rocksdb/utilities/transactions/transaction_util.cc
agiardullo 3bfd3d39a3 Use SST files for Transaction conflict detection
Summary:
Currently, transactions can fail even if there is no actual write conflict.  This is due to relying on only the memtables to check for write-conflicts.  Users have to tune memtable settings to try to avoid this, but it's hard to figure out exactly how to tune these settings.

With this diff, TransactionDB will use both memtables and SST files to determine if there are any write conflicts.  This relies on the fact that BlockBasedTable stores sequence numbers for all writes that happen after any open snapshot.  Also, D50295 is needed to prevent SingleDelete from disappearing writes (the TODOs in this test code will be fixed once the other diff is approved and merged).

Note that Optimistic transactions will still rely on tuning memtable settings as we do not want to read from SST while on the write thread.  Also, memtable settings can still be used to reduce how often TransactionDB needs to read SST files.

Test Plan: unit tests, db bench

Reviewers: rven, yhchiang, kradhakrishnan, IslamAbdelRahman, sdong

Reviewed By: sdong

Subscribers: dhruba, leveldb, yoshinorim

Differential Revision: https://reviews.facebook.net/D50475
2015-12-11 12:34:11 -08:00

163 lines
5.2 KiB
C++

// Copyright (c) 2015, 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.
#ifndef ROCKSDB_LITE
#ifndef __STDC_FORMAT_MACROS
#define __STDC_FORMAT_MACROS
#endif
#include "utilities/transactions/transaction_util.h"
#include <inttypes.h>
#include <string>
#include <vector>
#include "db/db_impl.h"
#include "rocksdb/status.h"
#include "rocksdb/utilities/write_batch_with_index.h"
#include "util/string_util.h"
namespace rocksdb {
Status TransactionUtil::CheckKeyForConflicts(DBImpl* db_impl,
ColumnFamilyHandle* column_family,
const std::string& key,
SequenceNumber key_seq,
bool cache_only) {
Status result;
auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
auto cfd = cfh->cfd();
SuperVersion* sv = db_impl->GetAndRefSuperVersion(cfd);
if (sv == nullptr) {
result = Status::InvalidArgument("Could not access column family " +
cfh->GetName());
}
if (result.ok()) {
SequenceNumber earliest_seq =
db_impl->GetEarliestMemTableSequenceNumber(sv, true);
result = CheckKey(db_impl, sv, earliest_seq, key_seq, key, cache_only);
db_impl->ReturnAndCleanupSuperVersion(cfd, sv);
}
return result;
}
Status TransactionUtil::CheckKey(DBImpl* db_impl, SuperVersion* sv,
SequenceNumber earliest_seq,
SequenceNumber key_seq, const std::string& key,
bool cache_only) {
Status result;
bool need_to_read_sst = false;
// Since it would be too slow to check the SST files, we will only use
// the memtables to check whether there have been any recent writes
// to this key after it was accessed in this transaction. But if the
// Memtables do not contain a long enough history, we must fail the
// transaction.
if (earliest_seq == kMaxSequenceNumber) {
// The age of this memtable is unknown. Cannot rely on it to check
// for recent writes. This error shouldn't happen often in practice as
// the Memtable should have a valid earliest sequence number except in some
// corner cases (such as error cases during recovery).
need_to_read_sst = true;
if (cache_only) {
result = Status::TryAgain(
"Transaction ould not check for conflicts as the MemTable does not "
"countain a long enough history to check write at SequenceNumber: ",
ToString(key_seq));
}
} else if (key_seq < earliest_seq) {
need_to_read_sst = true;
if (cache_only) {
// The age of this memtable is too new to use to check for recent
// writes.
char msg[255];
snprintf(msg, sizeof(msg),
"Transaction could not check for conflicts for operation at "
"SequenceNumber %" PRIu64
" as the MemTable only contains changes newer than "
"SequenceNumber %" PRIu64
". Increasing the value of the "
"max_write_buffer_number_to_maintain option could reduce the "
"frequency "
"of this error.",
key_seq, earliest_seq);
result = Status::TryAgain(msg);
}
}
if (result.ok()) {
SequenceNumber seq = kMaxSequenceNumber;
bool found_record_for_key = false;
Status s = db_impl->GetLatestSequenceForKey(sv, key, !need_to_read_sst,
&seq, &found_record_for_key);
if (!(s.ok() || s.IsNotFound() || s.IsMergeInProgress())) {
result = s;
} else if (found_record_for_key && (seq > key_seq)) {
// Write Conflict
result = Status::Busy();
}
}
return result;
}
Status TransactionUtil::CheckKeysForConflicts(DBImpl* db_impl,
const TransactionKeyMap& key_map,
bool cache_only) {
Status result;
for (auto& key_map_iter : key_map) {
uint32_t cf_id = key_map_iter.first;
const auto& keys = key_map_iter.second;
SuperVersion* sv = db_impl->GetAndRefSuperVersion(cf_id);
if (sv == nullptr) {
result = Status::InvalidArgument("Could not access column family " +
ToString(cf_id));
break;
}
SequenceNumber earliest_seq =
db_impl->GetEarliestMemTableSequenceNumber(sv, true);
// For each of the keys in this transaction, check to see if someone has
// written to this key since the start of the transaction.
for (const auto& key_iter : keys) {
const auto& key = key_iter.first;
const SequenceNumber key_seq = key_iter.second;
result = CheckKey(db_impl, sv, earliest_seq, key_seq, key, cache_only);
if (!result.ok()) {
break;
}
}
db_impl->ReturnAndCleanupSuperVersion(cf_id, sv);
if (!result.ok()) {
break;
}
}
return result;
}
} // namespace rocksdb
#endif // ROCKSDB_LITE