WritePrepared Txn: Disable GC during recovery

Summary:
Disables GC during recovery of a WritePrepared txn db to avoid GCing uncommitted key values.
Closes https://github.com/facebook/rocksdb/pull/2980

Differential Revision: D6000191

Pulled By: maysamyabandeh

fbshipit-source-id: fc4d522c643d24ebf043f811fe4ecd0dd0294675

db/db_impl.cc

@@ -182,8 +182,7 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname)
       unable_to_flush_oldest_log_(false),
       env_options_(BuildDBOptions(immutable_db_options_, mutable_db_options_)),
       env_options_for_compaction_(env_->OptimizeForCompactionTableWrite(
-                                    env_options_,
-                                    immutable_db_options_)),
+          env_options_, immutable_db_options_)),
       num_running_ingest_file_(0),
 #ifndef ROCKSDB_LITE
       wal_manager_(immutable_db_options_, env_options_),
@@ -195,7 +194,9 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname)
       opened_successfully_(false),
       concurrent_prepare_(options.concurrent_prepare),
       manual_wal_flush_(options.manual_wal_flush),
-      seq_per_batch_(options.seq_per_batch) {
+      seq_per_batch_(options.seq_per_batch),
+      // TODO(myabandeh): revise this when we change options.seq_per_batch
+      use_custom_gc_(options.seq_per_batch) {
   env_->GetAbsolutePath(dbname, &db_absolute_path_);
 
   // Reserve ten files or so for other uses and give the rest to TableCache.

db/db_impl.h

@@ -1300,6 +1300,7 @@ class DBImpl : public DB {
   const bool concurrent_prepare_;
   const bool manual_wal_flush_;
   const bool seq_per_batch_;
+  const bool use_custom_gc_;
 };
 
 extern Options SanitizeOptions(const std::string& db,

db/db_impl_compaction_flush.cc

@@ -86,10 +86,14 @@ Status DBImpl::FlushMemTableToOutputFile(
   std::vector<SequenceNumber> snapshot_seqs =
       snapshots_.GetAll(&earliest_write_conflict_snapshot);
 
+  auto snapshot_checker = snapshot_checker_.get();
+  if (use_custom_gc_ && snapshot_checker == nullptr) {
+    snapshot_checker = DisableGCSnapshotChecker::Instance();
+  }
   FlushJob flush_job(
       dbname_, cfd, immutable_db_options_, mutable_cf_options,
       env_options_for_compaction_, versions_.get(), &mutex_, &shutting_down_,
-      snapshot_seqs, earliest_write_conflict_snapshot, snapshot_checker_.get(),
+      snapshot_seqs, earliest_write_conflict_snapshot, snapshot_checker,
       job_context, log_buffer, directories_.GetDbDir(),
       directories_.GetDataDir(0U),
       GetCompressionFlush(*cfd->ioptions(), mutable_cf_options), stats_,
@@ -531,13 +535,17 @@ Status DBImpl::CompactFilesImpl(
   auto pending_outputs_inserted_elem =
       CaptureCurrentFileNumberInPendingOutputs();
 
+  auto snapshot_checker = snapshot_checker_.get();
+  if (use_custom_gc_ && snapshot_checker == nullptr) {
+    snapshot_checker = DisableGCSnapshotChecker::Instance();
+  }
   assert(is_snapshot_supported_ || snapshots_.empty());
   CompactionJob compaction_job(
       job_context->job_id, c.get(), immutable_db_options_,
       env_options_for_compaction_, versions_.get(), &shutting_down_, log_buffer,
       directories_.GetDbDir(), directories_.GetDataDir(c->output_path_id()),
       stats_, &mutex_, &bg_error_, snapshot_seqs,
-      earliest_write_conflict_snapshot, snapshot_checker_.get(), table_cache_,
+      earliest_write_conflict_snapshot, snapshot_checker, table_cache_,
       &event_logger_, c->mutable_cf_options()->paranoid_file_checks,
       c->mutable_cf_options()->report_bg_io_stats, dbname_,
       nullptr);  // Here we pass a nullptr for CompactionJobStats because
@@ -1678,6 +1686,10 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
     std::vector<SequenceNumber> snapshot_seqs =
         snapshots_.GetAll(&earliest_write_conflict_snapshot);
 
+    auto snapshot_checker = snapshot_checker_.get();
+    if (use_custom_gc_ && snapshot_checker == nullptr) {
+      snapshot_checker = DisableGCSnapshotChecker::Instance();
+    }
     assert(is_snapshot_supported_ || snapshots_.empty());
     CompactionJob compaction_job(
         job_context->job_id, c.get(), immutable_db_options_,
@@ -1685,7 +1697,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
         log_buffer, directories_.GetDbDir(),
         directories_.GetDataDir(c->output_path_id()), stats_, &mutex_,
         &bg_error_, snapshot_seqs, earliest_write_conflict_snapshot,
-        snapshot_checker_.get(), table_cache_, &event_logger_,
+        snapshot_checker, table_cache_, &event_logger_,
         c->mutable_cf_options()->paranoid_file_checks,
         c->mutable_cf_options()->report_bg_io_stats, dbname_,
         &compaction_job_stats);

db/db_impl_open.cc

@@ -883,10 +883,10 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
       SequenceNumber earliest_write_conflict_snapshot;
       std::vector<SequenceNumber> snapshot_seqs =
           snapshots_.GetAll(&earliest_write_conflict_snapshot);
-      // Only TransactionDB passes snapshot_checker and it creates it after db
-      // open. Just pass nullptr here.
-      SnapshotChecker* snapshot_checker = nullptr;
-
+      auto snapshot_checker = snapshot_checker_.get();
+      if (use_custom_gc_ && snapshot_checker == nullptr) {
+        snapshot_checker = DisableGCSnapshotChecker::Instance();
+      }
       s = BuildTable(
           dbname_, env_, *cfd->ioptions(), mutable_cf_options,
           env_options_for_compaction_, cfd->table_cache(), iter.get(),

db/repair.cc

@@ -400,11 +400,7 @@ class Repairer {
       int64_t _current_time = 0;
       status = env_->GetCurrentTime(&_current_time);  // ignore error
       const uint64_t current_time = static_cast<uint64_t>(_current_time);
-      // Only TransactionDB make use of snapshot_checker and repair doesn't
-      // currently support TransactionDB with uncommitted prepared keys in WAL.
-      // TODO(yiwu) Support repairing TransactionDB.
-      SnapshotChecker* snapshot_checker = nullptr;
-
+      SnapshotChecker* snapshot_checker = DisableGCSnapshotChecker::Instance();
       status = BuildTable(
           dbname_, env_, *cfd->ioptions(), *cfd->GetLatestMutableCFOptions(),
           env_options_, table_cache_, iter.get(),

db/snapshot_checker.h

@@ -8,16 +8,40 @@
 
 namespace rocksdb {
 
+// Callback class that control GC of duplicate keys in flush/compaction
+class SnapshotChecker {
+ public:
+  virtual ~SnapshotChecker() {}
+  virtual bool IsInSnapshot(SequenceNumber sequence,
+                            SequenceNumber snapshot_sequence) const = 0;
+};
+
+class DisableGCSnapshotChecker : public SnapshotChecker {
+ public:
+  virtual ~DisableGCSnapshotChecker() {}
+  virtual bool IsInSnapshot(SequenceNumber sequence,
+                            SequenceNumber snapshot_sequence) const {
+    // By returning false, we prevent all the values from being GCed
+    return false;
+  }
+  static DisableGCSnapshotChecker* Instance() { return &instance_; }
+
+ protected:
+  static DisableGCSnapshotChecker instance_;
+  explicit DisableGCSnapshotChecker() {}
+};
+
 class WritePreparedTxnDB;
 
 // Callback class created by WritePreparedTxnDB to check if a key
 // is visible by a snapshot.
-class SnapshotChecker {
+class WritePreparedSnapshotChecker : public SnapshotChecker {
  public:
-  explicit SnapshotChecker(WritePreparedTxnDB* txn_db);
+  explicit WritePreparedSnapshotChecker(WritePreparedTxnDB* txn_db);
+  virtual ~WritePreparedSnapshotChecker() {}
 
-  bool IsInSnapshot(SequenceNumber sequence,
-                    SequenceNumber snapshot_sequence) const;
+  virtual bool IsInSnapshot(SequenceNumber sequence,
+                            SequenceNumber snapshot_sequence) const override;
 
  private:
 #ifndef ROCKSDB_LITE

utilities/transactions/pessimistic_transaction_db.cc

@@ -148,7 +148,7 @@ Status WritePreparedTxnDB::Initialize(
   SequenceNumber last_seq = db_impl_->GetLatestSequenceNumber();
   AdvanceMaxEvictedSeq(prev_max, last_seq);
 
-  db_impl_->SetSnapshotChecker(new SnapshotChecker(this));
+  db_impl_->SetSnapshotChecker(new WritePreparedSnapshotChecker(this));
 
   auto s = PessimisticTransactionDB::Initialize(compaction_enabled_cf_indices,
                                                 handles);

utilities/transactions/snapshot_checker.cc

@@ -14,10 +14,11 @@
 namespace rocksdb {
 
 #ifdef ROCKSDB_LITE
-SnapshotChecker::SnapshotChecker(WritePreparedTxnDB* txn_db) {}
+WritePreparedSnapshotChecker::WritePreparedSnapshotChecker(
+    WritePreparedTxnDB* txn_db) {}
 
-bool SnapshotChecker::IsInSnapshot(SequenceNumber sequence,
-                                   SequenceNumber snapshot_sequence) const {
+bool WritePreparedSnapshotChecker::IsInSnapshot(
+    SequenceNumber sequence, SequenceNumber snapshot_sequence) const {
   // Should never be called in LITE mode.
   assert(false);
   return true;
@@ -25,13 +26,16 @@ bool SnapshotChecker::IsInSnapshot(SequenceNumber sequence,
 
 #else
 
-SnapshotChecker::SnapshotChecker(WritePreparedTxnDB* txn_db)
+WritePreparedSnapshotChecker::WritePreparedSnapshotChecker(
+    WritePreparedTxnDB* txn_db)
     : txn_db_(txn_db){};
 
-bool SnapshotChecker::IsInSnapshot(SequenceNumber sequence,
-                                   SequenceNumber snapshot_sequence) const {
+bool WritePreparedSnapshotChecker::IsInSnapshot(
+    SequenceNumber sequence, SequenceNumber snapshot_sequence) const {
   return txn_db_->IsInSnapshot(sequence, snapshot_sequence);
 }
+
 #endif  // ROCKSDB_LITE
+DisableGCSnapshotChecker DisableGCSnapshotChecker::instance_;
 
 }  // namespace rocksdb

utilities/transactions/write_prepared_transaction_test.cc

@@ -1349,6 +1349,28 @@ TEST_P(WritePreparedTransactionTest, DuplicateKeyTest) {
   }
 }
 
+TEST_P(WritePreparedTransactionTest, DisableGCDuringRecoveryTest) {
+  // Use large buffer to avoid memtable flush after 1024 insertions
+  options.write_buffer_size = 1024 * 1024;
+  ReOpen();
+  std::vector<KeyVersion> versions;
+  for (uint64_t i = 1; i <= 1024; i++) {
+    std::string v = "bar" + ToString(i);
+    ASSERT_OK(db->Put(WriteOptions(), "foo", v));
+    VerifyKeys({{"foo", v}});
+    KeyVersion kv = {"foo", v, i, kTypeValue};
+    versions.emplace_back(kv);
+  }
+  std::reverse(std::begin(versions), std::end(versions));
+  VerifyInternalKeys(versions);
+  DBImpl* db_impl = reinterpret_cast<DBImpl*>(db->GetRootDB());
+  db_impl->FlushWAL(true);
+  // Use small buffer to ensure memtable flush during recovery
+  options.write_buffer_size = 1024;
+  ReOpenNoDelete();
+  VerifyInternalKeys(versions);
+}
+
 TEST_P(WritePreparedTransactionTest, SequenceNumberZeroTest) {
   ASSERT_OK(db->Put(WriteOptions(), "foo", "bar"));
   VerifyKeys({{"foo", "bar"}});