Update HISTORY.md for PR 9273 (#9282)

Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/9282

Reviewed By: akankshamahajan15

Differential Revision: D33027844

Pulled By: ltamasi

fbshipit-source-id: 7540d36010414311bc39610fff92a6498be1570c

HISTORY.md

@@ -1,4 +1,17 @@
 # Rocksdb Change Log
+## 6.27.3 (2021-12-10)
+### Bug Fixes
+* Fixed a bug in TableOptions.prepopulate_block_cache which causes segmentation fault when used with TableOptions.partition_filters = true and TableOptions.cache_index_and_filter_blocks = true.
+* Fixed a bug affecting custom memtable factories which are not registered with the `ObjectRegistry`. The bug could result in failure to save the OPTIONS file.
+
+## 6.27.2 (2021-12-01)
+### Bug Fixes
+* Fixed a bug in rocksdb automatic implicit prefetching which got broken because of new feature adaptive_readahead and internal prefetching got disabled when iterator moves from one file to next.
+
+## 6.27.1 (2021-11-29)
+### Bug Fixes
+* Fixed a bug that could, with WAL enabled, cause backups, checkpoints, and `GetSortedWalFiles()` to fail randomly with an error like `IO error: 001234.log: No such file or directory`
+
 ## 6.27.0 (2021-11-19)
 ### New Features
 * Added new ChecksumType kXXH3 which is faster than kCRC32c on almost all x86\_64 hardware.

build_tools/gnu_parallel

@@ -1842,7 +1842,6 @@ sub start_another_job {
 }
 
 $opt::min_progress_interval = 0;
-$opt::progress_sep = "\r";
 
 sub init_progress {
     # Uses:
@@ -1852,7 +1851,6 @@ sub init_progress {
     $|=1;
     if (not $Global::is_terminal) {
       $opt::min_progress_interval = 30;
-      $opt::progress_sep = "\n";
     }
     if($opt::bar) {
 	return("","");
@@ -1878,7 +1876,9 @@ sub drain_job_queue {
     }
     my $last_header="";
     my $sleep = 0.2;
+    my $last_left = 1000000000;
     my $last_progress_time = 0;
+    my $ps_reported = 0;
     do {
         while($Global::total_running > 0) {
             debug($Global::total_running, "==", scalar
@@ -1889,14 +1889,37 @@ sub drain_job_queue {
 		    close $job->fh(0,"w");
 		}
 	    }
-            if($opt::progress and (time() - $last_progress_time) >= $opt::min_progress_interval) {
-                $last_progress_time = time();
+            # When not connected to terminal, assume CI (e.g. CircleCI). In
+            # that case we want occasional progress output to prevent abort
+            # due to timeout with no output, but we also need to stop sending
+            # progress output if there has been no actual progress, so that
+            # the job can time out appropriately (CirecleCI: 10m) in case of
+            # a hung test. But without special output, it is extremely
+            # annoying to diagnose which test is hung, so we add that using
+            # `ps` below.
+            if($opt::progress and
+               ($Global::is_terminal or (time() - $last_progress_time) >= 30)) {
                 my %progress = progress();
                 if($last_header ne $progress{'header'}) {
                     print $Global::original_stderr "\n", $progress{'header'}, "\n";
                     $last_header = $progress{'header'};
                 }
-                print $Global::original_stderr $opt::progress_sep,$progress{'status'};
+                if ($Global::is_terminal) {
+                    print $Global::original_stderr "\r",$progress{'status'};
+                }
+                if ($last_left > $Global::left) {
+                    if (not $Global::is_terminal) {
+                        print $Global::original_stderr $progress{'status'},"\n";
+                    }
+                    $last_progress_time = time();
+                    $ps_reported = 0;
+                } elsif (not $ps_reported and (time() - $last_progress_time) >= 60) {
+                    # No progress in at least 60 seconds: run ps
+                    print $Global::original_stderr "\n";
+                    system("ps", "-wf");
+                    $ps_reported = 1;
+                }
+                $last_left = $Global::left;
                 flush $Global::original_stderr;
             }
 	    if($Global::total_running < $Global::max_jobs_running
@@ -1968,6 +1991,7 @@ sub progress {
         compute_eta();
       $eta = sprintf("ETA: %ds Left: %d AVG: %.2fs  ",
         $this_eta, $left, $avgtime);
+      $Global::left = $left;
     }
     my $termcols = terminal_columns();
     my @workers = sort keys %Global::host;

db/column_family.cc

@@ -75,11 +75,6 @@ ColumnFamilyHandleImpl::~ColumnFamilyHandleImpl() {
       bool defer_purge =
           db_->immutable_db_options().avoid_unnecessary_blocking_io;
       db_->PurgeObsoleteFiles(job_context, defer_purge);
-      if (defer_purge) {
-        mutex_->Lock();
-        db_->SchedulePurge();
-        mutex_->Unlock();
-      }
     }
     job_context.Clean();
   }

db/db_block_cache_test.cc

@@ -650,13 +650,32 @@ TEST_F(DBBlockCacheTest, WarmCacheWithDataBlocksDuringFlush) {
 }
 
 // This test cache data, index and filter blocks during flush.
-TEST_F(DBBlockCacheTest, WarmCacheWithBlocksDuringFlush) {
+class DBBlockCacheTest1 : public DBTestBase,
+                          public ::testing::WithParamInterface<bool> {
+ public:
+  const size_t kNumBlocks = 10;
+  const size_t kValueSize = 100;
+  DBBlockCacheTest1() : DBTestBase("db_block_cache_test1", true) {}
+};
+
+INSTANTIATE_TEST_CASE_P(DBBlockCacheTest1, DBBlockCacheTest1,
+                        ::testing::Bool());
+
+TEST_P(DBBlockCacheTest1, WarmCacheWithBlocksDuringFlush) {
   Options options = CurrentOptions();
   options.create_if_missing = true;
   options.statistics = ROCKSDB_NAMESPACE::CreateDBStatistics();
 
   BlockBasedTableOptions table_options;
   table_options.block_cache = NewLRUCache(1 << 25, 0, false);
+
+  bool use_partition = GetParam();
+  if (use_partition) {
+    table_options.partition_filters = true;
+    table_options.index_type =
+        BlockBasedTableOptions::IndexType::kTwoLevelIndexSearch;
+  }
+
   table_options.cache_index_and_filter_blocks = true;
   table_options.prepopulate_block_cache =
       BlockBasedTableOptions::PrepopulateBlockCache::kFlushOnly;
@@ -669,9 +688,15 @@ TEST_F(DBBlockCacheTest, WarmCacheWithBlocksDuringFlush) {
     ASSERT_OK(Put(ToString(i), value));
     ASSERT_OK(Flush());
     ASSERT_EQ(i, options.statistics->getTickerCount(BLOCK_CACHE_DATA_ADD));
+    if (use_partition) {
+      ASSERT_EQ(2 * i,
+                options.statistics->getTickerCount(BLOCK_CACHE_INDEX_ADD));
+      ASSERT_EQ(2 * i,
+                options.statistics->getTickerCount(BLOCK_CACHE_FILTER_ADD));
+    } else {
       ASSERT_EQ(i, options.statistics->getTickerCount(BLOCK_CACHE_INDEX_ADD));
       ASSERT_EQ(i, options.statistics->getTickerCount(BLOCK_CACHE_FILTER_ADD));
-
+    }
     ASSERT_EQ(value, Get(ToString(i)));
 
     ASSERT_EQ(0, options.statistics->getTickerCount(BLOCK_CACHE_DATA_MISS));
@@ -679,11 +704,16 @@ TEST_F(DBBlockCacheTest, WarmCacheWithBlocksDuringFlush) {
 
     ASSERT_EQ(0, options.statistics->getTickerCount(BLOCK_CACHE_INDEX_MISS));
     ASSERT_EQ(i * 3, options.statistics->getTickerCount(BLOCK_CACHE_INDEX_HIT));
-
-    ASSERT_EQ(0, options.statistics->getTickerCount(BLOCK_CACHE_FILTER_MISS));
+    if (use_partition) {
+      ASSERT_EQ(i * 3,
+                options.statistics->getTickerCount(BLOCK_CACHE_FILTER_HIT));
+    } else {
       ASSERT_EQ(i * 2,
                 options.statistics->getTickerCount(BLOCK_CACHE_FILTER_HIT));
     }
+    ASSERT_EQ(0, options.statistics->getTickerCount(BLOCK_CACHE_FILTER_MISS));
+  }
+
   // Verify compaction not counted
   ASSERT_OK(db_->CompactRange(CompactRangeOptions(), /*begin=*/nullptr,
                               /*end=*/nullptr));
@@ -692,10 +722,17 @@ TEST_F(DBBlockCacheTest, WarmCacheWithBlocksDuringFlush) {
   // Index and filter blocks are automatically warmed when the new table file
   // is automatically opened at the end of compaction. This is not easily
   // disabled so results in the new index and filter blocks being warmed.
+  if (use_partition) {
+    EXPECT_EQ(2 * (1 + kNumBlocks),
+              options.statistics->getTickerCount(BLOCK_CACHE_INDEX_ADD));
+    EXPECT_EQ(2 * (1 + kNumBlocks),
+              options.statistics->getTickerCount(BLOCK_CACHE_FILTER_ADD));
+  } else {
     EXPECT_EQ(1 + kNumBlocks,
               options.statistics->getTickerCount(BLOCK_CACHE_INDEX_ADD));
     EXPECT_EQ(1 + kNumBlocks,
               options.statistics->getTickerCount(BLOCK_CACHE_FILTER_ADD));
+  }
 }
 
 TEST_F(DBBlockCacheTest, DynamicallyWarmCacheDuringFlush) {

db/db_filesnapshot.cc

@@ -127,31 +127,30 @@ Status DBImpl::GetLiveFiles(std::vector<std::string>& ret,
 }
 
 Status DBImpl::GetSortedWalFiles(VectorLogPtr& files) {
-  {
   // If caller disabled deletions, this function should return files that are
   // guaranteed not to be deleted until deletions are re-enabled. We need to
   // wait for pending purges to finish since WalManager doesn't know which
   // files are going to be purged. Additional purges won't be scheduled as
   // long as deletions are disabled (so the below loop must terminate).
+  // Also note that we disable deletions anyway to avoid the case where a
+  // file is deleted in the middle of the scan, causing IO error.
+  Status deletions_disabled = DisableFileDeletions();
+  {
     InstrumentedMutexLock l(&mutex_);
-    while (disable_delete_obsolete_files_ > 0 &&
-           (pending_purge_obsolete_files_ > 0 || bg_purge_scheduled_ > 0)) {
+    while (pending_purge_obsolete_files_ > 0 || bg_purge_scheduled_ > 0) {
       bg_cv_.Wait();
     }
   }
 
-  // Disable deletion in order to avoid the case where a file is deleted in
-  // the middle of the process so IO error is returned.
-  Status s = DisableFileDeletions();
-  bool file_deletion_supported = !s.IsNotSupported();
-  if (s.ok() || !file_deletion_supported) {
-    s = wal_manager_.GetSortedWalFiles(files);
-    if (file_deletion_supported) {
-      Status s2 = EnableFileDeletions(false);
-      if (!s2.ok() && s.ok()) {
-        s = s2;
-      }
-    }
+  Status s = wal_manager_.GetSortedWalFiles(files);
+
+  // DisableFileDeletions / EnableFileDeletions not supported in read-only DB
+  if (deletions_disabled.ok()) {
+    Status s2 = EnableFileDeletions(/*force*/ false);
+    assert(s2.ok());
+    s2.PermitUncheckedError();
+  } else {
+    assert(deletions_disabled.IsNotSupported());
   }
 
   return s;

db/db_impl/db_impl.cc

@@ -1546,6 +1546,8 @@ void DBImpl::BackgroundCallPurge() {
     mutex_.Lock();
   }
 
+  assert(bg_purge_scheduled_ > 0);
+
   // Can't use iterator to go over purge_files_ because inside the loop we're
   // unlocking the mutex that protects purge_files_.
   while (!purge_files_.empty()) {
@@ -1613,17 +1615,7 @@ static void CleanupIteratorState(void* arg1, void* /*arg2*/) {
       delete state->super_version;
     }
     if (job_context.HaveSomethingToDelete()) {
-      if (state->background_purge) {
-        // PurgeObsoleteFiles here does not delete files. Instead, it adds the
-        // files to be deleted to a job queue, and deletes it in a separate
-        // background thread.
-        state->db->PurgeObsoleteFiles(job_context, true /* schedule only */);
-        state->mu->Lock();
-        state->db->SchedulePurge();
-        state->mu->Unlock();
-      } else {
-        state->db->PurgeObsoleteFiles(job_context);
-      }
+      state->db->PurgeObsoleteFiles(job_context, state->background_purge);
     }
     job_context.Clean();
   }

db/db_impl/db_impl_files.cc

@@ -56,6 +56,8 @@ Status DBImpl::DisableFileDeletions() {
   return s;
 }
 
+// FIXME: can be inconsistent with DisableFileDeletions in cases like
+// DBImplReadOnly
 Status DBImpl::DisableFileDeletionsWithLock() {
   mutex_.AssertHeld();
   ++disable_delete_obsolete_files_;
@@ -642,6 +644,11 @@ void DBImpl::PurgeObsoleteFiles(JobContext& state, bool schedule_only) {
   InstrumentedMutexLock l(&mutex_);
   --pending_purge_obsolete_files_;
   assert(pending_purge_obsolete_files_ >= 0);
+  if (schedule_only) {
+    // Must change from pending_purge_obsolete_files_ to bg_purge_scheduled_
+    // while holding mutex (for GetSortedWalFiles() etc.)
+    SchedulePurge();
+  }
   if (pending_purge_obsolete_files_ == 0) {
     bg_cv_.SignalAll();
   }
@@ -657,11 +664,6 @@ void DBImpl::DeleteObsoleteFiles() {
   if (job_context.HaveSomethingToDelete()) {
     bool defer_purge = immutable_db_options_.avoid_unnecessary_blocking_io;
     PurgeObsoleteFiles(job_context, defer_purge);
-    if (defer_purge) {
-      mutex_.Lock();
-      SchedulePurge();
-      mutex_.Unlock();
-    }
   }
   job_context.Clean();
   mutex_.Lock();

db/db_options_test.cc

@@ -266,6 +266,50 @@ TEST_F(DBOptionsTest, SetMutableTableOptions) {
   ASSERT_EQ(c_bbto->block_restart_interval, 13);
 }
 
+TEST_F(DBOptionsTest, SetWithCustomMemTableFactory) {
+  class DummySkipListFactory : public SkipListFactory {
+   public:
+    static const char* kClassName() { return "DummySkipListFactory"; }
+    const char* Name() const override { return kClassName(); }
+    explicit DummySkipListFactory() : SkipListFactory(2) {}
+  };
+  {
+    // Verify the DummySkipList cannot be created
+    ConfigOptions config_options;
+    config_options.ignore_unsupported_options = false;
+    std::unique_ptr<MemTableRepFactory> factory;
+    ASSERT_NOK(MemTableRepFactory::CreateFromString(
+        config_options, DummySkipListFactory::kClassName(), &factory));
+  }
+  Options options;
+  options.create_if_missing = true;
+  // Try with fail_if_options_file_error=false/true to update the options
+  for (bool on_error : {false, true}) {
+    options.fail_if_options_file_error = on_error;
+    options.env = env_;
+    options.disable_auto_compactions = false;
+
+    options.memtable_factory.reset(new DummySkipListFactory());
+    Reopen(options);
+
+    ColumnFamilyHandle* cfh = dbfull()->DefaultColumnFamily();
+    ASSERT_OK(
+        dbfull()->SetOptions(cfh, {{"disable_auto_compactions", "true"}}));
+    ColumnFamilyDescriptor cfd;
+    ASSERT_OK(cfh->GetDescriptor(&cfd));
+    ASSERT_STREQ(cfd.options.memtable_factory->Name(),
+                 DummySkipListFactory::kClassName());
+    ColumnFamilyHandle* test = nullptr;
+    ASSERT_OK(dbfull()->CreateColumnFamily(options, "test", &test));
+    ASSERT_OK(test->GetDescriptor(&cfd));
+    ASSERT_STREQ(cfd.options.memtable_factory->Name(),
+                 DummySkipListFactory::kClassName());
+
+    ASSERT_OK(dbfull()->DropColumnFamily(test));
+    delete test;
+  }
+}
+
 TEST_F(DBOptionsTest, SetBytesPerSync) {
   const size_t kValueSize = 1024 * 1024;  // 1MB
   Options options;

db_stress_tool/db_stress_common.h

@@ -269,6 +269,8 @@ DECLARE_uint64(user_timestamp_size);
 DECLARE_string(secondary_cache_uri);
 DECLARE_int32(secondary_cache_fault_one_in);
 
+DECLARE_int32(prepopulate_block_cache);
+
 constexpr long KB = 1024;
 constexpr int kRandomValueMaxFactor = 3;
 constexpr int kValueMaxLen = 100;

db_stress_tool/db_stress_gflags.cc

@@ -860,5 +860,10 @@ DEFINE_int32(injest_error_severity, 1,
              "The severity of the injested IO Error. 1 is soft error (e.g. "
              "retryable error), 2 is fatal error, and the default is "
              "retryable error.");
+DEFINE_int32(prepopulate_block_cache,
+             static_cast<int32_t>(ROCKSDB_NAMESPACE::BlockBasedTableOptions::
+                                      PrepopulateBlockCache::kDisable),
+             "Options related to cache warming (see `enum "
+             "PrepopulateBlockCache` in table.h)");
 
 #endif  // GFLAGS

db_stress_tool/db_stress_test_base.cc

@@ -2226,6 +2226,9 @@ void StressTest::Open() {
         FLAGS_optimize_filters_for_memory;
     block_based_options.index_type =
         static_cast<BlockBasedTableOptions::IndexType>(FLAGS_index_type);
+    block_based_options.prepopulate_block_cache =
+        static_cast<BlockBasedTableOptions::PrepopulateBlockCache>(
+            FLAGS_prepopulate_block_cache);
     options_.table_factory.reset(
         NewBlockBasedTableFactory(block_based_options));
     options_.db_write_buffer_size = FLAGS_db_write_buffer_size;

env/io_posix.cc

@@ -1543,7 +1543,8 @@ PosixDirectory::PosixDirectory(int fd) : fd_(fd) {
 #ifdef OS_LINUX
   struct statfs buf;
   int ret = fstatfs(fd, &buf);
-  is_btrfs_ = (ret == 0 && buf.f_type == BTRFS_SUPER_MAGIC);
+  is_btrfs_ = (ret == 0 && buf.f_type == static_cast<decltype(buf.f_type)>(
+                                             BTRFS_SUPER_MAGIC));
 #endif
 }
 

file/file_prefetch_buffer.cc

@@ -123,6 +123,8 @@ bool FilePrefetchBuffer::TryReadFromCache(const IOOptions& opts,
   //    If readahead is enabled: prefetch the remaining bytes + readahead bytes
   //        and satisfy the request.
   //    If readahead is not enabled: return false.
+  TEST_SYNC_POINT_CALLBACK("FilePrefetchBuffer::TryReadFromCache",
+                           &readahead_size_);
   if (offset + n > buffer_offset_ + buffer_.CurrentSize()) {
     if (readahead_size_ > 0) {
       assert(reader != nullptr);
@@ -161,8 +163,6 @@ bool FilePrefetchBuffer::TryReadFromCache(const IOOptions& opts,
 #endif
         return false;
       }
-      TEST_SYNC_POINT_CALLBACK("FilePrefetchBuffer::TryReadFromCache",
-                               &readahead_size_);
       readahead_size_ = std::min(max_readahead_size_, readahead_size_ * 2);
     } else {
       return false;

file/file_prefetch_buffer.h

@@ -30,6 +30,8 @@ class RandomAccessFileReader;
 class FilePrefetchBuffer {
  public:
   static const int kMinNumFileReadsToStartAutoReadahead = 2;
+  static const size_t kInitAutoReadaheadSize = 8 * 1024;
+
   // Constructor.
   //
   // All arguments are optional.
@@ -57,7 +59,6 @@ class FilePrefetchBuffer {
       : buffer_offset_(0),
         readahead_size_(readahead_size),
         max_readahead_size_(max_readahead_size),
-        initial_readahead_size_(readahead_size),
         min_offset_read_(port::kMaxSizet),
         enable_(enable),
         track_min_offset_(track_min_offset),
@@ -95,6 +96,7 @@ class FilePrefetchBuffer {
   // tracked if track_min_offset = true.
   size_t min_offset_read() const { return min_offset_read_; }
 
+  // Called in case of implicit auto prefetching.
   void UpdateReadPattern(const uint64_t& offset, const size_t& len,
                          bool is_adaptive_readahead = false) {
     if (is_adaptive_readahead) {
@@ -111,9 +113,10 @@ class FilePrefetchBuffer {
     return (prev_len_ == 0 || (prev_offset_ + prev_len_ == offset));
   }
 
+  // Called in case of implicit auto prefetching.
   void ResetValues() {
     num_file_reads_ = 1;
-    readahead_size_ = initial_readahead_size_;
+    readahead_size_ = kInitAutoReadaheadSize;
   }
 
   void GetReadaheadState(ReadaheadFileInfo::ReadaheadInfo* readahead_info) {
@@ -136,8 +139,9 @@ class FilePrefetchBuffer {
       if ((offset + size > buffer_offset_ + buffer_.CurrentSize()) &&
           IsBlockSequential(offset) &&
           (num_file_reads_ + 1 > kMinNumFileReadsToStartAutoReadahead)) {
+        size_t initial_auto_readahead_size = kInitAutoReadaheadSize;
         readahead_size_ =
-            std::max(initial_readahead_size_,
+            std::max(initial_auto_readahead_size,
                      (readahead_size_ >= value ? readahead_size_ - value : 0));
       }
     }
@@ -150,7 +154,6 @@ class FilePrefetchBuffer {
   // FilePrefetchBuffer object won't be created from Iterator flow if
   // max_readahead_size_ = 0.
   size_t max_readahead_size_;
-  size_t initial_readahead_size_;
   // The minimum `offset` ever passed to TryReadFromCache().
   size_t min_offset_read_;
   // if false, TryReadFromCache() always return false, and we only take stats

file/prefetch_test.cc

@@ -670,13 +670,16 @@ TEST_P(PrefetchTest, PrefetchWhenReseekwithCache) {
   Close();
 }
 
-class PrefetchTest1 : public DBTestBase,
-                      public ::testing::WithParamInterface<bool> {
+class PrefetchTest1
+    : public DBTestBase,
+      public ::testing::WithParamInterface<std::tuple<bool, bool>> {
  public:
   PrefetchTest1() : DBTestBase("prefetch_test1", true) {}
 };
 
-INSTANTIATE_TEST_CASE_P(PrefetchTest1, PrefetchTest1, ::testing::Bool());
+INSTANTIATE_TEST_CASE_P(PrefetchTest1, PrefetchTest1,
+                        ::testing::Combine(::testing::Bool(),
+                                           ::testing::Bool()));
 
 #ifndef ROCKSDB_LITE
 TEST_P(PrefetchTest1, DBIterLevelReadAhead) {
@@ -686,12 +689,13 @@ TEST_P(PrefetchTest1, DBIterLevelReadAhead) {
       std::make_shared<MockFS>(env_->GetFileSystem(), false);
   std::unique_ptr<Env> env(new CompositeEnvWrapper(env_, fs));
 
+  bool is_adaptive_readahead = std::get<1>(GetParam());
   Options options = CurrentOptions();
   options.write_buffer_size = 1024;
   options.create_if_missing = true;
   options.compression = kNoCompression;
   options.env = env.get();
-  if (GetParam()) {
+  if (std::get<0>(GetParam())) {
     options.use_direct_reads = true;
     options.use_direct_io_for_flush_and_compaction = true;
   }
@@ -704,7 +708,8 @@ TEST_P(PrefetchTest1, DBIterLevelReadAhead) {
   options.table_factory.reset(NewBlockBasedTableFactory(table_options));
 
   Status s = TryReopen(options);
-  if (GetParam() && (s.IsNotSupported() || s.IsInvalidArgument())) {
+  if (std::get<0>(GetParam()) &&
+      (s.IsNotSupported() || s.IsInvalidArgument())) {
     // If direct IO is not supported, skip the test
     return;
   } else {
@@ -748,12 +753,15 @@ TEST_P(PrefetchTest1, DBIterLevelReadAhead) {
     SyncPoint::GetInstance()->SetCallBack(
         "FilePrefetchBuffer::TryReadFromCache", [&](void* arg) {
           current_readahead_size = *reinterpret_cast<size_t*>(arg);
+          ASSERT_GT(current_readahead_size, 0);
         });
 
     SyncPoint::GetInstance()->EnableProcessing();
 
     ReadOptions ro;
+    if (is_adaptive_readahead) {
       ro.adaptive_readahead = true;
+    }
     auto iter = std::unique_ptr<Iterator>(db_->NewIterator(ro));
     int num_keys = 0;
     for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
@@ -763,14 +771,28 @@ TEST_P(PrefetchTest1, DBIterLevelReadAhead) {
     ASSERT_GT(buff_prefetch_count, 0);
     buff_prefetch_count = 0;
     // For index and data blocks.
+    if (is_adaptive_readahead) {
       ASSERT_EQ(readahead_carry_over_count, 2 * (num_sst_files - 1));
+    } else {
+      ASSERT_EQ(readahead_carry_over_count, 0);
+    }
     SyncPoint::GetInstance()->DisableProcessing();
     SyncPoint::GetInstance()->ClearAllCallBacks();
   }
   Close();
 }
+#endif  //! ROCKSDB_LITE
 
-TEST_P(PrefetchTest1, NonSequentialReads) {
+class PrefetchTest2 : public DBTestBase,
+                      public ::testing::WithParamInterface<bool> {
+ public:
+  PrefetchTest2() : DBTestBase("prefetch_test2", true) {}
+};
+
+INSTANTIATE_TEST_CASE_P(PrefetchTest2, PrefetchTest2, ::testing::Bool());
+
+#ifndef ROCKSDB_LITE
+TEST_P(PrefetchTest2, NonSequentialReads) {
   const int kNumKeys = 1000;
   // Set options
   std::shared_ptr<MockFS> fs =
@@ -856,7 +878,7 @@ TEST_P(PrefetchTest1, NonSequentialReads) {
 }
 #endif  //! ROCKSDB_LITE
 
-TEST_P(PrefetchTest1, DecreaseReadAheadIfInCache) {
+TEST_P(PrefetchTest2, DecreaseReadAheadIfInCache) {
   const int kNumKeys = 2000;
   // Set options
   std::shared_ptr<MockFS> fs =

include/rocksdb/customizable.h

@@ -101,6 +101,20 @@ class Customizable : public Configurable {
     }
   }
 
+  const void* GetOptionsPtr(const std::string& name) const override {
+    const void* ptr = Configurable::GetOptionsPtr(name);
+    if (ptr != nullptr) {
+      return ptr;
+    } else {
+      const auto inner = Inner();
+      if (inner != nullptr) {
+        return inner->GetOptionsPtr(name);
+      } else {
+        return nullptr;
+      }
+    }
+  }
+
   // Returns the named instance of the Customizable as a T*, or nullptr if not
   // found. This method uses IsInstanceOf/Inner to find the appropriate class
   // instance and then casts it to the expected return type.

include/rocksdb/listener.h

@@ -467,7 +467,6 @@ struct IOErrorInfo {
   std::string file_path;
   size_t length;
   uint64_t offset;
-  ;
 };
 
 // EventListener class contains a set of callback functions that will

include/rocksdb/version.h

@@ -11,7 +11,7 @@
 
 #define ROCKSDB_MAJOR 6
 #define ROCKSDB_MINOR 27
-#define ROCKSDB_PATCH 0
+#define ROCKSDB_PATCH 3
 
 // 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

options/cf_options.cc

@@ -597,7 +597,7 @@ static std::unordered_map<std::string, OptionTypeInfo>
                 static_cast<std::shared_ptr<MemTableRepFactory>*>(addr);
             Status s =
                 MemTableRepFactory::CreateFromString(opts, value, &factory);
-            if (s.ok()) {
+            if (factory && s.ok()) {
               shared->reset(factory.release());
             }
             return s;
@@ -613,7 +613,7 @@ static std::unordered_map<std::string, OptionTypeInfo>
                 static_cast<std::shared_ptr<MemTableRepFactory>*>(addr);
             Status s =
                 MemTableRepFactory::CreateFromString(opts, value, &factory);
-            if (s.ok()) {
+            if (factory && s.ok()) {
               shared->reset(factory.release());
             }
             return s;

options/configurable.cc

@@ -43,6 +43,7 @@ Status Configurable::PrepareOptions(const ConfigOptions& opts) {
   Status status = Status::OK();
 #ifndef ROCKSDB_LITE
   for (auto opt_iter : options_) {
+    if (opt_iter.type_map != nullptr) {
       for (auto map_iter : *(opt_iter.type_map)) {
         auto& opt_info = map_iter.second;
         if (!opt_info.IsDeprecated() && !opt_info.IsAlias() &&
@@ -52,12 +53,13 @@ Status Configurable::PrepareOptions(const ConfigOptions& opts) {
                 opt_info.AsRawPointer<Configurable>(opt_iter.opt_ptr);
             if (config != nullptr) {
               status = config->PrepareOptions(opts);
+            } else if (!opt_info.CanBeNull()) {
+              status = Status::NotFound("Missing configurable object",
+                                        map_iter.first);
+            }
             if (!status.ok()) {
               return status;
             }
-          } else if (!opt_info.CanBeNull()) {
-            status =
-                Status::NotFound("Missing configurable object", map_iter.first);
           }
         }
       }
@@ -74,6 +76,7 @@ Status Configurable::ValidateOptions(const DBOptions& db_opts,
   Status status;
 #ifndef ROCKSDB_LITE
   for (auto opt_iter : options_) {
+    if (opt_iter.type_map != nullptr) {
       for (auto map_iter : *(opt_iter.type_map)) {
         auto& opt_info = map_iter.second;
         if (!opt_info.IsDeprecated() && !opt_info.IsAlias()) {
@@ -83,8 +86,8 @@ Status Configurable::ValidateOptions(const DBOptions& db_opts,
             if (config != nullptr) {
               status = config->ValidateOptions(db_opts, cf_opts);
             } else if (!opt_info.CanBeNull()) {
-            status =
-                Status::NotFound("Missing configurable object", map_iter.first);
+              status = Status::NotFound("Missing configurable object",
+                                        map_iter.first);
             }
             if (!status.ok()) {
               return status;
@@ -93,6 +96,7 @@ Status Configurable::ValidateOptions(const DBOptions& db_opts,
         }
       }
     }
+  }
 #else
   (void)db_opts;
   (void)cf_opts;
@@ -124,6 +128,7 @@ const OptionTypeInfo* ConfigurableHelper::FindOption(
     const std::vector<Configurable::RegisteredOptions>& options,
     const std::string& short_name, std::string* opt_name, void** opt_ptr) {
   for (auto iter : options) {
+    if (iter.type_map != nullptr) {
       const auto opt_info =
           OptionTypeInfo::Find(short_name, *(iter.type_map), opt_name);
       if (opt_info != nullptr) {
@@ -131,6 +136,7 @@ const OptionTypeInfo* ConfigurableHelper::FindOption(
         return opt_info;
       }
     }
+  }
   return nullptr;
 }
 #endif  // ROCKSDB_LITE
@@ -280,6 +286,7 @@ Status ConfigurableHelper::ConfigureOptions(
   if (!opts_map.empty()) {
 #ifndef ROCKSDB_LITE
     for (const auto& iter : configurable.options_) {
+      if (iter.type_map != nullptr) {
         s = ConfigureSomeOptions(config_options, configurable, *(iter.type_map),
                                  &remaining, iter.opt_ptr);
         if (remaining.empty()) {  // Are there more options left?
@@ -288,6 +295,7 @@ Status ConfigurableHelper::ConfigureOptions(
           break;
         }
       }
+    }
 #else
     (void)configurable;
     if (!config_options.ignore_unknown_options) {
@@ -573,6 +581,7 @@ Status ConfigurableHelper::SerializeOptions(const ConfigOptions& config_options,
                                             std::string* result) {
   assert(result);
   for (auto const& opt_iter : configurable.options_) {
+    if (opt_iter.type_map != nullptr) {
       for (const auto& map_iter : *(opt_iter.type_map)) {
         const auto& opt_name = map_iter.first;
         const auto& opt_info = map_iter.second;
@@ -607,6 +616,7 @@ Status ConfigurableHelper::SerializeOptions(const ConfigOptions& config_options,
         }
       }
     }
+  }
   return Status::OK();
 }
 #endif  // ROCKSDB_LITE
@@ -629,6 +639,7 @@ Status ConfigurableHelper::ListOptions(
     const std::string& prefix, std::unordered_set<std::string>* result) {
   Status status;
   for (auto const& opt_iter : configurable.options_) {
+    if (opt_iter.type_map != nullptr) {
       for (const auto& map_iter : *(opt_iter.type_map)) {
         const auto& opt_name = map_iter.first;
         const auto& opt_info = map_iter.second;
@@ -643,6 +654,7 @@ Status ConfigurableHelper::ListOptions(
         }
       }
     }
+  }
   return status;
 }
 #endif  // ROCKSDB_LITE
@@ -702,7 +714,7 @@ bool ConfigurableHelper::AreEquivalent(const ConfigOptions& config_options,
     if (this_offset != that_offset) {
       if (this_offset == nullptr || that_offset == nullptr) {
         return false;
-      } else {
+      } else if (o.type_map != nullptr) {
         for (const auto& map_iter : *(o.type_map)) {
           const auto& opt_info = map_iter.second;
           if (config_options.IsCheckEnabled(opt_info.GetSanityLevel())) {

options/configurable_test.cc

@@ -656,6 +656,30 @@ TEST_F(ConfigurableTest, TestNoCompare) {
   ASSERT_TRUE(base->AreEquivalent(config_options_, copy.get(), &mismatch));
   ASSERT_FALSE(copy->AreEquivalent(config_options_, base.get(), &mismatch));
 }
+
+TEST_F(ConfigurableTest, NullOptionMapTest) {
+  std::unique_ptr<Configurable> base;
+  std::unordered_set<std::string> names;
+  std::string str;
+
+  base.reset(
+      SimpleConfigurable::Create("c", TestConfigMode::kDefaultMode, nullptr));
+  ASSERT_NOK(base->ConfigureFromString(config_options_, "int=10"));
+  ASSERT_NOK(base->ConfigureFromString(config_options_, "int=20"));
+  ASSERT_NOK(base->ConfigureOption(config_options_, "int", "20"));
+  ASSERT_NOK(base->GetOption(config_options_, "int", &str));
+  ASSERT_NE(base->GetOptions<TestOptions>("c"), nullptr);
+  ASSERT_OK(base->GetOptionNames(config_options_, &names));
+  ASSERT_EQ(names.size(), 0UL);
+  ASSERT_OK(base->PrepareOptions(config_options_));
+  ASSERT_OK(base->ValidateOptions(DBOptions(), ColumnFamilyOptions()));
+  std::unique_ptr<Configurable> copy;
+  copy.reset(
+      SimpleConfigurable::Create("c", TestConfigMode::kDefaultMode, nullptr));
+  ASSERT_OK(base->GetOptionString(config_options_, &str));
+  ASSERT_OK(copy->ConfigureFromString(config_options_, str));
+  ASSERT_TRUE(base->AreEquivalent(config_options_, copy.get(), &str));
+}
 #endif
 
 static std::unordered_map<std::string, ConfigTestFactoryFunc> TestFactories = {

options/customizable_test.cc

@@ -612,11 +612,20 @@ static std::unordered_map<std::string, OptionTypeInfo> inner_option_info = {
 #endif  // ROCKSDB_LITE
 };
 
+struct InnerOptions {
+  static const char* kName() { return "InnerOptions"; }
+  std::shared_ptr<Customizable> inner;
+};
+
 class InnerCustomizable : public Customizable {
  public:
-  explicit InnerCustomizable(const std::shared_ptr<Customizable>& w)
-      : inner_(w) {}
+  explicit InnerCustomizable(const std::shared_ptr<Customizable>& w) {
+    iopts_.inner = w;
+    RegisterOptions(&iopts_, &inner_option_info);
+  }
   static const char* kClassName() { return "Inner"; }
+  const char* Name() const override { return kClassName(); }
+
   bool IsInstanceOf(const std::string& name) const override {
     if (name == kClassName()) {
       return true;
@@ -626,26 +635,51 @@ class InnerCustomizable : public Customizable {
   }
 
  protected:
-  const Customizable* Inner() const override { return inner_.get(); }
+  const Customizable* Inner() const override { return iopts_.inner.get(); }
 
  private:
-  std::shared_ptr<Customizable> inner_;
+  InnerOptions iopts_;
+};
+
+struct WrappedOptions1 {
+  static const char* kName() { return "WrappedOptions1"; }
+  int i = 42;
 };
 
 class WrappedCustomizable1 : public InnerCustomizable {
  public:
   explicit WrappedCustomizable1(const std::shared_ptr<Customizable>& w)
-      : InnerCustomizable(w) {}
+      : InnerCustomizable(w) {
+    RegisterOptions(&wopts_, nullptr);
+  }
   const char* Name() const override { return kClassName(); }
   static const char* kClassName() { return "Wrapped1"; }
+
+ private:
+  WrappedOptions1 wopts_;
 };
 
+struct WrappedOptions2 {
+  static const char* kName() { return "WrappedOptions2"; }
+  std::string s = "42";
+};
 class WrappedCustomizable2 : public InnerCustomizable {
  public:
   explicit WrappedCustomizable2(const std::shared_ptr<Customizable>& w)
       : InnerCustomizable(w) {}
+  const void* GetOptionsPtr(const std::string& name) const override {
+    if (name == WrappedOptions2::kName()) {
+      return &wopts_;
+    } else {
+      return InnerCustomizable::GetOptionsPtr(name);
+    }
+  }
+
   const char* Name() const override { return kClassName(); }
   static const char* kClassName() { return "Wrapped2"; }
+
+ private:
+  WrappedOptions2 wopts_;
 };
 }  // namespace
 
@@ -677,6 +711,29 @@ TEST_F(CustomizableTest, WrappedInnerTest) {
   ASSERT_EQ(wc2->CheckedCast<TestCustomizable>(), ac.get());
 }
 
+TEST_F(CustomizableTest, CustomizableInnerTest) {
+  std::shared_ptr<Customizable> c =
+      std::make_shared<InnerCustomizable>(std::make_shared<ACustomizable>("a"));
+  std::shared_ptr<Customizable> wc1 = std::make_shared<WrappedCustomizable1>(c);
+  std::shared_ptr<Customizable> wc2 = std::make_shared<WrappedCustomizable2>(c);
+  auto inner = c->GetOptions<InnerOptions>();
+  ASSERT_NE(inner, nullptr);
+
+  auto aopts = c->GetOptions<AOptions>();
+  ASSERT_NE(aopts, nullptr);
+  ASSERT_EQ(aopts, wc1->GetOptions<AOptions>());
+  ASSERT_EQ(aopts, wc2->GetOptions<AOptions>());
+  auto w1opts = wc1->GetOptions<WrappedOptions1>();
+  ASSERT_NE(w1opts, nullptr);
+  ASSERT_EQ(c->GetOptions<WrappedOptions1>(), nullptr);
+  ASSERT_EQ(wc2->GetOptions<WrappedOptions1>(), nullptr);
+
+  auto w2opts = wc2->GetOptions<WrappedOptions2>();
+  ASSERT_NE(w2opts, nullptr);
+  ASSERT_EQ(c->GetOptions<WrappedOptions2>(), nullptr);
+  ASSERT_EQ(wc1->GetOptions<WrappedOptions2>(), nullptr);
+}
+
 TEST_F(CustomizableTest, CopyObjectTest) {
   class CopyCustomizable : public Customizable {
    public:
@@ -714,20 +771,9 @@ TEST_F(CustomizableTest, CopyObjectTest) {
 }
 
 TEST_F(CustomizableTest, TestStringDepth) {
-  class ShallowCustomizable : public Customizable {
-   public:
-    ShallowCustomizable() {
-      inner_ = std::make_shared<ACustomizable>("a");
-      RegisterOptions("inner", &inner_, &inner_option_info);
-    }
-    static const char* kClassName() { return "shallow"; }
-    const char* Name() const override { return kClassName(); }
-
-   private:
-    std::shared_ptr<TestCustomizable> inner_;
-  };
   ConfigOptions shallow = config_options_;
-  std::unique_ptr<Configurable> c(new ShallowCustomizable());
+  std::unique_ptr<Configurable> c(
+      new InnerCustomizable(std::make_shared<ACustomizable>("a")));
   std::string opt_str;
   shallow.depth = ConfigOptions::Depth::kDepthShallow;
   ASSERT_OK(c->GetOptionString(shallow, &opt_str));

table/block_based/block_based_table_builder.cc

@@ -1221,7 +1221,8 @@ void BlockBasedTableBuilder::WriteRawBlock(const Slice& block_contents,
                                            CompressionType type,
                                            BlockHandle* handle,
                                            BlockType block_type,
-                                           const Slice* raw_block_contents) {
+                                           const Slice* raw_block_contents,
+                                           bool is_top_level_filter_block) {
   Rep* r = rep_;
   bool is_data_block = block_type == BlockType::kData;
   Status s = Status::OK();
@@ -1262,9 +1263,11 @@ void BlockBasedTableBuilder::WriteRawBlock(const Slice& block_contents,
       }
       if (warm_cache) {
         if (type == kNoCompression) {
-          s = InsertBlockInCacheHelper(block_contents, handle, block_type);
+          s = InsertBlockInCacheHelper(block_contents, handle, block_type,
+                                       is_top_level_filter_block);
         } else if (raw_block_contents != nullptr) {
-          s = InsertBlockInCacheHelper(*raw_block_contents, handle, block_type);
+          s = InsertBlockInCacheHelper(*raw_block_contents, handle, block_type,
+                                       is_top_level_filter_block);
         }
         if (!s.ok()) {
           r->SetStatus(s);
@@ -1472,12 +1475,12 @@ Status BlockBasedTableBuilder::InsertBlockInCompressedCache(
 
 Status BlockBasedTableBuilder::InsertBlockInCacheHelper(
     const Slice& block_contents, const BlockHandle* handle,
-    BlockType block_type) {
+    BlockType block_type, bool is_top_level_filter_block) {
   Status s;
   if (block_type == BlockType::kData || block_type == BlockType::kIndex) {
     s = InsertBlockInCache<Block>(block_contents, handle, block_type);
   } else if (block_type == BlockType::kFilter) {
-    if (rep_->filter_builder->IsBlockBased()) {
+    if (rep_->filter_builder->IsBlockBased() || is_top_level_filter_block) {
       s = InsertBlockInCache<Block>(block_contents, handle, block_type);
     } else {
       s = InsertBlockInCache<ParsedFullFilterBlock>(block_contents, handle,
@@ -1564,8 +1567,14 @@ void BlockBasedTableBuilder::WriteFilterBlock(
           rep_->filter_builder->Finish(filter_block_handle, &s, &filter_data);
       assert(s.ok() || s.IsIncomplete());
       rep_->props.filter_size += filter_content.size();
+      bool top_level_filter_block = false;
+      if (s.ok() && rep_->table_options.partition_filters &&
+          !rep_->filter_builder->IsBlockBased()) {
+        top_level_filter_block = true;
+      }
       WriteRawBlock(filter_content, kNoCompression, &filter_block_handle,
-                    BlockType::kFilter);
+                    BlockType::kFilter, nullptr /*raw_contents*/,
+                    top_level_filter_block);
     }
     rep_->filter_builder->ResetFilterBitsBuilder();
   }

table/block_based/block_based_table_builder.h

@@ -119,7 +119,8 @@ class BlockBasedTableBuilder : public TableBuilder {
                   BlockType block_type);
   // Directly write data to the file.
   void WriteRawBlock(const Slice& data, CompressionType, BlockHandle* handle,
-                     BlockType block_type, const Slice* raw_data = nullptr);
+                     BlockType block_type, const Slice* raw_data = nullptr,
+                     bool is_top_level_filter_block = false);
 
   void SetupCacheKeyPrefix(const TableBuilderOptions& tbo);
 
@@ -129,7 +130,8 @@ class BlockBasedTableBuilder : public TableBuilder {
 
   Status InsertBlockInCacheHelper(const Slice& block_contents,
                                   const BlockHandle* handle,
-                                  BlockType block_type);
+                                  BlockType block_type,
+                                  bool is_top_level_filter_block);
 
   Status InsertBlockInCompressedCache(const Slice& block_contents,
                                       const CompressionType type,

table/block_based/block_based_table_iterator.h

@@ -161,12 +161,14 @@ class BlockBasedTableIterator : public InternalIteratorBase<Slice> {
   }
 
   void SetReadaheadState(ReadaheadFileInfo* readahead_file_info) override {
+    if (read_options_.adaptive_readahead) {
       block_prefetcher_.SetReadaheadState(
           &(readahead_file_info->data_block_readahead_info));
       if (index_iter_) {
         index_iter_->SetReadaheadState(readahead_file_info);
       }
     }
+  }
 
   std::unique_ptr<InternalIteratorBase<IndexValue>> index_iter_;
 

table/block_based/block_prefetcher.h

@@ -30,8 +30,11 @@ class BlockPrefetcher {
 
   void ResetValues() {
     num_file_reads_ = 1;
-    readahead_size_ = BlockBasedTable::kInitAutoReadaheadSize;
-    initial_auto_readahead_size_ = readahead_size_;
+    // Since initial_auto_readahead_size_ can be different from
+    // kInitAutoReadaheadSize in case of adaptive_readahead, so fallback the
+    // readahead_size_ to kInitAutoReadaheadSize in case of reset.
+    initial_auto_readahead_size_ = BlockBasedTable::kInitAutoReadaheadSize;
+    readahead_size_ = initial_auto_readahead_size_;
     readahead_limit_ = 0;
     return;
   }

table/block_based/partitioned_index_iterator.h

@@ -123,9 +123,11 @@ class PartitionedIndexIterator : public InternalIteratorBase<IndexValue> {
   }
 
   void SetReadaheadState(ReadaheadFileInfo* readahead_file_info) override {
+    if (read_options_.adaptive_readahead) {
       block_prefetcher_.SetReadaheadState(
           &(readahead_file_info->index_block_readahead_info));
     }
+  }
 
   std::unique_ptr<InternalIteratorBase<IndexValue>> index_iter_;
 

tools/db_crashtest.py

@@ -154,6 +154,7 @@ default_params = {
         [0, 1024 * 1024, 2 * 1024 * 1024, 4 * 1024 * 1024, 8 * 1024 * 1024]),
     "user_timestamp_size": 0,
     "secondary_cache_fault_one_in" : lambda: random.choice([0, 0, 32]),
+    "prepopulate_block_cache" : lambda: random.choice([0, 1]),
 }
 
 _TEST_DIR_ENV_VAR = 'TEST_TMPDIR'