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 * @defgroup JobQueue JobQueue
22 * @author Aaron Schulz
26 * Class to handle enqueueing and running of background jobs
31 abstract class JobQueue
{
32 protected $wiki; // string; wiki ID
33 protected $type; // string; job type
34 protected $order; // string; job priority for pop()
35 protected $claimTTL; // integer; seconds
37 const QoS_Atomic
= 1; // integer; "all-or-nothing" job insertions
40 * @param $params array
42 protected function __construct( array $params ) {
43 $this->wiki
= $params['wiki'];
44 $this->type
= $params['type'];
45 $this->order
= isset( $params['order'] ) ?
$params['order'] : 'random';
46 $this->claimTTL
= isset( $params['claimTTL'] ) ?
$params['claimTTL'] : 0;
50 * Get a job queue object of the specified type.
52 * class : What job class to use (determines job type)
53 * wiki : wiki ID of the wiki the jobs are for (defaults to current wiki)
54 * type : The name of the job types this queue handles
55 * order : Order that pop() selects jobs, one of "fifo", "timestamp" or "random".
56 * If "fifo" is used, the queue will effectively be FIFO. Note that
57 * job completion will not appear to be exactly FIFO if there are multiple
58 * job runners since jobs can take different times to finish once popped.
59 * If "timestamp" is used, the queue will at least be loosely ordered
60 * by timestamp, allowing for some jobs to be popped off out of order.
61 * If "random" is used, pop() will pick jobs in random order. This might be
62 * useful for improving concurrency depending on the queue storage medium.
63 * claimTTL : If supported, the queue will recycle jobs that have been popped
64 * but not acknowledged as completed after this many seconds.
66 * @param $params array
70 final public static function factory( array $params ) {
71 $class = $params['class'];
72 if ( !MWInit
::classExists( $class ) ) {
73 throw new MWException( "Invalid job queue class '$class'." );
75 $obj = new $class( $params );
76 if ( !( $obj instanceof self
) ) {
77 throw new MWException( "Class '$class' is not a " . __CLASS__
. " class." );
83 * @return string Wiki ID
85 final public function getWiki() {
90 * @return string Job type that this queue handles
92 final public function getType() {
97 * Quickly check if the queue is empty.
98 * Queue classes should use caching if they are any slower without memcached.
102 final public function isEmpty() {
103 wfProfileIn( __METHOD__
);
104 $res = $this->doIsEmpty();
105 wfProfileOut( __METHOD__
);
110 * @see JobQueue::isEmpty()
113 abstract protected function doIsEmpty();
116 * Push a batch of jobs into the queue
118 * @param $jobs array List of Jobs
119 * @param $flags integer Bitfield (supports JobQueue::QoS_Atomic)
120 * @throws MWException
123 final public function batchPush( array $jobs, $flags = 0 ) {
124 foreach ( $jobs as $job ) {
125 if ( $job->getType() !== $this->type
) {
126 throw new MWException( "Got '{$job->getType()}' job; expected '{$this->type}'." );
129 wfProfileIn( __METHOD__
);
130 $ok = $this->doBatchPush( $jobs, $flags );
132 wfIncrStats( 'job-insert', count( $jobs ) );
134 wfProfileOut( __METHOD__
);
139 * @see JobQueue::batchPush()
142 abstract protected function doBatchPush( array $jobs, $flags );
145 * Pop a job off of the queue
147 * @return Job|bool Returns false on failure
149 final public function pop() {
150 wfProfileIn( __METHOD__
);
151 $job = $this->doPop();
153 wfIncrStats( 'job-pop' );
155 wfProfileOut( __METHOD__
);
160 * @see JobQueue::pop()
163 abstract protected function doPop();
166 * Acknowledge that a job was completed
169 * @throws MWException
172 final public function ack( Job
$job ) {
173 if ( $job->getType() !== $this->type
) {
174 throw new MWException( "Got '{$job->getType()}' job; expected '{$this->type}'." );
176 wfProfileIn( __METHOD__
);
177 $ok = $this->doAck( $job );
178 wfProfileOut( __METHOD__
);
183 * @see JobQueue::ack()
186 abstract protected function doAck( Job
$job );
189 * Register the "root job" of a given job into the queue for de-duplication.
190 * This should only be called right *after* all the new jobs have been inserted.
191 * This is used to turn older, duplicate, job entries into no-ops. The root job
192 * information will remain in the registry until it simply falls out of cache.
194 * This requires that $job has two special fields in the "params" array:
195 * - rootJobSignature : hash (e.g. SHA1) that identifies the task
196 * - rootJobTimestamp : TS_MW timestamp of this instance of the task
198 * A "root job" is a conceptual job that consist of potentially many smaller jobs
199 * that are actually inserted into the queue. For example, "refreshLinks" jobs are
200 * spawned when a template is edited. One can think of the task as "update links
201 * of pages that use template X" and an instance of that task as a "root job".
202 * However, what actually goes into the queue are potentially many refreshLinks2 jobs.
203 * Since these jobs include things like page ID ranges and DB master positions, and morph
204 * into smaller refreshLinks2 jobs recursively, simple duplicate detection (like job_sha1)
205 * for individual jobs being identical is not useful.
207 * In the case of "refreshLinks", if these jobs are still in the queue when the template
208 * is edited again, we want all of these old refreshLinks jobs for that template to become
209 * no-ops. This can greatly reduce server load, since refreshLinks jobs involves parsing.
210 * Essentially, the new batch of jobs belong to a new "root job" and the older ones to a
211 * previous "root job" for the same task of "update links of pages that use template X".
214 * @throws MWException
217 final public function deduplicateRootJob( Job
$job ) {
218 if ( $job->getType() !== $this->type
) {
219 throw new MWException( "Got '{$job->getType()}' job; expected '{$this->type}'." );
221 wfProfileIn( __METHOD__
);
222 $ok = $this->doDeduplicateRootJob( $job );
223 wfProfileOut( __METHOD__
);
228 * @see JobQueue::deduplicateRootJob()
232 protected function doDeduplicateRootJob( Job
$job ) {
237 * Wait for any slaves or backup servers to catch up
241 final public function waitForBackups() {
242 wfProfileIn( __METHOD__
);
243 $this->doWaitForBackups();
244 wfProfileOut( __METHOD__
);
248 * @see JobQueue::waitForBackups()
251 protected function doWaitForBackups() {}