3 * Database load balancing interface
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License along
16 * with this program; if not, write to the Free Software Foundation, Inc.,
17 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
18 * http://www.gnu.org/copyleft/gpl.html
23 namespace Wikimedia\Rdbms
;
26 use InvalidArgumentException
;
29 * Database cluster connection, tracking, load balancing, and transaction manager interface
31 * A "cluster" is considered to be one master database and zero or more replica databases.
32 * Typically, the replica DBs replicate from the master asynchronously. The first node in the
33 * "servers" configuration array is always considered the "master". However, this class can still
34 * be used when all or some of the "replica" DBs are multi-master peers of the master or even
35 * when all the DBs are non-replicating clones of each other holding read-only data. Thus, the
36 * role of "master" is in some cases merely nominal.
38 * By default, each DB server uses DBO_DEFAULT for its 'flags' setting, unless explicitly set
39 * otherwise in configuration. DBO_DEFAULT behavior depends on whether 'cliMode' is set:
40 * - In CLI mode, the flag has no effect with regards to LoadBalancer.
41 * - In non-CLI mode, the flag causes implicit transactions to be used; the first query on
42 * a database starts a transaction on that database. The transactions are meant to remain
43 * pending until either commitMasterChanges() or rollbackMasterChanges() is called. The
44 * application must have some point where it calls commitMasterChanges() near the end of
46 * Every iteration of beginMasterChanges()/commitMasterChanges() is called a "transaction round".
47 * Rounds are useful on the master DB connections because they make single-DB (and by and large
48 * multi-DB) updates in web requests all-or-nothing. Also, transactions on replica DBs are useful
49 * when REPEATABLE-READ or SERIALIZABLE isolation is used because all foriegn keys and constraints
50 * hold across separate queries in the DB transaction since the data appears within a consistent
51 * point-in-time snapshot.
53 * The typical caller will use LoadBalancer::getConnection( DB_* ) to yield a live database
54 * connection handle. The choice of which DB server to use is based on pre-defined loads for
55 * weighted random selection, adjustments thereof by LoadMonitor, and the amount of replication
56 * lag on each DB server. Lag checks might cause problems in certain setups, so they should be
57 * tuned in the server configuration maps as follows:
58 * - Master + N Replica(s): set 'max lag' to an appropriate threshold for avoiding any database
59 * lagged by this much or more. If all DBs are this lagged, then the load balancer considers
60 * the cluster to be read-only.
61 * - Galera Cluster: Seconds_Behind_Master will be 0, so there probably is nothing to tune.
62 * Note that lag is still possible depending on how wsrep-sync-wait is set server-side.
63 * - Read-only archive clones: set 'is static' in the server configuration maps. This will
64 * treat all such DBs as having 0 lag.
65 * - Externally updated dataset clones: set 'is static' in the server configuration maps.
66 * This will treat all such DBs as having 0 lag.
67 * - SQL load balancing proxy: any proxy should handle lag checks on its own, so the 'max lag'
68 * parameter should probably be set to INF in the server configuration maps. This will make
69 * the load balancer ignore whatever it detects as the lag of the logical replica is (which
70 * would probably just randomly bounce around).
72 * If using a SQL proxy service, it would probably be best to have two proxy hosts for the
73 * load balancer to talk to. One would be the 'host' of the master server entry and another for
74 * the (logical) replica server entry. The proxy could map the load balancer's "replica" DB to
75 * any number of physical replica DBs.
80 interface ILoadBalancer
{
81 /** @var int Request a replica DB connection */
82 const DB_REPLICA
= -1;
83 /** @var int Request a master DB connection */
86 /** @var string Domain specifier when no specific database needs to be selected */
87 const DOMAIN_ANY
= '';
89 /** @var int DB handle should have DBO_TRX disabled and the caller will leave it as such */
90 const CONN_TRX_AUTOCOMMIT
= 1;
91 /** @var int Return null on connection failure instead of throwing an exception */
92 const CONN_SILENCE_ERRORS
= 2;
94 /** @var string Manager of ILoadBalancer instances is running post-commit callbacks */
95 const STAGE_POSTCOMMIT_CALLBACKS
= 'stage-postcommit-callbacks';
96 /** @var string Manager of ILoadBalancer instances is running post-rollback callbacks */
97 const STAGE_POSTROLLBACK_CALLBACKS
= 'stage-postrollback-callbacks';
100 * Construct a manager of IDatabase connection objects
102 * @param array $params Parameter map with keys:
103 * - servers : Required. Array of server info structures.
104 * - localDomain: A DatabaseDomain or domain ID string.
105 * - loadMonitor : Name of a class used to fetch server lag and load.
106 * - readOnlyReason : Reason the master DB is read-only if so [optional]
107 * - waitTimeout : Maximum time to wait for replicas for consistency [optional]
108 * - maxLag: Try to avoid DB replicas with lag above this many seconds [optional]
109 * - srvCache : BagOStuff object for server cache [optional]
110 * - wanCache : WANObjectCache object [optional]
111 * - chronologyCallback: Callback to run before the first connection attempt [optional]
112 * - hostname : The name of the current server [optional]
113 * - cliMode: Whether the execution context is a CLI script. [optional]
114 * - profiler : Class name or instance with profileIn()/profileOut() methods. [optional]
115 * - trxProfiler: TransactionProfiler instance. [optional]
116 * - replLogger: PSR-3 logger instance. [optional]
117 * - connLogger: PSR-3 logger instance. [optional]
118 * - queryLogger: PSR-3 logger instance. [optional]
119 * - perfLogger: PSR-3 logger instance. [optional]
120 * - errorLogger : Callback that takes an Exception and logs it. [optional]
121 * - deprecationLogger: Callback to log a deprecation warning. [optional]
122 * - roundStage: STAGE_POSTCOMMIT_* class constant; for internal use [optional]
123 * - ownerId: integer ID of an LBFactory instance that manages this instance [optional]
124 * @throws InvalidArgumentException
126 public function __construct( array $params );
129 * Get the local (and default) database domain ID of connection handles
131 * @see DatabaseDomain
132 * @return string Database domain ID; this specifies DB name, schema, and table prefix
135 public function getLocalDomainID();
138 * @param DatabaseDomain|string|bool $domain Database domain
139 * @return string Value of $domain if it is foreign or the local domain otherwise
142 public function resolveDomainID( $domain );
145 * Close all connection and redefine the local domain for testing or schema creation
147 * @param DatabaseDomain|string $domain
150 public function redefineLocalDomain( $domain );
153 * Get the server index of the reader connection for a given group
155 * This takes into account load ratios and lag times. It should return a consistent
156 * index during the life time of the load balancer. This initially checks replica DBs
157 * for connectivity to avoid returning an unusable server. This means that connections
158 * might be attempted by calling this method (usally one at the most but possibly more).
159 * Subsequent calls with the same $group will not need to make new connection attempts
160 * since the acquired connection for each group is preserved.
162 * @param string|bool $group Query group, or false for the generic group
163 * @param string|bool $domain Domain ID, or false for the current domain
165 * @return bool|int|string
167 public function getReaderIndex( $group = false, $domain = false );
170 * Set the master position to reach before the next generic group DB handle query
172 * If a generic replica DB connection is already open then this immediately waits
173 * for that DB to catch up to the specified replication position. Otherwise, it will
174 * do so once such a connection is opened.
176 * If a timeout happens when waiting, then getLaggedReplicaMode()/laggedReplicaUsed()
179 * @param DBMasterPos|bool $pos Master position or false
181 public function waitFor( $pos );
184 * Set the master wait position and wait for a generic replica DB to catch up to it
186 * This can be used a faster proxy for waitForAll()
188 * @param DBMasterPos|bool $pos Master position or false
189 * @param int|null $timeout Max seconds to wait; default is mWaitTimeout
190 * @return bool Success (able to connect and no timeouts reached)
192 public function waitForOne( $pos, $timeout = null );
195 * Set the master wait position and wait for ALL replica DBs to catch up to it
197 * @param DBMasterPos|bool $pos Master position or false
198 * @param int|null $timeout Max seconds to wait; default is mWaitTimeout
199 * @return bool Success (able to connect and no timeouts reached)
201 public function waitForAll( $pos, $timeout = null );
204 * Get any open connection to a given server index, local or foreign
206 * Use CONN_TRX_AUTOCOMMIT to only look for connections opened with that flag.
207 * Avoid the use of begin() or startAtomic() on any such connections.
209 * @param int $i Server index or DB_MASTER/DB_REPLICA
210 * @param int $flags Bitfield of CONN_* class constants
211 * @return Database|bool False if no such connection is open
213 public function getAnyOpenConnection( $i, $flags = 0 );
216 * Get a connection handle by server index
218 * The CONN_TRX_AUTOCOMMIT flag is ignored for databases with ATTR_DB_LEVEL_LOCKING
219 * (e.g. sqlite) in order to avoid deadlocks. ILoadBalancer::getServerAttributes()
220 * can be used to check such flags beforehand.
222 * If the caller uses $domain or sets CONN_TRX_AUTOCOMMIT in $flags, then it must
223 * also call ILoadBalancer::reuseConnection() on the handle when finished using it.
224 * In all other cases, this is not necessary, though not harmful either.
225 * Avoid the use of begin() or startAtomic() on any such connections.
227 * @param int $i Server index (overrides $groups) or DB_MASTER/DB_REPLICA
228 * @param array|string|bool $groups Query group(s), or false for the generic reader
229 * @param string|bool $domain Domain ID, or false for the current domain
230 * @param int $flags Bitfield of CONN_* class constants
232 * @note This method throws DBAccessError if ILoadBalancer::disable() was called
234 * @throws DBError If any error occurs that prevents the yielding of a (live) IDatabase
235 * @return IDatabase|bool This returns false on failure if CONN_SILENCE_ERRORS is set
237 public function getConnection( $i, $groups = [], $domain = false, $flags = 0 );
240 * Mark a foreign connection as being available for reuse under a different DB domain
242 * This mechanism is reference-counted, and must be called the same number of times
243 * as getConnection() to work.
245 * @param IDatabase $conn
246 * @throws InvalidArgumentException
248 public function reuseConnection( IDatabase
$conn );
251 * Get a database connection handle reference
253 * The handle's methods simply wrap those of a Database handle
255 * The CONN_TRX_AUTOCOMMIT flag is ignored for databases with ATTR_DB_LEVEL_LOCKING
256 * (e.g. sqlite) in order to avoid deadlocks. ILoadBalancer::getServerAttributes()
257 * can be used to check such flags beforehand. Avoid the use of begin() or startAtomic()
258 * on any CONN_TRX_AUTOCOMMIT connections.
260 * @see ILoadBalancer::getConnection() for parameter information
262 * @param int $i Server index or DB_MASTER/DB_REPLICA
263 * @param array|string|bool $groups Query group(s), or false for the generic reader
264 * @param string|bool $domain Domain ID, or false for the current domain
265 * @param int $flags Bitfield of CONN_* class constants (e.g. CONN_TRX_AUTOCOMMIT)
268 public function getConnectionRef( $i, $groups = [], $domain = false, $flags = 0 );
271 * Get a database connection handle reference without connecting yet
273 * The handle's methods simply wrap those of a Database handle
275 * The CONN_TRX_AUTOCOMMIT flag is ignored for databases with ATTR_DB_LEVEL_LOCKING
276 * (e.g. sqlite) in order to avoid deadlocks. ILoadBalancer::getServerAttributes()
277 * can be used to check such flags beforehand. Avoid the use of begin() or startAtomic()
278 * on any CONN_TRX_AUTOCOMMIT connections.
280 * @see ILoadBalancer::getConnection() for parameter information
282 * @param int $i Server index or DB_MASTER/DB_REPLICA
283 * @param array|string|bool $groups Query group(s), or false for the generic reader
284 * @param string|bool $domain Domain ID, or false for the current domain
285 * @param int $flags Bitfield of CONN_* class constants (e.g. CONN_TRX_AUTOCOMMIT)
288 public function getLazyConnectionRef( $i, $groups = [], $domain = false, $flags = 0 );
291 * Get a maintenance database connection handle reference for migrations and schema changes
293 * The handle's methods simply wrap those of a Database handle
295 * The CONN_TRX_AUTOCOMMIT flag is ignored for databases with ATTR_DB_LEVEL_LOCKING
296 * (e.g. sqlite) in order to avoid deadlocks. ILoadBalancer::getServerAttributes()
297 * can be used to check such flags beforehand. Avoid the use of begin() or startAtomic()
298 * on any CONN_TRX_AUTOCOMMIT connections.
300 * @see ILoadBalancer::getConnection() for parameter information
302 * @param int $i Server index or DB_MASTER/DB_REPLICA
303 * @param array|string|bool $groups Query group(s), or false for the generic reader
304 * @param string|bool $domain Domain ID, or false for the current domain
305 * @param int $flags Bitfield of CONN_* class constants (e.g. CONN_TRX_AUTOCOMMIT)
306 * @return MaintainableDBConnRef
308 public function getMaintenanceConnectionRef( $i, $groups = [], $domain = false, $flags = 0 );
311 * Get the server index of the master server
315 public function getWriterIndex();
318 * Returns true if the specified index is a valid server index
323 public function haveIndex( $i );
326 * Returns true if the specified index is valid and has non-zero load
331 public function isNonZeroLoad( $i );
334 * Get the number of servers defined in configuration
338 public function getServerCount();
341 * Whether there are any replica servers configured
343 * This counts both servers using streaming replication from the master server and
344 * servers that just have a clone of the static dataset found on the master server
349 public function hasReplicaServers();
352 * Whether any replica servers use streaming replication from the master server
354 * Generally this is one less than getServerCount(), though it might otherwise
355 * return a lower number if some of the servers are configured with "is static".
356 * That flag is used when both the server has no active replication setup and the
357 * dataset is either read-only or occasionally updated out-of-band. For example,
358 * a script might import a new geographic information dataset each week by writing
359 * it to each server and later directing the application to use the new version.
361 * It is possible for some replicas to be configured with "is static" but not
362 * others, though it generally should either be set for all or none of the replicas.
364 * If this returns zero, this means that there is generally no reason to execute
365 * replication wait logic for session consistency and lag reduction.
370 public function hasStreamingReplicaServers();
373 * Get the host name or IP address of the server with the specified index
376 * @return string Readable name if available or IP/host otherwise
378 public function getServerName( $i );
381 * Return the server info structure for a given index, or false if the index is invalid.
386 public function getServerInfo( $i );
389 * Get DB type of the server with the specified index
392 * @return string One of (mysql,postgres,sqlite,...) or "unknown" for bad indexes
395 public function getServerType( $i );
398 * @param int $i Server index
399 * @return array (Database::ATTRIBUTE_* constant => value) for all such constants
402 public function getServerAttributes( $i );
405 * Get the current master position for chronology control purposes
406 * @return DBMasterPos|bool Returns false if not applicable
408 public function getMasterPos();
411 * Disable this load balancer. All connections are closed, and any attempt to
412 * open a new connection will result in a DBAccessError.
414 public function disable();
417 * Close all open connections
419 public function closeAll();
424 * Using this function makes sure the LoadBalancer knows the connection is closed.
425 * If you use $conn->close() directly, the load balancer won't update its state.
427 * @param IDatabase $conn
429 public function closeConnection( IDatabase
$conn );
432 * Commit transactions on all open connections
433 * @param string $fname Caller name
434 * @param int|null $owner ID of the calling instance (e.g. the LBFactory ID)
435 * @throws DBExpectedError
437 public function commitAll( $fname = __METHOD__
, $owner = null );
440 * Run pre-commit callbacks and defer execution of post-commit callbacks
442 * Use this only for mutli-database commits
444 * @param string $fname Caller name
445 * @param int|null $owner ID of the calling instance (e.g. the LBFactory ID)
446 * @return int Number of pre-commit callbacks run (since 1.32)
448 public function finalizeMasterChanges( $fname = __METHOD__
, $owner = null );
451 * Perform all pre-commit checks for things like replication safety
453 * Use this only for mutli-database commits
455 * @param array $options Includes:
456 * - maxWriteDuration : max write query duration time in seconds
457 * @param string $fname Caller name
458 * @param int|null $owner ID of the calling instance (e.g. the LBFactory ID)
459 * @throws DBTransactionError
461 public function approveMasterChanges( array $options, $fname, $owner = null );
464 * Flush any master transaction snapshots and set DBO_TRX (if DBO_DEFAULT is set)
466 * The DBO_TRX setting will be reverted to the default in each of these methods:
467 * - commitMasterChanges()
468 * - rollbackMasterChanges()
470 * This allows for custom transaction rounds from any outer transaction scope.
472 * @param string $fname Caller name
473 * @param int|null $owner ID of the calling instance (e.g. the LBFactory ID)
474 * @throws DBExpectedError
476 public function beginMasterChanges( $fname = __METHOD__
, $owner = null );
479 * Issue COMMIT on all open master connections to flush changes and view snapshots
480 * @param string $fname Caller name
481 * @param int|null $owner ID of the calling instance (e.g. the LBFactory ID)
482 * @throws DBExpectedError
484 public function commitMasterChanges( $fname = __METHOD__
, $owner = null );
487 * Consume and run all pending post-COMMIT/ROLLBACK callbacks and commit dangling transactions
489 * @param string $fname Caller name
490 * @param int|null $owner ID of the calling instance (e.g. the LBFactory ID)
491 * @return Exception|null The first exception or null if there were none
493 public function runMasterTransactionIdleCallbacks( $fname = __METHOD__
, $owner = null );
496 * Run all recurring post-COMMIT/ROLLBACK listener callbacks
498 * @param string $fname Caller name
499 * @param int|null $owner ID of the calling instance (e.g. the LBFactory ID)
500 * @return Exception|null The first exception or null if there were none
502 public function runMasterTransactionListenerCallbacks( $fname = __METHOD__
, $owner = null );
505 * Issue ROLLBACK only on master, only if queries were done on connection
506 * @param string $fname Caller name
507 * @param int|null $owner ID of the calling instance (e.g. the LBFactory ID)
508 * @throws DBExpectedError
510 public function rollbackMasterChanges( $fname = __METHOD__
, $owner = null );
513 * Commit all replica DB transactions so as to flush any REPEATABLE-READ or SSI snapshots
515 * @param string $fname Caller name
517 public function flushReplicaSnapshots( $fname = __METHOD__
);
520 * Commit all master DB transactions so as to flush any REPEATABLE-READ or SSI snapshots
522 * An error will be thrown if a connection has pending writes or callbacks
524 * @param string $fname Caller name
526 public function flushMasterSnapshots( $fname = __METHOD__
);
529 * @return bool Whether a master connection is already open
531 public function hasMasterConnection();
534 * Whether there are pending changes or callbacks in a transaction by this thread
537 public function hasMasterChanges();
540 * Get the timestamp of the latest write query done by this thread
541 * @return float|bool UNIX timestamp or false
543 public function lastMasterChangeTimestamp();
546 * Check if this load balancer object had any recent or still
547 * pending writes issued against it by this PHP thread
549 * @param float|null $age How many seconds ago is "recent" [defaults to mWaitTimeout]
552 public function hasOrMadeRecentMasterChanges( $age = null );
555 * Get the list of callers that have pending master changes
557 * @return string[] List of method names
559 public function pendingMasterChangeCallers();
562 * @note This method will trigger a DB connection if not yet done
563 * @param string|bool $domain Domain ID, or false for the current domain
564 * @return bool Whether the database for generic connections this request is highly "lagged"
566 public function getLaggedReplicaMode( $domain = false );
569 * Checks whether the database for generic connections this request was both:
570 * - a) Already choosen due to a prior connection attempt
571 * - b) Considered highly "lagged"
573 * @note This method will never cause a new DB connection
576 public function laggedReplicaUsed();
579 * @note This method may trigger a DB connection if not yet done
580 * @param string|bool $domain Domain ID, or false for the current domain
581 * @param IDatabase|null $conn DB master connection; used to avoid loops [optional]
582 * @return string|bool Reason the master is read-only or false if it is not
584 public function getReadOnlyReason( $domain = false, IDatabase
$conn = null );
587 * Disables/enables lag checks
588 * @param null|bool $mode
591 public function allowLagged( $mode = null );
596 public function pingAll();
599 * Call a function with each open connection object
600 * @param callable $callback
601 * @param array $params
603 public function forEachOpenConnection( $callback, array $params = [] );
606 * Call a function with each open connection object to a master
607 * @param callable $callback
608 * @param array $params
610 public function forEachOpenMasterConnection( $callback, array $params = [] );
613 * Call a function with each open replica DB connection object
614 * @param callable $callback
615 * @param array $params
617 public function forEachOpenReplicaConnection( $callback, array $params = [] );
620 * Get the hostname and lag time of the most-lagged replica server
622 * This is useful for maintenance scripts that need to throttle their updates.
623 * May attempt to open connections to replica DBs on the default DB. If there is
624 * no lag, the maximum lag will be reported as -1.
626 * @param bool|string $domain Domain ID, or false for the default database
627 * @return array ( host, max lag, index of max lagged host )
629 public function getMaxLag( $domain = false );
632 * Get an estimate of replication lag (in seconds) for each server
634 * Results are cached for a short time in memcached/process cache
636 * Values may be "false" if replication is too broken to estimate
638 * @param string|bool $domain
639 * @return int[] Map of (server index => float|int|bool)
641 public function getLagTimes( $domain = false );
644 * Wait for a replica DB to reach a specified master position
646 * This will connect to the master to get an accurate position if $pos is not given
648 * @param IDatabase $conn Replica DB
649 * @param DBMasterPos|bool $pos Master position; default: current position
650 * @param int $timeout Timeout in seconds [optional]
651 * @return bool Success
654 public function waitForMasterPos( IDatabase
$conn, $pos = false, $timeout = 10 );
657 * Set a callback via IDatabase::setTransactionListener() on
658 * all current and future master connections of this load balancer
660 * @param string $name Callback name
661 * @param callable|null $callback
663 public function setTransactionListener( $name, callable
$callback = null );
666 * Set a new table prefix for the existing local domain ID for testing
668 * @param string $prefix
671 public function setLocalDomainPrefix( $prefix );
674 * Make certain table names use their own database, schema, and table prefix
675 * when passed into SQL queries pre-escaped and without a qualified database name
677 * For example, "user" can be converted to "myschema.mydbname.user" for convenience.
678 * Appearances like `user`, somedb.user, somedb.someschema.user will used literally.
680 * Calling this twice will completely clear any old table aliases. Also, note that
681 * callers are responsible for making sure the schemas and databases actually exist.
683 * @param array[] $aliases Map of (table => (dbname, schema, prefix) map)
685 public function setTableAliases( array $aliases );
688 * Convert certain index names to alternative names before querying the DB
690 * Note that this applies to indexes regardless of the table they belong to.
692 * This can be employed when an index was renamed X => Y in code, but the new Y-named
693 * indexes were not yet built on all DBs. After all the Y-named ones are added by the DBA,
694 * the aliases can be removed, and then the old X-named indexes dropped.
696 * @param string[] $aliases
699 public function setIndexAliases( array $aliases );