Disable error as warning

The previuos change that use gcc-5 set USE_SSE to wrong flag by mistake. Fix it.

CMakeLists.txt

@@ -252,11 +252,11 @@ 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
@@ -391,7 +391,6 @@ set(SOURCES
         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/memory/memory_util.cc

HISTORY.md

@@ -1,4 +1,14 @@
 # Rocksdb Change Log
+## 4.13.5
+### Public API Change
+* Fix a regression in compaction performance.
+* Disallow calling IngestExternalFile() on a dropped column family.
+* Add EventListener::OnExternalFileIngested() event that will be called for files that are successfully ingested.
+
+## 4.13.4
+### Public API Change
+* Removed flashcache support.
+
 ## 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.

Makefile

@@ -216,10 +216,6 @@ default: all
 WARNING_FLAGS = -W -Wextra -Wall -Wsign-compare -Wshadow \
   -Wno-unused-parameter
 
-ifndef DISABLE_WARNING_AS_ERROR
-	WARNING_FLAGS += -Werror
-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
 

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

@@ -1,17 +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
+# 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"
 

db/column_family.cc

@@ -160,6 +160,10 @@ 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;
   }
@@ -545,14 +549,31 @@ 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
 

db/column_family_test.cc

@@ -941,6 +941,12 @@ 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) {

db/compaction_job.cc

@@ -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;
@@ -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);
 
@@ -846,9 +849,6 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
       }
     }
 
-    Status input_status = input->status();
-    c_iter->Next();
-
     // Close output file if it is big enough
     // TODO(aekmekji): determine if file should be closed earlier than this
     // during subcompactions (i.e. if output size, estimated by input size, is
@@ -857,6 +857,9 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
     if (sub_compact->compaction->output_level() != 0 &&
         sub_compact->current_output_file_size >=
             sub_compact->compaction->max_output_file_size()) {
+      Status input_status = input->status();
+      c_iter->Next();
+
       const Slice* next_key = nullptr;
       if (c_iter->Valid()) {
         next_key = &c_iter->key();
@@ -868,6 +871,8 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
         // files.
         sub_compact->compression_dict = std::move(compression_dict);
       }
+    } else {
+      c_iter->Next();
     }
   }
 

db/db_impl.cc

@@ -39,6 +39,7 @@
 #include "db/db_iter.h"
 #include "db/dbformat.h"
 #include "db/event_helpers.h"
+#include "db/external_sst_file_ingestion_job.h"
 #include "db/filename.h"
 #include "db/flush_job.h"
 #include "db/forward_iterator.h"
@@ -346,7 +347,7 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname)
       next_job_id_(1),
       has_unpersisted_data_(false),
       env_options_(BuildDBOptions(immutable_db_options_, mutable_db_options_)),
-      num_running_addfile_(0),
+      num_running_ingest_file_(0),
 #ifndef ROCKSDB_LITE
       wal_manager_(immutable_db_options_, env_options_),
 #endif  // ROCKSDB_LITE
@@ -1347,7 +1348,6 @@ Status DBImpl::Recover(
         }
       }
     }
-    SetTickerCount(stats_, SEQUENCE_NUMBER, versions_->LastSequence());
   }
 
   // Initial value
@@ -1609,10 +1609,16 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
       // we just ignore the update.
       // That's why we set ignore missing column families to true
       bool has_valid_writes = false;
+      // If we pass DB through and options.max_successive_merges is hit
+      // during recovery, Get() will be issued which will try to acquire
+      // DB mutex and cause deadlock, as DB mutex is already held.
+      // The DB pointer is not needed unless 2PC is used.
+      // TODO(sdong) fix the allow_2pc case too.
       status = WriteBatchInternal::InsertInto(
           &batch, column_family_memtables_.get(), &flush_scheduler_, true,
-          log_number, this, false /* concurrent_memtable_writes */,
-          next_sequence, &has_valid_writes);
+          log_number, immutable_db_options_.allow_2pc ? this : nullptr,
+          false /* concurrent_memtable_writes */, next_sequence,
+          &has_valid_writes);
       // If it is the first log file and there is no column family updated
       // after replaying the file, this file may be a stale file. We ignore
       // sequence IDs from the file. Otherwise, if a newer stale log file that
@@ -1687,6 +1693,9 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
     }
   }
 
+  // True if there's any data in the WALs; if not, we can skip re-processing
+  // them later
+  bool data_seen = false;
   if (!read_only) {
     // no need to refcount since client still doesn't have access
     // to the DB and can not drop column families while we iterate
@@ -1722,6 +1731,7 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
           cfd->CreateNewMemtable(*cfd->GetLatestMutableCFOptions(),
                                  *next_sequence);
         }
+        data_seen = true;
       }
 
       // write MANIFEST with update
@@ -1747,7 +1757,7 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
     }
   }
 
-  if (!flushed) {
+  if (data_seen && !flushed) {
     // Mark these as alive so they'll be considered for deletion later by
     // FindObsoleteFiles()
     for (auto log_number : log_numbers) {
@@ -2143,8 +2153,8 @@ Status DBImpl::CompactFiles(
     InstrumentedMutexLock l(&mutex_);
 
     // This call will unlock/lock the mutex to wait for current running
-    // AddFile() calls to finish.
-    WaitForAddFile();
+    // IngestExternalFile() calls to finish.
+    WaitForIngestFile();
 
     s = CompactFilesImpl(compact_options, cfd, sv->current,
                          input_file_names, output_level,
@@ -2899,7 +2909,8 @@ InternalIterator* DBImpl::NewInternalIterator(
 }
 
 Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
-                             const FlushOptions& flush_options) {
+                             const FlushOptions& flush_options,
+                             bool writes_stopped) {
   Status s;
   {
     WriteContext context;
@@ -2911,12 +2922,17 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
     }
 
     WriteThread::Writer w;
-    write_thread_.EnterUnbatched(&w, &mutex_);
+    if (!writes_stopped) {
+      write_thread_.EnterUnbatched(&w, &mutex_);
+    }
 
     // SwitchMemtable() will release and reacquire mutex
     // during execution
     s = SwitchMemtable(cfd, &context);
-    write_thread_.ExitUnbatched(&w);
+
+    if (!writes_stopped) {
+      write_thread_.ExitUnbatched(&w);
+    }
 
     cfd->imm()->FlushRequested();
 
@@ -2940,6 +2956,12 @@ Status DBImpl::WaitForFlushMemTable(ColumnFamilyData* cfd) {
     if (shutting_down_.load(std::memory_order_acquire)) {
       return Status::ShutdownInProgress();
     }
+    if (cfd->IsDropped()) {
+      // FlushJob cannot flush a dropped CF, if we did not break here
+      // we will loop forever since cfd->imm()->NumNotFlushed() will never
+      // drop to zero
+      return Status::InvalidArgument("Cannot flush a dropped CF");
+    }
     bg_cv_.Wait();
   }
   if (!bg_error_.ok()) {
@@ -3295,8 +3317,8 @@ void DBImpl::BackgroundCallCompaction(void* arg) {
     InstrumentedMutexLock l(&mutex_);
 
     // This call will unlock/lock the mutex to wait for current running
-    // AddFile() calls to finish.
-    WaitForAddFile();
+    // IngestExternalFile() calls to finish.
+    WaitForIngestFile();
 
     num_running_compactions_++;
 
@@ -3704,8 +3726,8 @@ void DBImpl::RemoveManualCompaction(DBImpl::ManualCompaction* m) {
 }
 
 bool DBImpl::ShouldntRunManualCompaction(ManualCompaction* m) {
-  if (num_running_addfile_ > 0) {
-    // We need to wait for other AddFile() calls to finish
+  if (num_running_ingest_file_ > 0) {
+    // We need to wait for other IngestExternalFile() calls to finish
     // before running a manual compaction.
     return true;
   }
@@ -3850,7 +3872,10 @@ InternalIterator* DBImpl::NewInternalIterator(const ReadOptions& read_options,
   InternalIterator* internal_iter;
   assert(arena != nullptr);
   // Need to create internal iterator from the arena.
-  MergeIteratorBuilder merge_iter_builder(&cfd->internal_comparator(), arena);
+  MergeIteratorBuilder merge_iter_builder(
+      &cfd->internal_comparator(), arena,
+      !read_options.total_order_seek &&
+          cfd->ioptions()->prefix_extractor != nullptr);
   // Collect iterator for mutable mem
   merge_iter_builder.AddIterator(
       super_version->mem->NewIterator(read_options, arena));
@@ -4610,7 +4635,6 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
     if (write_thread_.CompleteParallelWorker(&w)) {
       // we're responsible for early exit
       auto last_sequence = w.parallel_group->last_sequence;
-      SetTickerCount(stats_, SEQUENCE_NUMBER, last_sequence);
       versions_->SetLastSequence(last_sequence);
       write_thread_.EarlyExitParallelGroup(&w);
     }
@@ -4946,7 +4970,6 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
       }
 
       if (!exit_completed_early && w.status.ok()) {
-        SetTickerCount(stats_, SEQUENCE_NUMBER, last_sequence);
         versions_->SetLastSequence(last_sequence);
         if (!need_log_sync) {
           write_thread_.ExitAsBatchGroupLeader(&w, last_writer, w.status);
@@ -6364,6 +6387,136 @@ Status DBImpl::GetLatestSequenceForKey(SuperVersion* sv, const Slice& key,
 
   return Status::OK();
 }
+
+Status DBImpl::IngestExternalFile(
+    ColumnFamilyHandle* column_family,
+    const std::vector<std::string>& external_files,
+    const IngestExternalFileOptions& ingestion_options) {
+  Status status;
+  auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
+  auto cfd = cfh->cfd();
+
+  ExternalSstFileIngestionJob ingestion_job(env_, versions_.get(), cfd,
+                                            immutable_db_options_, env_options_,
+                                            &snapshots_, ingestion_options);
+
+  // Make sure that bg cleanup wont delete the files that we are ingesting
+  std::list<uint64_t>::iterator pending_output_elem;
+  {
+    InstrumentedMutexLock l(&mutex_);
+    pending_output_elem = CaptureCurrentFileNumberInPendingOutputs();
+  }
+
+  status = ingestion_job.Prepare(external_files);
+  if (!status.ok()) {
+    return status;
+  }
+
+  TEST_SYNC_POINT("DBImpl::AddFile:Start");
+  {
+    // Lock db mutex
+    InstrumentedMutexLock l(&mutex_);
+    TEST_SYNC_POINT("DBImpl::AddFile:MutexLock");
+
+    // Stop writes to the DB
+    WriteThread::Writer w;
+    write_thread_.EnterUnbatched(&w, &mutex_);
+
+    num_running_ingest_file_++;
+
+    // We cannot ingest a file into a dropped CF
+    if (cfd->IsDropped()) {
+      status = Status::InvalidArgument(
+          "Cannot ingest an external file into a dropped CF");
+    }
+
+    // Figure out if we need to flush the memtable first
+    if (status.ok()) {
+      bool need_flush = false;
+      status = ingestion_job.NeedsFlush(&need_flush);
+
+      if (status.ok() && need_flush) {
+        mutex_.Unlock();
+        status = FlushMemTable(cfd, FlushOptions(), true /* writes_stopped */);
+        mutex_.Lock();
+      }
+    }
+
+    // Run the ingestion job
+    if (status.ok()) {
+      status = ingestion_job.Run();
+    }
+
+    // Install job edit [Mutex will be unlocked here]
+    auto mutable_cf_options = cfd->GetLatestMutableCFOptions();
+    if (status.ok()) {
+      status =
+          versions_->LogAndApply(cfd, *mutable_cf_options, ingestion_job.edit(),
+                                 &mutex_, directories_.GetDbDir());
+    }
+    if (status.ok()) {
+      delete InstallSuperVersionAndScheduleWork(cfd, nullptr,
+                                                *mutable_cf_options);
+    }
+
+    // Resume writes to the DB
+    write_thread_.ExitUnbatched(&w);
+
+    // Update stats
+    if (status.ok()) {
+      ingestion_job.UpdateStats();
+    }
+
+    ReleaseFileNumberFromPendingOutputs(pending_output_elem);
+
+    num_running_ingest_file_--;
+    if (num_running_ingest_file_ == 0) {
+      bg_cv_.SignalAll();
+    }
+
+    TEST_SYNC_POINT("DBImpl::AddFile:MutexUnlock");
+  }
+  // mutex_ is unlocked here
+
+  // Cleanup
+  ingestion_job.Cleanup(status);
+
+  if (status.ok()) {
+    NotifyOnExternalFileIngested(cfd, ingestion_job);
+  }
+
+  return status;
+}
+
+void DBImpl::NotifyOnExternalFileIngested(
+    ColumnFamilyData* cfd, const ExternalSstFileIngestionJob& ingestion_job) {
+#ifndef ROCKSDB_LITE
+  if (immutable_db_options_.listeners.empty()) {
+    return;
+  }
+
+  for (const IngestedFileInfo& f : ingestion_job.files_to_ingest()) {
+    ExternalFileIngestionInfo info;
+    info.cf_name = cfd->GetName();
+    info.external_file_path = f.external_file_path;
+    info.internal_file_path = f.internal_file_path;
+    info.global_seqno = f.assigned_seqno;
+    info.table_properties = f.table_properties;
+    for (auto listener : immutable_db_options_.listeners) {
+      listener->OnExternalFileIngested(this, info);
+    }
+  }
+
+#endif
+}
+
+void DBImpl::WaitForIngestFile() {
+  mutex_.AssertHeld();
+  while (num_running_ingest_file_ > 0) {
+    bg_cv_.Wait();
+  }
+}
+
 #endif  // ROCKSDB_LITE
 
 }  // namespace rocksdb

db/db_impl.h

@@ -22,6 +22,7 @@
 #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"
@@ -260,13 +261,11 @@ 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
 
@@ -551,6 +550,9 @@ class DBImpl : public DB {
   void NotifyOnMemTableSealed(ColumnFamilyData* cfd,
                               const MemTableInfo& mem_table_info);
 
+  void NotifyOnExternalFileIngested(
+      ColumnFamilyData* cfd, const ExternalSstFileIngestionJob& ingestion_job);
+
   void NewThreadStatusCfInfo(ColumnFamilyData* cfd) const;
 
   void EraseThreadStatusCfInfo(ColumnFamilyData* cfd) const;
@@ -650,20 +652,13 @@ class DBImpl : public DB {
   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 +666,14 @@ 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);
@@ -752,7 +747,7 @@ 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>
@@ -994,9 +989,9 @@ 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_;

db/db_impl_add_file.cc

@@ -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

db/db_properties_test.cc

@@ -558,7 +558,7 @@ TEST_F(DBPropertiesTest, NumImmutableMemTable) {
     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
+    // break if we change the default skiplist implementation
     ASSERT_EQ(num, "384");
 
     uint64_t int_num;

db/db_tailing_iter_test.cc

@@ -699,6 +699,58 @@ 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)

db/db_test.cc

@@ -2646,15 +2646,11 @@ class ModelDB : public DB {
   }
 
 #ifndef ROCKSDB_LITE
-  using DB::AddFile;
-  virtual Status AddFile(ColumnFamilyHandle* column_family,
-                         const std::vector<ExternalSstFileInfo>& file_info_list,
-                         bool move_file, bool skip_snapshot_check) override {
-    return Status::NotSupported("Not implemented.");
-  }
-  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& options) override {
     return Status::NotSupported("Not implemented.");
   }
 

db/db_test2.cc

@@ -145,22 +145,6 @@ TEST_F(DBTest2, CacheIndexAndFilterWithDBRestart) {
   value = Get(1, "a");
 }
 
-TEST_F(DBTest2, MaxSuccessiveMergesChangeWithDBRecovery) {
-  Options options = CurrentOptions();
-  options.create_if_missing = true;
-  options.statistics = rocksdb::CreateDBStatistics();
-  options.max_successive_merges = 3;
-  options.merge_operator = MergeOperators::CreatePutOperator();
-  options.disable_auto_compactions = true;
-  DestroyAndReopen(options);
-  Put("poi", "Finch");
-  db_->Merge(WriteOptions(), "poi", "Reese");
-  db_->Merge(WriteOptions(), "poi", "Shaw");
-  db_->Merge(WriteOptions(), "poi", "Root");
-  options.max_successive_merges = 2;
-  Reopen(options);
-}
-
 #ifndef ROCKSDB_LITE
 class DBTestSharedWriteBufferAcrossCFs
     : public DBTestBase,
@@ -1905,6 +1889,23 @@ TEST_P(MergeOperatorPinningTest, TailingIterator) {
 }
 #endif  // ROCKSDB_LITE
 
+TEST_F(DBTest2, MaxSuccessiveMergesInRecovery) {
+  Options options;
+  options = CurrentOptions(options);
+  options.merge_operator = MergeOperators::CreatePutOperator();
+  DestroyAndReopen(options);
+
+  db_->Put(WriteOptions(), "foo", "bar");
+  ASSERT_OK(db_->Merge(WriteOptions(), "foo", "bar"));
+  ASSERT_OK(db_->Merge(WriteOptions(), "foo", "bar"));
+  ASSERT_OK(db_->Merge(WriteOptions(), "foo", "bar"));
+  ASSERT_OK(db_->Merge(WriteOptions(), "foo", "bar"));
+  ASSERT_OK(db_->Merge(WriteOptions(), "foo", "bar"));
+
+  options.max_successive_merges = 3;
+  Reopen(options);
+}
+
 size_t GetEncodedEntrySize(size_t key_size, size_t value_size) {
   std::string buffer;
 

db/db_test_util.cc

@@ -1077,7 +1077,7 @@ 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) {
   size_t total_reads = 0;
 
   for (auto& kv : true_data) {
@@ -1126,36 +1126,38 @@ void DBTestBase::VerifyDBFromMap(std::map<std::string, std::string> true_data,
     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;

db/db_test_util.h

@@ -822,7 +822,8 @@ 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);
+                       size_t* total_reads_res = nullptr,
+                       bool tailing_iter = false);
 
 #ifndef ROCKSDB_LITE
   uint64_t GetNumberOfSstFilesForColumnFamily(DB* db,

db/external_sst_file_ingestion_job.cc

@@ -0,0 +1,510 @@
+//  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/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/merger.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);
+      }
+    } else {
+      status = CopyFile(env_, path_outside_db, path_inside_db, 0);
+    }
+    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 {
+    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);
+      sv->current->AddIteratorsForLevel(ro, env_options_, &merge_iter_builder,
+                                        lvl);
+      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

db/external_sst_file_ingestion_job.h

@@ -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

db/fault_injection_test.cc

@@ -228,6 +228,11 @@ class FaultInjectionTest : public testing::Test,
     return Status::OK();
   }
 
+#if defined(__clang__)
+__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;

db/forward_iterator.cc

@@ -296,9 +296,15 @@ void ForwardIterator::SeekInternal(const Slice& internal_key,
   // an option to turn it off.
   if (seek_to_first || NeedToSeekImmutable(internal_key)) {
     immutable_status_ = Status::OK();
-    if ((has_iter_trimmed_for_upper_bound_) &&
-        (cfd_->internal_comparator().InternalKeyComparator::Compare(
-             prev_key_.GetKey(), internal_key) > 0)) {
+    if (has_iter_trimmed_for_upper_bound_ &&
+        (
+            // prev_ is not set yet
+            is_prev_set_ == false ||
+            // We are doing SeekToFirst() and internal_key.size() = 0
+            seek_to_first ||
+            // prev_key_ > internal_key
+            cfd_->internal_comparator().InternalKeyComparator::Compare(
+                prev_key_.GetKey(), internal_key) > 0)) {
       // Some iterators are trimmed. Need to rebuild.
       RebuildIterators(true);
       // Already seeked mutable iter, so seek again

db/memtable.cc

@@ -68,7 +68,7 @@ MemTable::MemTable(const InternalKeyComparator& cmp,
       table_(ioptions.memtable_factory->CreateMemTableRep(
           comparator_, &allocator_, ioptions.prefix_extractor,
           ioptions.info_log)),
-      range_del_table_(ioptions.memtable_factory->CreateMemTableRep(
+      range_del_table_(SkipListFactory().CreateMemTableRep(
           comparator_, &allocator_, nullptr /* transform */,
           ioptions.info_log)),
       data_size_(0),

db/plain_table_db_test.cc

@@ -386,10 +386,6 @@ TEST_P(PlainTableDBTest, Flush) {
       for (int total_order = 0; total_order <= 1; total_order++) {
         for (int store_index_in_file = 0; store_index_in_file <= 1;
              ++store_index_in_file) {
-          if (!bloom_bits && store_index_in_file) {
-            continue;
-          }
-
           Options options = CurrentOptions();
           options.create_if_missing = true;
           // Set only one bucket to force bucket conflict.

db/version_builder.cc

@@ -32,12 +32,12 @@
 namespace rocksdb {
 
 bool NewestFirstBySeqNo(FileMetaData* a, FileMetaData* b) {
-  if (a->smallest_seqno != b->smallest_seqno) {
-    return a->smallest_seqno > b->smallest_seqno;
-  }
   if (a->largest_seqno != b->largest_seqno) {
     return a->largest_seqno > b->largest_seqno;
   }
+  if (a->smallest_seqno != b->smallest_seqno) {
+    return a->smallest_seqno > b->smallest_seqno;
+  }
   // Break ties by file number
   return a->fd.GetNumber() > b->fd.GetNumber();
 }
@@ -146,13 +146,22 @@ class VersionBuilder::Rep {
             abort();
           }
 
-          if (!(f1->largest_seqno > f2->largest_seqno ||
-                // We can have multiple files with seqno = 0 as a result of
-                // using DB::AddFile()
-                (f1->largest_seqno == 0 && f2->largest_seqno == 0))) {
-            fprintf(stderr,
-                    "L0 files seqno missmatch %" PRIu64 " vs. %" PRIu64 "\n",
-                    f1->largest_seqno, f2->largest_seqno);
+          if (f2->smallest_seqno == f2->largest_seqno) {
+            // This is an external file that we ingested
+            SequenceNumber external_file_seqno = f2->smallest_seqno;
+            if (!(external_file_seqno < f1->largest_seqno ||
+                  external_file_seqno == 0)) {
+              fprintf(stderr, "L0 file with seqno %" PRIu64 " %" PRIu64
+                              " vs. file with global_seqno %" PRIu64 "\n",
+                      f1->smallest_seqno, f1->largest_seqno,
+                      external_file_seqno);
+              abort();
+            }
+          } else if (f1->smallest_seqno <= f2->smallest_seqno) {
+            fprintf(stderr, "L0 files seqno %" PRIu64 " %" PRIu64
+                            " vs. %" PRIu64 " %" PRIu64 "\n",
+                    f1->smallest_seqno, f1->largest_seqno, f2->smallest_seqno,
+                    f2->largest_seqno);
             abort();
           }
         } else {

db/version_set.cc

@@ -808,41 +808,51 @@ void Version::AddIterators(const ReadOptions& read_options,
                            MergeIteratorBuilder* merge_iter_builder) {
   assert(storage_info_.finalized_);
 
-  if (storage_info_.num_non_empty_levels() == 0) {
-    // No file in the Version.
+  for (int level = 0; level < storage_info_.num_non_empty_levels(); level++) {
+    AddIteratorsForLevel(read_options, soptions, merge_iter_builder, level);
+  }
+}
+
+void Version::AddIteratorsForLevel(const ReadOptions& read_options,
+                                   const EnvOptions& soptions,
+                                   MergeIteratorBuilder* merge_iter_builder,
+                                   int level) {
+  assert(storage_info_.finalized_);
+  if (level >= storage_info_.num_non_empty_levels()) {
+    // This is an empty level
+    return;
+  } else if (storage_info_.LevelFilesBrief(level).num_files == 0) {
+    // No files in this level
     return;
   }
 
   auto* arena = merge_iter_builder->GetArena();
-
-  // Merge all level zero files together since they may overlap
-  for (size_t i = 0; i < storage_info_.LevelFilesBrief(0).num_files; i++) {
-    const auto& file = storage_info_.LevelFilesBrief(0).files[i];
-    merge_iter_builder->AddIterator(cfd_->table_cache()->NewIterator(
-        read_options, soptions, cfd_->internal_comparator(), file.fd, nullptr,
-        cfd_->internal_stats()->GetFileReadHist(0), false, arena,
-        false /* skip_filters */, 0 /* level */));
-  }
-
-  // For levels > 0, we can use a concatenating iterator that sequentially
-  // walks through the non-overlapping files in the level, opening them
-  // lazily.
-  for (int level = 1; level < storage_info_.num_non_empty_levels(); level++) {
-    if (storage_info_.LevelFilesBrief(level).num_files != 0) {
-      auto* mem = arena->AllocateAligned(sizeof(LevelFileIteratorState));
-      auto* state = new (mem) LevelFileIteratorState(
-          cfd_->table_cache(), read_options, soptions,
-          cfd_->internal_comparator(),
-          cfd_->internal_stats()->GetFileReadHist(level),
-          false /* for_compaction */,
-          cfd_->ioptions()->prefix_extractor != nullptr, IsFilterSkipped(level),
-          level, nullptr /* range_del_agg */);
-      mem = arena->AllocateAligned(sizeof(LevelFileNumIterator));
-      auto* first_level_iter = new (mem) LevelFileNumIterator(
-          cfd_->internal_comparator(), &storage_info_.LevelFilesBrief(level));
-      merge_iter_builder->AddIterator(
-          NewTwoLevelIterator(state, first_level_iter, arena, false));
+  if (level == 0) {
+    // Merge all level zero files together since they may overlap
+    for (size_t i = 0; i < storage_info_.LevelFilesBrief(0).num_files; i++) {
+      const auto& file = storage_info_.LevelFilesBrief(0).files[i];
+      merge_iter_builder->AddIterator(cfd_->table_cache()->NewIterator(
+          read_options, soptions, cfd_->internal_comparator(), file.fd, nullptr,
+          cfd_->internal_stats()->GetFileReadHist(0), false, arena,
+          false /* skip_filters */, 0 /* level */));
     }
+  } else {
+    // For levels > 0, we can use a concatenating iterator that sequentially
+    // walks through the non-overlapping files in the level, opening them
+    // lazily.
+    auto* mem = arena->AllocateAligned(sizeof(LevelFileIteratorState));
+    auto* state = new (mem) LevelFileIteratorState(
+        cfd_->table_cache(), read_options, soptions,
+        cfd_->internal_comparator(),
+        cfd_->internal_stats()->GetFileReadHist(level),
+        false /* for_compaction */,
+        cfd_->ioptions()->prefix_extractor != nullptr, IsFilterSkipped(level),
+        level, nullptr /* range_del_agg */);
+    mem = arena->AllocateAligned(sizeof(LevelFileNumIterator));
+    auto* first_level_iter = new (mem) LevelFileNumIterator(
+        cfd_->internal_comparator(), &storage_info_.LevelFilesBrief(level));
+    merge_iter_builder->AddIterator(
+        NewTwoLevelIterator(state, first_level_iter, arena, false));
   }
 }
 

db/version_set.h

@@ -435,6 +435,10 @@ class Version {
   void AddIterators(const ReadOptions&, const EnvOptions& soptions,
                     MergeIteratorBuilder* merger_iter_builder);
 
+  void AddIteratorsForLevel(const ReadOptions&, const EnvOptions& soptions,
+                            MergeIteratorBuilder* merger_iter_builder,
+                            int level);
+
   // Lookup the value for key.  If found, store it in *val and
   // return OK.  Else return a non-OK status.
   // Uses *operands to store merge_operator operations to apply later.

db/write_batch.cc

@@ -885,13 +885,10 @@ class MemTableInserter : public WriteBatch::Handler {
         std::string merged_value;
 
         auto cf_handle = cf_mems_->GetColumnFamilyHandle();
-        Status s = Status::NotSupported();
-        if (db_ != nullptr && recovering_log_number_ != 0) {
-          if (cf_handle == nullptr) {
-            cf_handle = db_->DefaultColumnFamily();
-          }
-          s = db_->Get(ropts, cf_handle, key, &prev_value);
+        if (cf_handle == nullptr) {
+          cf_handle = db_->DefaultColumnFamily();
         }
+        Status s = db_->Get(ropts, cf_handle, key, &prev_value);
 
         char* prev_buffer = const_cast<char*>(prev_value.c_str());
         uint32_t prev_size = static_cast<uint32_t>(prev_value.size());
@@ -995,12 +992,7 @@ class MemTableInserter : public WriteBatch::Handler {
     auto* moptions = mem->GetMemTableOptions();
     bool perform_merge = false;
 
-    // If we pass DB through and options.max_successive_merges is hit
-    // during recovery, Get() will be issued which will try to acquire
-    // DB mutex and cause deadlock, as DB mutex is already held.
-    // So we disable merge in recovery
-    if (moptions->max_successive_merges > 0 && db_ != nullptr &&
-        recovering_log_number_ == 0) {
+    if (moptions->max_successive_merges > 0 && db_ != nullptr) {
       LookupKey lkey(key, sequence_);
 
       // Count the number of successive merges at the head

include/rocksdb/db.h

@@ -31,6 +31,11 @@
 #undef DeleteFile
 #endif
 
+#if defined(__GNUC__) || defined(__clang__)
+#define ROCKSDB_DEPRECATED_FUNC __attribute__((__deprecated__))
+#elif _WIN32
+#define ROCKSDB_DEPRECATED_FUNC __declspec(deprecated)
+#endif
 
 namespace rocksdb {
 
@@ -589,29 +594,20 @@ class DB {
     return CompactRange(options, DefaultColumnFamily(), begin, end);
   }
 
-#if defined(__GNUC__) || defined(__clang__)
-  __attribute__((__deprecated__))
-#elif _WIN32
-  __declspec(deprecated)
-#endif
-  virtual Status
-  CompactRange(ColumnFamilyHandle* column_family, const Slice* begin,
-               const Slice* end, bool change_level = false,
-               int target_level = -1, uint32_t target_path_id = 0) {
+  ROCKSDB_DEPRECATED_FUNC virtual Status CompactRange(
+      ColumnFamilyHandle* column_family, const Slice* begin, const Slice* end,
+      bool change_level = false, int target_level = -1,
+      uint32_t target_path_id = 0) {
     CompactRangeOptions options;
     options.change_level = change_level;
     options.target_level = target_level;
     options.target_path_id = target_path_id;
     return CompactRange(options, column_family, begin, end);
   }
-#if defined(__GNUC__) || defined(__clang__)
-  __attribute__((__deprecated__))
-#elif _WIN32
-  __declspec(deprecated)
-#endif
-  virtual Status
-  CompactRange(const Slice* begin, const Slice* end, bool change_level = false,
-               int target_level = -1, uint32_t target_path_id = 0) {
+
+  ROCKSDB_DEPRECATED_FUNC virtual Status CompactRange(
+      const Slice* begin, const Slice* end, bool change_level = false,
+      int target_level = -1, uint32_t target_path_id = 0) {
     CompactRangeOptions options;
     options.change_level = change_level;
     options.target_level = target_level;
@@ -803,79 +799,126 @@ class DB {
     GetColumnFamilyMetaData(DefaultColumnFamily(), metadata);
   }
 
-  // Batch load table files whose paths stored in  "file_path_list" into
-  // "column_family", a vector of  ExternalSstFileInfo can be used
-  // instead of "file_path_list" to do a blind batch add that wont
-  // need to read the file, move_file can be set to true to
-  // move the files instead of copying them, skip_snapshot_check can be set to
-  // true to ignore the snapshot, make sure that you know that when you use it,
-  // snapshots see the data that is added in the new files.
+  // IngestExternalFile() will load a list of external SST files (1) into the DB
+  // We will try to find the lowest possible level that the file can fit in, and
+  // ingest the file into this level (2). A file that have a key range that
+  // overlap with the memtable key range will require us to Flush the memtable
+  // first before ingesting the file.
   //
-  // Current Requirements:
-  // (1) The key ranges of the files don't overlap with each other
-  // (2) The key range of any file in list doesn't overlap with
-  //     existing keys or tombstones in DB.
-  // (3) No snapshots are held (check skip_snapshot_check to skip this check).
-  //
-  // Notes: We will try to ingest the files to the lowest possible level
-  //        even if the file compression dont match the level compression
-  virtual Status AddFile(ColumnFamilyHandle* column_family,
-                         const std::vector<std::string>& file_path_list,
-                         bool move_file = false, bool skip_snapshot_check = false) = 0;
-  virtual Status AddFile(const std::vector<std::string>& file_path_list,
-                         bool move_file = false, bool skip_snapshot_check = false) {
-    return AddFile(DefaultColumnFamily(), file_path_list, move_file, skip_snapshot_check);
+  // (1) External SST files can be created using SstFileWriter
+  // (2) We will try to ingest the files to the lowest possible level
+  //     even if the file compression dont match the level compression
+  virtual Status IngestExternalFile(
+      ColumnFamilyHandle* column_family,
+      const std::vector<std::string>& external_files,
+      const IngestExternalFileOptions& options) = 0;
+
+  virtual Status IngestExternalFile(
+      const std::vector<std::string>& external_files,
+      const IngestExternalFileOptions& options) {
+    return IngestExternalFile(DefaultColumnFamily(), external_files, options);
   }
-#if defined(__GNUC__) || defined(__clang__)
-  __attribute__((__deprecated__))
-#elif _WIN32
-  __declspec(deprecated)
-#endif
-  virtual Status
-  AddFile(ColumnFamilyHandle* column_family, const std::string& file_path,
-          bool move_file = false, bool skip_snapshot_check = false) {
-    return AddFile(column_family, std::vector<std::string>(1, file_path),
-                   move_file, skip_snapshot_check);
+
+  // AddFile() is deprecated, please use IngestExternalFile()
+  ROCKSDB_DEPRECATED_FUNC virtual Status AddFile(
+      ColumnFamilyHandle* column_family,
+      const std::vector<std::string>& file_path_list, bool move_file = false,
+      bool skip_snapshot_check = false) {
+    IngestExternalFileOptions ifo;
+    ifo.move_files = move_file;
+    ifo.snapshot_consistency = !skip_snapshot_check;
+    ifo.allow_global_seqno = false;
+    ifo.allow_blocking_flush = false;
+    return IngestExternalFile(column_family, file_path_list, ifo);
   }
-#if defined(__GNUC__) || defined(__clang__)
-  __attribute__((__deprecated__))
-#elif _WIN32
-  __declspec(deprecated)
-#endif
-  virtual Status
-  AddFile(const std::string& file_path, bool move_file = false, bool skip_snapshot_check = false) {
-    return AddFile(DefaultColumnFamily(),
-                   std::vector<std::string>(1, file_path), move_file, skip_snapshot_check);
+
+  ROCKSDB_DEPRECATED_FUNC virtual Status AddFile(
+      const std::vector<std::string>& file_path_list, bool move_file = false,
+      bool skip_snapshot_check = false) {
+    IngestExternalFileOptions ifo;
+    ifo.move_files = move_file;
+    ifo.snapshot_consistency = !skip_snapshot_check;
+    ifo.allow_global_seqno = false;
+    ifo.allow_blocking_flush = false;
+    return IngestExternalFile(DefaultColumnFamily(), file_path_list, ifo);
+  }
+
+  // AddFile() is deprecated, please use IngestExternalFile()
+  ROCKSDB_DEPRECATED_FUNC virtual Status AddFile(
+      ColumnFamilyHandle* column_family, const std::string& file_path,
+      bool move_file = false, bool skip_snapshot_check = false) {
+    IngestExternalFileOptions ifo;
+    ifo.move_files = move_file;
+    ifo.snapshot_consistency = !skip_snapshot_check;
+    ifo.allow_global_seqno = false;
+    ifo.allow_blocking_flush = false;
+    return IngestExternalFile(column_family, {file_path}, ifo);
+  }
+
+  ROCKSDB_DEPRECATED_FUNC virtual Status AddFile(
+      const std::string& file_path, bool move_file = false,
+      bool skip_snapshot_check = false) {
+    IngestExternalFileOptions ifo;
+    ifo.move_files = move_file;
+    ifo.snapshot_consistency = !skip_snapshot_check;
+    ifo.allow_global_seqno = false;
+    ifo.allow_blocking_flush = false;
+    return IngestExternalFile(DefaultColumnFamily(), {file_path}, ifo);
   }
 
   // Load table file with information "file_info" into "column_family"
-  virtual Status AddFile(ColumnFamilyHandle* column_family,
-                         const std::vector<ExternalSstFileInfo>& file_info_list,
-                         bool move_file = false, bool skip_snapshot_check = false) = 0;
-  virtual Status AddFile(const std::vector<ExternalSstFileInfo>& file_info_list,
-                         bool move_file = false, bool skip_snapshot_check = false) {
-    return AddFile(DefaultColumnFamily(), file_info_list, move_file, skip_snapshot_check);
+  ROCKSDB_DEPRECATED_FUNC virtual Status AddFile(
+      ColumnFamilyHandle* column_family,
+      const std::vector<ExternalSstFileInfo>& file_info_list,
+      bool move_file = false, bool skip_snapshot_check = false) {
+    std::vector<std::string> external_files;
+    for (const ExternalSstFileInfo& file_info : file_info_list) {
+      external_files.push_back(file_info.file_path);
+    }
+    IngestExternalFileOptions ifo;
+    ifo.move_files = move_file;
+    ifo.snapshot_consistency = !skip_snapshot_check;
+    ifo.allow_global_seqno = false;
+    ifo.allow_blocking_flush = false;
+    return IngestExternalFile(column_family, external_files, ifo);
   }
-#if defined(__GNUC__) || defined(__clang__)
-  __attribute__((__deprecated__))
-#elif _WIN32
-  __declspec(deprecated)
-#endif
-  virtual Status
-  AddFile(ColumnFamilyHandle* column_family,
-          const ExternalSstFileInfo* file_info, bool move_file = false, bool skip_snapshot_check = false) {
-    return AddFile(column_family,
-                   std::vector<ExternalSstFileInfo>(1, *file_info), move_file, skip_snapshot_check);
+
+  ROCKSDB_DEPRECATED_FUNC virtual Status AddFile(
+      const std::vector<ExternalSstFileInfo>& file_info_list,
+      bool move_file = false, bool skip_snapshot_check = false) {
+    std::vector<std::string> external_files;
+    for (const ExternalSstFileInfo& file_info : file_info_list) {
+      external_files.push_back(file_info.file_path);
+    }
+    IngestExternalFileOptions ifo;
+    ifo.move_files = move_file;
+    ifo.snapshot_consistency = !skip_snapshot_check;
+    ifo.allow_global_seqno = false;
+    ifo.allow_blocking_flush = false;
+    return IngestExternalFile(DefaultColumnFamily(), external_files, ifo);
   }
-#if defined(__GNUC__) || defined(__clang__)
-  __attribute__((__deprecated__))
-#elif _WIN32
-  __declspec(deprecated)
-#endif
-  virtual Status
-  AddFile(const ExternalSstFileInfo* file_info, bool move_file = false, bool skip_snapshot_check = false) {
-    return AddFile(DefaultColumnFamily(),
-                   std::vector<ExternalSstFileInfo>(1, *file_info), move_file, skip_snapshot_check);
+
+  ROCKSDB_DEPRECATED_FUNC virtual Status AddFile(
+      ColumnFamilyHandle* column_family, const ExternalSstFileInfo* file_info,
+      bool move_file = false, bool skip_snapshot_check = false) {
+    IngestExternalFileOptions ifo;
+    ifo.move_files = move_file;
+    ifo.snapshot_consistency = !skip_snapshot_check;
+    ifo.allow_global_seqno = false;
+    ifo.allow_blocking_flush = false;
+    return IngestExternalFile(column_family, {file_info->file_path}, ifo);
+  }
+
+  ROCKSDB_DEPRECATED_FUNC virtual Status AddFile(
+      const ExternalSstFileInfo* file_info, bool move_file = false,
+      bool skip_snapshot_check = false) {
+    IngestExternalFileOptions ifo;
+    ifo.move_files = move_file;
+    ifo.snapshot_consistency = !skip_snapshot_check;
+    ifo.allow_global_seqno = false;
+    ifo.allow_blocking_flush = false;
+    return IngestExternalFile(DefaultColumnFamily(), {file_info->file_path},
+                              ifo);
   }
 
 #endif  // ROCKSDB_LITE

include/rocksdb/listener.h

@@ -170,6 +170,20 @@ struct MemTableInfo {
 
 };
 
+struct ExternalFileIngestionInfo {
+  // the name of the column family
+  std::string cf_name;
+  // Path of the file outside the DB
+  std::string external_file_path;
+  // Path of the file inside the DB
+  std::string internal_file_path;
+  // The global sequence number assigned to keys in this file
+  SequenceNumber global_seqno;
+  // Table properties of the table being flushed
+  TableProperties table_properties;
+};
+
+
 // EventListener class contains a set of call-back functions that will
 // be called when specific RocksDB event happens such as flush.  It can
 // be used as a building block for developing custom features such as
@@ -291,6 +305,15 @@ class EventListener {
   virtual void OnColumnFamilyHandleDeletionStarted(ColumnFamilyHandle* handle) {
   }
 
+  // A call-back function for RocksDB which will be called after an external
+  // file is ingested using IngestExternalFile.
+  //
+  // Note that the this function will run on the same thread as
+  // IngestExternalFile(), if this function is blocked, IngestExternalFile()
+  // will be blocked from finishing.
+  virtual void OnExternalFileIngested(
+      DB* /*db*/, const ExternalFileIngestionInfo& /*info*/) {}
+
   virtual ~EventListener() {}
 };
 

include/rocksdb/options.h

@@ -1617,6 +1617,21 @@ struct CompactRangeOptions {
       BottommostLevelCompaction::kIfHaveCompactionFilter;
 };
 
+// IngestExternalFileOptions is used by IngestExternalFile()
+struct IngestExternalFileOptions {
+  // Can be set to true to move the files instead of copying them.
+  bool move_files = false;
+  // If set to false, an ingested file keys could appear in existing snapshots
+  // that where created before the file was ingested.
+  bool snapshot_consistency = true;
+  // If set to false, IngestExternalFile() will fail if the file key range
+  // overlaps with existing keys or tombstones in the DB.
+  bool allow_global_seqno = true;
+  // If set to false and the file key range overlaps with the memtable key range
+  // (memtable flush required), IngestExternalFile will fail.
+  bool allow_blocking_flush = true;
+};
+
 }  // namespace rocksdb
 
 #endif  // STORAGE_ROCKSDB_INCLUDE_OPTIONS_H_

include/rocksdb/sst_file_writer.h

@@ -7,6 +7,7 @@
 #include <string>
 #include "rocksdb/env.h"
 #include "rocksdb/options.h"
+#include "rocksdb/table_properties.h"
 #include "rocksdb/types.h"
 
 namespace rocksdb {
@@ -43,8 +44,12 @@ struct ExternalSstFileInfo {
 // All keys in files generated by SstFileWriter will have sequence number = 0
 class SstFileWriter {
  public:
+  // User can pass `column_family` to specify that the the generated file will
+  // be ingested into this column_family, note that passing nullptr means that
+  // the column_family is unknown.
   SstFileWriter(const EnvOptions& env_options, const Options& options,
-                const Comparator* user_comparator);
+                const Comparator* user_comparator,
+                ColumnFamilyHandle* column_family = nullptr);
 
   ~SstFileWriter();
 

include/rocksdb/statistics.h

@@ -152,7 +152,6 @@ enum Tickers : uint32_t {
   // written to storage because key does not exist
   NUMBER_FILTERED_DELETES,
   NUMBER_MERGE_FAILURES,
-  SEQUENCE_NUMBER,
 
   // number of times bloom was checked before creating iterator on a
   // file, and the number of times the check was useful in avoiding
@@ -280,7 +279,6 @@ const std::vector<std::pair<Tickers, std::string>> TickersNameMap = {
     {NUMBER_MULTIGET_BYTES_READ, "rocksdb.number.multiget.bytes.read"},
     {NUMBER_FILTERED_DELETES, "rocksdb.number.deletes.filtered"},
     {NUMBER_MERGE_FAILURES, "rocksdb.number.merge.failures"},
-    {SEQUENCE_NUMBER, "rocksdb.sequence.number"},
     {BLOOM_FILTER_PREFIX_CHECKED, "rocksdb.bloom.filter.prefix.checked"},
     {BLOOM_FILTER_PREFIX_USEFUL, "rocksdb.bloom.filter.prefix.useful"},
     {NUMBER_OF_RESEEKS_IN_ITERATION, "rocksdb.number.reseeks.iteration"},

include/rocksdb/utilities/flashcache.h

@@ -1,25 +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.
-
-#pragma once
-
-#include <string>
-#include "rocksdb/env.h"
-
-namespace rocksdb {
-
-// This API is experimental. We will mark it stable once we run it in production
-// for a while.
-// NewFlashcacheAwareEnv() creates and Env that blacklists all background
-// threads (used for flush and compaction) from using flashcache to cache their
-// reads. Reads from compaction thread don't need to be cached because they are
-// going to be soon made obsolete (due to nature of compaction)
-// Usually you would pass Env::Default() as base.
-// cachedev_fd is a file descriptor of the flashcache device. Caller has to
-// open flashcache device before calling this API.
-extern std::unique_ptr<Env> NewFlashcacheAwareEnv(
-    Env* base, const int cachedev_fd);
-
-}  // namespace rocksdb

include/rocksdb/utilities/stackable_db.h

@@ -68,16 +68,12 @@ class StackableDB : public DB {
     return db_->MultiGet(options, column_family, keys, values);
   }
 
-  using DB::AddFile;
-  virtual Status AddFile(ColumnFamilyHandle* column_family,
-                         const std::vector<ExternalSstFileInfo>& file_info_list,
-                         bool move_file, bool skip_snapshot_check) override {
-    return db_->AddFile(column_family, file_info_list, move_file, skip_snapshot_check);
-  }
-  virtual Status AddFile(ColumnFamilyHandle* column_family,
-                         const std::vector<std::string>& file_path_list,
-                         bool move_file, bool skip_snapshot_check) override {
-    return db_->AddFile(column_family, file_path_list, move_file, skip_snapshot_check);
+  using DB::IngestExternalFile;
+  virtual Status IngestExternalFile(
+      ColumnFamilyHandle* column_family,
+      const std::vector<std::string>& external_files,
+      const IngestExternalFileOptions& options) override {
+    return db_->IngestExternalFile(column_family, external_files, options);
   }
 
   using DB::KeyMayExist;

include/rocksdb/version.h

@@ -6,7 +6,7 @@
 
 #define ROCKSDB_MAJOR 4
 #define ROCKSDB_MINOR 13
-#define ROCKSDB_PATCH 0
+#define ROCKSDB_PATCH 5
 
 // 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

java/rocksjni/rocksjni.cc

@@ -16,8 +16,9 @@
 #include <algorithm>
 
 #include "include/org_rocksdb_RocksDB.h"
-#include "rocksdb/db.h"
 #include "rocksdb/cache.h"
+#include "rocksdb/db.h"
+#include "rocksdb/options.h"
 #include "rocksdb/types.h"
 #include "rocksjni/portal.h"
 
@@ -1757,17 +1758,6 @@ void add_file_helper(JNIEnv* env, const jobjectArray& jfile_path_list,
   }
 }
 
-void add_file_helper(
-    JNIEnv* env, jlongArray jfi_handle_list, int fi_handle_list_len,
-    std::vector<rocksdb::ExternalSstFileInfo>* file_info_list) {
-  jlong* jfih = env->GetLongArrayElements(jfi_handle_list, NULL);
-  for (int i = 0; i < fi_handle_list_len; i++) {
-    auto* file_info =
-        reinterpret_cast<rocksdb::ExternalSstFileInfo*>(*(jfih + i));
-    file_info_list->push_back(*file_info);
-  }
-}
-
 /*
  * Class:     org_rocksdb_RocksDB
  * Method:    addFile
@@ -1783,32 +1773,15 @@ void Java_org_rocksdb_RocksDB_addFile__JJ_3Ljava_lang_String_2IZ(
                   &file_path_list);
   auto* column_family =
       reinterpret_cast<rocksdb::ColumnFamilyHandle*>(jcf_handle);
+  rocksdb::IngestExternalFileOptions ifo;
+  ifo.move_files = static_cast<bool>(jmove_file);
+  ifo.snapshot_consistency = true;
+  ifo.allow_global_seqno = false;
+  ifo.allow_blocking_flush = false;
   rocksdb::Status s =
-      db->AddFile(column_family, file_path_list, static_cast<bool>(jmove_file));
+      db->IngestExternalFile(column_family, file_path_list, ifo);
   if (!s.ok()) {
     rocksdb::RocksDBExceptionJni::ThrowNew(env, s);
   }
 }
 
-/*
- * Class:     org_rocksdb_RocksDB
- * Method:    addFile
- * Signature: (JJ[JIZ)V
- */
-void Java_org_rocksdb_RocksDB_addFile__JJ_3JIZ(
-    JNIEnv* env, jobject jdb, jlong jdb_handle, jlong jcf_handle,
-    jlongArray jfile_info_handle_list, jint jfile_info_handle_list_len,
-    jboolean jmove_file) {
-  auto* db = reinterpret_cast<rocksdb::DB*>(jdb_handle);
-  std::vector<rocksdb::ExternalSstFileInfo> file_info_list;
-  add_file_helper(env, jfile_info_handle_list,
-                  static_cast<int>(jfile_info_handle_list_len),
-                  &file_info_list);
-  auto* column_family =
-      reinterpret_cast<rocksdb::ColumnFamilyHandle*>(jcf_handle);
-  rocksdb::Status s =
-      db->AddFile(column_family, file_info_list, static_cast<bool>(jmove_file));
-  if (!s.ok()) {
-    rocksdb::RocksDBExceptionJni::ThrowNew(env, s);
-  }
-}

java/src/main/java/org/rocksdb/TickerType.java

@@ -87,7 +87,6 @@ public enum TickerType {
   // written to storage because key does not exist
   NUMBER_FILTERED_DELETES(36),
   NUMBER_MERGE_FAILURES(37),
-  SEQUENCE_NUMBER(38),
 
   // number of times bloom was checked before creating iterator on a
   // file, and the number of times the check was useful in avoiding

port/win/io_win.h

@@ -26,7 +26,7 @@ namespace port {
 std::string GetWindowsErrSz(DWORD err);
 
 inline Status IOErrorFromWindowsError(const std::string& context, DWORD err) {
-  return (err == ERROR_HANDLE_DISK_FULL) ?
+  return ((err == ERROR_HANDLE_DISK_FULL) || (err == ERROR_DISK_FULL)) ?
       Status::NoSpace(context, GetWindowsErrSz(err)) :
       Status::IOError(context, GetWindowsErrSz(err));
 }

src.mk

@@ -17,9 +17,9 @@ LIB_SOURCES =                                                   \
   db/db_impl_debug.cc                                           \
   db/db_impl_readonly.cc                                        \
   db/db_impl_experimental.cc                                    \
-  db/db_impl_add_file.cc                                        \
   db/db_info_dumper.cc                                          \
   db/db_iter.cc                                                 \
+  db/external_sst_file_ingestion_job.cc                         \
   db/experimental.cc                                            \
   db/event_helpers.cc                                           \
   db/file_indexer.cc                                            \
@@ -154,7 +154,6 @@ LIB_SOURCES =                                                   \
   utilities/document/json_document.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/memory/memory_util.cc                               \

table/merger.cc

@@ -36,12 +36,13 @@ const size_t kNumIterReserve = 4;
 class MergingIterator : public InternalIterator {
  public:
   MergingIterator(const Comparator* comparator, InternalIterator** children,
-                  int n, bool is_arena_mode)
+                  int n, bool is_arena_mode, bool prefix_seek_mode)
       : is_arena_mode_(is_arena_mode),
         comparator_(comparator),
         current_(nullptr),
         direction_(kForward),
         minHeap_(comparator_),
+        prefix_seek_mode_(prefix_seek_mode),
         pinned_iters_mgr_(nullptr) {
     children_.resize(n);
     for (int i = 0; i < n; i++) {
@@ -204,9 +205,23 @@ class MergingIterator : public InternalIterator {
       InitMaxHeap();
       for (auto& child : children_) {
         if (&child != current_) {
-          child.SeekForPrev(key());
-          if (child.Valid() && comparator_->Equal(key(), child.key())) {
-            child.Prev();
+          if (!prefix_seek_mode_) {
+            child.Seek(key());
+            if (child.Valid()) {
+              // Child is at first entry >= key().  Step back one to be < key()
+              TEST_SYNC_POINT_CALLBACK("MergeIterator::Prev:BeforePrev",
+                                       &child);
+              child.Prev();
+            } else {
+              // Child has no entries >= key().  Position at last entry.
+              TEST_SYNC_POINT("MergeIterator::Prev:BeforeSeekToLast");
+              child.SeekToLast();
+            }
+          } else {
+            child.SeekForPrev(key());
+            if (child.Valid() && comparator_->Equal(key(), child.key())) {
+              child.Prev();
+            }
           }
         }
         if (child.Valid()) {
@@ -214,6 +229,13 @@ class MergingIterator : public InternalIterator {
         }
       }
       direction_ = kReverse;
+      if (!prefix_seek_mode_) {
+        // Note that we don't do assert(current_ == CurrentReverse()) here
+        // because it is possible to have some keys larger than the seek-key
+        // inserted between Seek() and SeekToLast(), which makes current_ not
+        // equal to CurrentReverse().
+        current_ = CurrentReverse();
+      }
       // The loop advanced all non-current children to be < key() so current_
       // should still be strictly the smallest key.
       assert(current_ == CurrentReverse());
@@ -299,6 +321,8 @@ class MergingIterator : public InternalIterator {
   };
   Direction direction_;
   MergerMinIterHeap minHeap_;
+  bool prefix_seek_mode_;
+
   // Max heap is used for reverse iteration, which is way less common than
   // forward.  Lazily initialize it to save memory.
   std::unique_ptr<MergerMaxIterHeap> maxHeap_;
@@ -331,7 +355,7 @@ void MergingIterator::InitMaxHeap() {
 
 InternalIterator* NewMergingIterator(const Comparator* cmp,
                                      InternalIterator** list, int n,
-                                     Arena* arena) {
+                                     Arena* arena, bool prefix_seek_mode) {
   assert(n >= 0);
   if (n == 0) {
     return NewEmptyInternalIterator(arena);
@@ -339,19 +363,20 @@ InternalIterator* NewMergingIterator(const Comparator* cmp,
     return list[0];
   } else {
     if (arena == nullptr) {
-      return new MergingIterator(cmp, list, n, false);
+      return new MergingIterator(cmp, list, n, false, prefix_seek_mode);
     } else {
       auto mem = arena->AllocateAligned(sizeof(MergingIterator));
-      return new (mem) MergingIterator(cmp, list, n, true);
+      return new (mem) MergingIterator(cmp, list, n, true, prefix_seek_mode);
     }
   }
 }
 
 MergeIteratorBuilder::MergeIteratorBuilder(const Comparator* comparator,
-                                           Arena* a)
+                                           Arena* a, bool prefix_seek_mode)
     : first_iter(nullptr), use_merging_iter(false), arena(a) {
   auto mem = arena->AllocateAligned(sizeof(MergingIterator));
-  merge_iter = new (mem) MergingIterator(comparator, nullptr, 0, true);
+  merge_iter =
+      new (mem) MergingIterator(comparator, nullptr, 0, true, prefix_seek_mode);
 }
 
 void MergeIteratorBuilder::AddIterator(InternalIterator* iter) {

table/merger.h

@@ -28,7 +28,8 @@ class Arena;
 // REQUIRES: n >= 0
 extern InternalIterator* NewMergingIterator(const Comparator* comparator,
                                             InternalIterator** children, int n,
-                                            Arena* arena = nullptr);
+                                            Arena* arena = nullptr,
+                                            bool prefix_seek_mode = false);
 
 class MergingIterator;
 
@@ -37,7 +38,8 @@ class MergeIteratorBuilder {
  public:
   // comparator: the comparator used in merging comparator
   // arena: where the merging iterator needs to be allocated from.
-  explicit MergeIteratorBuilder(const Comparator* comparator, Arena* arena);
+  explicit MergeIteratorBuilder(const Comparator* comparator, Arena* arena,
+                                bool prefix_seek_mode = false);
   ~MergeIteratorBuilder() {}
 
   // Add iter to the merging iterator.

table/plain_table_builder.cc

@@ -80,7 +80,6 @@ PlainTableBuilder::PlainTableBuilder(
     index_builder_.reset(
         new PlainTableIndexBuilder(&arena_, ioptions, index_sparseness,
                                    hash_table_ratio, huge_page_tlb_size_));
-    assert(bloom_bits_per_key_ > 0);
     properties_.user_collected_properties
         [PlainTablePropertyNames::kBloomVersion] = "1";  // For future use
   }
@@ -191,37 +190,40 @@ Status PlainTableBuilder::Finish() {
 
   if (store_index_in_file_ && (properties_.num_entries > 0)) {
     assert(properties_.num_entries <= std::numeric_limits<uint32_t>::max());
-    bloom_block_.SetTotalBits(
-        &arena_,
-        static_cast<uint32_t>(properties_.num_entries) * bloom_bits_per_key_,
-        ioptions_.bloom_locality, huge_page_tlb_size_, ioptions_.info_log);
-
-    PutVarint32(&properties_.user_collected_properties
-                     [PlainTablePropertyNames::kNumBloomBlocks],
-                bloom_block_.GetNumBlocks());
-
-    bloom_block_.AddKeysHashes(keys_or_prefixes_hashes_);
+    Status s;
     BlockHandle bloom_block_handle;
-    auto finish_result = bloom_block_.Finish();
+    if (bloom_bits_per_key_ > 0) {
+      bloom_block_.SetTotalBits(
+          &arena_,
+          static_cast<uint32_t>(properties_.num_entries) * bloom_bits_per_key_,
+          ioptions_.bloom_locality, huge_page_tlb_size_, ioptions_.info_log);
 
-    properties_.filter_size = finish_result.size();
-    auto s = WriteBlock(finish_result, file_, &offset_, &bloom_block_handle);
+      PutVarint32(&properties_.user_collected_properties
+                       [PlainTablePropertyNames::kNumBloomBlocks],
+                  bloom_block_.GetNumBlocks());
 
-    if (!s.ok()) {
-      return s;
+      bloom_block_.AddKeysHashes(keys_or_prefixes_hashes_);
+
+      Slice bloom_finish_result = bloom_block_.Finish();
+
+      properties_.filter_size = bloom_finish_result.size();
+      s = WriteBlock(bloom_finish_result, file_, &offset_, &bloom_block_handle);
+
+      if (!s.ok()) {
+        return s;
+      }
+      meta_index_builer.Add(BloomBlockBuilder::kBloomBlock, bloom_block_handle);
     }
-
     BlockHandle index_block_handle;
-    finish_result = index_builder_->Finish();
+    Slice index_finish_result = index_builder_->Finish();
 
-    properties_.index_size = finish_result.size();
-    s = WriteBlock(finish_result, file_, &offset_, &index_block_handle);
+    properties_.index_size = index_finish_result.size();
+    s = WriteBlock(index_finish_result, file_, &offset_, &index_block_handle);
 
     if (!s.ok()) {
       return s;
     }
 
-    meta_index_builer.Add(BloomBlockBuilder::kBloomBlock, bloom_block_handle);
     meta_index_builer.Add(PlainTableIndexBuilder::kPlainTableIndexBlock,
                           index_block_handle);
   }

table/plain_table_reader.cc

@@ -294,21 +294,25 @@ Status PlainTableReader::PopulateIndex(TableProperties* props,
   assert(props != nullptr);
   table_properties_.reset(props);
 
-  BlockContents bloom_block_contents;
-  auto s = ReadMetaBlock(file_info_.file.get(), file_size_,
-                         kPlainTableMagicNumber, ioptions_,
-                         BloomBlockBuilder::kBloomBlock, &bloom_block_contents);
-  bool index_in_file = s.ok();
-
   BlockContents index_block_contents;
-  s = ReadMetaBlock(
+  Status s = ReadMetaBlock(
       file_info_.file.get(), file_size_, kPlainTableMagicNumber, ioptions_,
       PlainTableIndexBuilder::kPlainTableIndexBlock, &index_block_contents);
 
-  index_in_file &= s.ok();
+  bool index_in_file = s.ok();
+
+  BlockContents bloom_block_contents;
+  bool bloom_in_file = false;
+  // We only need to read the bloom block if index block is in file.
+  if (index_in_file) {
+    s = ReadMetaBlock(file_info_.file.get(), file_size_, kPlainTableMagicNumber,
+                      ioptions_, BloomBlockBuilder::kBloomBlock,
+                      &bloom_block_contents);
+    bloom_in_file = s.ok() && bloom_block_contents.data.size() > 0;
+  }
 
   Slice* bloom_block;
-  if (index_in_file) {
+  if (bloom_in_file) {
     // If bloom_block_contents.allocation is not empty (which will be the case
     // for non-mmap mode), it holds the alloated memory for the bloom block.
     // It needs to be kept alive to keep `bloom_block` valid.
@@ -318,8 +322,6 @@ Status PlainTableReader::PopulateIndex(TableProperties* props,
     bloom_block = nullptr;
   }
 
-  // index_in_file == true only if there are kBloomBlock and
-  // kPlainTableIndexBlock in file
   Slice* index_block;
   if (index_in_file) {
     // If index_block_contents.allocation is not empty (which will be the case
@@ -355,7 +357,7 @@ Status PlainTableReader::PopulateIndex(TableProperties* props,
                             huge_page_tlb_size, ioptions_.info_log);
       }
     }
-  } else {
+  } else if (bloom_in_file) {
     enable_bloom_ = true;
     auto num_blocks_property = props->user_collected_properties.find(
         PlainTablePropertyNames::kNumBloomBlocks);
@@ -372,6 +374,10 @@ Status PlainTableReader::PopulateIndex(TableProperties* props,
         const_cast<unsigned char*>(
             reinterpret_cast<const unsigned char*>(bloom_block->data())),
         static_cast<uint32_t>(bloom_block->size()) * 8, num_blocks);
+  } else {
+    // Index in file but no bloom in file. Disable bloom filter in this case.
+    enable_bloom_ = false;
+    bloom_bits_per_key = 0;
   }
 
   PlainTableIndexBuilder index_builder(&arena_, ioptions_, index_sparseness,

table/sst_file_writer.cc

@@ -21,11 +21,12 @@ const std::string ExternalSstFilePropertyNames::kGlobalSeqno =
 
 struct SstFileWriter::Rep {
   Rep(const EnvOptions& _env_options, const Options& options,
-      const Comparator* _user_comparator)
+      const Comparator* _user_comparator, ColumnFamilyHandle* _cfh)
       : env_options(_env_options),
         ioptions(options),
         mutable_cf_options(options),
-        internal_comparator(_user_comparator) {}
+        internal_comparator(_user_comparator),
+        cfh(_cfh) {}
 
   std::unique_ptr<WritableFileWriter> file_writer;
   std::unique_ptr<TableBuilder> builder;
@@ -34,16 +35,26 @@ struct SstFileWriter::Rep {
   MutableCFOptions mutable_cf_options;
   InternalKeyComparator internal_comparator;
   ExternalSstFileInfo file_info;
-  std::string column_family_name;
   InternalKey ikey;
+  std::string column_family_name;
+  ColumnFamilyHandle* cfh;
 };
 
 SstFileWriter::SstFileWriter(const EnvOptions& env_options,
                              const Options& options,
-                             const Comparator* user_comparator)
-    : rep_(new Rep(env_options, options, user_comparator)) {}
+                             const Comparator* user_comparator,
+                             ColumnFamilyHandle* column_family)
+    : rep_(new Rep(env_options, options, user_comparator, column_family)) {}
 
-SstFileWriter::~SstFileWriter() { delete rep_; }
+SstFileWriter::~SstFileWriter() {
+  if (rep_->builder) {
+    // User did not call Finish() or Finish() failed, we need to
+    // abandon the builder.
+    rep_->builder->Abandon();
+  }
+
+  delete rep_;
+}
 
 Status SstFileWriter::Open(const std::string& file_path) {
   Rep* r = rep_;
@@ -81,6 +92,18 @@ Status SstFileWriter::Open(const std::string& file_path) {
             user_collector_factories[i]));
   }
   int unknown_level = -1;
+  uint32_t cf_id;
+
+  if (r->cfh != nullptr) {
+    // user explicitly specified that this file will be ingested into cfh,
+    // we can persist this information in the file.
+    cf_id = r->cfh->GetID();
+    r->column_family_name = r->cfh->GetName();
+  } else {
+    r->column_family_name = "";
+    cf_id = TablePropertiesCollectorFactory::Context::kUnknownColumnFamily;
+  }
+
   TableBuilderOptions table_builder_options(
       r->ioptions, r->internal_comparator, &int_tbl_prop_collector_factories,
       compression_type, r->ioptions.compression_opts,
@@ -92,9 +115,7 @@ Status SstFileWriter::Open(const std::string& file_path) {
   // TODO(tec) : If table_factory is using compressed block cache, we will
   // be adding the external sst file blocks into it, which is wasteful.
   r->builder.reset(r->ioptions.table_factory->NewTableBuilder(
-      table_builder_options,
-      TablePropertiesCollectorFactory::Context::kUnknownColumnFamily,
-      r->file_writer.get()));
+      table_builder_options, cf_id, r->file_writer.get()));
 
   r->file_info.file_path = file_path;
   r->file_info.file_size = 0;

third-party/flashcache/flashcache_ioctl.h

@@ -1,55 +0,0 @@
-/****************************************************************************
- *  flashcache_ioctl.h
- *  FlashCache: Device mapper target for block-level disk caching
- *
- *  Copyright 2010 Facebook, Inc.
- *  Author: Mohan Srinivasan (mohan@facebook.com)
- *
- *  Based on DM-Cache:
- *   Copyright (C) International Business Machines Corp., 2006
- *   Author: Ming Zhao (mingzhao@ufl.edu)
- *
- *  This program is free software; you can redistribute it and/or modify
- *  it under the terms of the GNU General Public License as published by
- *  the Free Software Foundation; under version 2 of the License.
- *
- *  This program is distributed in the hope that it will be useful,
- *  but WITHOUT ANY WARRANTY; without even the implied warranty of
- *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
- *  GNU General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program.  If not, see <http://www.gnu.org/licenses/>.
- ****************************************************************************/
-
-#ifdef OS_LINUX
-#ifndef FLASHCACHE_IOCTL_H
-#define FLASHCACHE_IOCTL_H
-
-#include <linux/types.h>
-
-#define FLASHCACHE_IOCTL 0xfe
-
-enum {
-	FLASHCACHEADDNCPID_CMD=200,
-	FLASHCACHEDELNCPID_CMD,
-	FLASHCACHEDELNCALL_CMD,
-	FLASHCACHEADDWHITELIST_CMD,
-	FLASHCACHEDELWHITELIST_CMD,
-	FLASHCACHEDELWHITELISTALL_CMD,
-};
-
-#define FLASHCACHEADDNCPID	_IOW(FLASHCACHE_IOCTL, FLASHCACHEADDNCPID_CMD, pid_t)
-#define FLASHCACHEDELNCPID	_IOW(FLASHCACHE_IOCTL, FLASHCACHEDELNCPID_CMD, pid_t)
-#define FLASHCACHEDELNCALL	_IOW(FLASHCACHE_IOCTL, FLASHCACHEDELNCALL_CMD, pid_t)
-
-#define FLASHCACHEADDBLACKLIST		FLASHCACHEADDNCPID
-#define FLASHCACHEDELBLACKLIST		FLASHCACHEDELNCPID
-#define FLASHCACHEDELALLBLACKLIST	FLASHCACHEDELNCALL
-
-#define FLASHCACHEADDWHITELIST		_IOW(FLASHCACHE_IOCTL, FLASHCACHEADDWHITELIST_CMD, pid_t)
-#define FLASHCACHEDELWHITELIST		_IOW(FLASHCACHE_IOCTL, FLASHCACHEDELWHITELIST_CMD, pid_t)
-#define FLASHCACHEDELALLWHITELIST	_IOW(FLASHCACHE_IOCTL, FLASHCACHEDELWHITELISTALL_CMD, pid_t)
-
-#endif /* FLASHCACHE_IOCTL_H */
-#endif /* OS_LINUX */

tools/db_bench_tool.cc

@@ -48,7 +48,6 @@
 #include "rocksdb/slice.h"
 #include "rocksdb/slice_transform.h"
 #include "rocksdb/utilities/env_registry.h"
-#include "rocksdb/utilities/flashcache.h"
 #include "rocksdb/utilities/optimistic_transaction_db.h"
 #include "rocksdb/utilities/options_util.h"
 #include "rocksdb/utilities/sim_cache.h"
@@ -573,8 +572,6 @@ DEFINE_string(
     "\t--row_cache_size\n"
     "\t--row_cache_numshardbits\n"
     "\t--enable_io_prio\n"
-    "\t--disable_flashcache_for_background_threads\n"
-    "\t--flashcache_dev\n"
     "\t--dump_malloc_stats\n"
     "\t--num_multi_db\n");
 #endif  // ROCKSDB_LITE
@@ -769,11 +766,6 @@ DEFINE_string(compaction_fadvice, "NORMAL",
 static auto FLAGS_compaction_fadvice_e =
   rocksdb::Options().access_hint_on_compaction_start;
 
-DEFINE_bool(disable_flashcache_for_background_threads, false,
-            "Disable flashcache for background threads");
-
-DEFINE_string(flashcache_dev, "", "Path to flashcache device");
-
 DEFINE_bool(use_tailing_iterator, false,
             "Use tailing iterator to access a series of keys instead of get");
 
@@ -1739,7 +1731,6 @@ class Benchmark {
   int64_t readwrites_;
   int64_t merge_keys_;
   bool report_file_operations_;
-  int cachedev_fd_;
 
   bool SanityCheck() {
     if (FLAGS_compression_ratio > 1) {
@@ -1995,8 +1986,7 @@ class Benchmark {
                 ? FLAGS_num
                 : ((FLAGS_writes > FLAGS_reads) ? FLAGS_writes : FLAGS_reads)),
         merge_keys_(FLAGS_merge_keys < 0 ? FLAGS_num : FLAGS_merge_keys),
-        report_file_operations_(FLAGS_report_file_operations),
-        cachedev_fd_(-1) {
+        report_file_operations_(FLAGS_report_file_operations) {
     // use simcache instead of cache
     if (FLAGS_simcache_size >= 0) {
       if (FLAGS_cache_numshardbits >= 1) {
@@ -2055,11 +2045,6 @@ class Benchmark {
       // this will leak, but we're shutting down so nobody cares
       cache_->DisownData();
     }
-    if (FLAGS_disable_flashcache_for_background_threads && cachedev_fd_ != -1) {
-      // Dtor for thiis env should run before cachedev_fd_ is closed
-      flashcache_aware_env_ = nullptr;
-      close(cachedev_fd_);
-    }
   }
 
   Slice AllocateKey(std::unique_ptr<const char[]>* key_guard) {
@@ -2415,7 +2400,6 @@ class Benchmark {
   }
 
  private:
-  std::unique_ptr<Env> flashcache_aware_env_;
   std::shared_ptr<TimestampEmulator> timestamp_emulator_;
 
   struct ThreadArg {
@@ -2994,23 +2978,7 @@ class Benchmark {
       FLAGS_env->LowerThreadPoolIOPriority(Env::LOW);
       FLAGS_env->LowerThreadPoolIOPriority(Env::HIGH);
     }
-    if (FLAGS_disable_flashcache_for_background_threads && cachedev_fd_ == -1) {
-      // Avoid creating the env twice when an use_existing_db is true
-      cachedev_fd_ = open(FLAGS_flashcache_dev.c_str(), O_RDONLY);
-      if (cachedev_fd_ < 0) {
-        fprintf(stderr, "Open flash device failed\n");
-        exit(1);
-      }
-      flashcache_aware_env_ = NewFlashcacheAwareEnv(FLAGS_env, cachedev_fd_);
-      if (flashcache_aware_env_.get() == nullptr) {
-        fprintf(stderr, "Failed to open flashcache device at %s\n",
-                FLAGS_flashcache_dev.c_str());
-        std::abort();
-      }
-      options.env = flashcache_aware_env_.get();
-    } else {
-      options.env = FLAGS_env;
-    }
+    options.env = FLAGS_env;
 
     if (FLAGS_num_multi_db <= 1) {
       OpenDb(options, FLAGS_db, &db_);

util/hash.cc

@@ -13,6 +13,12 @@
 
 namespace rocksdb {
 
+// This function may intentionally do a left shift on a -ve number
+#if defined(__clang__)
+__attribute__((__no_sanitize__("undefined")))
+#elif __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 9)
+__attribute__((__no_sanitize_undefined__))
+#endif
 uint32_t Hash(const char* data, size_t n, uint32_t seed) {
   // Similar to murmur hash
   const uint32_t m = 0xc6a4a793;

util/options_test.cc

@@ -305,8 +305,8 @@ TEST_F(OptionsTest, GetColumnFamilyOptionsFromStringTest) {
 
   // Units (k)
   ASSERT_OK(GetColumnFamilyOptionsFromString(
-      base_cf_opt, "max_write_buffer_number=-15K", &new_cf_opt));
-  ASSERT_EQ(new_cf_opt.max_write_buffer_number, -15 * kilo);
+      base_cf_opt, "max_write_buffer_number=15K", &new_cf_opt));
+  ASSERT_EQ(new_cf_opt.max_write_buffer_number, 15 * kilo);
   // Units (m)
   ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
             "max_write_buffer_number=16m;inplace_update_num_locks=17M",

util/statistics.cc

@@ -51,9 +51,11 @@ uint64_t StatisticsImpl::getTickerCount(uint32_t tickerType) const {
 
 std::unique_ptr<HistogramImpl>
 StatisticsImpl::HistogramInfo::getMergedHistogram() const {
-  MutexLock lock(&merge_lock);
   std::unique_ptr<HistogramImpl> res_hist(new HistogramImpl());
-  res_hist->Merge(merged_hist);
+  {
+    MutexLock lock(&merge_lock);
+    res_hist->Merge(merged_hist);
+  }
   thread_value->Fold(
       [](void* curr_ptr, void* res) {
         auto tmp_res_hist = static_cast<HistogramImpl*>(res);

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/col_buf_encoder.cc

@@ -46,6 +46,11 @@ ColBufEncoder *ColBufEncoder::NewColBufEncoder(
   return nullptr;
 }
 
+#if defined(__clang__)
+__attribute__((__no_sanitize__("undefined")))
+#elif __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 9)
+__attribute__((__no_sanitize_undefined__))
+#endif
 size_t FixedLengthColBufEncoder::Append(const char *buf) {
   if (nullable_) {
     if (buf == nullptr) {

utilities/flashcache/flashcache.cc

@@ -1,136 +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 "utilities/flashcache/flashcache.h"
-
-#include "rocksdb/utilities/flashcache.h"
-
-#ifdef OS_LINUX
-#include <fcntl.h>
-#include <sys/ioctl.h>
-#include <sys/stat.h>
-#include <sys/syscall.h>
-#include <unistd.h>
-
-#include "third-party/flashcache/flashcache_ioctl.h"
-#endif
-
-namespace rocksdb {
-
-#if !defined(ROCKSDB_LITE) && defined(OS_LINUX)
-// Most of the code that handles flashcache is copied from websql's branch of
-// mysql-5.6
-class FlashcacheAwareEnv : public EnvWrapper {
- public:
-  FlashcacheAwareEnv(Env* base, int cachedev_fd)
-      : EnvWrapper(base), cachedev_fd_(cachedev_fd) {
-    pid_t pid = getpid();
-    /* cleanup previous whitelistings */
-    if (ioctl(cachedev_fd_, FLASHCACHEDELALLWHITELIST, &pid) < 0) {
-      cachedev_fd_ = -1;
-      fprintf(stderr, "ioctl del-all-whitelist for flashcache failed\n");
-      return;
-    }
-    if (ioctl(cachedev_fd_, FLASHCACHEADDWHITELIST, &pid) < 0) {
-      fprintf(stderr, "ioctl add-whitelist for flashcache failed\n");
-    }
-  }
-
-  ~FlashcacheAwareEnv() {
-    // cachedev_fd_ is -1 if it's unitialized
-    if (cachedev_fd_ != -1) {
-      pid_t pid = getpid();
-      if (ioctl(cachedev_fd_, FLASHCACHEDELWHITELIST, &pid) < 0) {
-        fprintf(stderr, "ioctl del-whitelist for flashcache failed\n");
-      }
-    }
-  }
-
-  static int BlacklistCurrentThread(int cachedev_fd) {
-    pid_t pid = static_cast<pid_t>(syscall(SYS_gettid));
-    return ioctl(cachedev_fd, FLASHCACHEADDNCPID, &pid);
-  }
-
-  static int WhitelistCurrentThread(int cachedev_fd) {
-    pid_t pid = static_cast<pid_t>(syscall(SYS_gettid));
-    return ioctl(cachedev_fd, FLASHCACHEDELNCPID, &pid);
-  }
-
-  int GetFlashCacheFileDescriptor() { return cachedev_fd_; }
-
-  struct Arg {
-    Arg(void (*f)(void* arg), void* a, int _cachedev_fd)
-        : original_function_(f), original_arg_(a), cachedev_fd(_cachedev_fd) {}
-
-    void (*original_function_)(void* arg);
-    void* original_arg_;
-    int cachedev_fd;
-  };
-
-  static void BgThreadWrapper(void* a) {
-    Arg* arg = reinterpret_cast<Arg*>(a);
-    if (arg->cachedev_fd != -1) {
-      if (BlacklistCurrentThread(arg->cachedev_fd) < 0) {
-        fprintf(stderr, "ioctl add-nc-pid for flashcache failed\n");
-      }
-    }
-    arg->original_function_(arg->original_arg_);
-    if (arg->cachedev_fd != -1) {
-      if (WhitelistCurrentThread(arg->cachedev_fd) < 0) {
-        fprintf(stderr, "ioctl del-nc-pid for flashcache failed\n");
-      }
-    }
-    delete arg;
-  }
-
-  int UnSchedule(void* arg, Priority pri) override {
-    // no unschedule for you
-    return 0;
-  }
-
-  void Schedule(void (*f)(void* arg), void* a, Priority pri,
-                void* tag = nullptr, void (*u)(void* arg) = 0) override {
-    EnvWrapper::Schedule(&BgThreadWrapper, new Arg(f, a, cachedev_fd_), pri,
-                         tag);
-  }
-
- private:
-  int cachedev_fd_;
-};
-
-std::unique_ptr<Env> NewFlashcacheAwareEnv(Env* base,
-                                           const int cachedev_fd) {
-  std::unique_ptr<Env> ret(new FlashcacheAwareEnv(base, cachedev_fd));
-  return ret;
-}
-
-int FlashcacheBlacklistCurrentThread(Env* flashcache_aware_env) {
-  int fd = dynamic_cast<FlashcacheAwareEnv*>(flashcache_aware_env)
-               ->GetFlashCacheFileDescriptor();
-  if (fd == -1) {
-    return -1;
-  }
-  return FlashcacheAwareEnv::BlacklistCurrentThread(fd);
-}
-int FlashcacheWhitelistCurrentThread(Env* flashcache_aware_env) {
-  int fd = dynamic_cast<FlashcacheAwareEnv*>(flashcache_aware_env)
-               ->GetFlashCacheFileDescriptor();
-  if (fd == -1) {
-    return -1;
-  }
-  return FlashcacheAwareEnv::WhitelistCurrentThread(fd);
-}
-
-#else   // !defined(ROCKSDB_LITE) && defined(OS_LINUX)
-std::unique_ptr<Env> NewFlashcacheAwareEnv(Env* base,
-                                           const int cachedev_fd) {
-  return nullptr;
-}
-int FlashcacheBlacklistCurrentThread(Env* flashcache_aware_env) { return -1; }
-int FlashcacheWhitelistCurrentThread(Env* flashcache_aware_env) { return -1; }
-
-#endif  // !defined(ROCKSDB_LITE) && defined(OS_LINUX)
-
-}  // namespace rocksdb

utilities/flashcache/flashcache.h

@@ -1,18 +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.
-
-#pragma once
-
-#include <string>
-#include "rocksdb/env.h"
-
-namespace rocksdb {
-
-// This is internal API that will make hacking on flashcache easier. Not sure if
-// we need to expose this to public users, probably not
-extern int FlashcacheBlacklistCurrentThread(Env* flashcache_aware_env);
-extern int FlashcacheWhitelistCurrentThread(Env* flashcache_aware_env);
-
-}  // namespace rocksdb

utilities/option_change_migration/option_change_migration.cc

@@ -46,6 +46,7 @@ Status CompactToLevel(const Options& options, const std::string& dbname,
     // only one level. In this case, compacting to one file is also
     // optimal.
     no_compact_opts.target_file_size_base = 999999999999999;
+    no_compact_opts.max_compaction_bytes = 999999999999999;
   }
   Status s = OpenDb(no_compact_opts, dbname, &db);
   if (!s.ok()) {
@@ -54,6 +55,9 @@ Status CompactToLevel(const Options& options, const std::string& dbname,
   CompactRangeOptions cro;
   cro.change_level = true;
   cro.target_level = dest_level;
+  if (dest_level == 0) {
+    cro.bottommost_level_compaction = BottommostLevelCompaction::kForce;
+  }
   db->CompactRange(cro, nullptr, nullptr);
 
   if (need_reopen) {
@@ -72,7 +76,8 @@ Status CompactToLevel(const Options& options, const std::string& dbname,
 
 Status MigrateToUniversal(std::string dbname, const Options& old_opts,
                           const Options& new_opts) {
-  if (old_opts.num_levels <= new_opts.num_levels) {
+  if (old_opts.num_levels <= new_opts.num_levels ||
+      old_opts.compaction_style == CompactionStyle::kCompactionStyleFIFO) {
     return Status::OK();
   } else {
     bool need_compact = false;
@@ -132,11 +137,18 @@ Status MigrateToLevelBase(std::string dbname, const Options& old_opts,
 
 Status OptionChangeMigration(std::string dbname, const Options& old_opts,
                              const Options& new_opts) {
-  if (new_opts.compaction_style == CompactionStyle::kCompactionStyleUniversal) {
+  if (old_opts.compaction_style == CompactionStyle::kCompactionStyleFIFO) {
+    // LSM generated by FIFO compation can be opened by any compaction.
+    return Status::OK();
+  } else if (new_opts.compaction_style ==
+             CompactionStyle::kCompactionStyleUniversal) {
     return MigrateToUniversal(dbname, old_opts, new_opts);
   } else if (new_opts.compaction_style ==
              CompactionStyle::kCompactionStyleLevel) {
     return MigrateToLevelBase(dbname, old_opts, new_opts);
+  } else if (new_opts.compaction_style ==
+             CompactionStyle::kCompactionStyleFIFO) {
+    return CompactToLevel(old_opts, dbname, 0, true);
   } else {
     return Status::NotSupported(
         "Do not how to migrate to this compaction style");

utilities/option_change_migration/option_change_migration_test.cc

@@ -13,19 +13,19 @@
 #include "port/stack_trace.h"
 namespace rocksdb {
 
-class DBOptionChangeMigrationTest
+class DBOptionChangeMigrationTests
     : public DBTestBase,
       public testing::WithParamInterface<
-          std::tuple<int, bool, bool, int, bool, bool>> {
+          std::tuple<int, int, bool, int, int, bool>> {
  public:
-  DBOptionChangeMigrationTest()
+  DBOptionChangeMigrationTests()
       : DBTestBase("/db_option_change_migration_test") {
     level1_ = std::get<0>(GetParam());
-    is_universal1_ = std::get<1>(GetParam());
+    compaction_style1_ = std::get<1>(GetParam());
     is_dynamic1_ = std::get<2>(GetParam());
 
     level2_ = std::get<3>(GetParam());
-    is_universal2_ = std::get<4>(GetParam());
+    compaction_style2_ = std::get<4>(GetParam());
     is_dynamic2_ = std::get<5>(GetParam());
   }
 
@@ -34,23 +34,23 @@ class DBOptionChangeMigrationTest
   static void TearDownTestCase() {}
 
   int level1_;
-  bool is_universal1_;
+  int compaction_style1_;
   bool is_dynamic1_;
 
   int level2_;
-  bool is_universal2_;
+  int compaction_style2_;
   bool is_dynamic2_;
 };
 
 #ifndef ROCKSDB_LITE
-TEST_P(DBOptionChangeMigrationTest, Migrate1) {
+TEST_P(DBOptionChangeMigrationTests, Migrate1) {
   Options old_options = CurrentOptions();
-  if (is_universal1_) {
-    old_options.compaction_style = CompactionStyle::kCompactionStyleUniversal;
-  } else {
-    old_options.compaction_style = CompactionStyle::kCompactionStyleLevel;
+  old_options.compaction_style =
+      static_cast<CompactionStyle>(compaction_style1_);
+  if (old_options.compaction_style == CompactionStyle::kCompactionStyleLevel) {
     old_options.level_compaction_dynamic_level_bytes = is_dynamic1_;
   }
+
   old_options.level0_file_num_compaction_trigger = 3;
   old_options.write_buffer_size = 64 * 1024;
   old_options.target_file_size_base = 128 * 1024;
@@ -83,10 +83,9 @@ TEST_P(DBOptionChangeMigrationTest, Migrate1) {
   Close();
 
   Options new_options = old_options;
-  if (is_universal2_) {
-    new_options.compaction_style = CompactionStyle::kCompactionStyleUniversal;
-  } else {
-    new_options.compaction_style = CompactionStyle::kCompactionStyleLevel;
+  new_options.compaction_style =
+      static_cast<CompactionStyle>(compaction_style2_);
+  if (new_options.compaction_style == CompactionStyle::kCompactionStyleLevel) {
     new_options.level_compaction_dynamic_level_bytes = is_dynamic2_;
   }
   new_options.target_file_size_base = 256 * 1024;
@@ -113,12 +112,11 @@ TEST_P(DBOptionChangeMigrationTest, Migrate1) {
   }
 }
 
-TEST_P(DBOptionChangeMigrationTest, Migrate2) {
+TEST_P(DBOptionChangeMigrationTests, Migrate2) {
   Options old_options = CurrentOptions();
-  if (is_universal2_) {
-    old_options.compaction_style = CompactionStyle::kCompactionStyleUniversal;
-  } else {
-    old_options.compaction_style = CompactionStyle::kCompactionStyleLevel;
+  old_options.compaction_style =
+      static_cast<CompactionStyle>(compaction_style2_);
+  if (old_options.compaction_style == CompactionStyle::kCompactionStyleLevel) {
     old_options.level_compaction_dynamic_level_bytes = is_dynamic2_;
   }
   old_options.level0_file_num_compaction_trigger = 3;
@@ -154,10 +152,158 @@ TEST_P(DBOptionChangeMigrationTest, Migrate2) {
   Close();
 
   Options new_options = old_options;
-  if (is_universal1_) {
-    new_options.compaction_style = CompactionStyle::kCompactionStyleUniversal;
-  } else {
-    new_options.compaction_style = CompactionStyle::kCompactionStyleLevel;
+  new_options.compaction_style =
+      static_cast<CompactionStyle>(compaction_style1_);
+  if (new_options.compaction_style == CompactionStyle::kCompactionStyleLevel) {
+    new_options.level_compaction_dynamic_level_bytes = is_dynamic1_;
+  }
+  new_options.target_file_size_base = 256 * 1024;
+  new_options.num_levels = level1_;
+  new_options.max_bytes_for_level_base = 150 * 1024;
+  new_options.max_bytes_for_level_multiplier = 4;
+  ASSERT_OK(OptionChangeMigration(dbname_, old_options, new_options));
+  Reopen(new_options);
+  // Wait for compaction to finish and make sure it can reopen
+  dbfull()->TEST_WaitForFlushMemTable();
+  dbfull()->TEST_WaitForCompact();
+  Reopen(new_options);
+
+  {
+    std::unique_ptr<Iterator> it(db_->NewIterator(ReadOptions()));
+    it->SeekToFirst();
+    for (std::string key : keys) {
+      ASSERT_TRUE(it->Valid());
+      ASSERT_EQ(key, it->key().ToString());
+      it->Next();
+    }
+    ASSERT_TRUE(!it->Valid());
+  }
+}
+
+TEST_P(DBOptionChangeMigrationTests, Migrate3) {
+  Options old_options = CurrentOptions();
+  old_options.compaction_style =
+      static_cast<CompactionStyle>(compaction_style1_);
+  if (old_options.compaction_style == CompactionStyle::kCompactionStyleLevel) {
+    old_options.level_compaction_dynamic_level_bytes = is_dynamic1_;
+  }
+
+  old_options.level0_file_num_compaction_trigger = 3;
+  old_options.write_buffer_size = 64 * 1024;
+  old_options.target_file_size_base = 128 * 1024;
+  // Make level target of L1, L2 to be 200KB and 600KB
+  old_options.num_levels = level1_;
+  old_options.max_bytes_for_level_multiplier = 3;
+  old_options.max_bytes_for_level_base = 200 * 1024;
+
+  Reopen(old_options);
+  Random rnd(301);
+  for (int num = 0; num < 20; num++) {
+    for (int i = 0; i < 50; i++) {
+      ASSERT_OK(Put(Key(num * 100 + i), RandomString(&rnd, 900)));
+    }
+    Flush();
+    dbfull()->TEST_WaitForCompact();
+    if (num == 9) {
+      // Issue a full compaction to generate some zero-out files
+      CompactRangeOptions cro;
+      cro.bottommost_level_compaction = BottommostLevelCompaction::kForce;
+      dbfull()->CompactRange(cro, nullptr, nullptr);
+    }
+  }
+  dbfull()->TEST_WaitForFlushMemTable();
+  dbfull()->TEST_WaitForCompact();
+
+  // Will make sure exactly those keys are in the DB after migration.
+  std::set<std::string> keys;
+  {
+    std::unique_ptr<Iterator> it(db_->NewIterator(ReadOptions()));
+    it->SeekToFirst();
+    for (; it->Valid(); it->Next()) {
+      keys.insert(it->key().ToString());
+    }
+  }
+  Close();
+
+  Options new_options = old_options;
+  new_options.compaction_style =
+      static_cast<CompactionStyle>(compaction_style2_);
+  if (new_options.compaction_style == CompactionStyle::kCompactionStyleLevel) {
+    new_options.level_compaction_dynamic_level_bytes = is_dynamic2_;
+  }
+  new_options.target_file_size_base = 256 * 1024;
+  new_options.num_levels = level2_;
+  new_options.max_bytes_for_level_base = 150 * 1024;
+  new_options.max_bytes_for_level_multiplier = 4;
+  ASSERT_OK(OptionChangeMigration(dbname_, old_options, new_options));
+  Reopen(new_options);
+
+  // Wait for compaction to finish and make sure it can reopen
+  dbfull()->TEST_WaitForFlushMemTable();
+  dbfull()->TEST_WaitForCompact();
+  Reopen(new_options);
+
+  {
+    std::unique_ptr<Iterator> it(db_->NewIterator(ReadOptions()));
+    it->SeekToFirst();
+    for (std::string key : keys) {
+      ASSERT_TRUE(it->Valid());
+      ASSERT_EQ(key, it->key().ToString());
+      it->Next();
+    }
+    ASSERT_TRUE(!it->Valid());
+  }
+}
+
+TEST_P(DBOptionChangeMigrationTests, Migrate4) {
+  Options old_options = CurrentOptions();
+  old_options.compaction_style =
+      static_cast<CompactionStyle>(compaction_style2_);
+  if (old_options.compaction_style == CompactionStyle::kCompactionStyleLevel) {
+    old_options.level_compaction_dynamic_level_bytes = is_dynamic2_;
+  }
+  old_options.level0_file_num_compaction_trigger = 3;
+  old_options.write_buffer_size = 64 * 1024;
+  old_options.target_file_size_base = 128 * 1024;
+  // Make level target of L1, L2 to be 200KB and 600KB
+  old_options.num_levels = level2_;
+  old_options.max_bytes_for_level_multiplier = 3;
+  old_options.max_bytes_for_level_base = 200 * 1024;
+
+  Reopen(old_options);
+  Random rnd(301);
+  for (int num = 0; num < 20; num++) {
+    for (int i = 0; i < 50; i++) {
+      ASSERT_OK(Put(Key(num * 100 + i), RandomString(&rnd, 900)));
+    }
+    Flush();
+    dbfull()->TEST_WaitForCompact();
+    if (num == 9) {
+      // Issue a full compaction to generate some zero-out files
+      CompactRangeOptions cro;
+      cro.bottommost_level_compaction = BottommostLevelCompaction::kForce;
+      dbfull()->CompactRange(cro, nullptr, nullptr);
+    }
+  }
+  dbfull()->TEST_WaitForFlushMemTable();
+  dbfull()->TEST_WaitForCompact();
+
+  // Will make sure exactly those keys are in the DB after migration.
+  std::set<std::string> keys;
+  {
+    std::unique_ptr<Iterator> it(db_->NewIterator(ReadOptions()));
+    it->SeekToFirst();
+    for (; it->Valid(); it->Next()) {
+      keys.insert(it->key().ToString());
+    }
+  }
+
+  Close();
+
+  Options new_options = old_options;
+  new_options.compaction_style =
+      static_cast<CompactionStyle>(compaction_style1_);
+  if (new_options.compaction_style == CompactionStyle::kCompactionStyleLevel) {
     new_options.level_compaction_dynamic_level_bytes = is_dynamic1_;
   }
   new_options.target_file_size_base = 256 * 1024;
@@ -184,18 +330,90 @@ TEST_P(DBOptionChangeMigrationTest, Migrate2) {
 }
 
 INSTANTIATE_TEST_CASE_P(
-    DBOptionChangeMigrationTest, DBOptionChangeMigrationTest,
-    ::testing::Values(std::make_tuple(3, false, false, 4, false, false),
-                      std::make_tuple(3, false, true, 4, false, true),
-                      std::make_tuple(3, false, true, 4, false, false),
-                      std::make_tuple(3, false, false, 4, false, true),
-                      std::make_tuple(3, true, false, 4, true, false),
-                      std::make_tuple(1, true, false, 4, true, false),
-                      std::make_tuple(3, false, false, 4, true, false),
-                      std::make_tuple(3, false, false, 1, true, false),
-                      std::make_tuple(3, false, true, 4, true, false),
-                      std::make_tuple(3, false, true, 1, true, false),
-                      std::make_tuple(1, true, false, 4, false, false)));
+    DBOptionChangeMigrationTests, DBOptionChangeMigrationTests,
+    ::testing::Values(std::make_tuple(3, 0, false, 4, 0, false),
+                      std::make_tuple(3, 0, true, 4, 0, true),
+                      std::make_tuple(3, 0, true, 4, 0, false),
+                      std::make_tuple(3, 0, false, 4, 0, true),
+                      std::make_tuple(3, 1, false, 4, 1, false),
+                      std::make_tuple(1, 1, false, 4, 1, false),
+                      std::make_tuple(3, 0, false, 4, 1, false),
+                      std::make_tuple(3, 0, false, 1, 1, false),
+                      std::make_tuple(3, 0, true, 4, 1, false),
+                      std::make_tuple(3, 0, true, 1, 1, false),
+                      std::make_tuple(1, 1, false, 4, 0, false),
+                      std::make_tuple(4, 0, false, 1, 2, false),
+                      std::make_tuple(3, 0, true, 2, 2, false),
+                      std::make_tuple(3, 1, false, 3, 2, false),
+                      std::make_tuple(1, 1, false, 4, 2, false)));
+
+class DBOptionChangeMigrationTest : public DBTestBase {
+ public:
+  DBOptionChangeMigrationTest()
+      : DBTestBase("/db_option_change_migration_test2") {}
+};
+
+TEST_F(DBOptionChangeMigrationTest, CompactedSrcToUniversal) {
+  Options old_options = CurrentOptions();
+  old_options.compaction_style = CompactionStyle::kCompactionStyleLevel;
+  old_options.max_compaction_bytes = 200 * 1024;
+  old_options.level_compaction_dynamic_level_bytes = false;
+  old_options.level0_file_num_compaction_trigger = 3;
+  old_options.write_buffer_size = 64 * 1024;
+  old_options.target_file_size_base = 128 * 1024;
+  // Make level target of L1, L2 to be 200KB and 600KB
+  old_options.num_levels = 4;
+  old_options.max_bytes_for_level_multiplier = 3;
+  old_options.max_bytes_for_level_base = 200 * 1024;
+
+  Reopen(old_options);
+  Random rnd(301);
+  for (int num = 0; num < 20; num++) {
+    for (int i = 0; i < 50; i++) {
+      ASSERT_OK(Put(Key(num * 100 + i), RandomString(&rnd, 900)));
+    }
+  }
+  Flush();
+  CompactRangeOptions cro;
+  cro.bottommost_level_compaction = BottommostLevelCompaction::kForce;
+  dbfull()->CompactRange(cro, nullptr, nullptr);
+
+  // Will make sure exactly those keys are in the DB after migration.
+  std::set<std::string> keys;
+  {
+    std::unique_ptr<Iterator> it(db_->NewIterator(ReadOptions()));
+    it->SeekToFirst();
+    for (; it->Valid(); it->Next()) {
+      keys.insert(it->key().ToString());
+    }
+  }
+
+  Close();
+
+  Options new_options = old_options;
+  new_options.compaction_style = CompactionStyle::kCompactionStyleUniversal;
+  new_options.target_file_size_base = 256 * 1024;
+  new_options.num_levels = 1;
+  new_options.max_bytes_for_level_base = 150 * 1024;
+  new_options.max_bytes_for_level_multiplier = 4;
+  ASSERT_OK(OptionChangeMigration(dbname_, old_options, new_options));
+  Reopen(new_options);
+  // Wait for compaction to finish and make sure it can reopen
+  dbfull()->TEST_WaitForFlushMemTable();
+  dbfull()->TEST_WaitForCompact();
+  Reopen(new_options);
+
+  {
+    std::unique_ptr<Iterator> it(db_->NewIterator(ReadOptions()));
+    it->SeekToFirst();
+    for (std::string key : keys) {
+      ASSERT_TRUE(it->Valid());
+      ASSERT_EQ(key, it->key().ToString());
+      it->Next();
+    }
+    ASSERT_TRUE(!it->Valid());
+  }
+}
 
 #endif  // ROCKSDB_LITE
 }  // namespace rocksdb

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"