remove warning as error.

This reverts commit 61720850f6.

DEFAULT_OPTIONS_HISTORY.md

@@ -1,4 +1,8 @@
 # RocksDB default options change log
+## 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.

HISTORY.md

@@ -1,4 +1,12 @@
 # Rocksdb Change Log
+## Unreleased
+### Bug fixes
+* Fix the data corruption bug in the case that concurrent memtable write is enabled and 2PC is used.
+* OptimizeForPointLookup() doesn't work with the default DB setting of allow_concurrent_memtable_write=true. Fix it.
+* Fix a 2PC related bug where WAL files size grow too large.
+* 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.

Makefile

@@ -220,10 +220,6 @@ default: all
 WARNING_FLAGS = -W -Wextra -Wall -Wsign-compare -Wshadow \
   -Wno-unused-parameter
 
-ifndef DISABLE_WARNING_AS_ERROR
-	WARNING_FLAGS += -Werror
-endif
-
 
 ifdef LUA_PATH
 

build_tools/build_detect_platform

@@ -51,12 +51,7 @@ if [ -z "$ROCKSDB_NO_FBCODE" -a -d /mnt/gvfs/third-party ]; then
     FBCODE_BUILD="true"
     # If we're compiling with TSAN we need pic build
     PIC_BUILD=$COMPILE_WITH_TSAN
-    if [ -z "$ROCKSDB_FBCODE_BUILD_WITH_481" ]; then
-      source "$PWD/build_tools/fbcode_config.sh"
-    else
-      # we need this to build with MySQL. Don't use for other purposes.
-      source "$PWD/build_tools/fbcode_config4.8.1.sh"
-    fi
+    source "$PWD/build_tools/fbcode_config.sh"
 fi
 
 # Delete existing output, if it exists

build_tools/dependencies.sh

@@ -1,18 +1,19 @@
-GCC_BASE=/mnt/gvfs/third-party2/gcc/cf7d14c625ce30bae1a4661c2319c5a283e4dd22/4.9.x/centos6-native/108cf83
-CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/8598c375b0e94e1448182eb3df034704144a838d/stable/centos6-native/3f16ddd
-LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/d6e0a7da6faba45f5e5b1638f9edd7afc2f34e7d/4.9.x/gcc-4.9-glibc-2.20/024dbc3
-GLIBC_BASE=/mnt/gvfs/third-party2/glibc/d282e6e8f3d20f4e40a516834847bdc038e07973/2.20/gcc-4.9-glibc-2.20/500e281
-SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/8c38a4c1e52b4c2cc8a9cdc31b9c947ed7dbfcb4/1.1.3/gcc-4.9-glibc-2.20/e9936bf
-ZLIB_BASE=/mnt/gvfs/third-party2/zlib/0882df3713c7a84f15abe368dc004581f20b39d7/1.2.8/gcc-5-glibc-2.23/9bc6787
-BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/740325875f6729f42d28deaa2147b0854f3a347e/1.0.6/gcc-5-glibc-2.23/9bc6787
-LZ4_BASE=/mnt/gvfs/third-party2/lz4/0e790b441e2d9acd68d51e1d2e028f88c6a79ddf/r131/gcc-5-glibc-2.23/9bc6787
-ZSTD_BASE=/mnt/gvfs/third-party2/zstd/9455f75ff7f4831dc9fda02a6a0f8c68922fad8f/1.0.0/gcc-5-glibc-2.23/9bc6787
-GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/f001a51b2854957676d07306ef3abf67186b5c8b/2.1.1/gcc-4.8.1-glibc-2.17/c3f970a
-JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/fc8a13ca1fffa4d0765c716c5a0b49f0c107518f/master/gcc-5-glibc-2.23/1c32b4b
-NUMA_BASE=/mnt/gvfs/third-party2/numa/17c514c4d102a25ca15f4558be564eeed76f4b6a/2.0.8/gcc-5-glibc-2.23/9bc6787
-LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/ad576de2a1ea560c4d3434304f0fc4e079bede42/trunk/gcc-5-glibc-2.23/b1847cb
-TBB_BASE=/mnt/gvfs/third-party2/tbb/9d9a554877d0c5bef330fe818ab7178806dd316a/4.0_update2/gcc-4.9-glibc-2.20/e9936bf
-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
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
+GCC_BASE=/mnt/gvfs/third-party2/gcc/7331085db891a2ef4a88a48a751d834e8d68f4cb/7.x/centos7-native/b2ef2b6
+CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/963d9aeda70cc4779885b1277484fe7544a04e3e/9.0.0/platform007/9e92d53/
+LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/6ace84e956873d53638c738b6f65f3f469cca74c/7.x/platform007/5620abc
+GLIBC_BASE=/mnt/gvfs/third-party2/glibc/192b0f42d63dcf6210d6ceae387b49af049e6e0c/2.26/platform007/f259413
+SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/7f9bdaada18f59bc27ec2b0871eb8a6144343aef/1.1.3/platform007/ca4da3d
+ZLIB_BASE=/mnt/gvfs/third-party2/zlib/2d9f0b9a4274cc21f61272a9e89bdb859bce8f1f/1.2.8/platform007/ca4da3d
+BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/dc49a21c5fceec6456a7a28a94dcd16690af1337/1.0.6/platform007/ca4da3d
+LZ4_BASE=/mnt/gvfs/third-party2/lz4/0f607f8fc442ea7d6b876931b1898bb573d5e5da/1.9.1/platform007/ca4da3d
+ZSTD_BASE=/mnt/gvfs/third-party2/zstd/ca22bc441a4eb709e9e0b1f9fec9750fed7b31c5/1.4.x/platform007/15a3614
+GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/0b9929d2588991c65a57168bf88aff2db87c5d48/2.2.0/platform007/ca4da3d
+JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/c26f08f47ac35fc31da2633b7da92d6b863246eb/master/platform007/c26c002
+NUMA_BASE=/mnt/gvfs/third-party2/numa/3f3fb57a5ccc5fd21c66416c0b83e0aa76a05376/2.0.11/platform007/ca4da3d
+LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/40c73d874898b386a71847f1b99115d93822d11f/1.4/platform007/6f3e0a9
+TBB_BASE=/mnt/gvfs/third-party2/tbb/4ce8e8dba77cdbd81b75d6f0c32fd7a1b76a11ec/2018_U5/platform007/ca4da3d
+KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/fb251ecd2f5ae16f8671f7014c246e52a748fe0b/fb/platform007/da39a3e
+BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/ab9f09bba370e7066cafd4eb59752db93f2e8312/2.29.1/platform007/15a3614
+VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/d42d152a15636529b0861ec493927200ebebca8e/3.15.0/platform007/ca4da3d
+LUA_BASE=/mnt/gvfs/third-party2/lua/f0cd714433206d5139df61659eb7b28b1dea6683/5.3.4/platform007/5007832

build_tools/fbcode_config.sh

@@ -13,7 +13,7 @@ source "$BASEDIR/dependencies.sh"
 CFLAGS=""
 
 # libgcc
-LIBGCC_INCLUDE="$LIBGCC_BASE/include"
+LIBGCC_INCLUDE="$LIBGCC_BASE/include/c++/7.3.0"
 LIBGCC_LIBS=" -L $LIBGCC_BASE/lib"
 
 # glibc
@@ -43,12 +43,16 @@ if test -z $PIC_BUILD; then
   LZ4_INCLUDE=" -I $LZ4_BASE/include/"
   LZ4_LIBS=" $LZ4_BASE/lib/liblz4.a"
   CFLAGS+=" -DLZ4"
-
-  ZSTD_INCLUDE=" -I $ZSTD_BASE/include/"
-  ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd.a"
-  CFLAGS+=" -DZSTD"
 fi
 
+ZSTD_INCLUDE=" -I $ZSTD_BASE/include/"
+if test -z $PIC_BUILD; then
+  ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd.a"
+else
+  ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd_pic.a"
+fi
+CFLAGS+=" -DZSTD"
+
 # location of gflags headers and libraries
 GFLAGS_INCLUDE=" -I $GFLAGS_BASE/include/"
 if test -z $PIC_BUILD; then
@@ -56,7 +60,7 @@ if test -z $PIC_BUILD; then
 else
   GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags_pic.a"
 fi
-CFLAGS+=" -DGFLAGS=google"
+CFLAGS+=" -DGFLAGS=gflags"
 
 # location of jemalloc
 JEMALLOC_INCLUDE=" -I $JEMALLOC_BASE/include/"
@@ -104,8 +108,8 @@ if [ -z "$USE_CLANG" ]; then
   CXX="$GCC_BASE/bin/g++"
 
   CFLAGS+=" -B$BINUTILS/gold"
-  CFLAGS+=" -isystem $GLIBC_INCLUDE"
   CFLAGS+=" -isystem $LIBGCC_INCLUDE"
+  CFLAGS+=" -isystem $GLIBC_INCLUDE"
   JEMALLOC=1
 else
   # clang
@@ -116,8 +120,8 @@ else
   KERNEL_HEADERS_INCLUDE="$KERNEL_HEADERS_BASE/include"
 
   CFLAGS+=" -B$BINUTILS/gold -nostdinc -nostdlib"
-  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/4.9.x "
-  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/4.9.x/x86_64-facebook-linux "
+  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/7.x "
+  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/7.x/x86_64-facebook-linux "
   CFLAGS+=" -isystem $GLIBC_INCLUDE"
   CFLAGS+=" -isystem $LIBGCC_INCLUDE"
   CFLAGS+=" -isystem $CLANG_INCLUDE"
@@ -128,13 +132,14 @@ else
 fi
 
 CFLAGS+=" $DEPS_INCLUDE"
-CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE"
+CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DROCKSDB_SUPPORT_THREAD_LOCAL -DHAVE_SSE42"
 CXXFLAGS+=" $CFLAGS"
 
 EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS"
-EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/gcc-4.9-glibc-2.20/lib/ld.so"
+EXEC_LDFLAGS+=" -B$BINUTILS/gold"
+EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/platform007/lib/ld.so"
 EXEC_LDFLAGS+=" $LIBUNWIND"
-EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/gcc-4.9-glibc-2.20/lib"
+EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/platform007/lib"
 # required by libtbb
 EXEC_LDFLAGS+=" -ldl"
 
@@ -144,12 +149,4 @@ EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GF
 
 VALGRIND_VER="$VALGRIND_BASE/bin/"
 
-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
+export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD 

db/column_family.cc

@@ -370,7 +370,8 @@ 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.
@@ -492,6 +493,25 @@ ColumnFamilyOptions ColumnFamilyData::GetLatestCFOptions() const {
   return BuildColumnFamilyOptions(initial_cf_options_, mutable_cf_options_);
 }
 
+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;

db/column_family.h

@@ -239,6 +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);
@@ -404,6 +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

db/compaction_job.cc

@@ -772,9 +772,9 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
                    key, sub_compact->current_output_file_size) &&
                sub_compact->builder != nullptr) {
       CompactionIterationStats range_del_out_stats;
-      status =
-          FinishCompactionOutputFile(input->status(), sub_compact,
-                                     range_del_agg.get(), &range_del_out_stats);
+      status = FinishCompactionOutputFile(input->status(), sub_compact,
+                                          range_del_agg.get(),
+                                          &range_del_out_stats, &key);
       RecordDroppedKeys(range_del_out_stats,
                         &sub_compact->compaction_job_stats);
       if (!status.ok()) {

db/db_impl.cc

@@ -338,6 +338,7 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname)
       last_stats_dump_time_microsec_(0),
       next_job_id_(1),
       has_unpersisted_data_(false),
+      unable_to_flush_oldest_log_(false),
       env_options_(BuildDBOptions(immutable_db_options_, mutable_db_options_)),
       num_running_ingest_file_(0),
 #ifndef ROCKSDB_LITE
@@ -654,6 +655,10 @@ void DBImpl::MaybeDumpStats() {
 }
 
 uint64_t DBImpl::FindMinPrepLogReferencedByMemTable() {
+  if (!allow_2pc()) {
+    return 0;
+  }
+
   uint64_t min_log = 0;
 
   // we must look through the memtables for two phase transactions
@@ -698,6 +703,11 @@ void DBImpl::MarkLogAsContainingPrepSection(uint64_t log) {
 }
 
 uint64_t DBImpl::FindMinLogContainingOutstandingPrep() {
+
+  if (!allow_2pc()) {
+    return 0;
+  }
+
   std::lock_guard<std::mutex> lock(prep_heap_mutex_);
   uint64_t min_log = 0;
 
@@ -736,6 +746,34 @@ void DBImpl::ScheduleBgLogWriterClose(JobContext* job_context) {
   }
 }
 
+uint64_t DBImpl::MinLogNumberToKeep() {
+  uint64_t log_number = versions_->MinLogNumber();
+
+  if (allow_2pc()) {
+    // if are 2pc we must consider logs containing prepared
+    // sections of outstanding transactions.
+    //
+    // We must check min logs with outstanding prep before we check
+    // logs referneces by memtables because a log referenced by the
+    // first data structure could transition to the second under us.
+    //
+    // TODO(horuff): iterating over all column families under db mutex.
+    // should find more optimial solution
+    auto min_log_in_prep_heap = FindMinLogContainingOutstandingPrep();
+
+    if (min_log_in_prep_heap != 0 && min_log_in_prep_heap < log_number) {
+      log_number = min_log_in_prep_heap;
+    }
+
+    auto min_log_refed_by_mem = FindMinPrepLogReferencedByMemTable();
+
+    if (min_log_refed_by_mem != 0 && min_log_refed_by_mem < log_number) {
+      log_number = min_log_refed_by_mem;
+    }
+  }
+  return log_number;
+}
+
 // * Returns the list of live files in 'sst_live'
 // If it's doing full scan:
 // * Returns the list of all files in the filesystem in
@@ -794,32 +832,7 @@ void DBImpl::FindObsoleteFiles(JobContext* job_context, bool force,
   job_context->manifest_file_number = versions_->manifest_file_number();
   job_context->pending_manifest_file_number =
       versions_->pending_manifest_file_number();
-  job_context->log_number = versions_->MinLogNumber();
-
-  if (allow_2pc()) {
-    // if are 2pc we must consider logs containing prepared
-    // sections of outstanding transactions.
-    //
-    // We must check min logs with outstanding prep before we check
-    // logs referneces by memtables because a log referenced by the
-    // first data structure could transition to the second under us.
-    //
-    // TODO(horuff): iterating over all column families under db mutex.
-    // should find more optimial solution
-    auto min_log_in_prep_heap = FindMinLogContainingOutstandingPrep();
-
-    if (min_log_in_prep_heap != 0 &&
-        min_log_in_prep_heap < job_context->log_number) {
-      job_context->log_number = min_log_in_prep_heap;
-    }
-
-    auto min_log_refed_by_mem = FindMinPrepLogReferencedByMemTable();
-
-    if (min_log_refed_by_mem != 0 &&
-        min_log_refed_by_mem < job_context->log_number) {
-      job_context->log_number = min_log_refed_by_mem;
-    }
-  }
+  job_context->log_number = MinLogNumberToKeep();
 
   job_context->prev_log_number = versions_->prev_log_number();
 
@@ -2493,7 +2506,7 @@ Status DBImpl::SetDBOptions(
       mutable_db_options_ = new_options;
 
       if (total_log_size_ > GetMaxTotalWalSize()) {
-        FlushColumnFamilies();
+        MaybeFlushColumnFamilies();
       }
 
       persist_options_status = PersistOptions();
@@ -4686,9 +4699,8 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
          versions_->GetColumnFamilySet()->NumberOfColumnFamilies() == 1);
 
   if (UNLIKELY(!single_column_family_mode_ &&
-               !alive_log_files_.begin()->getting_flushed &&
                total_log_size_ > GetMaxTotalWalSize())) {
-    FlushColumnFamilies();
+    MaybeFlushColumnFamilies();
   } else if (UNLIKELY(write_buffer_manager_->ShouldFlush())) {
     // Before a new memtable is added in SwitchMemtable(),
     // write_buffer_manager_->ShouldFlush() will keep returning true. If another
@@ -5006,28 +5018,40 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
   return status;
 }
 
-void DBImpl::FlushColumnFamilies() {
+void DBImpl::MaybeFlushColumnFamilies() {
   mutex_.AssertHeld();
 
-  WriteContext context;
-
   if (alive_log_files_.begin()->getting_flushed) {
     return;
   }
 
-  uint64_t flush_column_family_if_log_file = alive_log_files_.begin()->number;
-  alive_log_files_.begin()->getting_flushed = true;
+  auto oldest_alive_log = alive_log_files_.begin()->number;
+  auto oldest_log_with_uncommited_prep = FindMinLogContainingOutstandingPrep();
+
+  if (allow_2pc() &&
+      unable_to_flush_oldest_log_ &&
+      oldest_log_with_uncommited_prep > 0 &&
+      oldest_log_with_uncommited_prep <= oldest_alive_log) {
+    // we already attempted to flush all column families dependent on
+    // the oldest alive log but the log still contained uncommited transactions.
+    // the oldest alive log STILL contains uncommited transaction so there
+    // is still nothing that we can do.
+    return;
+  }
+
+  WriteContext context;
+
   Log(InfoLogLevel::INFO_LEVEL, immutable_db_options_.info_log,
       "Flushing all column families with data in WAL number %" PRIu64
       ". Total log size is %" PRIu64 " while max_total_wal_size is %" PRIu64,
-      flush_column_family_if_log_file, total_log_size_, GetMaxTotalWalSize());
+      oldest_alive_log, total_log_size_, GetMaxTotalWalSize());
   // no need to refcount because drop is happening in write thread, so can't
   // happen while we're in the write thread
   for (auto cfd : *versions_->GetColumnFamilySet()) {
     if (cfd->IsDropped()) {
       continue;
     }
-    if (cfd->GetLogNumber() <= flush_column_family_if_log_file) {
+    if (cfd->OldestLogToKeep() <= oldest_alive_log) {
       auto status = SwitchMemtable(cfd, &context);
       if (!status.ok()) {
         break;
@@ -5037,6 +5061,26 @@ void DBImpl::FlushColumnFamilies() {
     }
   }
   MaybeScheduleFlushOrCompaction();
+
+  // we only mark this log as getting flushed if we have successfully
+  // flushed all data in this log. If this log contains outstanding prepred
+  // transactions then we cannot flush this log until those transactions are commited.
+
+  unable_to_flush_oldest_log_ = false;
+
+  if (allow_2pc()) {
+    if (oldest_log_with_uncommited_prep == 0 ||
+        oldest_log_with_uncommited_prep > oldest_alive_log) {
+      // this log contains no outstanding prepared transactions
+      alive_log_files_.begin()->getting_flushed = true;
+    } else {
+      Log(InfoLogLevel::WARN_LEVEL, immutable_db_options_.info_log,
+          "Unable to release oldest log due to uncommited transaction");
+      unable_to_flush_oldest_log_ = true;
+    }
+  } else {
+    alive_log_files_.begin()->getting_flushed = true;
+  }
 }
 
 uint64_t DBImpl::GetMaxTotalWalSize() const {

db/db_impl.h

@@ -308,6 +308,16 @@ 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);
@@ -380,6 +390,8 @@ 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
@@ -732,7 +744,7 @@ class DBImpl : public DB {
   // REQUIRES: mutex locked
   Status PersistOptions();
 
-  void FlushColumnFamilies();
+  void MaybeFlushColumnFamilies();
 
   uint64_t GetMaxTotalWalSize() const;
 
@@ -991,6 +1003,15 @@ class DBImpl : public DB {
   // Used when disableWAL is true.
   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()
   static const uint64_t kNoTimeOut = port::kMaxUint64;

db/db_impl_debug.cc

@@ -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;

db/db_range_del_test.cc

@@ -112,6 +112,59 @@ TEST_F(DBRangeDelTest, CompactionOutputFilesExactlyFilled) {
   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, FlushRangeDelsSameStartKey) {
   db_->Put(WriteOptions(), "b1", "val");
   ASSERT_OK(

db/db_test2.cc

@@ -2212,6 +2212,18 @@ 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
 
 }  // namespace rocksdb

db/external_sst_file_ingestion_job.cc

@@ -96,10 +96,12 @@ Status ExternalSstFileIngestionJob::Prepare(
       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);
+        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);
+      status = CopyFile(env_, path_outside_db, path_inside_db, 0,
+                        db_options_.use_fsync);
     }
     TEST_SYNC_POINT("DBImpl::AddFile:FileCopied");
     if (!status.ok()) {

db/internal_stats.cc

@@ -221,6 +221,7 @@ static const std::string num_live_versions = "num-live-versions";
 static const std::string current_version_number =
     "current-super-version-number";
 static const std::string estimate_live_data_size = "estimate-live-data-size";
+static const std::string min_log_number_to_keep = "min-log-number-to-keep";
 static const std::string base_level = "base-level";
 static const std::string total_sst_files_size = "total-sst-files-size";
 static const std::string estimate_pending_comp_bytes =
@@ -285,6 +286,8 @@ const std::string DB::Properties::kCurrentSuperVersionNumber =
     rocksdb_prefix + current_version_number;
 const std::string DB::Properties::kEstimateLiveDataSize =
                       rocksdb_prefix + estimate_live_data_size;
+const std::string DB::Properties::kMinLogNumberToKeep =
+    rocksdb_prefix + min_log_number_to_keep;
 const std::string DB::Properties::kTotalSstFilesSize =
                       rocksdb_prefix + total_sst_files_size;
 const std::string DB::Properties::kBaseLevel = rocksdb_prefix + base_level;
@@ -368,6 +371,8 @@ const std::unordered_map<std::string, DBPropertyInfo>
           nullptr}},
         {DB::Properties::kEstimateLiveDataSize,
          {true, nullptr, &InternalStats::HandleEstimateLiveDataSize, nullptr}},
+        {DB::Properties::kMinLogNumberToKeep,
+         {false, nullptr, &InternalStats::HandleMinLogNumberToKeep, nullptr}},
         {DB::Properties::kBaseLevel,
          {false, nullptr, &InternalStats::HandleBaseLevel, nullptr}},
         {DB::Properties::kTotalSstFilesSize,
@@ -705,6 +710,12 @@ bool InternalStats::HandleEstimateLiveDataSize(uint64_t* value, DBImpl* db,
   return true;
 }
 
+bool InternalStats::HandleMinLogNumberToKeep(uint64_t* value, DBImpl* db,
+                                             Version* version) {
+  *value = db->MinLogNumberToKeep();
+  return true;
+}
+
 void InternalStats::DumpDBStats(std::string* value) {
   char buf[1000];
   // DB-level stats, only available from default column family

db/internal_stats.h

@@ -401,6 +401,7 @@ class InternalStats {
                                      Version* version);
   bool HandleEstimateLiveDataSize(uint64_t* value, DBImpl* db,
                                   Version* version);
+  bool HandleMinLogNumberToKeep(uint64_t* value, DBImpl* db, Version* version);
 
   // Total number of background errors encountered. Every time a flush task
   // or compaction task fails, this counter is incremented. The failure can

db/memtable_list.h

@@ -221,6 +221,8 @@ class MemTableList {
   // PickMemtablesToFlush() is called.
   void FlushRequested() { flush_requested_ = true; }
 
+  bool HasFlushRequested() { return flush_requested_; }
+
   // Copying allowed
   // MemTableList(const MemTableList&);
   // void operator=(const MemTableList&);

db/write_thread.cc

@@ -317,7 +317,7 @@ void WriteThread::LaunchParallelFollowers(ParallelGroup* pg,
 
   while (w != pg->last_writer) {
     // Writers that won't write don't get sequence allotment
-    if (!w->CallbackFailed()) {
+    if (!w->CallbackFailed() && w->ShouldWriteToMemtable()) {
       sequence += WriteBatchInternal::Count(w->batch);
     }
     w = w->link_newer;

include/rocksdb/db.h

@@ -500,6 +500,10 @@ class DB {
     //      live data in bytes.
     static const std::string kEstimateLiveDataSize;
 
+    //  "rocksdb.min-log-number-to-keep" - return the minmum log number of the
+    //      log files that should be kept.
+    static const std::string kMinLogNumberToKeep;
+
     //  "rocksdb.total-sst-files-size" - returns total size (bytes) of all SST
     //      files.
     //  WARNING: may slow down online queries if there are too many files.
@@ -565,6 +569,7 @@ class DB {
   //  "rocksdb.num-live-versions"
   //  "rocksdb.current-super-version-number"
   //  "rocksdb.estimate-live-data-size"
+  //  "rocksdb.min-log-number-to-keep"
   //  "rocksdb.total-sst-files-size"
   //  "rocksdb.base-level"
   //  "rocksdb.estimate-pending-compaction-bytes"

include/rocksdb/version.h

@@ -6,7 +6,7 @@
 
 #define ROCKSDB_MAJOR 5
 #define ROCKSDB_MINOR 0
-#define ROCKSDB_PATCH 1
+#define ROCKSDB_PATCH 2
 
 // Do not use these. We made the mistake of declaring macros starting with
 // double underscore. Now we have to live with our choice. We'll deprecate these

util/file_util.cc

@@ -16,7 +16,7 @@ namespace rocksdb {
 
 // Utility function to copy a file up to a specified length
 Status CopyFile(Env* env, const std::string& source,
-                const std::string& destination, uint64_t size) {
+                const std::string& destination, uint64_t size, bool use_fsync) {
   const EnvOptions soptions;
   Status s;
   unique_ptr<SequentialFileReader> src_reader;
@@ -62,6 +62,7 @@ Status CopyFile(Env* env, const std::string& source,
     }
     size -= slice.size();
   }
+  dest_writer->Sync(use_fsync);
   return Status::OK();
 }
 

util/file_util.h

@@ -12,9 +12,11 @@
 #include "util/db_options.h"
 
 namespace rocksdb {
-
+// use_fsync maps to options.use_fsync, which determines the way that
+// the file is synced after copying.
 extern Status CopyFile(Env* env, const std::string& source,
-                       const std::string& destination, uint64_t size = 0);
+                       const std::string& destination, uint64_t size,
+                       bool use_fsync);
 
 extern Status CreateFile(Env* env, const std::string& destination,
                          const std::string& contents);

util/options.cc

@@ -706,7 +706,7 @@ ColumnFamilyOptions* ColumnFamilyOptions::OptimizeForPointLookup(
   block_based_options.block_cache =
       NewLRUCache(static_cast<size_t>(block_cache_size_mb * 1024 * 1024));
   table_factory.reset(new BlockBasedTableFactory(block_based_options));
-  memtable_factory.reset(NewHashLinkListRepFactory());
+  memtable_prefix_bloom_size_ratio = 0.02;
   return this;
 }
 

util/sync_point.h

@@ -6,6 +6,7 @@
 
 #include <assert.h>
 #include <condition_variable>
+#include <functional>
 #include <mutex>
 #include <string>
 #include <thread>

util/thread_local.h

@@ -10,6 +10,7 @@
 #pragma once
 
 #include <atomic>
+#include <functional>
 #include <memory>
 #include <unordered_map>
 #include <vector>

utilities/checkpoint/checkpoint.cc

@@ -62,6 +62,8 @@ Status CheckpointImpl::CreateCheckpoint(const std::string& checkpoint_dir) {
   Status s;
   std::vector<std::string> live_files;
   uint64_t manifest_file_size = 0;
+  DBOptions db_options = db_->GetDBOptions();
+  uint64_t min_log_num = port::kMaxUint64;
   uint64_t sequence_number = db_->GetLatestSequenceNumber();
   bool same_fs = true;
   VectorLogPtr live_wal_files;
@@ -78,6 +80,35 @@ Status CheckpointImpl::CreateCheckpoint(const std::string& checkpoint_dir) {
   if (s.ok()) {
     // this will return live_files prefixed with "/"
     s = db_->GetLiveFiles(live_files, &manifest_file_size);
+
+    if (s.ok() && db_options.allow_2pc) {
+      // If 2PC is enabled, we need to get minimum log number after the flush.
+      // Need to refetch the live files to recapture the snapshot.
+      if (!db_->GetIntProperty(DB::Properties::kMinLogNumberToKeep,
+                               &min_log_num)) {
+        db_->EnableFileDeletions(false);
+        return Status::InvalidArgument(
+            "2PC enabled but cannot fine the min log number to keep.");
+      }
+      // We need to refetch live files with flush to handle this case:
+      // A previous 000001.log contains the prepare record of transaction tnx1.
+      // The current log file is 000002.log, and sequence_number points to this
+      // file.
+      // After calling GetLiveFiles(), 000003.log is created.
+      // Then tnx1 is committed. The commit record is written to 000003.log.
+      // Now we fetch min_log_num, which will be 3.
+      // Then only 000002.log and 000003.log will be copied, and 000001.log will
+      // be skipped. 000003.log contains commit message of tnx1, but we don't
+      // have respective prepare record for it.
+      // In order to avoid this situation, we need to force flush to make sure
+      // all transactions commited before getting min_log_num will be flushed
+      // to SST files.
+      // We cannot get min_log_num before calling the GetLiveFiles() for the
+      // first time, because if we do that, all the logs files will be included,
+      // far more than needed.
+      s = db_->GetLiveFiles(live_files, &manifest_file_size, /* flush */ true);
+    }
+
     TEST_SYNC_POINT("CheckpointImpl::CreateCheckpoint:SavedLiveFiles1");
     TEST_SYNC_POINT("CheckpointImpl::CreateCheckpoint:SavedLiveFiles2");
   }
@@ -91,7 +122,7 @@ Status CheckpointImpl::CreateCheckpoint(const std::string& checkpoint_dir) {
   }
 
   size_t wal_size = live_wal_files.size();
-  Log(db_->GetOptions().info_log,
+  Log(db_options.info_log,
       "Started the snapshot process -- creating snapshot in directory %s",
       checkpoint_dir.c_str());
 
@@ -130,7 +161,7 @@ Status CheckpointImpl::CreateCheckpoint(const std::string& checkpoint_dir) {
     // * if it's kDescriptorFile, limit the size to manifest_file_size
     // * always copy if cross-device link
     if ((type == kTableFile) && same_fs) {
-      Log(db_->GetOptions().info_log, "Hard Linking %s", src_fname.c_str());
+      Log(db_options.info_log, "Hard Linking %s", src_fname.c_str());
       s = db_->GetEnv()->LinkFile(db_->GetName() + src_fname,
                                   full_private_path + src_fname);
       if (s.IsNotSupported()) {
@@ -139,39 +170,41 @@ Status CheckpointImpl::CreateCheckpoint(const std::string& checkpoint_dir) {
       }
     }
     if ((type != kTableFile) || (!same_fs)) {
-      Log(db_->GetOptions().info_log, "Copying %s", src_fname.c_str());
+      Log(db_options.info_log, "Copying %s", src_fname.c_str());
       s = CopyFile(db_->GetEnv(), db_->GetName() + src_fname,
                    full_private_path + src_fname,
-                   (type == kDescriptorFile) ? manifest_file_size : 0);
+                   (type == kDescriptorFile) ? manifest_file_size : 0,
+                   db_options.use_fsync);
     }
   }
   if (s.ok() && !current_fname.empty() && !manifest_fname.empty()) {
     s = CreateFile(db_->GetEnv(), full_private_path + current_fname,
                    manifest_fname.substr(1) + "\n");
   }
-  Log(db_->GetOptions().info_log, "Number of log files %" ROCKSDB_PRIszt,
+  Log(db_options.info_log, "Number of log files %" ROCKSDB_PRIszt,
       live_wal_files.size());
 
   // Link WAL files. Copy exact size of last one because it is the only one
   // that has changes after the last flush.
   for (size_t i = 0; s.ok() && i < wal_size; ++i) {
     if ((live_wal_files[i]->Type() == kAliveLogFile) &&
-        (live_wal_files[i]->StartSequence() >= sequence_number)) {
+        (live_wal_files[i]->StartSequence() >= sequence_number ||
+         live_wal_files[i]->LogNumber() >= min_log_num)) {
       if (i + 1 == wal_size) {
-        Log(db_->GetOptions().info_log, "Copying %s",
+        Log(db_options.info_log, "Copying %s",
             live_wal_files[i]->PathName().c_str());
         s = CopyFile(db_->GetEnv(),
-                     db_->GetOptions().wal_dir + live_wal_files[i]->PathName(),
+                     db_options.wal_dir + live_wal_files[i]->PathName(),
                      full_private_path + live_wal_files[i]->PathName(),
-                     live_wal_files[i]->SizeFileBytes());
+                     live_wal_files[i]->SizeFileBytes(), db_options.use_fsync);
         break;
       }
       if (same_fs) {
         // we only care about live log files
-        Log(db_->GetOptions().info_log, "Hard Linking %s",
+        Log(db_options.info_log, "Hard Linking %s",
             live_wal_files[i]->PathName().c_str());
         s = db_->GetEnv()->LinkFile(
-            db_->GetOptions().wal_dir + live_wal_files[i]->PathName(),
+            db_options.wal_dir + live_wal_files[i]->PathName(),
             full_private_path + live_wal_files[i]->PathName());
         if (s.IsNotSupported()) {
           same_fs = false;
@@ -179,11 +212,12 @@ Status CheckpointImpl::CreateCheckpoint(const std::string& checkpoint_dir) {
         }
       }
       if (!same_fs) {
-        Log(db_->GetOptions().info_log, "Copying %s",
+        Log(db_options.info_log, "Copying %s",
             live_wal_files[i]->PathName().c_str());
         s = CopyFile(db_->GetEnv(),
-                     db_->GetOptions().wal_dir + live_wal_files[i]->PathName(),
-                     full_private_path + live_wal_files[i]->PathName(), 0);
+                     db_options.wal_dir + live_wal_files[i]->PathName(),
+                     full_private_path + live_wal_files[i]->PathName(), 0,
+                     db_options.use_fsync);
       }
     }
   }
@@ -205,27 +239,26 @@ Status CheckpointImpl::CreateCheckpoint(const std::string& checkpoint_dir) {
 
   if (!s.ok()) {
     // clean all the files we might have created
-    Log(db_->GetOptions().info_log, "Snapshot failed -- %s",
-        s.ToString().c_str());
+    Log(db_options.info_log, "Snapshot failed -- %s", s.ToString().c_str());
     // we have to delete the dir and all its children
     std::vector<std::string> subchildren;
     db_->GetEnv()->GetChildren(full_private_path, &subchildren);
     for (auto& subchild : subchildren) {
       std::string subchild_path = full_private_path + "/" + subchild;
       Status s1 = db_->GetEnv()->DeleteFile(subchild_path);
-      Log(db_->GetOptions().info_log, "Delete file %s -- %s",
-          subchild_path.c_str(), s1.ToString().c_str());
+      Log(db_options.info_log, "Delete file %s -- %s", subchild_path.c_str(),
+          s1.ToString().c_str());
     }
     // finally delete the private dir
     Status s1 = db_->GetEnv()->DeleteDir(full_private_path);
-    Log(db_->GetOptions().info_log, "Delete dir %s -- %s",
-        full_private_path.c_str(), s1.ToString().c_str());
+    Log(db_options.info_log, "Delete dir %s -- %s", full_private_path.c_str(),
+        s1.ToString().c_str());
     return s;
   }
 
   // here we know that we succeeded and installed the new snapshot
-  Log(db_->GetOptions().info_log, "Snapshot DONE. All is good");
-  Log(db_->GetOptions().info_log, "Snapshot sequence number: %" PRIu64,
+  Log(db_options.info_log, "Snapshot DONE. All is good");
+  Log(db_options.info_log, "Snapshot sequence number: %" PRIu64,
       sequence_number);
 
   return s;

utilities/checkpoint/checkpoint_test.cc

@@ -21,6 +21,7 @@
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
 #include "rocksdb/utilities/checkpoint.h"
+#include "rocksdb/utilities/transaction_db.h"
 #include "util/sync_point.h"
 #include "util/testharness.h"
 #include "util/xfunc.h"
@@ -390,6 +391,120 @@ TEST_F(DBTest, CurrentFileModifiedWhileCheckpointing) {
   snapshotDB = nullptr;
 }
 
+TEST_F(DBTest, CurrentFileModifiedWhileCheckpointing2PC) {
+  const std::string kSnapshotName = test::TmpDir(env_) + "/snapshot";
+  const std::string dbname = test::TmpDir() + "/transaction_testdb";
+  ASSERT_OK(DestroyDB(kSnapshotName, CurrentOptions()));
+  ASSERT_OK(DestroyDB(dbname, CurrentOptions()));
+  env_->DeleteDir(kSnapshotName);
+  env_->DeleteDir(dbname);
+  Close();
+
+  Options options = CurrentOptions();
+  // allow_2pc is implicitly set with tx prepare
+  // options.allow_2pc = true;
+  TransactionDBOptions txn_db_options;
+  TransactionDB* txdb;
+  Status s = TransactionDB::Open(options, txn_db_options, dbname, &txdb);
+  assert(s.ok());
+  ColumnFamilyHandle* cfa;
+  ColumnFamilyHandle* cfb;
+  ColumnFamilyOptions cf_options;
+  ASSERT_OK(txdb->CreateColumnFamily(cf_options, "CFA", &cfa));
+
+  WriteOptions write_options;
+  // Insert something into CFB so lots of log files will be kept
+  // before creating the checkpoint.
+  ASSERT_OK(txdb->CreateColumnFamily(cf_options, "CFB", &cfb));
+  ASSERT_OK(txdb->Put(write_options, cfb, "", ""));
+
+  ReadOptions read_options;
+  std::string value;
+  TransactionOptions txn_options;
+  Transaction* txn = txdb->BeginTransaction(write_options, txn_options);
+  s = txn->SetName("xid");
+  ASSERT_OK(s);
+  ASSERT_EQ(txdb->GetTransactionByName("xid"), txn);
+
+  s = txn->Put(Slice("foo"), Slice("bar"));
+  s = txn->Put(cfa, Slice("foocfa"), Slice("barcfa"));
+  ASSERT_OK(s);
+  // Writing prepare into middle of first WAL, then flush WALs many times
+  for (int i = 1; i <= 100000; i++) {
+    Transaction* tx = txdb->BeginTransaction(write_options, txn_options);
+    ASSERT_OK(tx->SetName("x"));
+    ASSERT_OK(tx->Put(Slice(std::to_string(i)), Slice("val")));
+    ASSERT_OK(tx->Put(cfa, Slice("aaa"), Slice("111")));
+    ASSERT_OK(tx->Prepare());
+    ASSERT_OK(tx->Commit());
+    if (i % 10000 == 0) {
+      txdb->Flush(FlushOptions());
+    }
+    if (i == 88888) {
+      ASSERT_OK(txn->Prepare());
+    }
+  }
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      {{"CheckpointImpl::CreateCheckpoint:SavedLiveFiles1",
+        "DBTest::CurrentFileModifiedWhileCheckpointing2PC:PreCommit"},
+       {"DBTest::CurrentFileModifiedWhileCheckpointing2PC:PostCommit",
+        "CheckpointImpl::CreateCheckpoint:SavedLiveFiles2"}});
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+  std::thread t([&]() {
+    Checkpoint* checkpoint;
+    ASSERT_OK(Checkpoint::Create(txdb, &checkpoint));
+    ASSERT_OK(checkpoint->CreateCheckpoint(kSnapshotName));
+    delete checkpoint;
+  });
+  TEST_SYNC_POINT("DBTest::CurrentFileModifiedWhileCheckpointing2PC:PreCommit");
+  ASSERT_OK(txn->Commit());
+  TEST_SYNC_POINT(
+      "DBTest::CurrentFileModifiedWhileCheckpointing2PC:PostCommit");
+  t.join();
+
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+
+  // No more than two logs files should exist.
+  std::vector<std::string> files;
+  env_->GetChildren(kSnapshotName, &files);
+  int num_log_files = 0;
+  for (auto& file : files) {
+    uint64_t num;
+    FileType type;
+    WalFileType log_type;
+    if (ParseFileName(file, &num, &type, &log_type) && type == kLogFile) {
+      num_log_files++;
+    }
+  }
+  // One flush after preapare + one outstanding file before checkpoint + one log
+  // file generated after checkpoint.
+  ASSERT_LE(num_log_files, 3);
+
+  TransactionDB* snapshotDB;
+  std::vector<ColumnFamilyDescriptor> column_families;
+  column_families.push_back(
+      ColumnFamilyDescriptor(kDefaultColumnFamilyName, ColumnFamilyOptions()));
+  column_families.push_back(
+      ColumnFamilyDescriptor("CFA", ColumnFamilyOptions()));
+  column_families.push_back(
+      ColumnFamilyDescriptor("CFB", ColumnFamilyOptions()));
+  std::vector<rocksdb::ColumnFamilyHandle*> cf_handles;
+  ASSERT_OK(TransactionDB::Open(options, txn_db_options, kSnapshotName,
+                                column_families, &cf_handles, &snapshotDB));
+  ASSERT_OK(snapshotDB->Get(read_options, "foo", &value));
+  ASSERT_EQ(value, "bar");
+  ASSERT_OK(snapshotDB->Get(read_options, cf_handles[1], "foocfa", &value));
+  ASSERT_EQ(value, "barcfa");
+
+  delete cfa;
+  delete cfb;
+  delete cf_handles[0];
+  delete cf_handles[1];
+  delete cf_handles[2];
+  delete snapshotDB;
+  snapshotDB = nullptr;
+}
+
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

utilities/persistent_cache/block_cache_tier_file.cc

@@ -9,6 +9,7 @@
 #ifndef OS_WIN
 #include <unistd.h>
 #endif
+#include <functional>
 #include <memory>
 #include <vector>
 

utilities/persistent_cache/block_cache_tier_file.h

@@ -6,6 +6,7 @@
 
 #ifndef ROCKSDB_LITE
 
+#include <functional>
 #include <list>
 #include <memory>
 #include <string>

utilities/persistent_cache/hash_table_evictable.h

@@ -7,6 +7,7 @@
 
 #ifndef ROCKSDB_LITE
 
+#include <functional>
 #include "util/random.h"
 #include "utilities/persistent_cache/hash_table.h"
 #include "utilities/persistent_cache/lrulist.h"

utilities/transactions/transaction_test.cc

@@ -1178,6 +1178,108 @@ TEST_P(TransactionTest, TwoPhaseLogRollingTest) {
   delete cfb;
 }
 
+TEST_P(TransactionTest, TwoPhaseLogRollingTest2) {
+  DBImpl* db_impl = reinterpret_cast<DBImpl*>(db->GetRootDB());
+
+  Status s;
+  ColumnFamilyHandle *cfa, *cfb;
+
+  ColumnFamilyOptions cf_options;
+  s = db->CreateColumnFamily(cf_options, "CFA", &cfa);
+  ASSERT_OK(s);
+  s = db->CreateColumnFamily(cf_options, "CFB", &cfb);
+  ASSERT_OK(s);
+
+  WriteOptions wopts;
+  wopts.disableWAL = false;
+  wopts.sync = true;
+
+  auto cfh_a = reinterpret_cast<ColumnFamilyHandleImpl*>(cfa);
+  auto cfh_b = reinterpret_cast<ColumnFamilyHandleImpl*>(cfb);
+
+  TransactionOptions topts1;
+  Transaction* txn1 = db->BeginTransaction(wopts, topts1);
+  s = txn1->SetName("xid1");
+  ASSERT_OK(s);
+  s = txn1->Put(cfa, "boys", "girls1");
+  ASSERT_OK(s);
+
+  Transaction* txn2 = db->BeginTransaction(wopts, topts1);
+  s = txn2->SetName("xid2");
+  ASSERT_OK(s);
+  s = txn2->Put(cfb, "up", "down1");
+  ASSERT_OK(s);
+
+  // prepre transaction in LOG A
+  s = txn1->Prepare();
+  ASSERT_OK(s);
+
+  // prepre transaction in LOG A
+  s = txn2->Prepare();
+  ASSERT_OK(s);
+
+  // regular put so that mem table can actually be flushed for log rolling
+  s = db->Put(wopts, "cats", "dogs1");
+  ASSERT_OK(s);
+
+  auto prepare_log_no = txn1->GetLogNumber();
+
+  // roll to LOG B
+  s = db_impl->TEST_FlushMemTable(true);
+  ASSERT_OK(s);
+
+  // now we pause background work so that
+  // imm()s are not flushed before we can check their status
+  s = db_impl->PauseBackgroundWork();
+  ASSERT_OK(s);
+
+  ASSERT_GT(db_impl->TEST_LogfileNumber(), prepare_log_no);
+  ASSERT_GT(cfh_a->cfd()->GetLogNumber(), prepare_log_no);
+  ASSERT_EQ(cfh_a->cfd()->GetLogNumber(), db_impl->TEST_LogfileNumber());
+  ASSERT_EQ(db_impl->TEST_FindMinLogContainingOutstandingPrep(),
+            prepare_log_no);
+  ASSERT_EQ(db_impl->TEST_FindMinPrepLogReferencedByMemTable(), 0);
+
+  // commit in LOG B
+  s = txn1->Commit();
+  ASSERT_OK(s);
+
+  ASSERT_EQ(db_impl->TEST_FindMinPrepLogReferencedByMemTable(), prepare_log_no);
+
+  ASSERT_TRUE(!db_impl->TEST_UnableToFlushOldestLog());
+
+  // request a flush for all column families such that the earliest
+  // alive log file can be killed
+  db_impl->TEST_MaybeFlushColumnFamilies();
+  // log cannot be flushed because txn2 has not been commited
+  ASSERT_TRUE(!db_impl->TEST_IsLogGettingFlushed());
+  ASSERT_TRUE(db_impl->TEST_UnableToFlushOldestLog());
+
+  // assert that cfa has a flush requested
+  ASSERT_TRUE(cfh_a->cfd()->imm()->HasFlushRequested());
+
+  // cfb should not be flushed becuse it has no data from LOG A
+  ASSERT_TRUE(!cfh_b->cfd()->imm()->HasFlushRequested());
+
+  // cfb now has data from LOG A
+  s = txn2->Commit();
+  ASSERT_OK(s);
+
+  db_impl->TEST_MaybeFlushColumnFamilies();
+  ASSERT_TRUE(!db_impl->TEST_UnableToFlushOldestLog());
+
+  // we should see that cfb now has a flush requested
+  ASSERT_TRUE(cfh_b->cfd()->imm()->HasFlushRequested());
+
+  // all data in LOG A resides in a memtable that has been
+  // requested for a flush
+  ASSERT_TRUE(db_impl->TEST_IsLogGettingFlushed());
+
+  delete txn1;
+  delete txn2;
+  delete cfa;
+  delete cfb;
+}
 /*
  * 1) use prepare to keep first log around to determine starting sequence
  * during recovery.