rocksdb/utilities/transactions/pessimistic_transaction_db.h

345 lines
13 KiB
C
Raw Normal View History

// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
// This source code is licensed under both the GPLv2 (found in the
// COPYING file in the root directory) and Apache 2.0 License
// (found in the LICENSE.Apache file in the root directory).
#pragma once
#include <mutex>
#include <queue>
#include <set>
#include <string>
#include <unordered_map>
#include <vector>
#include "db/db_iter.h"
#include "db/read_callback.h"
#include "db/snapshot_checker.h"
#include "rocksdb/db.h"
#include "rocksdb/options.h"
#include "rocksdb/utilities/transaction_db.h"
#include "util/cast_util.h"
#include "utilities/transactions/lock/lock_manager.h"
#include "utilities/transactions/lock/range/range_lock_manager.h"
#include "utilities/transactions/pessimistic_transaction.h"
#include "utilities/transactions/write_prepared_txn.h"
namespace ROCKSDB_NAMESPACE {
class PessimisticTransactionDB : public TransactionDB {
public:
explicit PessimisticTransactionDB(DB* db,
const TransactionDBOptions& txn_db_options);
explicit PessimisticTransactionDB(StackableDB* db,
const TransactionDBOptions& txn_db_options);
virtual ~PessimisticTransactionDB();
const Snapshot* GetSnapshot() override { return db_->GetSnapshot(); }
virtual Status Initialize(
const std::vector<size_t>& compaction_enabled_cf_indices,
const std::vector<ColumnFamilyHandle*>& handles);
Transaction* BeginTransaction(const WriteOptions& write_options,
const TransactionOptions& txn_options,
Transaction* old_txn) override = 0;
using StackableDB::Put;
Status Put(const WriteOptions& options, ColumnFamilyHandle* column_family,
const Slice& key, const Slice& val) override;
Status PutEntity(const WriteOptions& options,
ColumnFamilyHandle* column_family, const Slice& key,
const WideColumns& columns) override;
Status PutEntity(const WriteOptions& /* options */, const Slice& /* key */,
const AttributeGroups& attribute_groups) override {
if (attribute_groups.empty()) {
return Status::InvalidArgument(
"Cannot call this method without attribute groups");
}
return Status::NotSupported(
"PutEntity with AttributeGroups not supported by "
"PessimisticTransactionDB");
}
using StackableDB::Delete;
Status Delete(const WriteOptions& wopts, ColumnFamilyHandle* column_family,
const Slice& key) override;
using StackableDB::SingleDelete;
Status SingleDelete(const WriteOptions& wopts,
ColumnFamilyHandle* column_family,
const Slice& key) override;
using StackableDB::Merge;
Status Merge(const WriteOptions& options, ColumnFamilyHandle* column_family,
const Slice& key, const Slice& value) override;
using TransactionDB::Write;
Status Write(const WriteOptions& opts, WriteBatch* updates) override;
inline Status WriteWithConcurrencyControl(const WriteOptions& opts,
WriteBatch* updates) {
Status s;
if (opts.protection_bytes_per_key > 0) {
s = WriteBatchInternal::UpdateProtectionInfo(
updates, opts.protection_bytes_per_key);
}
if (s.ok()) {
// Need to lock all keys in this batch to prevent write conflicts with
// concurrent transactions.
Transaction* txn = BeginInternalTransaction(opts);
txn->DisableIndexing();
auto txn_impl = static_cast_with_check<PessimisticTransaction>(txn);
// Since commitBatch sorts the keys before locking, concurrent Write()
// operations will not cause a deadlock.
// In order to avoid a deadlock with a concurrent Transaction,
// Transactions should use a lock timeout.
s = txn_impl->CommitBatch(updates);
delete txn;
}
return s;
}
using StackableDB::CreateColumnFamily;
Status CreateColumnFamily(const ColumnFamilyOptions& options,
const std::string& column_family_name,
ColumnFamilyHandle** handle) override;
Status CreateColumnFamilies(
const ColumnFamilyOptions& options,
const std::vector<std::string>& column_family_names,
std::vector<ColumnFamilyHandle*>* handles) override;
Status CreateColumnFamilies(
const std::vector<ColumnFamilyDescriptor>& column_families,
std::vector<ColumnFamilyHandle*>* handles) override;
using StackableDB::CreateColumnFamilyWithImport;
Status CreateColumnFamilyWithImport(
const ColumnFamilyOptions& options, const std::string& column_family_name,
const ImportColumnFamilyOptions& import_options,
const ExportImportFilesMetaData& metadata,
ColumnFamilyHandle** handle) override {
const std::vector<const ExportImportFilesMetaData*>& metadatas{&metadata};
return CreateColumnFamilyWithImport(options, column_family_name,
import_options, metadatas, handle);
}
Status CreateColumnFamilyWithImport(
const ColumnFamilyOptions& options, const std::string& column_family_name,
const ImportColumnFamilyOptions& import_options,
const std::vector<const ExportImportFilesMetaData*>& metadatas,
ColumnFamilyHandle** handle) override;
using StackableDB::DropColumnFamily;
Status DropColumnFamily(ColumnFamilyHandle* column_family) override;
Status DropColumnFamilies(
const std::vector<ColumnFamilyHandle*>& column_families) override;
Status TryLock(PessimisticTransaction* txn, uint32_t cfh_id,
const std::string& key, bool exclusive);
Status TryRangeLock(PessimisticTransaction* txn, uint32_t cfh_id,
const Endpoint& start_endp, const Endpoint& end_endp);
Replace tracked_keys with a new LockTracker interface in TransactionDB (#7013) Summary: We're going to support more locking protocols such as range lock in transaction. However, in current design, `TransactionBase` has a member `tracked_keys` which assumes that point lock (lock a single key) is used, and is used in snapshot checking (isolation protocol). When using range lock, we may use read committed instead of snapshot checking as the isolation protocol. The most significant usage scenarios of `tracked_keys` are: 1. pessimistic transaction uses it to track the locked keys, and unlock these keys when commit or rollback. 2. optimistic transaction does not lock keys upfront, it only tracks the lock intentions in tracked_keys, and do write conflict checking when commit. 3. each `SavePoint` tracks the keys that are locked since the `SavePoint`, `RollbackToSavePoint` or `PopSavePoint` relies on both the tracked keys in `SavePoint`s and `tracked_keys`. Based on these scenarios, if we can abstract out a `LockTracker` interface to hold a set of tracked locks (can be keys or key ranges), and have methods that can be composed together to implement the scenarios, then `tracked_keys` can be an internal data structure of one implementation of `LockTracker`. See `utilities/transactions/lock/lock_tracker.h` for the detailed interface design, and `utilities/transactions/lock/point_lock_tracker.cc` for the implementation. In the future, a `RangeLockTracker` can be implemented to track range locks without affecting other components. After this PR, a clean interface for lock manager should be possible, and then ideally, we can have pluggable locking protocols. Pull Request resolved: https://github.com/facebook/rocksdb/pull/7013 Test Plan: Run `transaction_test` and `optimistic_transaction_test`. Reviewed By: ajkr Differential Revision: D22163706 Pulled By: cheng-chang fbshipit-source-id: f2860577b5334e31dd2994f5bc6d7c40d502b1b4
2020-08-06 19:36:48 +00:00
void UnLock(PessimisticTransaction* txn, const LockTracker& keys);
void UnLock(PessimisticTransaction* txn, uint32_t cfh_id,
const std::string& key);
void AddColumnFamily(const ColumnFamilyHandle* handle);
static TransactionDBOptions ValidateTxnDBOptions(
const TransactionDBOptions& txn_db_options);
const TransactionDBOptions& GetTxnDBOptions() const {
return txn_db_options_;
}
void InsertExpirableTransaction(TransactionID tx_id,
PessimisticTransaction* tx);
void RemoveExpirableTransaction(TransactionID tx_id);
// If transaction is no longer available, locks can be stolen
// If transaction is available, try stealing locks directly from transaction
// It is the caller's responsibility to ensure that the referred transaction
// is expirable (GetExpirationTime() > 0) and that it is expired.
bool TryStealingExpiredTransactionLocks(TransactionID tx_id);
Transaction* GetTransactionByName(const TransactionName& name) override;
void RegisterTransaction(Transaction* txn);
void UnregisterTransaction(Transaction* txn);
// not thread safe. current use case is during recovery (single thread)
void GetAllPreparedTransactions(std::vector<Transaction*>* trans) override;
LockManager::PointLockStatus GetLockStatusData() override;
std::vector<DeadlockPath> GetDeadlockInfoBuffer() override;
void SetDeadlockInfoBufferSize(uint32_t target_size) override;
// The default implementation does nothing. The actual implementation is moved
// to the child classes that actually need this information. This was due to
// an odd performance drop we observed when the added std::atomic member to
// the base class even when the subclass do not read it in the fast path.
virtual void UpdateCFComparatorMap(const std::vector<ColumnFamilyHandle*>&) {}
virtual void UpdateCFComparatorMap(ColumnFamilyHandle*) {}
// Use the returned factory to create LockTrackers in transactions.
const LockTrackerFactory& GetLockTrackerFactory() const {
return lock_manager_->GetLockTrackerFactory();
}
Snapshots with user-specified timestamps (#9879) Summary: In RocksDB, keys are associated with (internal) sequence numbers which denote when the keys are written to the database. Sequence numbers in different RocksDB instances are unrelated, thus not comparable. It is nice if we can associate sequence numbers with their corresponding actual timestamps. One thing we can do is to support user-defined timestamp, which allows the applications to specify the format of custom timestamps and encode a timestamp with each key. More details can be found at https://github.com/facebook/rocksdb/wiki/User-defined-Timestamp-%28Experimental%29. This PR provides a different but complementary approach. We can associate rocksdb snapshots (defined in https://github.com/facebook/rocksdb/blob/7.2.fb/include/rocksdb/snapshot.h#L20) with **user-specified** timestamps. Since a snapshot is essentially an object representing a sequence number, this PR establishes a bi-directional mapping between sequence numbers and timestamps. In the past, snapshots are usually taken by readers. The current super-version is grabbed, and a `rocksdb::Snapshot` object is created with the last published sequence number of the super-version. You can see that the reader actually has no good idea of what timestamp to assign to this snapshot, because by the time the `GetSnapshot()` is called, an arbitrarily long period of time may have already elapsed since the last write, which is when the last published sequence number is written. This observation motivates the creation of "timestamped" snapshots on the write path. Currently, this functionality is exposed only to the layer of `TransactionDB`. Application can tell RocksDB to create a snapshot when a transaction commits, effectively associating the last sequence number with a timestamp. It is also assumed that application will ensure any two snapshots with timestamps should satisfy the following: ``` snapshot1.seq < snapshot2.seq iff. snapshot1.ts < snapshot2.ts ``` If the application can guarantee that when a reader takes a timestamped snapshot, there is no active writes going on in the database, then we also allow the user to use a new API `TransactionDB::CreateTimestampedSnapshot()` to create a snapshot with associated timestamp. Code example ```cpp // Create a timestamped snapshot when committing transaction. txn->SetCommitTimestamp(100); txn->SetSnapshotOnNextOperation(); txn->Commit(); // A wrapper API for convenience Status Transaction::CommitAndTryCreateSnapshot( std::shared_ptr<TransactionNotifier> notifier, TxnTimestamp ts, std::shared_ptr<const Snapshot>* ret); // Create a timestamped snapshot if caller guarantees no concurrent writes std::pair<Status, std::shared_ptr<const Snapshot>> snapshot = txn_db->CreateTimestampedSnapshot(100); ``` The snapshots created in this way will be managed by RocksDB with ref-counting and potentially shared with other readers. We provide the following APIs for readers to retrieve a snapshot given a timestamp. ```cpp // Return the timestamped snapshot correponding to given timestamp. If ts is // kMaxTxnTimestamp, then we return the latest timestamped snapshot if present. // Othersise, we return the snapshot whose timestamp is equal to `ts`. If no // such snapshot exists, then we return null. std::shared_ptr<const Snapshot> TransactionDB::GetTimestampedSnapshot(TxnTimestamp ts) const; // Return the latest timestamped snapshot if present. std::shared_ptr<const Snapshot> TransactionDB::GetLatestTimestampedSnapshot() const; ``` We also provide two additional APIs for stats collection and reporting purposes. ```cpp Status TransactionDB::GetAllTimestampedSnapshots( std::vector<std::shared_ptr<const Snapshot>>& snapshots) const; // Return timestamped snapshots whose timestamps fall in [ts_lb, ts_ub) and store them in `snapshots`. Status TransactionDB::GetTimestampedSnapshots( TxnTimestamp ts_lb, TxnTimestamp ts_ub, std::vector<std::shared_ptr<const Snapshot>>& snapshots) const; ``` To prevent the number of timestamped snapshots from growing infinitely, we provide the following API to release timestamped snapshots whose timestamps are older than or equal to a given threshold. ```cpp void TransactionDB::ReleaseTimestampedSnapshotsOlderThan(TxnTimestamp ts); ``` Before shutdown, RocksDB will release all timestamped snapshots. Comparison with user-defined timestamp and how they can be combined: User-defined timestamp persists every key with a timestamp, while timestamped snapshots maintain a volatile mapping between snapshots (sequence numbers) and timestamps. Different internal keys with the same user key but different timestamps will be treated as different by compaction, thus a newer version will not hide older versions (with smaller timestamps) unless they are eligible for garbage collection. In contrast, taking a timestamped snapshot at a certain sequence number and timestamp prevents all the keys visible in this snapshot from been dropped by compaction. Here, visible means (seq < snapshot and most recent). The timestamped snapshot supports the semantics of reading at an exact point in time. Timestamped snapshots can also be used with user-defined timestamp. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9879 Test Plan: ``` make check TEST_TMPDIR=/dev/shm make crash_test_with_txn ``` Reviewed By: siying Differential Revision: D35783919 Pulled By: riversand963 fbshipit-source-id: 586ad905e169189e19d3bfc0cb0177a7239d1bd4
2022-06-10 23:07:03 +00:00
std::pair<Status, std::shared_ptr<const Snapshot>> CreateTimestampedSnapshot(
TxnTimestamp ts) override;
std::shared_ptr<const Snapshot> GetTimestampedSnapshot(
TxnTimestamp ts) const override;
void ReleaseTimestampedSnapshotsOlderThan(TxnTimestamp ts) override;
Status GetTimestampedSnapshots(TxnTimestamp ts_lb, TxnTimestamp ts_ub,
std::vector<std::shared_ptr<const Snapshot>>&
timestamped_snapshots) const override;
protected:
DBImpl* db_impl_;
std::shared_ptr<Logger> info_log_;
const TransactionDBOptions txn_db_options_;
Support user-defined timestamps in write-committed txns (#9629) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/9629 Pessimistic transactions use pessimistic concurrency control, i.e. locking. Keys are locked upon first operation that writes the key or has the intention of writing. For example, `PessimisticTransaction::Put()`, `PessimisticTransaction::Delete()`, `PessimisticTransaction::SingleDelete()` will write to or delete a key, while `PessimisticTransaction::GetForUpdate()` is used by application to indicate to RocksDB that the transaction has the intention of performing write operation later in the same transaction. Pessimistic transactions support two-phase commit (2PC). A transaction can be `Prepared()`'ed and then `Commit()`. The prepare phase is similar to a promise: once `Prepare()` succeeds, the transaction has acquired the necessary resources to commit. The resources include locks, persistence of WAL, etc. Write-committed transaction is the default pessimistic transaction implementation. In RocksDB write-committed transaction, `Prepare()` will write data to the WAL as a prepare section. `Commit()` will write a commit marker to the WAL and then write data to the memtables. While writing to the memtables, different keys in the transaction's write batch will be assigned different sequence numbers in ascending order. Until commit/rollback, the transaction holds locks on the keys so that no other transaction can write to the same keys. Furthermore, the keys' sequence numbers represent the order in which they are committed and should be made visible. This is convenient for us to implement support for user-defined timestamps. Since column families with and without timestamps can co-exist in the same database, a transaction may or may not involve timestamps. Based on this observation, we add two optional members to each `PessimisticTransaction`, `read_timestamp_` and `commit_timestamp_`. If no key in the transaction's write batch has timestamp, then setting these two variables do not have any effect. For the rest of this commit, we discuss only the cases when these two variables are meaningful. read_timestamp_ is used mainly for validation, and should be set before first call to `GetForUpdate()`. Otherwise, the latter will return non-ok status. `GetForUpdate()` calls `TryLock()` that can verify if another transaction has written the same key since `read_timestamp_` till this call to `GetForUpdate()`. If another transaction has indeed written the same key, then validation fails, and RocksDB allows this transaction to refine `read_timestamp_` by increasing it. Note that a transaction can still use `Get()` with a different timestamp to read, but the result of the read should not be used to determine data that will be written later. commit_timestamp_ must be set after finishing writing and before transaction commit. This applies to both 2PC and non-2PC cases. In the case of 2PC, it's usually set after prepare phase succeeds. We currently require that the commit timestamp be chosen after all keys are locked. This means we disallow the `TransactionDB`-level APIs if user-defined timestamp is used by the transaction. Specifically, calling `PessimisticTransactionDB::Put()`, `PessimisticTransactionDB::Delete()`, `PessimisticTransactionDB::SingleDelete()`, etc. will return non-ok status because they specify timestamps before locking the keys. Users are also prompted to use the `Transaction` APIs when they receive the non-ok status. Reviewed By: ltamasi Differential Revision: D31822445 fbshipit-source-id: b82abf8e230216dc89cc519564a588224a88fd43
2022-03-09 00:20:59 +00:00
static Status FailIfBatchHasTs(const WriteBatch* wb);
static Status FailIfCfEnablesTs(const DB* db,
const ColumnFamilyHandle* column_family);
void ReinitializeTransaction(
Transaction* txn, const WriteOptions& write_options,
const TransactionOptions& txn_options = TransactionOptions());
virtual Status VerifyCFOptions(const ColumnFamilyOptions& cf_options);
private:
friend class WritePreparedTxnDB;
friend class WritePreparedTxnDBMock;
friend class WriteUnpreparedTxn;
friend class TransactionTest_DoubleCrashInRecovery_Test;
friend class TransactionTest_DoubleEmptyWrite_Test;
friend class TransactionTest_DuplicateKeys_Test;
friend class TransactionTest_PersistentTwoPhaseTransactionTest_Test;
friend class TransactionTest_TwoPhaseDoubleRecoveryTest_Test;
friend class TransactionTest_TwoPhaseOutOfOrderDelete_Test;
friend class TransactionStressTest_TwoPhaseLongPrepareTest_Test;
friend class WriteUnpreparedTransactionTest_RecoveryTest_Test;
friend class WriteUnpreparedTransactionTest_MarkLogWithPrepSection_Test;
Support user-defined timestamps in write-committed txns (#9629) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/9629 Pessimistic transactions use pessimistic concurrency control, i.e. locking. Keys are locked upon first operation that writes the key or has the intention of writing. For example, `PessimisticTransaction::Put()`, `PessimisticTransaction::Delete()`, `PessimisticTransaction::SingleDelete()` will write to or delete a key, while `PessimisticTransaction::GetForUpdate()` is used by application to indicate to RocksDB that the transaction has the intention of performing write operation later in the same transaction. Pessimistic transactions support two-phase commit (2PC). A transaction can be `Prepared()`'ed and then `Commit()`. The prepare phase is similar to a promise: once `Prepare()` succeeds, the transaction has acquired the necessary resources to commit. The resources include locks, persistence of WAL, etc. Write-committed transaction is the default pessimistic transaction implementation. In RocksDB write-committed transaction, `Prepare()` will write data to the WAL as a prepare section. `Commit()` will write a commit marker to the WAL and then write data to the memtables. While writing to the memtables, different keys in the transaction's write batch will be assigned different sequence numbers in ascending order. Until commit/rollback, the transaction holds locks on the keys so that no other transaction can write to the same keys. Furthermore, the keys' sequence numbers represent the order in which they are committed and should be made visible. This is convenient for us to implement support for user-defined timestamps. Since column families with and without timestamps can co-exist in the same database, a transaction may or may not involve timestamps. Based on this observation, we add two optional members to each `PessimisticTransaction`, `read_timestamp_` and `commit_timestamp_`. If no key in the transaction's write batch has timestamp, then setting these two variables do not have any effect. For the rest of this commit, we discuss only the cases when these two variables are meaningful. read_timestamp_ is used mainly for validation, and should be set before first call to `GetForUpdate()`. Otherwise, the latter will return non-ok status. `GetForUpdate()` calls `TryLock()` that can verify if another transaction has written the same key since `read_timestamp_` till this call to `GetForUpdate()`. If another transaction has indeed written the same key, then validation fails, and RocksDB allows this transaction to refine `read_timestamp_` by increasing it. Note that a transaction can still use `Get()` with a different timestamp to read, but the result of the read should not be used to determine data that will be written later. commit_timestamp_ must be set after finishing writing and before transaction commit. This applies to both 2PC and non-2PC cases. In the case of 2PC, it's usually set after prepare phase succeeds. We currently require that the commit timestamp be chosen after all keys are locked. This means we disallow the `TransactionDB`-level APIs if user-defined timestamp is used by the transaction. Specifically, calling `PessimisticTransactionDB::Put()`, `PessimisticTransactionDB::Delete()`, `PessimisticTransactionDB::SingleDelete()`, etc. will return non-ok status because they specify timestamps before locking the keys. Users are also prompted to use the `Transaction` APIs when they receive the non-ok status. Reviewed By: ltamasi Differential Revision: D31822445 fbshipit-source-id: b82abf8e230216dc89cc519564a588224a88fd43
2022-03-09 00:20:59 +00:00
Transaction* BeginInternalTransaction(const WriteOptions& options);
std::shared_ptr<LockManager> lock_manager_;
// Must be held when adding/dropping column families.
InstrumentedMutex column_family_mutex_;
// Used to ensure that no locks are stolen from an expirable transaction
// that has started a commit. Only transactions with an expiration time
// should be in this map.
std::mutex map_mutex_;
std::unordered_map<TransactionID, PessimisticTransaction*>
expirable_transactions_map_;
// map from name to two phase transaction instance
std::mutex name_map_mutex_;
std::unordered_map<TransactionName, Transaction*> transactions_;
// Signal that we are testing a crash scenario. Some asserts could be relaxed
// in such cases.
virtual void TEST_Crash() {}
};
// A PessimisticTransactionDB that writes the data to the DB after the commit.
// In this way the DB only contains the committed data.
class WriteCommittedTxnDB : public PessimisticTransactionDB {
public:
explicit WriteCommittedTxnDB(DB* db,
const TransactionDBOptions& txn_db_options)
: PessimisticTransactionDB(db, txn_db_options) {}
explicit WriteCommittedTxnDB(StackableDB* db,
const TransactionDBOptions& txn_db_options)
: PessimisticTransactionDB(db, txn_db_options) {}
virtual ~WriteCommittedTxnDB() {}
Transaction* BeginTransaction(const WriteOptions& write_options,
const TransactionOptions& txn_options,
Transaction* old_txn) override;
// Optimized version of ::Write that makes use of skip_concurrency_control
// hint
using TransactionDB::Write;
Status Write(const WriteOptions& opts,
const TransactionDBWriteOptimizations& optimizations,
WriteBatch* updates) override;
Status Write(const WriteOptions& opts, WriteBatch* updates) override;
};
Support user-defined timestamps in write-committed txns (#9629) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/9629 Pessimistic transactions use pessimistic concurrency control, i.e. locking. Keys are locked upon first operation that writes the key or has the intention of writing. For example, `PessimisticTransaction::Put()`, `PessimisticTransaction::Delete()`, `PessimisticTransaction::SingleDelete()` will write to or delete a key, while `PessimisticTransaction::GetForUpdate()` is used by application to indicate to RocksDB that the transaction has the intention of performing write operation later in the same transaction. Pessimistic transactions support two-phase commit (2PC). A transaction can be `Prepared()`'ed and then `Commit()`. The prepare phase is similar to a promise: once `Prepare()` succeeds, the transaction has acquired the necessary resources to commit. The resources include locks, persistence of WAL, etc. Write-committed transaction is the default pessimistic transaction implementation. In RocksDB write-committed transaction, `Prepare()` will write data to the WAL as a prepare section. `Commit()` will write a commit marker to the WAL and then write data to the memtables. While writing to the memtables, different keys in the transaction's write batch will be assigned different sequence numbers in ascending order. Until commit/rollback, the transaction holds locks on the keys so that no other transaction can write to the same keys. Furthermore, the keys' sequence numbers represent the order in which they are committed and should be made visible. This is convenient for us to implement support for user-defined timestamps. Since column families with and without timestamps can co-exist in the same database, a transaction may or may not involve timestamps. Based on this observation, we add two optional members to each `PessimisticTransaction`, `read_timestamp_` and `commit_timestamp_`. If no key in the transaction's write batch has timestamp, then setting these two variables do not have any effect. For the rest of this commit, we discuss only the cases when these two variables are meaningful. read_timestamp_ is used mainly for validation, and should be set before first call to `GetForUpdate()`. Otherwise, the latter will return non-ok status. `GetForUpdate()` calls `TryLock()` that can verify if another transaction has written the same key since `read_timestamp_` till this call to `GetForUpdate()`. If another transaction has indeed written the same key, then validation fails, and RocksDB allows this transaction to refine `read_timestamp_` by increasing it. Note that a transaction can still use `Get()` with a different timestamp to read, but the result of the read should not be used to determine data that will be written later. commit_timestamp_ must be set after finishing writing and before transaction commit. This applies to both 2PC and non-2PC cases. In the case of 2PC, it's usually set after prepare phase succeeds. We currently require that the commit timestamp be chosen after all keys are locked. This means we disallow the `TransactionDB`-level APIs if user-defined timestamp is used by the transaction. Specifically, calling `PessimisticTransactionDB::Put()`, `PessimisticTransactionDB::Delete()`, `PessimisticTransactionDB::SingleDelete()`, etc. will return non-ok status because they specify timestamps before locking the keys. Users are also prompted to use the `Transaction` APIs when they receive the non-ok status. Reviewed By: ltamasi Differential Revision: D31822445 fbshipit-source-id: b82abf8e230216dc89cc519564a588224a88fd43
2022-03-09 00:20:59 +00:00
inline Status PessimisticTransactionDB::FailIfBatchHasTs(
const WriteBatch* batch) {
if (batch != nullptr && WriteBatchInternal::HasKeyWithTimestamp(*batch)) {
return Status::NotSupported(
"Writes with timestamp must go through transaction API instead of "
"TransactionDB.");
}
return Status::OK();
}
inline Status PessimisticTransactionDB::FailIfCfEnablesTs(
const DB* db, const ColumnFamilyHandle* column_family) {
assert(db);
column_family = column_family ? column_family : db->DefaultColumnFamily();
assert(column_family);
const Comparator* const ucmp = column_family->GetComparator();
assert(ucmp);
if (ucmp->timestamp_size() > 0) {
return Status::NotSupported(
"Write operation with user timestamp must go through the transaction "
"API instead of TransactionDB.");
}
return Status::OK();
}
Snapshots with user-specified timestamps (#9879) Summary: In RocksDB, keys are associated with (internal) sequence numbers which denote when the keys are written to the database. Sequence numbers in different RocksDB instances are unrelated, thus not comparable. It is nice if we can associate sequence numbers with their corresponding actual timestamps. One thing we can do is to support user-defined timestamp, which allows the applications to specify the format of custom timestamps and encode a timestamp with each key. More details can be found at https://github.com/facebook/rocksdb/wiki/User-defined-Timestamp-%28Experimental%29. This PR provides a different but complementary approach. We can associate rocksdb snapshots (defined in https://github.com/facebook/rocksdb/blob/7.2.fb/include/rocksdb/snapshot.h#L20) with **user-specified** timestamps. Since a snapshot is essentially an object representing a sequence number, this PR establishes a bi-directional mapping between sequence numbers and timestamps. In the past, snapshots are usually taken by readers. The current super-version is grabbed, and a `rocksdb::Snapshot` object is created with the last published sequence number of the super-version. You can see that the reader actually has no good idea of what timestamp to assign to this snapshot, because by the time the `GetSnapshot()` is called, an arbitrarily long period of time may have already elapsed since the last write, which is when the last published sequence number is written. This observation motivates the creation of "timestamped" snapshots on the write path. Currently, this functionality is exposed only to the layer of `TransactionDB`. Application can tell RocksDB to create a snapshot when a transaction commits, effectively associating the last sequence number with a timestamp. It is also assumed that application will ensure any two snapshots with timestamps should satisfy the following: ``` snapshot1.seq < snapshot2.seq iff. snapshot1.ts < snapshot2.ts ``` If the application can guarantee that when a reader takes a timestamped snapshot, there is no active writes going on in the database, then we also allow the user to use a new API `TransactionDB::CreateTimestampedSnapshot()` to create a snapshot with associated timestamp. Code example ```cpp // Create a timestamped snapshot when committing transaction. txn->SetCommitTimestamp(100); txn->SetSnapshotOnNextOperation(); txn->Commit(); // A wrapper API for convenience Status Transaction::CommitAndTryCreateSnapshot( std::shared_ptr<TransactionNotifier> notifier, TxnTimestamp ts, std::shared_ptr<const Snapshot>* ret); // Create a timestamped snapshot if caller guarantees no concurrent writes std::pair<Status, std::shared_ptr<const Snapshot>> snapshot = txn_db->CreateTimestampedSnapshot(100); ``` The snapshots created in this way will be managed by RocksDB with ref-counting and potentially shared with other readers. We provide the following APIs for readers to retrieve a snapshot given a timestamp. ```cpp // Return the timestamped snapshot correponding to given timestamp. If ts is // kMaxTxnTimestamp, then we return the latest timestamped snapshot if present. // Othersise, we return the snapshot whose timestamp is equal to `ts`. If no // such snapshot exists, then we return null. std::shared_ptr<const Snapshot> TransactionDB::GetTimestampedSnapshot(TxnTimestamp ts) const; // Return the latest timestamped snapshot if present. std::shared_ptr<const Snapshot> TransactionDB::GetLatestTimestampedSnapshot() const; ``` We also provide two additional APIs for stats collection and reporting purposes. ```cpp Status TransactionDB::GetAllTimestampedSnapshots( std::vector<std::shared_ptr<const Snapshot>>& snapshots) const; // Return timestamped snapshots whose timestamps fall in [ts_lb, ts_ub) and store them in `snapshots`. Status TransactionDB::GetTimestampedSnapshots( TxnTimestamp ts_lb, TxnTimestamp ts_ub, std::vector<std::shared_ptr<const Snapshot>>& snapshots) const; ``` To prevent the number of timestamped snapshots from growing infinitely, we provide the following API to release timestamped snapshots whose timestamps are older than or equal to a given threshold. ```cpp void TransactionDB::ReleaseTimestampedSnapshotsOlderThan(TxnTimestamp ts); ``` Before shutdown, RocksDB will release all timestamped snapshots. Comparison with user-defined timestamp and how they can be combined: User-defined timestamp persists every key with a timestamp, while timestamped snapshots maintain a volatile mapping between snapshots (sequence numbers) and timestamps. Different internal keys with the same user key but different timestamps will be treated as different by compaction, thus a newer version will not hide older versions (with smaller timestamps) unless they are eligible for garbage collection. In contrast, taking a timestamped snapshot at a certain sequence number and timestamp prevents all the keys visible in this snapshot from been dropped by compaction. Here, visible means (seq < snapshot and most recent). The timestamped snapshot supports the semantics of reading at an exact point in time. Timestamped snapshots can also be used with user-defined timestamp. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9879 Test Plan: ``` make check TEST_TMPDIR=/dev/shm make crash_test_with_txn ``` Reviewed By: siying Differential Revision: D35783919 Pulled By: riversand963 fbshipit-source-id: 586ad905e169189e19d3bfc0cb0177a7239d1bd4
2022-06-10 23:07:03 +00:00
class SnapshotCreationCallback : public PostMemTableCallback {
public:
explicit SnapshotCreationCallback(
DBImpl* dbi, TxnTimestamp commit_ts,
const std::shared_ptr<TransactionNotifier>& notifier,
std::shared_ptr<const Snapshot>& snapshot)
: db_impl_(dbi),
commit_ts_(commit_ts),
snapshot_notifier_(notifier),
snapshot_(snapshot) {
assert(db_impl_);
}
~SnapshotCreationCallback() override {
snapshot_creation_status_.PermitUncheckedError();
}
Status operator()(SequenceNumber seq, bool disable_memtable) override;
private:
DBImpl* const db_impl_;
const TxnTimestamp commit_ts_;
std::shared_ptr<TransactionNotifier> snapshot_notifier_;
std::shared_ptr<const Snapshot>& snapshot_;
Status snapshot_creation_status_;
};
} // namespace ROCKSDB_NAMESPACE