Revert "Snapshot release triggered compaction without multiple tombstones" (#396)

 

Signed-off-by: Yang Zhang <[email protected]>

Author: v01dstar

HISTORY.md

@@ -1030,8 +1030,6 @@ Note: The next release will be major release 7.0. See https://github.com/faceboo
 * BackupEngineOptions::sync (default true) now applies to restoring backups in addition to creating backups. This could slow down restores, but ensures they are fully persisted before returning OK. (Consider increasing max_background_operations to improve performance.)
 
 ## 6.23.0 (2021-07-16)
-### Behavior Changes
-* Obsolete keys in the bottommost level that were preserved for a snapshot will now be cleaned upon snapshot release in all cases. This form of compaction (snapshot release triggered compaction) previously had an artificial limitation that multiple tombstones needed to be present.
 ### Bug Fixes
 * Blob file checksums are now printed in hexadecimal format when using the `manifest_dump` `ldb` command.
 * `GetLiveFilesMetaData()` now populates the `temperature`, `oldest_ancester_time`, and `file_creation_time` fields of its `LiveFileMetaData` results when the information is available. Previously these fields always contained zero indicating unknown.
@@ -1056,6 +1054,7 @@ Note: The next release will be major release 7.0. See https://github.com/faceboo
 ### Behavior Changes
 * Added two additional tickers, MEMTABLE_PAYLOAD_BYTES_AT_FLUSH and MEMTABLE_GARBAGE_BYTES_AT_FLUSH. These stats can be used to estimate the ratio of "garbage" (outdated) bytes in the memtable that are discarded at flush time.
 * Added API comments clarifying safe usage of Disable/EnableManualCompaction and EventListener callbacks for compaction.
+
 ### Bug Fixes
 * fs_posix.cc GetFreeSpace() always report disk space available to root even when running as non-root.  Linux defaults often have disk mounts with 5 to 10 percent of total space reserved only for root.  Out of space could result for non-root users.
 * Subcompactions are now disabled when user-defined timestamps are used, since the subcompaction boundary picking logic is currently not timestamp-aware, which could lead to incorrect results when different subcompactions process keys that only differ by timestamp.

db/version_set.cc

@@ -4216,7 +4216,8 @@ void VersionStorageInfo::ComputeBottommostFilesMarkedForCompaction(
 
   for (auto& level_and_file : bottommost_files_) {
     if (!level_and_file.second->being_compacted &&
-        level_and_file.second->fd.largest_seqno != 0) {
+        level_and_file.second->fd.largest_seqno != 0 &&
+        level_and_file.second->num_deletions > 1) {
       // largest_seqno might be nonzero due to containing the final key in an
       // earlier compaction, whose seqnum we didn't zero out.
       if (level_and_file.second->fd.largest_seqno < oldest_snapshot_seqnum_) {

utilities/blob_db/blob_db_test.cc

@@ -570,6 +570,7 @@ TEST_F(BlobDBTest, EnableDisableCompressionGC) {
   Random rnd(301);
   BlobDBOptions bdb_options;
   bdb_options.min_blob_size = 0;
+  bdb_options.enable_garbage_collection = true;
   bdb_options.garbage_collection_cutoff = 1.0;
   bdb_options.disable_background_tasks = true;
   bdb_options.compression = kSnappyCompression;
@@ -598,11 +599,6 @@ TEST_F(BlobDBTest, EnableDisableCompressionGC) {
   ASSERT_EQ(2, blob_files.size());
   ASSERT_EQ(kNoCompression, blob_files[1]->GetCompressionType());
 
-  // Enable GC. If we do it earlier the snapshot release triggered compaction
-  // may compact files and trigger GC before we can verify there are two files.
-  bdb_options.enable_garbage_collection = true;
-  Reopen(bdb_options);
-
   // Trigger compaction
   ASSERT_OK(blob_db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
   blob_db_impl()->TEST_DeleteObsoleteFiles();
@@ -641,6 +637,7 @@ TEST_F(BlobDBTest, ChangeCompressionGC) {
   Random rnd(301);
   BlobDBOptions bdb_options;
   bdb_options.min_blob_size = 0;
+  bdb_options.enable_garbage_collection = true;
   bdb_options.garbage_collection_cutoff = 1.0;
   bdb_options.disable_background_tasks = true;
   bdb_options.compression = kLZ4Compression;
@@ -670,11 +667,6 @@ TEST_F(BlobDBTest, ChangeCompressionGC) {
   ASSERT_EQ(2, blob_files.size());
   ASSERT_EQ(kSnappyCompression, blob_files[1]->GetCompressionType());
 
-  // Enable GC. If we do it earlier the snapshot release triggered compaction
-  // may compact files and trigger GC before we can verify there are two files.
-  bdb_options.enable_garbage_collection = true;
-  Reopen(bdb_options);
-
   ASSERT_OK(blob_db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
   VerifyDB(data);
 

utilities/transactions/write_prepared_transaction_test.cc

@@ -2582,7 +2582,6 @@ TEST_P(WritePreparedTransactionTest, ReleaseSnapshotDuringCompaction) {
   const size_t snapshot_cache_bits = 7;  // same as default
   const size_t commit_cache_bits = 0;    // minimum commit cache
   UpdateTransactionDBOptions(snapshot_cache_bits, commit_cache_bits);
-  options.disable_auto_compactions = true;
   ASSERT_OK(ReOpen());
 
   ASSERT_OK(db->Put(WriteOptions(), "key1", "value1_1"));
@@ -2602,13 +2601,7 @@ TEST_P(WritePreparedTransactionTest, ReleaseSnapshotDuringCompaction) {
   VerifyKeys({{"key1", "value1_1"}}, snapshot2);
   // Add a flush to avoid compaction to fallback to trivial move.
 
-  // The callback might be called twice, record the calling state to
-  // prevent double calling.
-  bool callback_finished = false;
   auto callback = [&](void*) {
-    if (callback_finished) {
-      return;
-    }
     // Release snapshot1 after CompactionIterator init.
     // CompactionIterator need to figure out the earliest snapshot
     // that can see key1:value1_2 is kMaxSequenceNumber, not
@@ -2617,7 +2610,6 @@ TEST_P(WritePreparedTransactionTest, ReleaseSnapshotDuringCompaction) {
     // Add some keys to advance max_evicted_seq.
     ASSERT_OK(db->Put(WriteOptions(), "key3", "value3"));
     ASSERT_OK(db->Put(WriteOptions(), "key4", "value4"));
-    callback_finished = true;
   };
   SyncPoint::GetInstance()->SetCallBack("CompactionIterator:AfterInit",
                                         callback);
@@ -2639,7 +2631,6 @@ TEST_P(WritePreparedTransactionTest, ReleaseSnapshotDuringCompaction2) {
   const size_t snapshot_cache_bits = 7;  // same as default
   const size_t commit_cache_bits = 0;    // minimum commit cache
   UpdateTransactionDBOptions(snapshot_cache_bits, commit_cache_bits);
-  options.disable_auto_compactions = true;
   ASSERT_OK(ReOpen());
 
   ASSERT_OK(db->Put(WriteOptions(), "key1", "value1"));
@@ -2690,7 +2681,6 @@ TEST_P(WritePreparedTransactionTest, ReleaseSnapshotDuringCompaction3) {
   const size_t snapshot_cache_bits = 7;  // same as default
   const size_t commit_cache_bits = 1;    // commit cache size = 2
   UpdateTransactionDBOptions(snapshot_cache_bits, commit_cache_bits);
-  options.disable_auto_compactions = true;
   ASSERT_OK(ReOpen());
 
   // Add a dummy key to evict v2 commit cache, but keep v1 commit cache.
@@ -2720,18 +2710,11 @@ TEST_P(WritePreparedTransactionTest, ReleaseSnapshotDuringCompaction3) {
   add_dummy();
   auto* s2 = db->GetSnapshot();
 
-  // The callback might be called twice, record the calling state to
-  // prevent double calling.
-  bool callback_finished = false;
   auto callback = [&](void*) {
-    if (callback_finished) {
-      return;
-    }
     db->ReleaseSnapshot(s1);
     // Add some dummy entries to trigger s1 being cleanup from old_commit_map.
     add_dummy();
     add_dummy();
-    callback_finished = true;
   };
   SyncPoint::GetInstance()->SetCallBack("CompactionIterator:AfterInit",
                                         callback);
@@ -2749,7 +2732,6 @@ TEST_P(WritePreparedTransactionTest, ReleaseEarliestSnapshotDuringCompaction) {
   const size_t snapshot_cache_bits = 7;  // same as default
   const size_t commit_cache_bits = 0;    // minimum commit cache
   UpdateTransactionDBOptions(snapshot_cache_bits, commit_cache_bits);
-  options.disable_auto_compactions = true;
   ASSERT_OK(ReOpen());
 
   ASSERT_OK(db->Put(WriteOptions(), "key1", "value1"));