+
+}  // namespace rocksdb
+
+int main(int argc, char** argv) {
+  rocksdb::port::InstallStackTraceHandler();
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}
@@ -221,8 +221,6 @@ TEST_F(DBSSTTest, DeleteObsoleteFilesPendingOutputs) {
  listener->VerifyMatchedCount(1);
 }

-#endif  // ROCKSDB_LITE
-
 TEST_F(DBSSTTest, DBWithSstFileManager) {
  std::shared_ptr<SstFileManager> sst_file_manager(NewSstFileManager(env_));
  auto sfm = static_cast<SstFileManagerImpl*>(sst_file_manager.get());
@@ -287,7 +285,6 @@ TEST_F(DBSSTTest, DBWithSstFileManager) {
  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }

-#ifndef ROCKSDB_LITE
 TEST_F(DBSSTTest, RateLimitedDelete) {
  Destroy(last_options_);
  rocksdb::SyncPoint::GetInstance()->LoadDependency({
@@ -299,8 +296,31 @@ TEST_F(DBSSTTest, RateLimitedDelete) {
  rocksdb::SyncPoint::GetInstance()->SetCallBack(
      "DeleteScheduler::BackgroundEmptyTrash:Wait",
      [&](void* arg) { penalties.push_back(*(static_cast<int*>(arg))); });
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "InstrumentedCondVar::TimedWaitInternal", [&](void* arg) {
+        // Turn timed wait into a simulated sleep
+        uint64_t* abs_time_us = static_cast<uint64_t*>(arg);
+        int64_t cur_time = 0;
+        env_->GetCurrentTime(&cur_time);
+        if (*abs_time_us > static_cast<uint64_t>(cur_time)) {
+          env_->addon_time_.fetch_add(*abs_time_us -
+                                      static_cast<uint64_t>(cur_time));
+        }
+
+        // Randomly sleep shortly
+        env_->addon_time_.fetch_add(
+            static_cast<uint64_t>(Random::GetTLSInstance()->Uniform(10)));
+
+        // Set wait until time to before current to force not to sleep.
+        int64_t real_cur_time = 0;
+        Env::Default()->GetCurrentTime(&real_cur_time);
+        *abs_time_us = static_cast<uint64_t>(real_cur_time);
+      });
+
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

+  env_->no_slowdown_ = true;
+  env_->time_elapse_only_sleep_ = true;
  Options options = CurrentOptions();
  options.disable_auto_compactions = true;
  options.env = env_;
@@ -308,9 +328,10 @@ TEST_F(DBSSTTest, RateLimitedDelete) {
  std::string trash_dir = test::TmpDir(env_) + "/trash";
  int64_t rate_bytes_per_sec = 1024 * 10;  // 10 Kbs / Sec
  Status s;
-  options.sst_file_manager.reset(NewSstFileManager(
-      env_, nullptr, trash_dir, rate_bytes_per_sec, false, &s));
+  options.sst_file_manager.reset(
+      NewSstFileManager(env_, nullptr, trash_dir, 0, false, &s));
  ASSERT_OK(s);
+  options.sst_file_manager->SetDeleteRateBytesPerSecond(rate_bytes_per_sec);
  auto sfm = static_cast<SstFileManagerImpl*>(options.sst_file_manager.get());

  ASSERT_OK(TryReopen(options));
@@ -348,6 +369,7 @@ TEST_F(DBSSTTest, RateLimitedDelete) {
    ASSERT_EQ(expected_penlty, penalties[i]);
  }
  ASSERT_GT(time_spent_deleting, expected_penlty * 0.9);
+  ASSERT_LT(time_spent_deleting, expected_penlty * 1.1);

  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }
@@ -458,7 +480,6 @@ TEST_F(DBSSTTest, DestroyDBWithRateLimitedDelete) {
  // We have deleted the 4 sst files in the delete_scheduler
  ASSERT_EQ(bg_delete_file, 4);
 }
-#endif  // ROCKSDB_LITE

 TEST_F(DBSSTTest, DBWithMaxSpaceAllowed) {
  std::shared_ptr<SstFileManager> sst_file_manager(NewSstFileManager(env_));
@@ -496,18 +517,25 @@ TEST_F(DBSSTTest, DBWithMaxSpaceAllowedRandomized) {
  // than the limit.

  std::vector<int> max_space_limits_mbs = {1, 2, 4, 8, 10};
-
+  decltype(max_space_limits_mbs)::value_type limit_mb_cb;
  bool bg_error_set = false;
  uint64_t total_sst_files_size = 0;

+  std::atomic<int> estimate_multiplier(1);
  int reached_max_space_on_flush = 0;
  int reached_max_space_on_compaction = 0;
  rocksdb::SyncPoint::GetInstance()->SetCallBack(
      "DBImpl::FlushMemTableToOutputFile:MaxAllowedSpaceReached",
      [&](void* arg) {
+        Status* bg_error = static_cast<Status*>(arg);
        bg_error_set = true;
        GetAllSSTFiles(&total_sst_files_size);
        reached_max_space_on_flush++;
+        // low limit for size calculated using sst files
+        ASSERT_GE(total_sst_files_size, limit_mb_cb * 1024 * 1024);
+        // clear error to ensure compaction callback is called
+        *bg_error = Status::OK();
+        estimate_multiplier++;  // used in the main loop assert
      });

  rocksdb::SyncPoint::GetInstance()->SetCallBack(
@@ -521,6 +549,8 @@ TEST_F(DBSSTTest, DBWithMaxSpaceAllowedRandomized) {
  for (auto limit_mb : max_space_limits_mbs) {
    bg_error_set = false;
    total_sst_files_size = 0;
+    estimate_multiplier = 1;
+    limit_mb_cb = limit_mb;
    rocksdb::SyncPoint::GetInstance()->ClearTrace();
    rocksdb::SyncPoint::GetInstance()->EnableProcessing();
    std::shared_ptr<SstFileManager> sst_file_manager(NewSstFileManager(env_));
@@ -545,7 +575,8 @@ TEST_F(DBSSTTest, DBWithMaxSpaceAllowedRandomized) {
      // Check the estimated db size vs the db limit just to make sure we
      // dont run into an infinite loop
      estimated_db_size = keys_written * 60;  // ~60 bytes per key
-      ASSERT_LT(estimated_db_size, limit_mb * 1024 * 1024 * 2);
+      ASSERT_LT(estimated_db_size,
+                estimate_multiplier * limit_mb * 1024 * 1024 * 2);
    }
    ASSERT_TRUE(bg_error_set);
    ASSERT_GE(total_sst_files_size, limit_mb * 1024 * 1024);
@@ -556,7 +587,6 @@ TEST_F(DBSSTTest, DBWithMaxSpaceAllowedRandomized) {
  ASSERT_GT(reached_max_space_on_compaction, 0);
 }

-#ifndef ROCKSDB_LITE
 TEST_F(DBSSTTest, OpenDBWithInfiniteMaxOpenFiles) {
  // Open DB with infinite max open files
  //  - First iteration use 1 thread to open files
@@ -594,7 +624,7 @@ TEST_F(DBSSTTest, OpenDBWithInfiniteMaxOpenFiles) {
    }
    Close();

-    // Reopening the DB will load all exisitng files
+    // Reopening the DB will load all existing files
    Reopen(options);
    ASSERT_EQ("12,0,12", FilesPerLevel(0));
    std::vector<std::vector<FileMetaData>> files;
@@ -294,8 +294,6 @@ TEST_F(DBTestTailingIterator, TailingIteratorDeletes) {
 }

 TEST_F(DBTestTailingIterator, TailingIteratorPrefixSeek) {
-  XFUNC_TEST("", "dbtest_prefix", prefix_skip1, XFuncPoint::SetSkip,
-             kSkipNoPrefix);
  ReadOptions read_options;
  read_options.tailing = true;

@@ -305,6 +303,7 @@ TEST_F(DBTestTailingIterator, TailingIteratorPrefixSeek) {
  options.disable_auto_compactions = true;
  options.prefix_extractor.reset(NewFixedPrefixTransform(2));
  options.memtable_factory.reset(NewHashSkipListRepFactory(16));
+  options.allow_concurrent_memtable_write = false;
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);

@@ -325,7 +324,6 @@ TEST_F(DBTestTailingIterator, TailingIteratorPrefixSeek) {

  iter->Next();
  ASSERT_TRUE(!iter->Valid());
-  XFUNC_TEST("", "dbtest_prefix", prefix_skip1, XFuncPoint::SetSkip, 0);
 }

 TEST_F(DBTestTailingIterator, TailingIteratorIncomplete) {
@@ -428,6 +426,59 @@ TEST_F(DBTestTailingIterator, TailingIteratorUpperBound) {
  ASSERT_EQ(0, immutable_seeks);
 }

+TEST_F(DBTestTailingIterator, TailingIteratorGap) {
+  // level 1:            [20, 25]  [35, 40]
+  // level 2:  [10 - 15]                    [45 - 50]
+  // level 3:            [20,    30,    40]
+  // Previously there is a bug in tailing_iterator that if there is a gap in
+  // lower level, the key will be skipped if it is within the range between
+  // the largest key of index n file and the smallest key of index n+1 file
+  // if both file fit in that gap. In this example, 25 < key < 35
+  // https://github.com/facebook/rocksdb/issues/1372
+  CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
+
+  ReadOptions read_options;
+  read_options.tailing = true;
+
+  ASSERT_OK(Put(1, "20", "20"));
+  ASSERT_OK(Put(1, "30", "30"));
+  ASSERT_OK(Put(1, "40", "40"));
+  ASSERT_OK(Flush(1));
+  MoveFilesToLevel(3, 1);
+
+  ASSERT_OK(Put(1, "10", "10"));
+  ASSERT_OK(Put(1, "15", "15"));
+  ASSERT_OK(Flush(1));
+  ASSERT_OK(Put(1, "45", "45"));
+  ASSERT_OK(Put(1, "50", "50"));
+  ASSERT_OK(Flush(1));
+  MoveFilesToLevel(2, 1);
+
+  ASSERT_OK(Put(1, "20", "20"));
+  ASSERT_OK(Put(1, "25", "25"));
+  ASSERT_OK(Flush(1));
+  ASSERT_OK(Put(1, "35", "35"));
+  ASSERT_OK(Put(1, "40", "40"));
+  ASSERT_OK(Flush(1));
+  MoveFilesToLevel(1, 1);
+
+  ColumnFamilyMetaData meta;
+  db_->GetColumnFamilyMetaData(handles_[1], &meta);
+
+  std::unique_ptr<Iterator> it(db_->NewIterator(read_options, handles_[1]));
+  it->Seek("30");
+  ASSERT_TRUE(it->Valid());
+  ASSERT_EQ("30", it->key().ToString());
+
+  it->Next();
+  ASSERT_TRUE(it->Valid());
+  ASSERT_EQ("35", it->key().ToString());
+
+  it->Next();
+  ASSERT_TRUE(it->Valid());
+  ASSERT_EQ("40", it->key().ToString());
+}
+
 TEST_F(DBTestTailingIterator, ManagedTailingIteratorSingle) {
  ReadOptions read_options;
  read_options.tailing = true;
@@ -560,8 +611,6 @@ TEST_F(DBTestTailingIterator, ManagedTailingIteratorDeletes) {
 }

 TEST_F(DBTestTailingIterator, ManagedTailingIteratorPrefixSeek) {
-  XFUNC_TEST("", "dbtest_prefix", prefix_skip1, XFuncPoint::SetSkip,
-             kSkipNoPrefix);
  ReadOptions read_options;
  read_options.tailing = true;
  read_options.managed = true;
@@ -572,6 +621,7 @@ TEST_F(DBTestTailingIterator, ManagedTailingIteratorPrefixSeek) {
  options.disable_auto_compactions = true;
  options.prefix_extractor.reset(NewFixedPrefixTransform(2));
  options.memtable_factory.reset(NewHashSkipListRepFactory(16));
+  options.allow_concurrent_memtable_write = false;
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);

@@ -592,7 +642,6 @@ TEST_F(DBTestTailingIterator, ManagedTailingIteratorPrefixSeek) {

  iter->Next();
  ASSERT_TRUE(!iter->Valid());
-  XFUNC_TEST("", "dbtest_prefix", prefix_skip1, XFuncPoint::SetSkip, 0);
 }

 TEST_F(DBTestTailingIterator, ManagedTailingIteratorIncomplete) {
@@ -699,6 +748,58 @@ TEST_F(DBTestTailingIterator, ForwardIteratorVersionProperty) {
  }
  ASSERT_EQ(v3, v4);
 }
+
+TEST_F(DBTestTailingIterator, SeekWithUpperBoundBug) {
+  ReadOptions read_options;
+  read_options.tailing = true;
+  const Slice upper_bound("cc", 3);
+  read_options.iterate_upper_bound = &upper_bound;
+
+
+  // 1st L0 file
+  ASSERT_OK(db_->Put(WriteOptions(), "aa", "SEEN"));
+  ASSERT_OK(Flush());
+
+  // 2nd L0 file
+  ASSERT_OK(db_->Put(WriteOptions(), "zz", "NOT-SEEN"));
+  ASSERT_OK(Flush());
+
+  std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+
+  iter->Seek("aa");
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ(iter->key().ToString(), "aa");
+}
+
+TEST_F(DBTestTailingIterator, SeekToFirstWithUpperBoundBug) {
+  ReadOptions read_options;
+  read_options.tailing = true;
+  const Slice upper_bound("cc", 3);
+  read_options.iterate_upper_bound = &upper_bound;
+
+
+  // 1st L0 file
+  ASSERT_OK(db_->Put(WriteOptions(), "aa", "SEEN"));
+  ASSERT_OK(Flush());
+
+  // 2nd L0 file
+  ASSERT_OK(db_->Put(WriteOptions(), "zz", "NOT-SEEN"));
+  ASSERT_OK(Flush());
+
+  std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+
+  iter->SeekToFirst();
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ(iter->key().ToString(), "aa");
+
+  iter->Next();
+  ASSERT_FALSE(iter->Valid());
+
+  iter->SeekToFirst();
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ(iter->key().ToString(), "aa");
+}
+
 }  // namespace rocksdb

 #endif  // !defined(ROCKSDB_LITE)
@@ -11,6 +11,7 @@
 #include <functional>

 #include "db/db_test_util.h"
+#include "port/port.h"
 #include "port/stack_trace.h"
 #include "rocksdb/persistent_cache.h"
 #include "rocksdb/wal_filter.h"
@@ -956,6 +957,7 @@ TEST_F(DBTest2, PresetCompressionDict) {
  const int kNumL0Files = 5;

  Options options;
+  options.allow_concurrent_memtable_write = false;
  options.arena_block_size = kBlockSizeBytes;
  options.compaction_style = kCompactionStyleUniversal;
  options.create_if_missing = true;
@@ -1469,6 +1471,10 @@ class MockPersistentCache : public PersistentCache {

  virtual ~MockPersistentCache() {}

+  PersistentCache::StatsType Stats() override {
+    return PersistentCache::StatsType();
+  }
+
  Status Insert(const Slice& page_key, const char* data,
                const size_t size) override {
    MutexLock _(&lock_);
@@ -1500,6 +1506,10 @@ class MockPersistentCache : public PersistentCache {

  bool IsCompressed() override { return is_compressed_; }

+  std::string GetPrintableOptions() const override {
+    return "MockPersistentCache";
+  }
+
  port::Mutex lock_;
  std::map<std::string, std::string> data_;
  const bool is_compressed_ = true;
@@ -1607,8 +1617,8 @@ TEST_F(DBTest2, SyncPointMarker) {
    CountSyncPoint();
  };

-  auto thread1 = std::thread(func1);
-  auto thread2 = std::thread(func2);
+  auto thread1 = port::Thread(func1);
+  auto thread2 = port::Thread(func2);
  thread1.join();
  thread2.join();

@@ -1897,8 +1907,8 @@ TEST_P(MergeOperatorPinningTest, TailingIterator) {
    delete iter;
  };

-  std::thread writer_thread(writer_func);
-  std::thread reader_thread(reader_func);
+  rocksdb::port::Thread writer_thread(writer_func);
+  rocksdb::port::Thread reader_thread(reader_func);

  writer_thread.join();
  reader_thread.join();
@@ -2085,8 +2095,26 @@ TEST_F(DBTest2, AutomaticCompactionOverlapManualCompaction) {
  cro.target_level = 2;
  ASSERT_OK(db_->CompactRange(cro, nullptr, nullptr));

+  auto get_stat = [](std::string level_str, LevelStatType type,
+                     std::map<std::string, double> props) {
+    auto prop_str =
+        level_str + "." +
+        InternalStats::compaction_level_stats.at(type).property_name.c_str();
+    auto prop_item = props.find(prop_str);
+    return prop_item == props.end() ? 0 : prop_item->second;
+  };
+
  // Trivial move 2 files to L2
  ASSERT_EQ("0,0,2", FilesPerLevel());
+  // Also test that the stats GetMapProperty API reporting the same result
+  {
+    std::map<std::string, double> prop;
+    ASSERT_TRUE(dbfull()->GetMapProperty("rocksdb.cfstats", &prop));
+    ASSERT_EQ(0, get_stat("L0", LevelStatType::NUM_FILES, prop));
+    ASSERT_EQ(0, get_stat("L1", LevelStatType::NUM_FILES, prop));
+    ASSERT_EQ(2, get_stat("L2", LevelStatType::NUM_FILES, prop));
+    ASSERT_EQ(2, get_stat("Sum", LevelStatType::NUM_FILES, prop));
+  }

  // While the compaction is running, we will create 2 new files that
  // can fit in L2, these 2 files will be moved to L2 and overlap with
@@ -2113,6 +2141,13 @@ TEST_F(DBTest2, AutomaticCompactionOverlapManualCompaction) {
  ASSERT_OK(db_->CompactRange(cro, nullptr, nullptr));

  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+
+  // Test that the stats GetMapProperty API reporting 1 file in L2
+  {
+    std::map<std::string, double> prop;
+    ASSERT_TRUE(dbfull()->GetMapProperty("rocksdb.cfstats", &prop));
+    ASSERT_EQ(1, get_stat("L2", LevelStatType::NUM_FILES, prop));
+  }
 }

 TEST_F(DBTest2, ManualCompactionOverlapManualCompaction) {
@@ -2144,7 +2179,7 @@ TEST_F(DBTest2, ManualCompactionOverlapManualCompaction) {
    cro.exclusive_manual_compaction = false;
    ASSERT_OK(db_->CompactRange(cro, &k1s, &k2s));
  };
-  std::thread bg_thread;
+  rocksdb::port::Thread bg_thread;

  // While the compaction is running, we will create 2 new files that
  // can fit in L1, these 2 files will be moved to L1 and overlap with
@@ -2165,7 +2200,7 @@ TEST_F(DBTest2, ManualCompactionOverlapManualCompaction) {
        ASSERT_OK(Flush());

        // Start a non-exclusive manual compaction in a bg thread
-        bg_thread = std::thread(bg_manual_compact);
+        bg_thread = port::Thread(bg_manual_compact);
        // This manual compaction conflict with the other manual compaction
        // so it should wait until the first compaction finish
        env_->SleepForMicroseconds(1000000);
@@ -2182,8 +2217,151 @@ TEST_F(DBTest2, ManualCompactionOverlapManualCompaction) {

  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }
+
+TEST_F(DBTest2, OptimizeForPointLookup) {
+  Options options = CurrentOptions();
+  Close();
+  options.OptimizeForPointLookup(2);
+  ASSERT_OK(DB::Open(options, dbname_, &db_));
+
+  ASSERT_OK(Put("foo", "v1"));
+  ASSERT_EQ("v1", Get("foo"));
+  Flush();
+  ASSERT_EQ("v1", Get("foo"));
+}
+
 #endif  // ROCKSDB_LITE

+TEST_F(DBTest2, GetRaceFlush1) {
+  ASSERT_OK(Put("foo", "v1"));
+
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      {{"DBImpl::GetImpl:1", "DBTest2::GetRaceFlush:1"},
+       {"DBTest2::GetRaceFlush:2", "DBImpl::GetImpl:2"}});
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  rocksdb::port::Thread t1([&] {
+    TEST_SYNC_POINT("DBTest2::GetRaceFlush:1");
+    ASSERT_OK(Put("foo", "v2"));
+    Flush();
+    TEST_SYNC_POINT("DBTest2::GetRaceFlush:2");
+  });
+
+  // Get() is issued after the first Put(), so it should see either
+  // "v1" or "v2".
+  ASSERT_NE("NOT_FOUND", Get("foo"));
+  t1.join();
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+}
+
+TEST_F(DBTest2, GetRaceFlush2) {
+  ASSERT_OK(Put("foo", "v1"));
+
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      {{"DBImpl::GetImpl:3", "DBTest2::GetRaceFlush:1"},
+       {"DBTest2::GetRaceFlush:2", "DBImpl::GetImpl:4"}});
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  port::Thread t1([&] {
+    TEST_SYNC_POINT("DBTest2::GetRaceFlush:1");
+    ASSERT_OK(Put("foo", "v2"));
+    Flush();
+    TEST_SYNC_POINT("DBTest2::GetRaceFlush:2");
+  });
+
+  // Get() is issued after the first Put(), so it should see either
+  // "v1" or "v2".
+  ASSERT_NE("NOT_FOUND", Get("foo"));
+  t1.join();
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+}
+
+TEST_F(DBTest2, DirectIO) {
+  if (!IsDirectIOSupported()) {
+    return;
+  }
+  Options options = CurrentOptions();
+  options.use_direct_reads = options.use_direct_writes = true;
+  options.allow_mmap_reads = options.allow_mmap_writes = false;
+  DestroyAndReopen(options);
+
+  ASSERT_OK(Put(Key(0), "a"));
+  ASSERT_OK(Put(Key(5), "a"));
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(Put(Key(10), "a"));
+  ASSERT_OK(Put(Key(15), "a"));
+  ASSERT_OK(Flush());
+
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  Reopen(options);
+}
+
+TEST_F(DBTest2, MemtableOnlyIterator) {
+  Options options = CurrentOptions();
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  ASSERT_OK(Put(1, "foo", "first"));
+  ASSERT_OK(Put(1, "bar", "second"));
+
+  ReadOptions ropt;
+  ropt.read_tier = kMemtableTier;
+  std::string value;
+  Iterator* it = nullptr;
+
+  // Before flushing
+  // point lookups
+  ASSERT_OK(db_->Get(ropt, handles_[1], "foo", &value));
+  ASSERT_EQ("first", value);
+  ASSERT_OK(db_->Get(ropt, handles_[1], "bar", &value));
+  ASSERT_EQ("second", value);
+
+  // Memtable-only iterator (read_tier=kMemtableTier); data not flushed yet.
+  it = db_->NewIterator(ropt, handles_[1]);
+  int count = 0;
+  for (it->SeekToFirst(); it->Valid(); it->Next()) {
+    ASSERT_TRUE(it->Valid());
+    count++;
+  }
+  ASSERT_TRUE(!it->Valid());
+  ASSERT_EQ(2, count);
+  delete it;
+
+  Flush(1);
+
+  // After flushing
+  // point lookups
+  ASSERT_OK(db_->Get(ropt, handles_[1], "foo", &value));
+  ASSERT_EQ("first", value);
+  ASSERT_OK(db_->Get(ropt, handles_[1], "bar", &value));
+  ASSERT_EQ("second", value);
+  // nothing should be returned using memtable-only iterator after flushing.
+  it = db_->NewIterator(ropt, handles_[1]);
+  count = 0;
+  for (it->SeekToFirst(); it->Valid(); it->Next()) {
+    ASSERT_TRUE(it->Valid());
+    count++;
+  }
+  ASSERT_TRUE(!it->Valid());
+  ASSERT_EQ(0, count);
+  delete it;
+
+  // Add a key to memtable
+  ASSERT_OK(Put(1, "foobar", "third"));
+  it = db_->NewIterator(ropt, handles_[1]);
+  count = 0;
+  for (it->SeekToFirst(); it->Valid(); it->Next()) {
+    ASSERT_TRUE(it->Valid());
+    ASSERT_EQ("foobar", it->key().ToString());
+    ASSERT_EQ("third", it->value().ToString());
+    count++;
+  }
+  ASSERT_TRUE(!it->Valid());
+  ASSERT_EQ(1, count);
+  delete it;
+}
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -8,6 +8,7 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

 #include "db/db_test_util.h"
+#include "db/forward_iterator.h"

 namespace rocksdb {

@@ -19,7 +20,7 @@ SpecialEnv::SpecialEnv(Env* base)
      sleep_counter_(this),
      addon_time_(0),
      time_elapse_only_sleep_(false),
-      no_sleep_(false) {
+      no_slowdown_(false) {
  delay_sstable_sync_.store(false, std::memory_order_release);
  drop_writes_.store(false, std::memory_order_release);
  no_space_.store(false, std::memory_order_release);
@@ -71,7 +72,12 @@ DBTestBase::~DBTestBase() {
  options.db_paths.emplace_back(dbname_ + "_2", 0);
  options.db_paths.emplace_back(dbname_ + "_3", 0);
  options.db_paths.emplace_back(dbname_ + "_4", 0);
-  EXPECT_OK(DestroyDB(dbname_, options));
+
+  if (getenv("KEEP_DB")) {
+    printf("DB is still at %s\n", dbname_.c_str());
+  } else {
+    EXPECT_OK(DestroyDB(dbname_, options));
+  }
  delete env_;
 }

@@ -185,6 +191,36 @@ bool DBTestBase::ChangeCompactOptions() {
  }
 }

+// Switch between different WAL settings
+bool DBTestBase::ChangeWalOptions() {
+  if (option_config_ == kDefault) {
+    option_config_ = kDBLogDir;
+    Destroy(last_options_);
+    auto options = CurrentOptions();
+    Destroy(options);
+    options.create_if_missing = true;
+    TryReopen(options);
+    return true;
+  } else if (option_config_ == kDBLogDir) {
+    option_config_ = kWalDirAndMmapReads;
+    Destroy(last_options_);
+    auto options = CurrentOptions();
+    Destroy(options);
+    options.create_if_missing = true;
+    TryReopen(options);
+    return true;
+  } else if (option_config_ == kWalDirAndMmapReads) {
+    option_config_ = kRecycleLogFiles;
+    Destroy(last_options_);
+    auto options = CurrentOptions();
+    Destroy(options);
+    TryReopen(options);
+    return true;
+  } else {
+    return false;
+  }
+}
+
 // Switch between different filter policy
 // Jump from kDefault to kFilter to kFullFilter
 bool DBTestBase::ChangeFilterOptions() {
@@ -223,9 +259,6 @@ Options DBTestBase::CurrentOptions(
    const anon::OptionsOverride& options_override) {
  // this redundant copy is to minimize code change w/o having lint error.
  Options options = defaultOptions;
-  XFUNC_TEST("", "dbtest_options", inplace_options1, GetXFTestOptions,
-             reinterpret_cast<Options*>(&options),
-             options_override.skip_policy);
  BlockBasedTableOptions table_options;
  bool set_block_based_table_factory = true;
  switch (option_config_) {
@@ -233,6 +266,7 @@ Options DBTestBase::CurrentOptions(
    case kHashSkipList:
      options.prefix_extractor.reset(NewFixedPrefixTransform(1));
      options.memtable_factory.reset(NewHashSkipListRepFactory(16));
+      options.allow_concurrent_memtable_write = false;
      break;
    case kPlainTableFirstBytePrefix:
      options.table_factory.reset(new PlainTableFactory());
@@ -264,15 +298,18 @@ Options DBTestBase::CurrentOptions(
      break;
    case kVectorRep:
      options.memtable_factory.reset(new VectorRepFactory(100));
+      options.allow_concurrent_memtable_write = false;
      break;
    case kHashLinkList:
      options.prefix_extractor.reset(NewFixedPrefixTransform(1));
      options.memtable_factory.reset(
          NewHashLinkListRepFactory(4, 0, 3, true, 4));
+      options.allow_concurrent_memtable_write = false;
      break;
    case kHashCuckoo:
      options.memtable_factory.reset(
          NewHashCuckooRepFactory(options.write_buffer_size));
+      options.allow_concurrent_memtable_write = false;
      break;
 #endif  // ROCKSDB_LITE
    case kMergePut:
@@ -303,6 +340,7 @@ Options DBTestBase::CurrentOptions(
      break;
    case kManifestFileSize:
      options.max_manifest_file_size = 50;  // 50 bytes
+      break;
    case kPerfOptions:
      options.soft_rate_limit = 2.0;
      options.delayed_write_rate = 8 * 1024 * 1024;
@@ -342,6 +380,11 @@ Options DBTestBase::CurrentOptions(
      options.prefix_extractor.reset(NewNoopTransform());
      break;
    }
+    case kBlockBasedTableWithPartitionedIndex: {
+      table_options.index_type = BlockBasedTableOptions::kTwoLevelIndexSearch;
+      options.prefix_extractor.reset(NewNoopTransform());
+      break;
+    }
    case kBlockBasedTableWithIndexRestartInterval: {
      table_options.index_block_restart_interval = 8;
      break;
@@ -468,10 +511,33 @@ Status DBTestBase::ReadOnlyReopen(const Options& options) {

 Status DBTestBase::TryReopen(const Options& options) {
  Close();
+  last_options_.table_factory.reset();
+  // Note: operator= is an unsafe approach here since it destructs shared_ptr in
+  // the same order of their creation, in contrast to destructors which
+  // destructs them in the opposite order of creation. One particular problme is
+  // that the cache destructor might invoke callback functions that use Option
+  // members such as statistics. To work around this problem, we manually call
+  // destructor of table_facotry which eventually clears the block cache.
  last_options_ = options;
  return DB::Open(options, dbname_, &db_);
 }

+bool DBTestBase::IsDirectIOSupported() {
+  EnvOptions env_options;
+  env_options.use_mmap_writes = false;
+  env_options.use_direct_writes = true;
+  std::string tmp = TempFileName(dbname_, 999);
+  Status s;
+  {
+    unique_ptr<WritableFile> file;
+    s = env_->NewWritableFile(tmp, &file, env_options);
+  }
+  if (s.ok()) {
+    s = env_->DeleteFile(tmp);
+  }
+  return s.ok();
+}
+
 Status DBTestBase::Flush(int cf) {
  if (cf == 0) {
    return db_->Flush(FlushOptions());
@@ -497,6 +563,15 @@ Status DBTestBase::Put(int cf, const Slice& k, const Slice& v,
  }
 }

+Status DBTestBase::Merge(const Slice& k, const Slice& v, WriteOptions wo) {
+  return db_->Merge(wo, k, v);
+}
+
+Status DBTestBase::Merge(int cf, const Slice& k, const Slice& v,
+                         WriteOptions wo) {
+  return db_->Merge(wo, handles_[cf], k, v);
+}
+
 Status DBTestBase::Delete(const std::string& k) {
  return db_->Delete(WriteOptions(), k);
 }
@@ -585,11 +660,15 @@ std::string DBTestBase::Contents(int cf) {

 std::string DBTestBase::AllEntriesFor(const Slice& user_key, int cf) {
  Arena arena;
+  auto options = CurrentOptions();
+  InternalKeyComparator icmp(options.comparator);
+  RangeDelAggregator range_del_agg(icmp, {} /* snapshots */);
  ScopedArenaIterator iter;
  if (cf == 0) {
-    iter.set(dbfull()->NewInternalIterator(&arena));
+    iter.set(dbfull()->NewInternalIterator(&arena, &range_del_agg));
  } else {
-    iter.set(dbfull()->NewInternalIterator(&arena, handles_[cf]));
+    iter.set(
+        dbfull()->NewInternalIterator(&arena, &range_del_agg, handles_[cf]));
  }
  InternalKey target(user_key, kMaxSequenceNumber, kTypeValue);
  iter->Seek(target.Encode());
@@ -990,10 +1069,14 @@ UpdateStatus DBTestBase::updateInPlaceNoAction(char* prevValue,
 void DBTestBase::validateNumberOfEntries(int numValues, int cf) {
  ScopedArenaIterator iter;
  Arena arena;
+  auto options = CurrentOptions();
+  InternalKeyComparator icmp(options.comparator);
+  RangeDelAggregator range_del_agg(icmp, {} /* snapshots */);
  if (cf != 0) {
-    iter.set(dbfull()->NewInternalIterator(&arena, handles_[cf]));
+    iter.set(
+        dbfull()->NewInternalIterator(&arena, &range_del_agg, handles_[cf]));
  } else {
-    iter.set(dbfull()->NewInternalIterator(&arena));
+    iter.set(dbfull()->NewInternalIterator(&arena, &range_del_agg));
  }
  iter->SeekToFirst();
  ASSERT_EQ(iter->status().ok(), true);
@@ -1077,11 +1160,18 @@ std::vector<std::uint64_t> DBTestBase::ListTableFiles(Env* env,
 }

 void DBTestBase::VerifyDBFromMap(std::map<std::string, std::string> true_data,
-                                 size_t* total_reads_res) {
+                                 size_t* total_reads_res, bool tailing_iter,
+                                 std::map<std::string, Status> status) {
  size_t total_reads = 0;

  for (auto& kv : true_data) {
-    ASSERT_EQ(Get(kv.first), kv.second);
+    Status s = status[kv.first];
+    if (s.ok()) {
+      ASSERT_EQ(Get(kv.first), kv.second);
+    } else {
+      std::string value;
+      ASSERT_EQ(s, db_->Get(ReadOptions(), kv.first, &value));
+    }
    total_reads++;
  }

@@ -1094,21 +1184,40 @@ void DBTestBase::VerifyDBFromMap(std::map<std::string, std::string> true_data,
    // Verify Iterator::Next()
    iter_cnt = 0;
    auto data_iter = true_data.begin();
+    Status s;
    for (iter->SeekToFirst(); iter->Valid(); iter->Next(), data_iter++) {
      ASSERT_EQ(iter->key().ToString(), data_iter->first);
-      ASSERT_EQ(iter->value().ToString(), data_iter->second);
+      Status current_status = status[data_iter->first];
+      if (!current_status.ok()) {
+        s = current_status;
+      }
+      ASSERT_EQ(iter->status(), s);
+      if (current_status.ok()) {
+        ASSERT_EQ(iter->value().ToString(), data_iter->second);
+      }
      iter_cnt++;
      total_reads++;
    }
    ASSERT_EQ(data_iter, true_data.end()) << iter_cnt << " / "
                                          << true_data.size();
+    delete iter;

    // Verify Iterator::Prev()
+    // Use a new iterator to make sure its status is clean.
+    iter = db_->NewIterator(ro);
    iter_cnt = 0;
+    s = Status::OK();
    auto data_rev = true_data.rbegin();
    for (iter->SeekToLast(); iter->Valid(); iter->Prev(), data_rev++) {
      ASSERT_EQ(iter->key().ToString(), data_rev->first);
-      ASSERT_EQ(iter->value().ToString(), data_rev->second);
+      Status current_status = status[data_rev->first];
+      if (!current_status.ok()) {
+        s = current_status;
+      }
+      ASSERT_EQ(iter->status(), s);
+      if (current_status.ok()) {
+        ASSERT_EQ(iter->value().ToString(), data_rev->second);
+      }
      iter_cnt++;
      total_reads++;
    }
@@ -1122,46 +1231,66 @@ void DBTestBase::VerifyDBFromMap(std::map<std::string, std::string> true_data,
      ASSERT_EQ(kv.second, iter->value().ToString());
      total_reads++;
    }
-
    delete iter;
  }

+  if (tailing_iter) {
 #ifndef ROCKSDB_LITE
-  // Tailing iterator
-  int iter_cnt = 0;
-  ReadOptions ro;
-  ro.tailing = true;
-  ro.total_order_seek = true;
-  Iterator* iter = db_->NewIterator(ro);
+    // Tailing iterator
+    int iter_cnt = 0;
+    ReadOptions ro;
+    ro.tailing = true;
+    ro.total_order_seek = true;
+    Iterator* iter = db_->NewIterator(ro);

-  // Verify ForwardIterator::Next()
-  iter_cnt = 0;
-  auto data_iter = true_data.begin();
-  for (iter->SeekToFirst(); iter->Valid(); iter->Next(), data_iter++) {
-    ASSERT_EQ(iter->key().ToString(), data_iter->first);
-    ASSERT_EQ(iter->value().ToString(), data_iter->second);
-    iter_cnt++;
-    total_reads++;
-  }
-  ASSERT_EQ(data_iter, true_data.end()) << iter_cnt << " / "
-                                        << true_data.size();
+    // Verify ForwardIterator::Next()
+    iter_cnt = 0;
+    auto data_iter = true_data.begin();
+    for (iter->SeekToFirst(); iter->Valid(); iter->Next(), data_iter++) {
+      ASSERT_EQ(iter->key().ToString(), data_iter->first);
+      ASSERT_EQ(iter->value().ToString(), data_iter->second);
+      iter_cnt++;
+      total_reads++;
+    }
+    ASSERT_EQ(data_iter, true_data.end()) << iter_cnt << " / "
+                                          << true_data.size();

-  // Verify ForwardIterator::Seek()
-  for (auto kv : true_data) {
-    iter->Seek(kv.first);
-    ASSERT_EQ(kv.first, iter->key().ToString());
-    ASSERT_EQ(kv.second, iter->value().ToString());
-    total_reads++;
-  }
+    // Verify ForwardIterator::Seek()
+    for (auto kv : true_data) {
+      iter->Seek(kv.first);
+      ASSERT_EQ(kv.first, iter->key().ToString());
+      ASSERT_EQ(kv.second, iter->value().ToString());
+      total_reads++;
+    }

-  delete iter;
+    delete iter;
 #endif  // ROCKSDB_LITE
+  }

  if (total_reads_res) {
    *total_reads_res = total_reads;
  }
 }

+void DBTestBase::VerifyDBInternal(
+    std::vector<std::pair<std::string, std::string>> true_data) {
+  Arena arena;
+  InternalKeyComparator icmp(last_options_.comparator);
+  RangeDelAggregator range_del_agg(icmp, {});
+  auto iter = dbfull()->NewInternalIterator(&arena, &range_del_agg);
+  iter->SeekToFirst();
+  for (auto p : true_data) {
+    ASSERT_TRUE(iter->Valid());
+    ParsedInternalKey ikey;
+    ASSERT_TRUE(ParseInternalKey(iter->key(), &ikey));
+    ASSERT_EQ(p.first, ikey.user_key);
+    ASSERT_EQ(p.second, iter->value());
+    iter->Next();
+  };
+  ASSERT_FALSE(iter->Valid());
+  iter->~InternalIterator();
+}
+
 #ifndef ROCKSDB_LITE

 uint64_t DBTestBase::GetNumberOfSstFilesForColumnFamily(
@@ -58,7 +58,6 @@
 #endif  // !(defined NDEBUG) || !defined(OS_WIN)
 #include "util/testharness.h"
 #include "util/testutil.h"
-#include "util/xfunc.h"
 #include "utilities/merge_operators.h"

 namespace rocksdb {
@@ -120,6 +119,8 @@ struct OptionsOverride {

 }  // namespace anon

+enum SkipPolicy { kSkipNone = 0, kSkipNoSnapshot = 1, kSkipNoPrefix = 2 };
+
 // A hacky skip list mem table that triggers flush after number of entries.
 class SpecialMemTableRep : public MemTableRep {
 public:
@@ -193,6 +194,10 @@ class SpecialSkipListFactory : public MemTableRepFactory {
  }
  virtual const char* Name() const override { return "SkipListFactory"; }

+  bool IsInsertConcurrentlySupported() const override {
+    return factory_.IsInsertConcurrentlySupported();
+  }
+
 private:
  SkipListFactory factory_;
  int num_entries_flush_;
@@ -227,6 +232,20 @@ class SpecialEnv : public EnvWrapper {
          return base_->Append(data);
        }
      }
+      Status PositionedAppend(const Slice& data, uint64_t offset) override {
+        if (env_->table_write_callback_) {
+          (*env_->table_write_callback_)();
+        }
+        if (env_->drop_writes_.load(std::memory_order_acquire)) {
+          // Drop writes on the floor
+          return Status::OK();
+        } else if (env_->no_space_.load(std::memory_order_acquire)) {
+          return Status::NoSpace("No space left on device");
+        } else {
+          env_->bytes_written_ += data.size();
+          return base_->PositionedAppend(data, offset);
+        }
+      }
      Status Truncate(uint64_t size) override { return base_->Truncate(size); }
      Status Close() override {
 // SyncPoint is not supported in Released Windows Mode.
@@ -253,6 +272,9 @@ class SpecialEnv : public EnvWrapper {
      Env::IOPriority GetIOPriority() override {
        return base_->GetIOPriority();
      }
+      bool use_direct_io() const override {
+        return base_->use_direct_io();
+      }
    };
    class ManifestFile : public WritableFile {
     public:
@@ -354,7 +376,14 @@ class SpecialEnv : public EnvWrapper {
      return Status::IOError("simulated write error");
    }

-    Status s = target()->NewWritableFile(f, r, soptions);
+    EnvOptions optimized = soptions;
+    if (strstr(f.c_str(), "MANIFEST") != nullptr ||
+        strstr(f.c_str(), "log") != nullptr) {
+      optimized.use_mmap_writes = false;
+      optimized.use_direct_writes = false;
+    }
+
+    Status s = target()->NewWritableFile(f, r, optimized);
    if (s.ok()) {
      if (strstr(f.c_str(), ".sst") != nullptr) {
        r->reset(new SSTableFile(this, std::move(*r)));
@@ -428,10 +457,10 @@ class SpecialEnv : public EnvWrapper {

  virtual void SleepForMicroseconds(int micros) override {
    sleep_counter_.Increment();
-    if (no_sleep_ || time_elapse_only_sleep_) {
+    if (no_slowdown_ || time_elapse_only_sleep_) {
      addon_time_.fetch_add(micros);
    }
-    if (!no_sleep_) {
+    if (!no_slowdown_) {
      target()->SleepForMicroseconds(micros);
    }
  }
@@ -520,7 +549,7 @@ class SpecialEnv : public EnvWrapper {

  bool time_elapse_only_sleep_;

-  bool no_sleep_;
+  bool no_slowdown_;

  std::atomic<bool> is_wal_sync_thread_safe_{true};
 };
@@ -590,6 +619,7 @@ class DBTestBase : public testing::Test {
    kLevelSubcompactions = 31,
    kUniversalSubcompactions = 32,
    kBlockBasedTableWithIndexRestartInterval = 33,
+    kBlockBasedTableWithPartitionedIndex = 34,
  };
  int option_config_;

@@ -641,9 +671,12 @@ class DBTestBase : public testing::Test {
  // test.  Return false if there are no more configurations to test.
  bool ChangeOptions(int skip_mask = kNoSkip);

-  // Switch between different compaction styles (we have only 2 now).
+  // Switch between different compaction styles.
  bool ChangeCompactOptions();

+  // Switch between different WAL-realted options.
+  bool ChangeWalOptions();
+
  // Switch between different filter policy
  // Jump from kDefault to kFilter to kFullFilter
  bool ChangeFilterOptions();
@@ -688,6 +721,8 @@ class DBTestBase : public testing::Test {

  Status TryReopen(const Options& options);

+  bool IsDirectIOSupported();
+
  Status Flush(int cf = 0);

  Status Put(const Slice& k, const Slice& v, WriteOptions wo = WriteOptions());
@@ -695,6 +730,12 @@ class DBTestBase : public testing::Test {
  Status Put(int cf, const Slice& k, const Slice& v,
             WriteOptions wo = WriteOptions());

+  Status Merge(const Slice& k, const Slice& v,
+               WriteOptions wo = WriteOptions());
+
+  Status Merge(int cf, const Slice& k, const Slice& v,
+               WriteOptions wo = WriteOptions());
+
  Status Delete(const std::string& k);

  Status Delete(int cf, const std::string& k);
@@ -821,8 +862,13 @@ class DBTestBase : public testing::Test {

  std::vector<std::uint64_t> ListTableFiles(Env* env, const std::string& path);

-  void VerifyDBFromMap(std::map<std::string, std::string> true_data,
-                       size_t* total_reads_res = nullptr);
+  void VerifyDBFromMap(
+      std::map<std::string, std::string> true_data,
+      size_t* total_reads_res = nullptr, bool tailing_iter = false,
+      std::map<std::string, Status> status = std::map<std::string, Status>());
+
+  void VerifyDBInternal(
+      std::vector<std::pair<std::string, std::string>> true_data);

 #ifndef ROCKSDB_LITE
  uint64_t GetNumberOfSstFilesForColumnFamily(DB* db,
@@ -1028,13 +1028,15 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest1) {
 // Test that checks trivial move in universal compaction
 TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest2) {
  int32_t trivial_move = 0;
-  int32_t non_trivial_move = 0;
  rocksdb::SyncPoint::GetInstance()->SetCallBack(
      "DBImpl::BackgroundCompaction:TrivialMove",
      [&](void* arg) { trivial_move++; });
  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "DBImpl::BackgroundCompaction:NonTrivial",
-      [&](void* arg) { non_trivial_move++; });
+      "DBImpl::BackgroundCompaction:NonTrivial", [&](void* arg) {
+        ASSERT_TRUE(arg != nullptr);
+        int output_level = *(static_cast<int*>(arg));
+        ASSERT_EQ(output_level, 0);
+      });

  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

@@ -1044,7 +1046,7 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest2) {
  options.num_levels = 15;
  options.write_buffer_size = 100 << 10;  // 100KB
  options.level0_file_num_compaction_trigger = 8;
-  options.max_background_compactions = 4;
+  options.max_background_compactions = 2;
  options.target_file_size_base = 64 * 1024;
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);
@@ -1065,7 +1067,6 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest2) {
  dbfull()->TEST_WaitForCompact();

  ASSERT_GT(trivial_move, 0);
-  ASSERT_EQ(non_trivial_move, 0);

  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }
@@ -8,6 +8,7 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

 #include "db/db_test_util.h"
+#include "port/port.h"
 #include "port/stack_trace.h"
 #include "util/fault_injection_test_env.h"
 #include "util/options_helper.h"
@@ -50,7 +51,7 @@ TEST_F(DBWALTest, WAL) {
    // again both values should be present.
    ASSERT_EQ("v3", Get(1, "foo"));
    ASSERT_EQ("v3", Get(1, "bar"));
-  } while (ChangeCompactOptions());
+  } while (ChangeWalOptions());
 }

 TEST_F(DBWALTest, RollLog) {
@@ -67,7 +68,7 @@ TEST_F(DBWALTest, RollLog) {
    for (int i = 0; i < 10; i++) {
      ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
    }
-  } while (ChangeOptions());
+  } while (ChangeWalOptions());
 }

 TEST_F(DBWALTest, SyncWALNotBlockWrite) {
@@ -86,7 +87,7 @@ TEST_F(DBWALTest, SyncWALNotBlockWrite) {
  });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  std::thread thread([&]() { ASSERT_OK(db_->SyncWAL()); });
+  rocksdb::port::Thread thread([&]() { ASSERT_OK(db_->SyncWAL()); });

  TEST_SYNC_POINT("DBWALTest::SyncWALNotBlockWrite:1");
  ASSERT_OK(Put("foo2", "bar2"));
@@ -118,7 +119,7 @@ TEST_F(DBWALTest, SyncWALNotWaitWrite) {
  });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  std::thread thread([&]() { ASSERT_OK(Put("foo2", "bar2")); });
+  rocksdb::port::Thread thread([&]() { ASSERT_OK(Put("foo2", "bar2")); });
  TEST_SYNC_POINT("DBWALTest::SyncWALNotWaitWrite:1");
  ASSERT_OK(db_->SyncWAL());
  TEST_SYNC_POINT("DBWALTest::SyncWALNotWaitWrite:2");
@@ -150,7 +151,7 @@ TEST_F(DBWALTest, Recover) {
    ASSERT_EQ("v4", Get(1, "foo"));
    ASSERT_EQ("v2", Get(1, "bar"));
    ASSERT_EQ("v5", Get(1, "baz"));
-  } while (ChangeOptions());
+  } while (ChangeWalOptions());
 }

 TEST_F(DBWALTest, RecoverWithTableHandle) {
@@ -187,23 +188,22 @@ TEST_F(DBWALTest, RecoverWithTableHandle) {
        }
      }
    }
-  } while (ChangeOptions());
+  } while (ChangeWalOptions());
 }

 TEST_F(DBWALTest, IgnoreRecoveredLog) {
  std::string backup_logs = dbname_ + "/backup_logs";

-  // delete old files in backup_logs directory
-  env_->CreateDirIfMissing(backup_logs);
-  std::vector<std::string> old_files;
-  env_->GetChildren(backup_logs, &old_files);
-  for (auto& file : old_files) {
-    if (file != "." && file != "..") {
-      env_->DeleteFile(backup_logs + "/" + file);
-    }
-  }
-
  do {
+    // delete old files in backup_logs directory
+    env_->CreateDirIfMissing(backup_logs);
+    std::vector<std::string> old_files;
+    env_->GetChildren(backup_logs, &old_files);
+    for (auto& file : old_files) {
+      if (file != "." && file != "..") {
+        env_->DeleteFile(backup_logs + "/" + file);
+      }
+    }
    Options options = CurrentOptions();
    options.create_if_missing = true;
    options.merge_operator = MergeOperators::CreateUInt64AddOperator();
@@ -276,7 +276,8 @@ TEST_F(DBWALTest, IgnoreRecoveredLog) {
    }
    Status s = TryReopen(options);
    ASSERT_TRUE(!s.ok());
-  } while (ChangeOptions(kSkipHashCuckoo));
+    Destroy(options);
+  } while (ChangeWalOptions());
 }

 TEST_F(DBWALTest, RecoveryWithEmptyLog) {
@@ -289,7 +290,7 @@ TEST_F(DBWALTest, RecoveryWithEmptyLog) {
    ASSERT_OK(Put(1, "foo", "v3"));
    ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
    ASSERT_EQ("v3", Get(1, "foo"));
-  } while (ChangeOptions());
+  } while (ChangeWalOptions());
 }

 #if !(defined NDEBUG) || !defined(OS_WIN)
@@ -428,7 +429,7 @@ TEST_F(DBWALTest, GetSortedWalFiles) {
    ASSERT_OK(Put(1, "foo", "v1"));
    ASSERT_OK(dbfull()->GetSortedWalFiles(log_files));
    ASSERT_EQ(1, log_files.size());
-  } while (ChangeOptions());
+  } while (ChangeWalOptions());
 }

 TEST_F(DBWALTest, RecoveryWithLogDataForSomeCFs) {
@@ -453,7 +454,7 @@ TEST_F(DBWALTest, RecoveryWithLogDataForSomeCFs) {
    }
    // Check at least the first WAL was cleaned up during the recovery.
    ASSERT_LT(earliest_log_nums[0], earliest_log_nums[1]);
-  } while (ChangeOptions());
+  } while (ChangeWalOptions());
 }

 TEST_F(DBWALTest, RecoverWithLargeLog) {
@@ -480,7 +481,7 @@ TEST_F(DBWALTest, RecoverWithLargeLog) {
    ASSERT_EQ(std::string(10, '3'), Get(1, "small3"));
    ASSERT_EQ(std::string(10, '4'), Get(1, "small4"));
    ASSERT_GT(NumTableFilesAtLevel(0, 1), 1);
-  } while (ChangeCompactOptions());
+  } while (ChangeWalOptions());
 }

 // In https://reviews.facebook.net/D20661 we change
@@ -645,8 +646,6 @@ TEST_F(DBWALTest, PartOfWritesWithWALDisabled) {
  Options options = CurrentOptions();
  options.env = fault_env.get();
  options.disable_auto_compactions = true;
-  // TODO(yiwu): fix for 2PC.
-  options.allow_2pc = false;
  WriteOptions wal_on, wal_off;
  wal_on.sync = true;
  wal_on.disableWAL = false;
@@ -680,9 +679,9 @@ class RecoveryTestHelper {
  // Starting number for the WAL file name like 00010.log
  static const int kWALFileOffset = 10;
  // Keys to be written per WAL file
-  static const int kKeysPerWALFile = 1024;
+  static const int kKeysPerWALFile = 133;
  // Size of the value
-  static const int kValueSize = 10;
+  static const int kValueSize = 96;

  // Create WAL files with values filled in
  static void FillData(DBWALTest* test, const Options& options,
@@ -691,7 +690,7 @@ class RecoveryTestHelper {

    *count = 0;

-    shared_ptr<Cache> table_cache = NewLRUCache(50000, 16);
+    shared_ptr<Cache> table_cache = NewLRUCache(50, 0);
    EnvOptions env_options;
    WriteBufferManager write_buffer_manager(db_options.db_write_buffer_size);

@@ -718,12 +717,13 @@ class RecoveryTestHelper {
          new log::Writer(std::move(file_writer), current_log_number,
                          db_options.recycle_log_file_num > 0));

+      WriteBatch batch;
      for (int i = 0; i < kKeysPerWALFile; i++) {
        std::string key = "key" + ToString((*count)++);
        std::string value = test->DummyString(kValueSize);
        assert(current_log_writer.get() != nullptr);
        uint64_t seq = versions->LastSequence() + 1;
-        WriteBatch batch;
+        batch.Clear();
        batch.Put(key, value);
        WriteBatchInternal::SetSequence(&batch, seq);
        current_log_writer->AddRecord(WriteBatchInternal::Contents(&batch));
@@ -774,7 +774,7 @@ class RecoveryTestHelper {
    if (trunc) {
      ASSERT_EQ(0, truncate(fname.c_str(), static_cast<int64_t>(size * off)));
    } else {
-      InduceCorruption(fname, static_cast<size_t>(size * off),
+      InduceCorruption(fname, static_cast<size_t>(size * off + 8),
                       static_cast<size_t>(size * len));
    }
  }
@@ -793,7 +793,7 @@ class RecoveryTestHelper {
    ASSERT_EQ(offset, lseek(fd, static_cast<long>(offset), SEEK_SET));

    void* buf = alloca(len);
-    memset(buf, 'a', len);
+    memset(buf, 'b', len);
    ASSERT_EQ(len, write(fd, buf, static_cast<unsigned int>(len)));

    close(fd);
@@ -809,7 +809,7 @@ TEST_F(DBWALTest, kTolerateCorruptedTailRecords) {
  const int jend = jstart + RecoveryTestHelper::kWALFilesCount;

  for (auto trunc : {true, false}) {        /* Corruption style */
-    for (int i = 0; i < 4; i++) {           /* Corruption offset position */
+    for (int i = 0; i < 3; i++) {           /* Corruption offset position */
      for (int j = jstart; j < jend; j++) { /* WAL file */
        // Fill data for testing
        Options options = CurrentOptions();
@@ -48,6 +48,11 @@ void AppendInternalKey(std::string* result, const ParsedInternalKey& key) {
  PutFixed64(result, PackSequenceAndType(key.sequence, key.type));
 }

+void AppendInternalKeyFooter(std::string* result, SequenceNumber s,
+                             ValueType t) {
+  PutFixed64(result, PackSequenceAndType(s, t));
+}
+
 std::string ParsedInternalKey::DebugString(bool hex) const {
  char buf[50];
  snprintf(buf, sizeof(buf), "' @ %" PRIu64 ": %d", sequence,
@@ -99,6 +99,11 @@ extern void UnPackSequenceAndType(uint64_t packed, uint64_t* seq, ValueType* t);
 // Append the serialization of "key" to *result.
 extern void AppendInternalKey(std::string* result,
                              const ParsedInternalKey& key);
+// Serialized internal key consists of user key followed by footer.
+// This function appends the footer to *result, assuming that *result already
+// contains the user key at the end.
+extern void AppendInternalKeyFooter(std::string* result, SequenceNumber s,
+                                    ValueType t);

 // Attempt to parse an internal key from "internal_key".  On success,
 // stores the parsed data in "*result", and returns true.
@@ -197,6 +202,16 @@ class InternalKey {

  void Clear() { rep_.clear(); }

+  // The underlying representation.
+  // Intended only to be used together with ConvertFromUserKey().
+  std::string* rep() { return &rep_; }
+
+  // Assuming that *rep() contains a user key, this method makes internal key
+  // out of it in-place. This saves a memcpy compared to Set()/SetFrom().
+  void ConvertFromUserKey(SequenceNumber s, ValueType t) {
+    AppendInternalKeyFooter(&rep_, s, t);
+  }
+
  std::string DebugString(bool hex = false) const;
 };

@@ -513,10 +528,11 @@ struct RangeTombstone {
  Slice start_key_;
  Slice end_key_;
  SequenceNumber seq_;
-  explicit RangeTombstone(Slice sk, Slice ek, SequenceNumber sn)
+  RangeTombstone() = default;
+  RangeTombstone(Slice sk, Slice ek, SequenceNumber sn)
      : start_key_(sk), end_key_(ek), seq_(sn) {}

-  explicit RangeTombstone(ParsedInternalKey parsed_key, Slice value) {
+  RangeTombstone(ParsedInternalKey parsed_key, Slice value) {
    start_key_ = parsed_key.user_key;
    seq_ = parsed_key.sequence;
    end_key_ = value;
@@ -128,7 +128,7 @@ void EventHelpers::LogAndNotifyTableFileDeletion(
  info.job_id = job_id;
  info.file_path = file_path;
  info.status = status;
-  for (auto listener : listeners) {
+  for (auto& listener : listeners) {
    listener->OnTableFileDeleted(info);
  }
 #endif  // !ROCKSDB_LITE
@@ -0,0 +1,323 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#include <functional>
+
+#include "db/db_test_util.h"
+#include "port/port.h"
+#include "port/stack_trace.h"
+#include "rocksdb/sst_file_writer.h"
+#include "util/testutil.h"
+
+namespace rocksdb {
+
+#ifndef ROCKSDB_LITE
+class ExternalSSTFileBasicTest : public DBTestBase {
+ public:
+  ExternalSSTFileBasicTest() : DBTestBase("/external_sst_file_test") {
+    sst_files_dir_ = dbname_ + "/sst_files/";
+    DestroyAndRecreateExternalSSTFilesDir();
+  }
+
+  void DestroyAndRecreateExternalSSTFilesDir() {
+    test::DestroyDir(env_, sst_files_dir_);
+    env_->CreateDir(sst_files_dir_);
+  }
+
+  Status DeprecatedAddFile(const std::vector<std::string>& files,
+                           bool move_files = false,
+                           bool skip_snapshot_check = false) {
+    IngestExternalFileOptions opts;
+    opts.move_files = move_files;
+    opts.snapshot_consistency = !skip_snapshot_check;
+    opts.allow_global_seqno = false;
+    opts.allow_blocking_flush = false;
+    return db_->IngestExternalFile(files, opts);
+  }
+
+  Status GenerateAndAddExternalFile(
+      const Options options, std::vector<int> keys, int file_id,
+      std::map<std::string, std::string>* true_data) {
+    std::string file_path = sst_files_dir_ + ToString(file_id);
+    SstFileWriter sst_file_writer(EnvOptions(), options, options.comparator,
+                                  nullptr);
+
+    Status s = sst_file_writer.Open(file_path);
+    if (!s.ok()) {
+      return s;
+    }
+    for (int k : keys) {
+      std::string key = Key(k);
+      std::string value = Key(k) + ToString(file_id);
+      s = sst_file_writer.Add(key, value);
+      (*true_data)[key] = value;
+      if (!s.ok()) {
+        sst_file_writer.Finish();
+        return s;
+      }
+    }
+    s = sst_file_writer.Finish();
+
+    if (s.ok()) {
+      IngestExternalFileOptions ifo;
+      ifo.allow_global_seqno = true;
+      s = db_->IngestExternalFile({file_path}, ifo);
+    }
+
+    return s;
+  }
+
+  ~ExternalSSTFileBasicTest() { test::DestroyDir(env_, sst_files_dir_); }
+
+ protected:
+  std::string sst_files_dir_;
+};
+
+TEST_F(ExternalSSTFileBasicTest, Basic) {
+  Options options = CurrentOptions();
+
+  SstFileWriter sst_file_writer(EnvOptions(), options, options.comparator);
+
+  // Current file size should be 0 after sst_file_writer init and before open a
+  // file.
+  ASSERT_EQ(sst_file_writer.FileSize(), 0);
+
+  // file1.sst (0 => 99)
+  std::string file1 = sst_files_dir_ + "file1.sst";
+  ASSERT_OK(sst_file_writer.Open(file1));
+  for (int k = 0; k < 100; k++) {
+    ASSERT_OK(sst_file_writer.Add(Key(k), Key(k) + "_val"));
+  }
+  ExternalSstFileInfo file1_info;
+  Status s = sst_file_writer.Finish(&file1_info);
+  ASSERT_TRUE(s.ok()) << s.ToString();
+
+  // Current file size should be non-zero after success write.
+  ASSERT_GT(sst_file_writer.FileSize(), 0);
+
+  ASSERT_EQ(file1_info.file_path, file1);
+  ASSERT_EQ(file1_info.num_entries, 100);
+  ASSERT_EQ(file1_info.smallest_key, Key(0));
+  ASSERT_EQ(file1_info.largest_key, Key(99));
+  // sst_file_writer already finished, cannot add this value
+  s = sst_file_writer.Add(Key(100), "bad_val");
+  ASSERT_FALSE(s.ok()) << s.ToString();
+
+  DestroyAndReopen(options);
+  // Add file using file path
+  s = DeprecatedAddFile({file1});
+  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_EQ(db_->GetLatestSequenceNumber(), 0U);
+  for (int k = 0; k < 100; k++) {
+    ASSERT_EQ(Get(Key(k)), Key(k) + "_val");
+  }
+
+  DestroyAndRecreateExternalSSTFilesDir();
+}
+
+TEST_F(ExternalSSTFileBasicTest, NoCopy) {
+  Options options = CurrentOptions();
+  const ImmutableCFOptions ioptions(options);
+
+  SstFileWriter sst_file_writer(EnvOptions(), options, options.comparator);
+
+  // file1.sst (0 => 99)
+  std::string file1 = sst_files_dir_ + "file1.sst";
+  ASSERT_OK(sst_file_writer.Open(file1));
+  for (int k = 0; k < 100; k++) {
+    ASSERT_OK(sst_file_writer.Add(Key(k), Key(k) + "_val"));
+  }
+  ExternalSstFileInfo file1_info;
+  Status s = sst_file_writer.Finish(&file1_info);
+  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_EQ(file1_info.file_path, file1);
+  ASSERT_EQ(file1_info.num_entries, 100);
+  ASSERT_EQ(file1_info.smallest_key, Key(0));
+  ASSERT_EQ(file1_info.largest_key, Key(99));
+
+  // file2.sst (100 => 299)
+  std::string file2 = sst_files_dir_ + "file2.sst";
+  ASSERT_OK(sst_file_writer.Open(file2));
+  for (int k = 100; k < 300; k++) {
+    ASSERT_OK(sst_file_writer.Add(Key(k), Key(k) + "_val"));
+  }
+  ExternalSstFileInfo file2_info;
+  s = sst_file_writer.Finish(&file2_info);
+  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_EQ(file2_info.file_path, file2);
+  ASSERT_EQ(file2_info.num_entries, 200);
+  ASSERT_EQ(file2_info.smallest_key, Key(100));
+  ASSERT_EQ(file2_info.largest_key, Key(299));
+
+  // file3.sst (110 => 124) .. overlap with file2.sst
+  std::string file3 = sst_files_dir_ + "file3.sst";
+  ASSERT_OK(sst_file_writer.Open(file3));
+  for (int k = 110; k < 125; k++) {
+    ASSERT_OK(sst_file_writer.Add(Key(k), Key(k) + "_val_overlap"));
+  }
+  ExternalSstFileInfo file3_info;
+  s = sst_file_writer.Finish(&file3_info);
+  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_EQ(file3_info.file_path, file3);
+  ASSERT_EQ(file3_info.num_entries, 15);
+  ASSERT_EQ(file3_info.smallest_key, Key(110));
+  ASSERT_EQ(file3_info.largest_key, Key(124));
+  s = DeprecatedAddFile({file1}, true /* move file */);
+  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_EQ(Status::NotFound(), env_->FileExists(file1));
+
+  s = DeprecatedAddFile({file2}, false /* copy file */);
+  ASSERT_TRUE(s.ok()) << s.ToString();
+  ASSERT_OK(env_->FileExists(file2));
+
+  // This file have overlapping values with the existing data
+  s = DeprecatedAddFile({file2}, true /* move file */);
+  ASSERT_FALSE(s.ok()) << s.ToString();
+  ASSERT_OK(env_->FileExists(file3));
+
+  for (int k = 0; k < 300; k++) {
+    ASSERT_EQ(Get(Key(k)), Key(k) + "_val");
+  }
+}
+
+TEST_F(ExternalSSTFileBasicTest, IngestFileWithGlobalSeqnoPickedSeqno) {
+  Options options = CurrentOptions();
+  DestroyAndReopen(options);
+  std::map<std::string, std::string> true_data;
+
+  int file_id = 1;
+
+  ASSERT_OK(GenerateAndAddExternalFile(options, {1, 2, 3, 4, 5, 6}, file_id++,
+                                       &true_data));
+  // File dont overwrite any keys, No seqno needed
+  ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), 0);
+
+  ASSERT_OK(GenerateAndAddExternalFile(options, {10, 11, 12, 13}, file_id++,
+                                       &true_data));
+  // File dont overwrite any keys, No seqno needed
+  ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), 0);
+
+  ASSERT_OK(
+      GenerateAndAddExternalFile(options, {1, 4, 6}, file_id++, &true_data));
+  // File overwrite some keys, a seqno will be assigned
+  ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), 1);
+
+  ASSERT_OK(
+      GenerateAndAddExternalFile(options, {11, 15, 19}, file_id++, &true_data));
+  // File overwrite some keys, a seqno will be assigned
+  ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), 2);
+
+  ASSERT_OK(
+      GenerateAndAddExternalFile(options, {120, 130}, file_id++, &true_data));
+  // File dont overwrite any keys, No seqno needed
+  ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), 2);
+
+  ASSERT_OK(
+      GenerateAndAddExternalFile(options, {1, 130}, file_id++, &true_data));
+  // File overwrite some keys, a seqno will be assigned
+  ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), 3);
+
+  // Write some keys through normal write path
+  for (int i = 0; i < 50; i++) {
+    ASSERT_OK(Put(Key(i), "memtable"));
+    true_data[Key(i)] = "memtable";
+  }
+  SequenceNumber last_seqno = dbfull()->GetLatestSequenceNumber();
+
+  ASSERT_OK(
+      GenerateAndAddExternalFile(options, {60, 61, 62}, file_id++, &true_data));
+  // File dont overwrite any keys, No seqno needed
+  ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), last_seqno);
+
+  ASSERT_OK(
+      GenerateAndAddExternalFile(options, {40, 41, 42}, file_id++, &true_data));
+  // File overwrite some keys, a seqno will be assigned
+  ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), last_seqno + 1);
+
+  ASSERT_OK(
+      GenerateAndAddExternalFile(options, {20, 30, 40}, file_id++, &true_data));
+  // File overwrite some keys, a seqno will be assigned
+  ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), last_seqno + 2);
+
+  const Snapshot* snapshot = db_->GetSnapshot();
+
+  // We will need a seqno for the file regardless if the file overwrite
+  // keys in the DB or not because we have a snapshot
+  ASSERT_OK(
+      GenerateAndAddExternalFile(options, {1000, 1002}, file_id++, &true_data));
+  // A global seqno will be assigned anyway because of the snapshot
+  ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), last_seqno + 3);
+
+  ASSERT_OK(
+      GenerateAndAddExternalFile(options, {2000, 3002}, file_id++, &true_data));
+  // A global seqno will be assigned anyway because of the snapshot
+  ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), last_seqno + 4);
+
+  ASSERT_OK(GenerateAndAddExternalFile(options, {1, 20, 40, 100, 150},
+                                       file_id++, &true_data));
+  // A global seqno will be assigned anyway because of the snapshot
+  ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), last_seqno + 5);
+
+  db_->ReleaseSnapshot(snapshot);
+
+  ASSERT_OK(
+      GenerateAndAddExternalFile(options, {5000, 5001}, file_id++, &true_data));
+  // No snapshot anymore, no need to assign a seqno
+  ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), last_seqno + 5);
+
+  size_t kcnt = 0;
+  VerifyDBFromMap(true_data, &kcnt, false);
+}
+
+TEST_F(ExternalSSTFileBasicTest, FadviseTrigger) {
+  Options options = CurrentOptions();
+  const int kNumKeys = 10000;
+
+  size_t total_fadvised_bytes = 0;
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "SstFileWriter::InvalidatePageCache", [&](void* arg) {
+        size_t fadvise_size = *(reinterpret_cast<size_t*>(arg));
+        total_fadvised_bytes += fadvise_size;
+      });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  std::unique_ptr<SstFileWriter> sst_file_writer;
+
+  std::string sst_file_path = sst_files_dir_ + "file_fadvise_disable.sst";
+  sst_file_writer.reset(new SstFileWriter(EnvOptions(), options,
+                                          options.comparator, nullptr, false));
+  ASSERT_OK(sst_file_writer->Open(sst_file_path));
+  for (int i = 0; i < kNumKeys; i++) {
+    ASSERT_OK(sst_file_writer->Add(Key(i), Key(i)));
+  }
+  ASSERT_OK(sst_file_writer->Finish());
+  // fadvise disabled
+  ASSERT_EQ(total_fadvised_bytes, 0);
+
+
+  sst_file_path = sst_files_dir_ + "file_fadvise_enable.sst";
+  sst_file_writer.reset(new SstFileWriter(EnvOptions(), options,
+                                          options.comparator, nullptr, true));
+  ASSERT_OK(sst_file_writer->Open(sst_file_path));
+  for (int i = 0; i < kNumKeys; i++) {
+    ASSERT_OK(sst_file_writer->Add(Key(i), Key(i)));
+  }
+  ASSERT_OK(sst_file_writer->Finish());
+  // fadvise enabled
+  ASSERT_EQ(total_fadvised_bytes, sst_file_writer->FileSize());
+  ASSERT_GT(total_fadvised_bytes, 0);
+
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+}
+
+#endif  // ROCKSDB_LITE
+
+}  // namespace rocksdb
+
+int main(int argc, char** argv) {
+  rocksdb::port::InstallStackTraceHandler();
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}
@@ -0,0 +1,531 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#ifndef ROCKSDB_LITE
+
+#include "db/external_sst_file_ingestion_job.h"
+
+#define __STDC_FORMAT_MACROS
+
+#include <inttypes.h>
+#include <algorithm>
+#include <string>
+#include <vector>
+
+#include "db/version_edit.h"
+#include "table/merging_iterator.h"
+#include "table/scoped_arena_iterator.h"
+#include "table/sst_file_writer_collectors.h"
+#include "table/table_builder.h"
+#include "util/file_reader_writer.h"
+#include "util/file_util.h"
+#include "util/stop_watch.h"
+#include "util/sync_point.h"
+
+namespace rocksdb {
+
+Status ExternalSstFileIngestionJob::Prepare(
+    const std::vector<std::string>& external_files_paths) {
+  Status status;
+
+  // Read the information of files we are ingesting
+  for (const std::string& file_path : external_files_paths) {
+    IngestedFileInfo file_to_ingest;
+    status = GetIngestedFileInfo(file_path, &file_to_ingest);
+    if (!status.ok()) {
+      return status;
+    }
+    files_to_ingest_.push_back(file_to_ingest);
+  }
+
+  for (const IngestedFileInfo& f : files_to_ingest_) {
+    if (f.cf_id !=
+            TablePropertiesCollectorFactory::Context::kUnknownColumnFamily &&
+        f.cf_id != cfd_->GetID()) {
+      return Status::InvalidArgument(
+          "External file column family id dont match");
+    }
+  }
+
+  const Comparator* ucmp = cfd_->internal_comparator().user_comparator();
+  auto num_files = files_to_ingest_.size();
+  if (num_files == 0) {
+    return Status::InvalidArgument("The list of files is empty");
+  } else if (num_files > 1) {
+    // Verify that passed files dont have overlapping ranges
+    autovector<const IngestedFileInfo*> sorted_files;
+    for (size_t i = 0; i < num_files; i++) {
+      sorted_files.push_back(&files_to_ingest_[i]);
+    }
+
+    std::sort(
+        sorted_files.begin(), sorted_files.end(),
+        [&ucmp](const IngestedFileInfo* info1, const IngestedFileInfo* info2) {
+          return ucmp->Compare(info1->smallest_user_key,
+                               info2->smallest_user_key) < 0;
+        });
+
+    for (size_t i = 0; i < num_files - 1; i++) {
+      if (ucmp->Compare(sorted_files[i]->largest_user_key,
+                        sorted_files[i + 1]->smallest_user_key) >= 0) {
+        return Status::NotSupported("Files have overlapping ranges");
+      }
+    }
+  }
+
+  for (IngestedFileInfo& f : files_to_ingest_) {
+    if (f.num_entries == 0) {
+      return Status::InvalidArgument("File contain no entries");
+    }
+
+    if (!f.smallest_internal_key().Valid() ||
+        !f.largest_internal_key().Valid()) {
+      return Status::Corruption("Generated table have corrupted keys");
+    }
+  }
+
+  // Copy/Move external files into DB
+  for (IngestedFileInfo& f : files_to_ingest_) {
+    f.fd = FileDescriptor(versions_->NewFileNumber(), 0, f.file_size);
+
+    const std::string path_outside_db = f.external_file_path;
+    const std::string path_inside_db =
+        TableFileName(db_options_.db_paths, f.fd.GetNumber(), f.fd.GetPathId());
+
+    if (ingestion_options_.move_files) {
+      status = env_->LinkFile(path_outside_db, path_inside_db);
+      if (status.IsNotSupported()) {
+        // Original file is on a different FS, use copy instead of hard linking
+        status = CopyFile(env_, path_outside_db, path_inside_db, 0,
+                          db_options_.use_fsync);
+      }
+    } else {
+      status = CopyFile(env_, path_outside_db, path_inside_db, 0,
+                        db_options_.use_fsync);
+    }
+    TEST_SYNC_POINT("DBImpl::AddFile:FileCopied");
+    if (!status.ok()) {
+      break;
+    }
+    f.internal_file_path = path_inside_db;
+  }
+
+  if (!status.ok()) {
+    // We failed, remove all files that we copied into the db
+    for (IngestedFileInfo& f : files_to_ingest_) {
+      if (f.internal_file_path == "") {
+        break;
+      }
+      Status s = env_->DeleteFile(f.internal_file_path);
+      if (!s.ok()) {
+        Log(InfoLogLevel::WARN_LEVEL, db_options_.info_log,
+            "AddFile() clean up for file %s failed : %s",
+            f.internal_file_path.c_str(), s.ToString().c_str());
+      }
+    }
+  }
+
+  return status;
+}
+
+Status ExternalSstFileIngestionJob::NeedsFlush(bool* flush_needed) {
+  SuperVersion* super_version = cfd_->GetSuperVersion();
+  Status status =
+      IngestedFilesOverlapWithMemtables(super_version, flush_needed);
+
+  if (status.ok() && *flush_needed &&
+      !ingestion_options_.allow_blocking_flush) {
+    status = Status::InvalidArgument("External file requires flush");
+  }
+  return status;
+}
+
+Status ExternalSstFileIngestionJob::Run() {
+  Status status;
+#ifndef NDEBUG
+  // We should never run the job with a memtable that is overlapping
+  // with the files we are ingesting
+  bool need_flush = false;
+  status = NeedsFlush(&need_flush);
+  assert(status.ok() && need_flush == false);
+#endif
+
+  bool consumed_seqno = false;
+  bool force_global_seqno = false;
+  const SequenceNumber last_seqno = versions_->LastSequence();
+  if (ingestion_options_.snapshot_consistency && !db_snapshots_->empty()) {
+    // We need to assign a global sequence number to all the files even
+    // if the dont overlap with any ranges since we have snapshots
+    force_global_seqno = true;
+  }
+
+  SuperVersion* super_version = cfd_->GetSuperVersion();
+  edit_.SetColumnFamily(cfd_->GetID());
+  // The levels that the files will be ingested into
+  for (IngestedFileInfo& f : files_to_ingest_) {
+    bool overlap_with_db = false;
+    status = AssignLevelForIngestedFile(super_version, &f, &overlap_with_db);
+    if (!status.ok()) {
+      return status;
+    }
+
+    if (overlap_with_db || force_global_seqno) {
+      status = AssignGlobalSeqnoForIngestedFile(&f, last_seqno + 1);
+      consumed_seqno = true;
+    } else {
+      status = AssignGlobalSeqnoForIngestedFile(&f, 0);
+    }
+
+    if (!status.ok()) {
+      return status;
+    }
+
+    edit_.AddFile(f.picked_level, f.fd.GetNumber(), f.fd.GetPathId(),
+                  f.fd.GetFileSize(), f.smallest_internal_key(),
+                  f.largest_internal_key(), f.assigned_seqno, f.assigned_seqno,
+                  false);
+  }
+
+  if (consumed_seqno) {
+    versions_->SetLastSequence(last_seqno + 1);
+  }
+
+  return status;
+}
+
+void ExternalSstFileIngestionJob::UpdateStats() {
+  // Update internal stats for new ingested files
+  uint64_t total_keys = 0;
+  uint64_t total_l0_files = 0;
+  uint64_t total_time = env_->NowMicros() - job_start_time_;
+  for (IngestedFileInfo& f : files_to_ingest_) {
+    InternalStats::CompactionStats stats(1);
+    stats.micros = total_time;
+    stats.bytes_written = f.fd.GetFileSize();
+    stats.num_output_files = 1;
+    cfd_->internal_stats()->AddCompactionStats(f.picked_level, stats);
+    cfd_->internal_stats()->AddCFStats(InternalStats::BYTES_INGESTED_ADD_FILE,
+                                       f.fd.GetFileSize());
+    total_keys += f.num_entries;
+    if (f.picked_level == 0) {
+      total_l0_files += 1;
+    }
+    Log(InfoLogLevel::INFO_LEVEL, db_options_.info_log,
+        "[AddFile] External SST file %s was ingested in L%d with path %s "
+        "(global_seqno=%" PRIu64 ")\n",
+        f.external_file_path.c_str(), f.picked_level,
+        f.internal_file_path.c_str(), f.assigned_seqno);
+  }
+  cfd_->internal_stats()->AddCFStats(InternalStats::INGESTED_NUM_KEYS_TOTAL,
+                                     total_keys);
+  cfd_->internal_stats()->AddCFStats(InternalStats::INGESTED_NUM_FILES_TOTAL,
+                                     files_to_ingest_.size());
+  cfd_->internal_stats()->AddCFStats(
+      InternalStats::INGESTED_LEVEL0_NUM_FILES_TOTAL, total_l0_files);
+}
+
+void ExternalSstFileIngestionJob::Cleanup(const Status& status) {
+  if (!status.ok()) {
+    // We failed to add the files to the database
+    // remove all the files we copied
+    for (IngestedFileInfo& f : files_to_ingest_) {
+      Status s = env_->DeleteFile(f.internal_file_path);
+      if (!s.ok()) {
+        Log(InfoLogLevel::WARN_LEVEL, db_options_.info_log,
+            "AddFile() clean up for file %s failed : %s",
+            f.internal_file_path.c_str(), s.ToString().c_str());
+      }
+    }
+  } else if (status.ok() && ingestion_options_.move_files) {
+    // The files were moved and added successfully, remove original file links
+    for (IngestedFileInfo& f : files_to_ingest_) {
+      Status s = env_->DeleteFile(f.external_file_path);
+      if (!s.ok()) {
+        Log(InfoLogLevel::WARN_LEVEL, db_options_.info_log,
+            "%s was added to DB successfully but failed to remove original "
+            "file link : %s",
+            f.external_file_path.c_str(), s.ToString().c_str());
+      }
+    }
+  }
+}
+
+Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
+    const std::string& external_file, IngestedFileInfo* file_to_ingest) {
+  file_to_ingest->external_file_path = external_file;
+
+  // Get external file size
+  Status status = env_->GetFileSize(external_file, &file_to_ingest->file_size);
+  if (!status.ok()) {
+    return status;
+  }
+
+  // Create TableReader for external file
+  std::unique_ptr<TableReader> table_reader;
+  std::unique_ptr<RandomAccessFile> sst_file;
+  std::unique_ptr<RandomAccessFileReader> sst_file_reader;
+
+  status = env_->NewRandomAccessFile(external_file, &sst_file, env_options_);
+  if (!status.ok()) {
+    return status;
+  }
+  sst_file_reader.reset(new RandomAccessFileReader(std::move(sst_file)));
+
+  status = cfd_->ioptions()->table_factory->NewTableReader(
+      TableReaderOptions(*cfd_->ioptions(), env_options_,
+                         cfd_->internal_comparator()),
+      std::move(sst_file_reader), file_to_ingest->file_size, &table_reader);
+  if (!status.ok()) {
+    return status;
+  }
+
+  // Get the external file properties
+  auto props = table_reader->GetTableProperties();
+  const auto& uprops = props->user_collected_properties;
+
+  // Get table version
+  auto version_iter = uprops.find(ExternalSstFilePropertyNames::kVersion);
+  if (version_iter == uprops.end()) {
+    return Status::Corruption("External file version not found");
+  }
+  file_to_ingest->version = DecodeFixed32(version_iter->second.c_str());
+
+  auto seqno_iter = uprops.find(ExternalSstFilePropertyNames::kGlobalSeqno);
+  if (file_to_ingest->version == 2) {
+    // version 2 imply that we have global sequence number
+    if (seqno_iter == uprops.end()) {
+      return Status::Corruption(
+          "External file global sequence number not found");
+    }
+
+    // Set the global sequence number
+    file_to_ingest->original_seqno = DecodeFixed64(seqno_iter->second.c_str());
+    file_to_ingest->global_seqno_offset = props->properties_offsets.at(
+        ExternalSstFilePropertyNames::kGlobalSeqno);
+
+    if (file_to_ingest->global_seqno_offset == 0) {
+      return Status::Corruption("Was not able to find file global seqno field");
+    }
+  } else if (file_to_ingest->version == 1) {
+    // SST file V1 should not have global seqno field
+    assert(seqno_iter == uprops.end());
+    file_to_ingest->original_seqno = 0;
+    if (ingestion_options_.allow_blocking_flush ||
+            ingestion_options_.allow_global_seqno) {
+      return Status::InvalidArgument(
+            "External SST file V1 does not support global seqno");
+    }
+  } else {
+    return Status::InvalidArgument("External file version is not supported");
+  }
+  // Get number of entries in table
+  file_to_ingest->num_entries = props->num_entries;
+
+  ParsedInternalKey key;
+  ReadOptions ro;
+  // During reading the external file we can cache blocks that we read into
+  // the block cache, if we later change the global seqno of this file, we will
+  // have block in cache that will include keys with wrong seqno.
+  // We need to disable fill_cache so that we read from the file without
+  // updating the block cache.
+  ro.fill_cache = false;
+  std::unique_ptr<InternalIterator> iter(table_reader->NewIterator(ro));
+
+  // Get first (smallest) key from file
+  iter->SeekToFirst();
+  if (!ParseInternalKey(iter->key(), &key)) {
+    return Status::Corruption("external file have corrupted keys");
+  }
+  if (key.sequence != 0) {
+    return Status::Corruption("external file have non zero sequence number");
+  }
+  file_to_ingest->smallest_user_key = key.user_key.ToString();
+
+  // Get last (largest) key from file
+  iter->SeekToLast();
+  if (!ParseInternalKey(iter->key(), &key)) {
+    return Status::Corruption("external file have corrupted keys");
+  }
+  if (key.sequence != 0) {
+    return Status::Corruption("external file have non zero sequence number");
+  }
+  file_to_ingest->largest_user_key = key.user_key.ToString();
+
+  file_to_ingest->cf_id = static_cast<uint32_t>(props->column_family_id);
+
+  file_to_ingest->table_properties = *props;
+
+  return status;
+}
+
+Status ExternalSstFileIngestionJob::IngestedFilesOverlapWithMemtables(
+    SuperVersion* sv, bool* overlap) {
+  // Create an InternalIterator over all memtables
+  Arena arena;
+  ReadOptions ro;
+  ro.total_order_seek = true;
+  MergeIteratorBuilder merge_iter_builder(&cfd_->internal_comparator(), &arena);
+  merge_iter_builder.AddIterator(sv->mem->NewIterator(ro, &arena));
+  sv->imm->AddIterators(ro, &merge_iter_builder);
+  ScopedArenaIterator memtable_iter(merge_iter_builder.Finish());
+
+  Status status;
+  *overlap = false;
+  for (IngestedFileInfo& f : files_to_ingest_) {
+    status =
+        IngestedFileOverlapWithIteratorRange(&f, memtable_iter.get(), overlap);
+    if (!status.ok() || *overlap == true) {
+      break;
+    }
+  }
+
+  return status;
+}
+
+Status ExternalSstFileIngestionJob::AssignLevelForIngestedFile(
+    SuperVersion* sv, IngestedFileInfo* file_to_ingest, bool* overlap_with_db) {
+  *overlap_with_db = false;
+
+  Arena arena;
+  ReadOptions ro;
+  ro.total_order_seek = true;
+
+  Status status;
+  int target_level = 0;
+  auto* vstorage = cfd_->current()->storage_info();
+  for (int lvl = 0; lvl < cfd_->NumberLevels(); lvl++) {
+    if (lvl > 0 && lvl < vstorage->base_level()) {
+      continue;
+    }
+
+    if (vstorage->NumLevelFiles(lvl) > 0) {
+      bool overlap_with_level = false;
+      MergeIteratorBuilder merge_iter_builder(&cfd_->internal_comparator(),
+                                              &arena);
+      RangeDelAggregator range_del_agg(cfd_->internal_comparator(),
+                                       {} /* snapshots */);
+      sv->current->AddIteratorsForLevel(ro, env_options_, &merge_iter_builder,
+                                        lvl, &range_del_agg);
+      if (!range_del_agg.IsEmpty()) {
+        return Status::NotSupported(
+            "file ingestion with range tombstones is currently unsupported");
+      }
+      ScopedArenaIterator level_iter(merge_iter_builder.Finish());
+
+      status = IngestedFileOverlapWithIteratorRange(
+          file_to_ingest, level_iter.get(), &overlap_with_level);
+      if (!status.ok()) {
+        return status;
+      }
+
+      if (overlap_with_level) {
+        // We must use L0 or any level higher than `lvl` to be able to overwrite
+        // the keys that we overlap with in this level, We also need to assign
+        // this file a seqno to overwrite the existing keys in level `lvl`
+        *overlap_with_db = true;
+        break;
+      }
+    }
+
+    // We dont overlap with any keys in this level, but we still need to check
+    // if our file can fit in it
+
+    if (IngestedFileFitInLevel(file_to_ingest, lvl)) {
+      target_level = lvl;
+    }
+  }
+  file_to_ingest->picked_level = target_level;
+  return status;
+}
+
+Status ExternalSstFileIngestionJob::AssignGlobalSeqnoForIngestedFile(
+    IngestedFileInfo* file_to_ingest, SequenceNumber seqno) {
+  if (file_to_ingest->original_seqno == seqno) {
+    // This file already have the correct global seqno
+    return Status::OK();
+  } else if (!ingestion_options_.allow_global_seqno) {
+    return Status::InvalidArgument("Global seqno is required, but disabled");
+  } else if (file_to_ingest->global_seqno_offset == 0) {
+    return Status::InvalidArgument(
+        "Trying to set global seqno for a file that dont have a global seqno "
+        "field");
+  }
+
+  std::unique_ptr<RandomRWFile> rwfile;
+  Status status = env_->NewRandomRWFile(file_to_ingest->internal_file_path,
+                                        &rwfile, env_options_);
+  if (!status.ok()) {
+    return status;
+  }
+
+  // Write the new seqno in the global sequence number field in the file
+  std::string seqno_val;
+  PutFixed64(&seqno_val, seqno);
+  status = rwfile->Write(file_to_ingest->global_seqno_offset, seqno_val);
+  if (status.ok()) {
+    file_to_ingest->assigned_seqno = seqno;
+  }
+  return status;
+}
+
+Status ExternalSstFileIngestionJob::IngestedFileOverlapWithIteratorRange(
+    const IngestedFileInfo* file_to_ingest, InternalIterator* iter,
+    bool* overlap) {
+  auto* vstorage = cfd_->current()->storage_info();
+  auto* ucmp = vstorage->InternalComparator()->user_comparator();
+  InternalKey range_start(file_to_ingest->smallest_user_key, kMaxSequenceNumber,
+                          kValueTypeForSeek);
+  iter->Seek(range_start.Encode());
+  if (!iter->status().ok()) {
+    return iter->status();
+  }
+
+  *overlap = false;
+  if (iter->Valid()) {
+    ParsedInternalKey seek_result;
+    if (!ParseInternalKey(iter->key(), &seek_result)) {
+      return Status::Corruption("DB have corrupted keys");
+    }
+
+    if (ucmp->Compare(seek_result.user_key, file_to_ingest->largest_user_key) <=
+        0) {
+      *overlap = true;
+    }
+  }
+
+  return iter->status();
+}
+
+bool ExternalSstFileIngestionJob::IngestedFileFitInLevel(
+    const IngestedFileInfo* file_to_ingest, int level) {
+  if (level == 0) {
+    // Files can always fit in L0
+    return true;
+  }
+
+  auto* vstorage = cfd_->current()->storage_info();
+  Slice file_smallest_user_key(file_to_ingest->smallest_user_key);
+  Slice file_largest_user_key(file_to_ingest->largest_user_key);
+
+  if (vstorage->OverlapInLevel(level, &file_smallest_user_key,
+                               &file_largest_user_key)) {
+    // File overlap with another files in this level, we cannot
+    // add it to this level
+    return false;
+  }
+  if (cfd_->RangeOverlapWithCompaction(file_smallest_user_key,
+                                       file_largest_user_key, level)) {
+    // File overlap with a running compaction output that will be stored
+    // in this level, we cannot add this file to this level
+    return false;
+  }
+
+  // File did not overlap with level files, our compaction output
+  return true;
+}
+
+}  // namespace rocksdb
+
+#endif  // !ROCKSDB_LITE
@@ -0,0 +1,151 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#pragma once
+#include <string>
+#include <unordered_set>
+#include <vector>
+
+#include "db/column_family.h"
+#include "db/dbformat.h"
+#include "db/internal_stats.h"
+#include "db/snapshot_impl.h"
+#include "rocksdb/db.h"
+#include "rocksdb/env.h"
+#include "rocksdb/sst_file_writer.h"
+#include "util/autovector.h"
+#include "util/db_options.h"
+
+namespace rocksdb {
+
+struct IngestedFileInfo {
+  // External file path
+  std::string external_file_path;
+  // Smallest user key in external file
+  std::string smallest_user_key;
+  // Largest user key in external file
+  std::string largest_user_key;
+  // Sequence number for keys in external file
+  SequenceNumber original_seqno;
+  // Offset of the global sequence number field in the file, will
+  // be zero if version is 1 (global seqno is not supported)
+  size_t global_seqno_offset;
+  // External file size
+  uint64_t file_size;
+  // total number of keys in external file
+  uint64_t num_entries;
+  // Id of column family this file shoule be ingested into
+  uint32_t cf_id;
+  // TableProperties read from external file
+  TableProperties table_properties;
+  // Version of external file
+  int version;
+
+  // FileDescriptor for the file inside the DB
+  FileDescriptor fd;
+  // file path that we picked for file inside the DB
+  std::string internal_file_path = "";
+  // Global sequence number that we picked for the file inside the DB
+  SequenceNumber assigned_seqno = 0;
+  // Level inside the DB we picked for the external file.
+  int picked_level = 0;
+
+  InternalKey smallest_internal_key() const {
+    return InternalKey(smallest_user_key, assigned_seqno,
+                       ValueType::kTypeValue);
+  }
+
+  InternalKey largest_internal_key() const {
+    return InternalKey(largest_user_key, assigned_seqno, ValueType::kTypeValue);
+  }
+};
+
+class ExternalSstFileIngestionJob {
+ public:
+  ExternalSstFileIngestionJob(
+      Env* env, VersionSet* versions, ColumnFamilyData* cfd,
+      const ImmutableDBOptions& db_options, const EnvOptions& env_options,
+      SnapshotList* db_snapshots,
+      const IngestExternalFileOptions& ingestion_options)
+      : env_(env),
+        versions_(versions),
+        cfd_(cfd),
+        db_options_(db_options),
+        env_options_(env_options),
+        db_snapshots_(db_snapshots),
+        ingestion_options_(ingestion_options),
+        job_start_time_(env_->NowMicros()) {}
+
+  // Prepare the job by copying external files into the DB.
+  Status Prepare(const std::vector<std::string>& external_files_paths);
+
+  // Check if we need to flush the memtable before running the ingestion job
+  // This will be true if the files we are ingesting are overlapping with any
+  // key range in the memtable.
+  // REQUIRES: Mutex held
+  Status NeedsFlush(bool* flush_needed);
+
+  // Will execute the ingestion job and prepare edit() to be applied.
+  // REQUIRES: Mutex held
+  Status Run();
+
+  // Update column family stats.
+  // REQUIRES: Mutex held
+  void UpdateStats();
+
+  // Cleanup after successfull/failed job
+  void Cleanup(const Status& status);
+
+  VersionEdit* edit() { return &edit_; }
+
+  const autovector<IngestedFileInfo>& files_to_ingest() const {
+    return files_to_ingest_;
+  }
+
+ private:
+  // Open the external file and populate `file_to_ingest` with all the
+  // external information we need to ingest this file.
+  Status GetIngestedFileInfo(const std::string& external_file,
+                             IngestedFileInfo* file_to_ingest);
+
+  // Check if the files we are ingesting overlap with any memtable.
+  // REQUIRES: Mutex held
+  Status IngestedFilesOverlapWithMemtables(SuperVersion* sv, bool* overlap);
+
+  // Assign `file_to_ingest` the lowest possible level that it can
+  // be ingested to.
+  // REQUIRES: Mutex held
+  Status AssignLevelForIngestedFile(SuperVersion* sv,
+                                    IngestedFileInfo* file_to_ingest,
+                                    bool* overlap_with_db);
+
+  // Set the file global sequence number to `seqno`
+  Status AssignGlobalSeqnoForIngestedFile(IngestedFileInfo* file_to_ingest,
+                                          SequenceNumber seqno);
+
+  // Check if `file_to_ingest` key range overlap with the range `iter` represent
+  // REQUIRES: Mutex held
+  Status IngestedFileOverlapWithIteratorRange(
+      const IngestedFileInfo* file_to_ingest, InternalIterator* iter,
+      bool* overlap);
+
+  // Check if `file_to_ingest` can fit in level `level`
+  // REQUIRES: Mutex held
+  bool IngestedFileFitInLevel(const IngestedFileInfo* file_to_ingest,
+                              int level);
+
+  Env* env_;
+  VersionSet* versions_;
+  ColumnFamilyData* cfd_;
+  const ImmutableDBOptions& db_options_;
+  const EnvOptions& env_options_;
+  SnapshotList* db_snapshots_;
+  autovector<IngestedFileInfo> files_to_ingest_;
+  const IngestExternalFileOptions& ingestion_options_;
+  VersionEdit edit_;
+  uint64_t job_start_time_;
+};
+
+}  // namespace rocksdb
@@ -228,6 +228,11 @@ class FaultInjectionTest : public testing::Test,
    return Status::OK();
  }

+#if __clang_major__ > 3 || (__clang_major__ == 3 && __clang_minor__ >= 9)
+__attribute__((__no_sanitize__("undefined")))
+#elif __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 9)
+__attribute__((__no_sanitize_undefined__))
+#endif
  // Return the ith key
  Slice Key(int i, std::string* storage) const {
    int num = i;
@@ -9,8 +9,9 @@

 #include "db/file_indexer.h"
 #include <algorithm>
-#include "rocksdb/comparator.h"
+#include <functional>
 #include "db/version_edit.h"
+#include "rocksdb/comparator.h"

 namespace rocksdb {

@@ -389,12 +389,8 @@ Status SetIdentityFile(Env* env, const std::string& dbname) {
 Status SyncManifest(Env* env, const ImmutableDBOptions* db_options,
                    WritableFileWriter* file) {
  TEST_KILL_RANDOM("SyncManifest:0", rocksdb_kill_odds * REDUCE_ODDS2);
-  if (db_options->disable_data_sync) {
-    return Status::OK();
-  } else {
-    StopWatch sw(env, db_options->statistics.get(), MANIFEST_FILE_SYNC_MICROS);
-    return file->Sync(db_options->use_fsync);
-  }
+  StopWatch sw(env, db_options->statistics.get(), MANIFEST_FILE_SYNC_MICROS);
+  return file->Sync(db_options->use_fsync);
 }

 }  // namespace rocksdb
@@ -38,7 +38,7 @@
 #include "rocksdb/table.h"
 #include "table/block.h"
 #include "table/block_based_table_factory.h"
-#include "table/merger.h"
+#include "table/merging_iterator.h"
 #include "table/table_builder.h"
 #include "table/two_level_iterator.h"
 #include "util/coding.h"
@@ -230,6 +230,12 @@ Status FlushJob::Run(FileMetaData* file_meta) {
  return s;
 }

+void FlushJob::Cancel() {
+  db_mutex_->AssertHeld();
+  assert(base_ != nullptr);
+  base_->Unref();
+}
+
 Status FlushJob::WriteLevel0Table() {
  AutoThreadOperationStageUpdater stage_updater(
      ThreadStatus::STAGE_FLUSH_WRITE_L0);
@@ -241,7 +247,11 @@ Status FlushJob::WriteLevel0Table() {
    if (log_buffer_) {
      log_buffer_->FlushBufferToLog();
    }
+    // memtables and range_del_iters store internal iterators over each data
+    // memtable and its associated range deletion memtable, respectively, at
+    // corresponding indexes.
    std::vector<InternalIterator*> memtables;
+    std::vector<InternalIterator*> range_del_iters;
    ReadOptions ro;
    ro.total_order_seek = true;
    Arena arena;
@@ -252,6 +262,10 @@ Status FlushJob::WriteLevel0Table() {
          "[%s] [JOB %d] Flushing memtable with next log file: %" PRIu64 "\n",
          cfd_->GetName().c_str(), job_context_->job_id, m->GetNextLogNumber());
      memtables.push_back(m->NewIterator(ro, &arena));
+      auto* range_del_iter = m->NewRangeTombstoneIterator(ro);
+      if (range_del_iter != nullptr) {
+        range_del_iters.push_back(range_del_iter);
+      }
      total_num_entries += m->num_entries();
      total_num_deletes += m->num_deletes();
      total_memory_usage += m->ApproximateMemoryUsage();
@@ -268,6 +282,10 @@ Status FlushJob::WriteLevel0Table() {
      ScopedArenaIterator iter(
          NewMergingIterator(&cfd_->internal_comparator(), &memtables[0],
                             static_cast<int>(memtables.size()), &arena));
+      std::unique_ptr<InternalIterator> range_del_iter(NewMergingIterator(
+          &cfd_->internal_comparator(),
+          range_del_iters.empty() ? nullptr : &range_del_iters[0],
+          static_cast<int>(range_del_iters.size())));
      Log(InfoLogLevel::INFO_LEVEL, db_options_.info_log,
          "[%s] [JOB %d] Level-0 flush table #%" PRIu64 ": started",
          cfd_->GetName().c_str(), job_context_->job_id, meta_.fd.GetNumber());
@@ -276,9 +294,10 @@ Status FlushJob::WriteLevel0Table() {
                               &output_compression_);
      s = BuildTable(
          dbname_, db_options_.env, *cfd_->ioptions(), mutable_cf_options_,
-          env_options_, cfd_->table_cache(), iter.get(), &meta_,
-          cfd_->internal_comparator(), cfd_->int_tbl_prop_collector_factories(),
-          cfd_->GetID(), cfd_->GetName(), existing_snapshots_,
+          env_options_, cfd_->table_cache(), iter.get(),
+          std::move(range_del_iter), &meta_, cfd_->internal_comparator(),
+          cfd_->int_tbl_prop_collector_factories(), cfd_->GetID(),
+          cfd_->GetName(), existing_snapshots_,
          earliest_write_conflict_snapshot_, output_compression_,
          cfd_->ioptions()->compression_opts,
          mutable_cf_options_.paranoid_file_checks, cfd_->internal_stats(),
@@ -294,7 +313,7 @@ Status FlushJob::WriteLevel0Table() {
        meta_.fd.GetFileSize(), s.ToString().c_str(),
        meta_.marked_for_compaction ? " (needs compaction)" : "");

-    if (!db_options_.disable_data_sync && output_file_directory_ != nullptr) {
+    if (output_file_directory_ != nullptr) {
      output_file_directory_->Fsync();
    }
    TEST_SYNC_POINT("FlushJob::WriteLevel0Table");
@@ -67,9 +67,11 @@ class FlushJob {

  ~FlushJob();

-  // Require db_mutex held
+  // Require db_mutex held.
+  // Once PickMemTable() is called, either Run() or Cancel() has to be call.
  void PickMemTable();
  Status Run(FileMetaData* file_meta = nullptr);
+  void Cancel();
  TableProperties GetTableProperties() const { return table_properties_; }

 private: