3 * Redis-backed job queue code.
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
21 * @author Aaron Schulz
25 * Class to handle job queues stored in Redis
27 * This is faster, less resource intensive, queue that JobQueueDB.
28 * All data for a queue using this class is placed into one redis server.
30 * There are eight main redis keys used to track jobs:
31 * - l-unclaimed : A list of job IDs used for ready unclaimed jobs
32 * - z-claimed : A sorted set of (job ID, UNIX timestamp as score) used for job retries
33 * - z-abandoned : A sorted set of (job ID, UNIX timestamp as score) used for broken jobs
34 * - z-delayed : A sorted set of (job ID, UNIX timestamp as score) used for delayed jobs
35 * - h-idBySha1 : A hash of (SHA1 => job ID) for unclaimed jobs used for de-duplication
36 * - h-sha1ById : A hash of (job ID => SHA1) for unclaimed jobs used for de-duplication
37 * - h-attempts : A hash of (job ID => attempt count) used for job claiming/retries
38 * - h-data : A hash of (job ID => serialized blobs) for job storage
39 * A job ID can be in only one of z-delayed, l-unclaimed, z-claimed, and z-abandoned.
40 * If an ID appears in any of those lists, it should have a h-data entry for its ID.
41 * If a job has a SHA1 de-duplication value and its ID is in l-unclaimed or z-delayed, then
42 * there should be no other such jobs with that SHA1. Every h-idBySha1 entry has an h-sha1ById
43 * entry and every h-sha1ById must refer to an ID that is l-unclaimed. If a job has its
44 * ID in z-claimed or z-abandoned, then it must also have an h-attempts entry for its ID.
46 * Additionally, "rootjob:* keys track "root jobs" used for additional de-duplication.
47 * Aside from root job keys, all keys have no expiry, and are only removed when jobs are run.
48 * All the keys are prefixed with the relevant wiki ID information.
50 * This class requires Redis 2.6 as it makes use Lua scripts for fast atomic operations.
51 * Additionally, it should be noted that redis has different persistence modes, such
52 * as rdb snapshots, journaling, and no persistent. Appropriate configuration should be
53 * made on the servers based on what queues are using it and what tolerance they have.
59 class JobQueueRedis
extends JobQueue
{
60 /** @var RedisConnectionPool */
63 protected $server; // string; server address
65 const MAX_AGE_PRUNE
= 604800; // integer; seconds a job can live once claimed (7 days)
67 protected $key; // string; key to prefix the queue keys with (used for testing)
71 * - redisConfig : An array of parameters to RedisConnectionPool::__construct().
72 * Note that the serializer option is ignored "none" is always used.
73 * - redisServer : A hostname/port combination or the absolute path of a UNIX socket.
74 * If a hostname is specified but no port, the standard port number
75 * 6379 will be used. Required.
76 * @param array $params
78 public function __construct( array $params ) {
79 parent
::__construct( $params );
80 $params['redisConfig']['serializer'] = 'none'; // make it easy to use Lua
81 $this->server
= $params['redisServer'];
82 $this->redisPool
= RedisConnectionPool
::singleton( $params['redisConfig'] );
85 protected function supportedOrders() {
86 return array( 'timestamp', 'fifo' );
89 protected function optimalOrder() {
93 protected function supportsDelayedJobs() {
98 * @see JobQueue::doIsEmpty()
100 * @throws MWException
102 protected function doIsEmpty() {
103 return $this->doGetSize() == 0;
107 * @see JobQueue::doGetSize()
109 * @throws MWException
111 protected function doGetSize() {
112 $conn = $this->getConnection();
114 return $conn->lSize( $this->getQueueKey( 'l-unclaimed' ) );
115 } catch ( RedisException
$e ) {
116 $this->throwRedisException( $this->server
, $conn, $e );
121 * @see JobQueue::doGetAcquiredCount()
123 * @throws MWException
125 protected function doGetAcquiredCount() {
126 if ( $this->claimTTL
<= 0 ) {
127 return 0; // no acknowledgements
129 $conn = $this->getConnection();
131 $conn->multi( Redis
::PIPELINE
);
132 $conn->zSize( $this->getQueueKey( 'z-claimed' ) );
133 $conn->zSize( $this->getQueueKey( 'z-abandoned' ) );
134 return array_sum( $conn->exec() );
135 } catch ( RedisException
$e ) {
136 $this->throwRedisException( $this->server
, $conn, $e );
141 * @see JobQueue::doGetDelayedCount()
143 * @throws MWException
145 protected function doGetDelayedCount() {
146 if ( !$this->checkDelay
) {
147 return 0; // no delayed jobs
149 $conn = $this->getConnection();
151 return $conn->zSize( $this->getQueueKey( 'z-delayed' ) );
152 } catch ( RedisException
$e ) {
153 $this->throwRedisException( $this->server
, $conn, $e );
158 * @see JobQueue::doGetAbandonedCount()
160 * @throws MWException
162 protected function doGetAbandonedCount() {
163 if ( $this->claimTTL
<= 0 ) {
164 return 0; // no acknowledgements
166 $conn = $this->getConnection();
168 return $conn->zSize( $this->getQueueKey( 'z-abandoned' ) );
169 } catch ( RedisException
$e ) {
170 $this->throwRedisException( $this->server
, $conn, $e );
175 * @see JobQueue::doBatchPush()
179 * @throws MWException
181 protected function doBatchPush( array $jobs, $flags ) {
182 // Convert the jobs into field maps (de-duplicated against each other)
183 $items = array(); // (job ID => job fields map)
184 foreach ( $jobs as $job ) {
185 $item = $this->getNewJobFields( $job );
186 if ( strlen( $item['sha1'] ) ) { // hash identifier => de-duplicate
187 $items[$item['sha1']] = $item;
189 $items[$item['uuid']] = $item;
193 if ( !count( $items ) ) {
194 return true; // nothing to do
197 $conn = $this->getConnection();
199 // Actually push the non-duplicate jobs into the queue...
200 if ( $flags & self
::QOS_ATOMIC
) {
201 $batches = array( $items ); // all or nothing
203 $batches = array_chunk( $items, 500 ); // avoid tying up the server
207 foreach ( $batches as $itemBatch ) {
208 $added = $this->pushBlobs( $conn, $itemBatch );
209 if ( is_int( $added ) ) {
212 $failed +
= count( $itemBatch );
216 wfDebugLog( 'JobQueueRedis', "Could not insert {$failed} {$this->type} job(s)." );
219 JobQueue
::incrStats( 'job-insert', $this->type
, count( $items ) );
220 JobQueue
::incrStats( 'job-insert-duplicate', $this->type
,
221 count( $items ) - $failed - $pushed );
222 } catch ( RedisException
$e ) {
223 $this->throwRedisException( $this->server
, $conn, $e );
230 * @param RedisConnRef $conn
231 * @param array $items List of results from JobQueueRedis::getNewJobFields()
232 * @return integer Number of jobs inserted (duplicates are ignored)
233 * @throws RedisException
235 protected function pushBlobs( RedisConnRef
$conn, array $items ) {
236 $args = array(); // ([id, sha1, rtime, blob [, id, sha1, rtime, blob ... ] ] )
237 foreach ( $items as $item ) {
238 $args[] = (string)$item['uuid'];
239 $args[] = (string)$item['sha1'];
240 $args[] = (string)$item['rtimestamp'];
241 $args[] = (string)serialize( $item );
245 if #ARGV % 4 ~= 0 then return redis.error_reply('Unmatched arguments') end
248 local id,sha1,rtimestamp,blob = ARGV[i],ARGV[i+1],ARGV[i+2],ARGV[i+3]
249 if sha1 == '' or redis.call('hExists',KEYS[3],sha1) == 0 then
250 if 1*rtimestamp > 0 then
251 -- Insert into delayed queue (release time as score)
252 redis.call('zAdd',KEYS[4],rtimestamp,id)
254 -- Insert into unclaimed queue
255 redis.call('lPush',KEYS[1],id)
258 redis.call('hSet',KEYS[2],id,sha1)
259 redis.call('hSet',KEYS[3],sha1,id)
261 redis.call('hSet',KEYS[5],id,blob)
267 return $conn->luaEval( $script,
270 $this->getQueueKey( 'l-unclaimed' ), # KEYS[1]
271 $this->getQueueKey( 'h-sha1ById' ), # KEYS[2]
272 $this->getQueueKey( 'h-idBySha1' ), # KEYS[3]
273 $this->getQueueKey( 'z-delayed' ), # KEYS[4]
274 $this->getQueueKey( 'h-data' ), # KEYS[5]
278 5 # number of first argument(s) that are keys
283 * @see JobQueue::doPop()
285 * @throws MWException
287 protected function doPop() {
290 // Push ready delayed jobs into the queue every 10 jobs to spread the load.
291 // This is also done as a periodic task, but we don't want too much done at once.
292 if ( $this->checkDelay
&& mt_rand( 0, 9 ) == 0 ) {
293 $this->releaseReadyDelayedJobs();
296 $conn = $this->getConnection();
299 if ( $this->claimTTL
> 0 ) {
300 // Keep the claimed job list down for high-traffic queues
301 if ( mt_rand( 0, 99 ) == 0 ) {
302 $this->recycleAndDeleteStaleJobs();
304 $blob = $this->popAndAcquireBlob( $conn );
306 $blob = $this->popAndDeleteBlob( $conn );
308 if ( $blob === false ) {
309 break; // no jobs; nothing to do
312 JobQueue
::incrStats( 'job-pop', $this->type
);
313 $item = unserialize( $blob );
314 if ( $item === false ) {
315 wfDebugLog( 'JobQueueRedis', "Could not unserialize {$this->type} job." );
319 // If $item is invalid, recycleAndDeleteStaleJobs() will cleanup as needed
320 $job = $this->getJobFromFields( $item ); // may be false
321 } while ( !$job ); // job may be false if invalid
322 } catch ( RedisException
$e ) {
323 $this->throwRedisException( $this->server
, $conn, $e );
330 * @param RedisConnRef $conn
331 * @return array serialized string or false
332 * @throws RedisException
334 protected function popAndDeleteBlob( RedisConnRef
$conn ) {
337 -- Pop an item off the queue
338 local id = redis.call('rpop',KEYS[1])
339 if not id then return false end
340 -- Get the job data and remove it
341 local item = redis.call('hGet',KEYS[4],id)
342 redis.call('hDel',KEYS[4],id)
343 -- Allow new duplicates of this job
344 local sha1 = redis.call('hGet',KEYS[2],id)
345 if sha1 then redis.call('hDel',KEYS[3],sha1) end
346 redis.call('hDel',KEYS[2],id)
347 -- Return the job data
350 return $conn->luaEval( $script,
352 $this->getQueueKey( 'l-unclaimed' ), # KEYS[1]
353 $this->getQueueKey( 'h-sha1ById' ), # KEYS[2]
354 $this->getQueueKey( 'h-idBySha1' ), # KEYS[3]
355 $this->getQueueKey( 'h-data' ), # KEYS[4]
357 4 # number of first argument(s) that are keys
362 * @param RedisConnRef $conn
363 * @return array serialized string or false
364 * @throws RedisException
366 protected function popAndAcquireBlob( RedisConnRef
$conn ) {
369 -- Pop an item off the queue
370 local id = redis.call('rPop',KEYS[1])
371 if not id then return false end
372 -- Allow new duplicates of this job
373 local sha1 = redis.call('hGet',KEYS[2],id)
374 if sha1 then redis.call('hDel',KEYS[3],sha1) end
375 redis.call('hDel',KEYS[2],id)
376 -- Mark the jobs as claimed and return it
377 redis.call('zAdd',KEYS[4],ARGV[1],id)
378 redis.call('hIncrBy',KEYS[5],id,1)
379 return redis.call('hGet',KEYS[6],id)
381 return $conn->luaEval( $script,
383 $this->getQueueKey( 'l-unclaimed' ), # KEYS[1]
384 $this->getQueueKey( 'h-sha1ById' ), # KEYS[2]
385 $this->getQueueKey( 'h-idBySha1' ), # KEYS[3]
386 $this->getQueueKey( 'z-claimed' ), # KEYS[4]
387 $this->getQueueKey( 'h-attempts' ), # KEYS[5]
388 $this->getQueueKey( 'h-data' ), # KEYS[6]
389 time(), # ARGV[1] (injected to be replication-safe)
391 6 # number of first argument(s) that are keys
396 * @see JobQueue::doAck()
399 * @throws MWException
401 protected function doAck( Job
$job ) {
402 if ( !isset( $job->metadata
['uuid'] ) ) {
403 throw new MWException( "Job of type '{$job->getType()}' has no UUID." );
405 if ( $this->claimTTL
> 0 ) {
406 $conn = $this->getConnection();
410 -- Unmark the job as claimed
411 redis.call('zRem',KEYS[1],ARGV[1])
412 redis.call('hDel',KEYS[2],ARGV[1])
413 -- Delete the job data itself
414 return redis.call('hDel',KEYS[3],ARGV[1])
416 $res = $conn->luaEval( $script,
418 $this->getQueueKey( 'z-claimed' ), # KEYS[1]
419 $this->getQueueKey( 'h-attempts' ), # KEYS[2]
420 $this->getQueueKey( 'h-data' ), # KEYS[3]
421 $job->metadata
['uuid'] # ARGV[1]
423 3 # number of first argument(s) that are keys
427 wfDebugLog( 'JobQueueRedis', "Could not acknowledge {$this->type} job." );
430 } catch ( RedisException
$e ) {
431 $this->throwRedisException( $this->server
, $conn, $e );
438 * @see JobQueue::doDeduplicateRootJob()
441 * @throws MWException
443 protected function doDeduplicateRootJob( Job
$job ) {
444 if ( !$job->hasRootJobParams() ) {
445 throw new MWException( "Cannot register root job; missing parameters." );
447 $params = $job->getRootJobParams();
449 $key = $this->getRootJobCacheKey( $params['rootJobSignature'] );
451 $conn = $this->getConnection();
453 $timestamp = $conn->get( $key ); // current last timestamp of this job
454 if ( $timestamp && $timestamp >= $params['rootJobTimestamp'] ) {
455 return true; // a newer version of this root job was enqueued
457 // Update the timestamp of the last root job started at the location...
458 return $conn->set( $key, $params['rootJobTimestamp'], self
::ROOTJOB_TTL
); // 2 weeks
459 } catch ( RedisException
$e ) {
460 $this->throwRedisException( $this->server
, $conn, $e );
465 * @see JobQueue::doIsRootJobOldDuplicate()
469 protected function doIsRootJobOldDuplicate( Job
$job ) {
470 if ( !$job->hasRootJobParams() ) {
471 return false; // job has no de-deplication info
473 $params = $job->getRootJobParams();
475 $conn = $this->getConnection();
477 // Get the last time this root job was enqueued
478 $timestamp = $conn->get( $this->getRootJobCacheKey( $params['rootJobSignature'] ) );
479 } catch ( RedisException
$e ) {
480 $this->throwRedisException( $this->server
, $conn, $e );
483 // Check if a new root job was started at the location after this one's...
484 return ( $timestamp && $timestamp > $params['rootJobTimestamp'] );
488 * @see JobQueue::doDelete()
491 protected function doDelete() {
492 static $props = array( 'l-unclaimed', 'z-claimed', 'z-abandoned',
493 'z-delayed', 'h-idBySha1', 'h-sha1ById', 'h-attempts', 'h-data' );
495 $conn = $this->getConnection();
498 foreach ( $props as $prop ) {
499 $keys[] = $this->getQueueKey( $prop );
501 $res = ( $conn->delete( $keys ) !== false );
502 } catch ( RedisException
$e ) {
503 $this->throwRedisException( $this->server
, $conn, $e );
508 * @see JobQueue::getAllQueuedJobs()
511 public function getAllQueuedJobs() {
512 $conn = $this->getConnection();
515 return new MappedIterator(
516 $conn->lRange( $this->getQueueKey( 'l-unclaimed' ), 0, -1 ),
517 function( $uid ) use ( $that, $conn ) {
518 return $that->getJobFromUidInternal( $uid, $conn );
521 } catch ( RedisException
$e ) {
522 $this->throwRedisException( $this->server
, $conn, $e );
527 * @see JobQueue::getAllQueuedJobs()
530 public function getAllDelayedJobs() {
531 $conn = $this->getConnection();
534 return new MappedIterator( // delayed jobs
535 $conn->zRange( $this->getQueueKey( 'z-delayed' ), 0, -1 ),
536 function( $uid ) use ( $that, $conn ) {
537 return $that->getJobFromUidInternal( $uid, $conn );
540 } catch ( RedisException
$e ) {
541 $this->throwRedisException( $this->server
, $conn, $e );
546 * This function should not be called outside JobQueueRedis
549 * @param $conn RedisConnRef
551 * @throws MWException
553 public function getJobFromUidInternal( $uid, RedisConnRef
$conn ) {
555 $item = unserialize( $conn->hGet( $this->getQueueKey( 'h-data' ), $uid ) );
556 if ( !is_array( $item ) ) { // this shouldn't happen
557 throw new MWException( "Could not find job with ID '$uid'." );
559 $title = Title
::makeTitle( $item['namespace'], $item['title'] );
560 $job = Job
::factory( $item['type'], $title, $item['params'] );
561 $job->metadata
['uuid'] = $item['uuid'];
563 } catch ( RedisException
$e ) {
564 $this->throwRedisException( $this->server
, $conn, $e );
569 * Release any ready delayed jobs into the queue
571 * @return integer Number of jobs released
572 * @throws MWException
574 public function releaseReadyDelayedJobs() {
577 $conn = $this->getConnection();
581 -- Get the list of ready delayed jobs, sorted by readiness
582 local ids = redis.call('zRangeByScore',KEYS[1],0,ARGV[1])
583 -- Migrate the jobs from the "delayed" set to the "unclaimed" list
584 for k,id in ipairs(ids) do
585 redis.call('lPush',KEYS[2],id)
586 redis.call('zRem',KEYS[1],id)
590 $count +
= (int)$conn->luaEval( $script,
592 $this->getQueueKey( 'z-delayed' ), // KEYS[1]
593 $this->getQueueKey( 'l-unclaimed' ), // KEYS[2]
594 time() // ARGV[1]; max "delay until" UNIX timestamp
596 2 # first two arguments are keys
598 } catch ( RedisException
$e ) {
599 $this->throwRedisException( $this->server
, $conn, $e );
606 * Recycle or destroy any jobs that have been claimed for too long
608 * @return integer Number of jobs recycled/deleted
609 * @throws MWException
611 public function recycleAndDeleteStaleJobs() {
612 if ( $this->claimTTL
<= 0 ) { // sanity
613 throw new MWException( "Cannot recycle jobs since acknowledgements are disabled." );
616 // For each job item that can be retried, we need to add it back to the
617 // main queue and remove it from the list of currenty claimed job items.
618 // For those that cannot, they are marked as dead and kept around for
619 // investigation and manual job restoration but are eventually deleted.
620 $conn = $this->getConnection();
625 local released,abandoned,pruned = 0,0,0
626 -- Get all non-dead jobs that have an expired claim on them.
627 -- The score for each item is the last claim timestamp (UNIX).
628 local staleClaims = redis.call('zRangeByScore',KEYS[1],0,ARGV[1])
629 for k,id in ipairs(staleClaims) do
630 local timestamp = redis.call('zScore',KEYS[1],id)
631 local attempts = redis.call('hGet',KEYS[2],id)
632 if attempts < ARGV[3] then
633 -- Claim expired and retries left: re-enqueue the job
634 redis.call('lPush',KEYS[3],id)
635 redis.call('hIncrBy',KEYS[2],id,1)
636 released = released + 1
638 -- Claim expired and no retries left: mark the job as dead
639 redis.call('zAdd',KEYS[5],timestamp,id)
640 abandoned = abandoned + 1
642 redis.call('zRem',KEYS[1],id)
644 -- Get all of the dead jobs that have been marked as dead for too long.
645 -- The score for each item is the last claim timestamp (UNIX).
646 local deadClaims = redis.call('zRangeByScore',KEYS[5],0,ARGV[2])
647 for k,id in ipairs(deadClaims) do
648 -- Stale and out of retries: remove any traces of the job
649 redis.call('zRem',KEYS[5],id)
650 redis.call('hDel',KEYS[2],id)
651 redis.call('hDel',KEYS[4],id)
654 return {released,abandoned,pruned}
656 $res = $conn->luaEval( $script,
658 $this->getQueueKey( 'z-claimed' ), # KEYS[1]
659 $this->getQueueKey( 'h-attempts' ), # KEYS[2]
660 $this->getQueueKey( 'l-unclaimed' ), # KEYS[3]
661 $this->getQueueKey( 'h-data' ), # KEYS[4]
662 $this->getQueueKey( 'z-abandoned' ), # KEYS[5]
663 $now - $this->claimTTL
, # ARGV[1]
664 $now - self
::MAX_AGE_PRUNE
, # ARGV[2]
665 $this->maxTries
# ARGV[3]
667 5 # number of first argument(s) that are keys
670 list( $released, $abandoned, $pruned ) = $res;
671 $count +
= $released +
$pruned;
672 JobQueue
::incrStats( 'job-recycle', $this->type
, $released );
673 JobQueue
::incrStats( 'job-abandon', $this->type
, $abandoned );
675 } catch ( RedisException
$e ) {
676 $this->throwRedisException( $this->server
, $conn, $e );
685 protected function doGetPeriodicTasks() {
687 if ( $this->claimTTL
> 0 ) {
688 $tasks['recycleAndDeleteStaleJobs'] = array(
689 'callback' => array( $this, 'recycleAndDeleteStaleJobs' ),
690 'period' => ceil( $this->claimTTL
/ 2 )
693 if ( $this->checkDelay
) {
694 $tasks['releaseReadyDelayedJobs'] = array(
695 'callback' => array( $this, 'releaseReadyDelayedJobs' ),
696 'period' => 300 // 5 minutes
706 protected function getNewJobFields( Job
$job ) {
708 // Fields that describe the nature of the job
709 'type' => $job->getType(),
710 'namespace' => $job->getTitle()->getNamespace(),
711 'title' => $job->getTitle()->getDBkey(),
712 'params' => $job->getParams(),
713 // Some jobs cannot run until a "release timestamp"
714 'rtimestamp' => $job->getReleaseTimestamp() ?
: 0,
715 // Additional job metadata
716 'uuid' => UIDGenerator
::newRawUUIDv4( UIDGenerator
::QUICK_RAND
),
717 'sha1' => $job->ignoreDuplicates()
718 ?
wfBaseConvert( sha1( serialize( $job->getDeduplicationInfo() ) ), 16, 36, 31 )
720 'timestamp' => time() // UNIX timestamp
725 * @param $fields array
728 protected function getJobFromFields( array $fields ) {
729 $title = Title
::makeTitleSafe( $fields['namespace'], $fields['title'] );
731 $job = Job
::factory( $fields['type'], $title, $fields['params'] );
732 $job->metadata
['uuid'] = $fields['uuid'];
739 * Get a connection to the server that handles all sub-queues for this queue
741 * @return Array (server name, Redis instance)
742 * @throws MWException
744 protected function getConnection() {
745 $conn = $this->redisPool
->getConnection( $this->server
);
747 throw new MWException( "Unable to connect to redis server." );
753 * @param $server string
754 * @param $conn RedisConnRef
755 * @param $e RedisException
756 * @throws MWException
758 protected function throwRedisException( $server, RedisConnRef
$conn, $e ) {
759 $this->redisPool
->handleException( $server, $conn, $e );
760 throw new MWException( "Redis server error: {$e->getMessage()}\n" );
764 * @param $prop string
767 private function getQueueKey( $prop ) {
768 list( $db, $prefix ) = wfSplitWikiID( $this->wiki
);
769 if ( strlen( $this->key
) ) { // namespaced queue (for testing)
770 return wfForeignMemcKey( $db, $prefix, 'jobqueue', $this->type
, $this->key
, $prop );
772 return wfForeignMemcKey( $db, $prefix, 'jobqueue', $this->type
, $prop );
780 public function setTestingPrefix( $key ) {