Merge "RCFilters: Clarify 'hidden' and 'sticky' filters"
[lhc/web/wiklou.git] / includes / libs / objectcache / WANObjectCache.php
1 <?php
2 /**
3 * This program is free software; you can redistribute it and/or modify
4 * it under the terms of the GNU General Public License as published by
5 * the Free Software Foundation; either version 2 of the License, or
6 * (at your option) any later version.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 * GNU General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public License along
14 * with this program; if not, write to the Free Software Foundation, Inc.,
15 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
16 * http://www.gnu.org/copyleft/gpl.html
17 *
18 * @file
19 * @ingroup Cache
20 */
21
22 use Liuggio\StatsdClient\Factory\StatsdDataFactoryInterface;
23 use Psr\Log\LoggerAwareInterface;
24 use Psr\Log\LoggerInterface;
25 use Psr\Log\NullLogger;
26
27 /**
28 * Multi-datacenter aware caching interface
29 *
30 * All operations go to the local datacenter cache, except for delete(),
31 * touchCheckKey(), and resetCheckKey(), which broadcast to all datacenters.
32 *
33 * This class is intended for caching data from primary stores.
34 * If the get() method does not return a value, then the caller
35 * should query the new value and backfill the cache using set().
36 * The preferred way to do this logic is through getWithSetCallback().
37 * When querying the store on cache miss, the closest DB replica
38 * should be used. Try to avoid heavyweight DB master or quorum reads.
39 * When the source data changes, a purge method should be called.
40 * Since purges are expensive, they should be avoided. One can do so if:
41 * - a) The object cached is immutable; or
42 * - b) Validity is checked against the source after get(); or
43 * - c) Using a modest TTL is reasonably correct and performant
44 *
45 * The simplest purge method is delete().
46 *
47 * There are three supported ways to handle broadcasted operations:
48 * - a) Configure the 'purge' EventRelayer to point to a valid PubSub endpoint
49 * that has subscribed listeners on the cache servers applying the cache updates.
50 * - b) Ignore the 'purge' EventRelayer configuration (default is NullEventRelayer)
51 * and set up mcrouter as the underlying cache backend, using one of the memcached
52 * BagOStuff classes as 'cache'. Use OperationSelectorRoute in the mcrouter settings
53 * to configure 'set' and 'delete' operations to go to all DCs via AllAsyncRoute and
54 * configure other operations to go to the local DC via PoolRoute (for reference,
55 * see https://github.com/facebook/mcrouter/wiki/List-of-Route-Handles).
56 * - c) Ignore the 'purge' EventRelayer configuration (default is NullEventRelayer)
57 * and set up dynomite as cache middleware between the web servers and either
58 * memcached or redis. This will also broadcast all key setting operations, not just purges,
59 * which can be useful for cache warming. Writes are eventually consistent via the
60 * Dynamo replication model (see https://github.com/Netflix/dynomite).
61 *
62 * Broadcasted operations like delete() and touchCheckKey() are done asynchronously
63 * in all datacenters this way, though the local one should likely be near immediate.
64 *
65 * This means that callers in all datacenters may see older values for however many
66 * milliseconds that the purge took to reach that datacenter. As with any cache, this
67 * should not be relied on for cases where reads are used to determine writes to source
68 * (e.g. non-cache) data stores, except when reading immutable data.
69 *
70 * All values are wrapped in metadata arrays. Keys use a "WANCache:" prefix
71 * to avoid collisions with keys that are not wrapped as metadata arrays. The
72 * prefixes are as follows:
73 * - a) "WANCache:v" : used for regular value keys
74 * - b) "WANCache:i" : used for temporarily storing values of tombstoned keys
75 * - c) "WANCache:t" : used for storing timestamp "check" keys
76 * - d) "WANCache:m" : used for temporary mutex keys to avoid cache stampedes
77 *
78 * @ingroup Cache
79 * @since 1.26
80 */
81 class WANObjectCache implements IExpiringStore, LoggerAwareInterface {
82 /** @var BagOStuff The local datacenter cache */
83 protected $cache;
84 /** @var HashBagOStuff[] Map of group PHP instance caches */
85 protected $processCaches = [];
86 /** @var string Purge channel name */
87 protected $purgeChannel;
88 /** @var EventRelayer Bus that handles purge broadcasts */
89 protected $purgeRelayer;
90 /** @var LoggerInterface */
91 protected $logger;
92 /** @var StatsdDataFactoryInterface */
93 protected $stats;
94 /** @var bool Whether to use "interim" caching while keys are tombstoned */
95 protected $useInterimHoldOffCaching = true;
96 /** @var callable|null Function that takes a WAN cache callback and runs it later */
97 protected $asyncHandler;
98
99 /** @var int ERR_* constant for the "last error" registry */
100 protected $lastRelayError = self::ERR_NONE;
101
102 /** @var int Callback stack depth for getWithSetCallback() */
103 private $callbackDepth = 0;
104 /** @var mixed[] Temporary warm-up cache */
105 private $warmupCache = [];
106 /** @var int Key fetched */
107 private $warmupKeyMisses = 0;
108
109 /** Max time expected to pass between delete() and DB commit finishing */
110 const MAX_COMMIT_DELAY = 3;
111 /** Max replication+snapshot lag before applying TTL_LAGGED or disallowing set() */
112 const MAX_READ_LAG = 7;
113 /** Seconds to tombstone keys on delete() */
114 const HOLDOFF_TTL = 11; // MAX_COMMIT_DELAY + MAX_READ_LAG + 1
115
116 /** Seconds to keep dependency purge keys around */
117 const CHECK_KEY_TTL = self::TTL_YEAR;
118 /** Seconds to keep interim value keys for tombstoned keys around */
119 const INTERIM_KEY_TTL = 1;
120
121 /** Seconds to keep lock keys around */
122 const LOCK_TTL = 10;
123 /** Default remaining TTL at which to consider pre-emptive regeneration */
124 const LOW_TTL = 30;
125 /** Default time-since-expiry on a miss that makes a key "hot" */
126 const LOCK_TSE = 1;
127
128 /** Never consider performing "popularity" refreshes until a key reaches this age */
129 const AGE_NEW = 60;
130 /** The time length of the "popularity" refresh window for hot keys */
131 const HOT_TTR = 900;
132 /** Hits/second for a refresh to be expected within the "popularity" window */
133 const HIT_RATE_HIGH = 1;
134 /** Seconds to ramp up to the "popularity" refresh chance after a key is no longer new */
135 const RAMPUP_TTL = 30;
136
137 /** Idiom for getWithSetCallback() callbacks to avoid calling set() */
138 const TTL_UNCACHEABLE = -1;
139 /** Idiom for getWithSetCallback() callbacks to 'lockTSE' logic */
140 const TSE_NONE = -1;
141 /** Max TTL to store keys when a data sourced is lagged */
142 const TTL_LAGGED = 30;
143 /** Idiom for delete() for "no hold-off" */
144 const HOLDOFF_NONE = 0;
145 /** Idiom for set()/getWithSetCallback() for "do not augment the storage medium TTL" */
146 const STALE_TTL_NONE = 0;
147 /** Idiom for set()/getWithSetCallback() for "no post-expired grace period" */
148 const GRACE_TTL_NONE = 0;
149
150 /** Idiom for getWithSetCallback() for "no minimum required as-of timestamp" */
151 const MIN_TIMESTAMP_NONE = 0.0;
152
153 /** Tiny negative float to use when CTL comes up >= 0 due to clock skew */
154 const TINY_NEGATIVE = -0.000001;
155
156 /** Cache format version number */
157 const VERSION = 1;
158
159 const FLD_VERSION = 0; // key to cache version number
160 const FLD_VALUE = 1; // key to the cached value
161 const FLD_TTL = 2; // key to the original TTL
162 const FLD_TIME = 3; // key to the cache time
163 const FLD_FLAGS = 4; // key to the flags bitfield
164 const FLD_HOLDOFF = 5; // key to any hold-off TTL
165
166 /** @var int Treat this value as expired-on-arrival */
167 const FLG_STALE = 1;
168
169 const ERR_NONE = 0; // no error
170 const ERR_NO_RESPONSE = 1; // no response
171 const ERR_UNREACHABLE = 2; // can't connect
172 const ERR_UNEXPECTED = 3; // response gave some error
173 const ERR_RELAY = 4; // relay broadcast failed
174
175 const VALUE_KEY_PREFIX = 'WANCache:v:';
176 const INTERIM_KEY_PREFIX = 'WANCache:i:';
177 const TIME_KEY_PREFIX = 'WANCache:t:';
178 const MUTEX_KEY_PREFIX = 'WANCache:m:';
179
180 const PURGE_VAL_PREFIX = 'PURGED:';
181
182 const VFLD_DATA = 'WOC:d'; // key to the value of versioned data
183 const VFLD_VERSION = 'WOC:v'; // key to the version of the value present
184
185 const PC_PRIMARY = 'primary:1000'; // process cache name and max key count
186
187 const DEFAULT_PURGE_CHANNEL = 'wancache-purge';
188
189 /**
190 * @param array $params
191 * - cache : BagOStuff object for a persistent cache
192 * - channels : Map of (action => channel string). Actions include "purge".
193 * - relayers : Map of (action => EventRelayer object). Actions include "purge".
194 * - logger : LoggerInterface object
195 * - stats : LoggerInterface object
196 * - asyncHandler : A function that takes a callback and runs it later. If supplied,
197 * whenever a preemptive refresh would be triggered in getWithSetCallback(), the
198 * current cache value is still used instead. However, the async-handler function
199 * receives a WAN cache callback that, when run, will execute the value generation
200 * callback supplied by the getWithSetCallback() caller. The result will be saved
201 * as normal. The handler is expected to call the WAN cache callback at an opportune
202 * time (e.g. HTTP post-send), though generally within a few 100ms. [optional]
203 */
204 public function __construct( array $params ) {
205 $this->cache = $params['cache'];
206 $this->purgeChannel = isset( $params['channels']['purge'] )
207 ? $params['channels']['purge']
208 : self::DEFAULT_PURGE_CHANNEL;
209 $this->purgeRelayer = isset( $params['relayers']['purge'] )
210 ? $params['relayers']['purge']
211 : new EventRelayerNull( [] );
212 $this->setLogger( isset( $params['logger'] ) ? $params['logger'] : new NullLogger() );
213 $this->stats = isset( $params['stats'] ) ? $params['stats'] : new NullStatsdDataFactory();
214 $this->asyncHandler = isset( $params['asyncHandler'] ) ? $params['asyncHandler'] : null;
215 }
216
217 public function setLogger( LoggerInterface $logger ) {
218 $this->logger = $logger;
219 }
220
221 /**
222 * Get an instance that wraps EmptyBagOStuff
223 *
224 * @return WANObjectCache
225 */
226 public static function newEmpty() {
227 return new static( [
228 'cache' => new EmptyBagOStuff(),
229 'pool' => 'empty'
230 ] );
231 }
232
233 /**
234 * Fetch the value of a key from cache
235 *
236 * If supplied, $curTTL is set to the remaining TTL (current time left):
237 * - a) INF; if $key exists, has no TTL, and is not expired by $checkKeys
238 * - b) float (>=0); if $key exists, has a TTL, and is not expired by $checkKeys
239 * - c) float (<0); if $key is tombstoned, stale, or existing but expired by $checkKeys
240 * - d) null; if $key does not exist and is not tombstoned
241 *
242 * If a key is tombstoned, $curTTL will reflect the time since delete().
243 *
244 * The timestamp of $key will be checked against the last-purge timestamp
245 * of each of $checkKeys. Those $checkKeys not in cache will have the last-purge
246 * initialized to the current timestamp. If any of $checkKeys have a timestamp
247 * greater than that of $key, then $curTTL will reflect how long ago $key
248 * became invalid. Callers can use $curTTL to know when the value is stale.
249 * The $checkKeys parameter allow mass invalidations by updating a single key:
250 * - a) Each "check" key represents "last purged" of some source data
251 * - b) Callers pass in relevant "check" keys as $checkKeys in get()
252 * - c) When the source data that "check" keys represent changes,
253 * the touchCheckKey() method is called on them
254 *
255 * Source data entities might exists in a DB that uses snapshot isolation
256 * (e.g. the default REPEATABLE-READ in innoDB). Even for mutable data, that
257 * isolation can largely be maintained by doing the following:
258 * - a) Calling delete() on entity change *and* creation, before DB commit
259 * - b) Keeping transaction duration shorter than delete() hold-off TTL
260 *
261 * However, pre-snapshot values might still be seen if an update was made
262 * in a remote datacenter but the purge from delete() didn't relay yet.
263 *
264 * Consider using getWithSetCallback() instead of get() and set() cycles.
265 * That method has cache slam avoiding features for hot/expensive keys.
266 *
267 * @param string $key Cache key made from makeKey() or makeGlobalKey()
268 * @param mixed &$curTTL Approximate TTL left on the key if present/tombstoned [returned]
269 * @param array $checkKeys List of "check" keys
270 * @param float &$asOf UNIX timestamp of cached value; null on failure [returned]
271 * @return mixed Value of cache key or false on failure
272 */
273 final public function get( $key, &$curTTL = null, array $checkKeys = [], &$asOf = null ) {
274 $curTTLs = [];
275 $asOfs = [];
276 $values = $this->getMulti( [ $key ], $curTTLs, $checkKeys, $asOfs );
277 $curTTL = isset( $curTTLs[$key] ) ? $curTTLs[$key] : null;
278 $asOf = isset( $asOfs[$key] ) ? $asOfs[$key] : null;
279
280 return isset( $values[$key] ) ? $values[$key] : false;
281 }
282
283 /**
284 * Fetch the value of several keys from cache
285 *
286 * @see WANObjectCache::get()
287 *
288 * @param array $keys List of cache keys made from makeKey() or makeGlobalKey()
289 * @param array &$curTTLs Map of (key => approximate TTL left) for existing keys [returned]
290 * @param array $checkKeys List of check keys to apply to all $keys. May also apply "check"
291 * keys to specific cache keys only by using cache keys as keys in the $checkKeys array.
292 * @param float[] &$asOfs Map of (key => UNIX timestamp of cached value; null on failure)
293 * @return array Map of (key => value) for keys that exist and are not tombstoned
294 */
295 final public function getMulti(
296 array $keys, &$curTTLs = [], array $checkKeys = [], array &$asOfs = []
297 ) {
298 $result = [];
299 $curTTLs = [];
300 $asOfs = [];
301
302 $vPrefixLen = strlen( self::VALUE_KEY_PREFIX );
303 $valueKeys = self::prefixCacheKeys( $keys, self::VALUE_KEY_PREFIX );
304
305 $checkKeysForAll = [];
306 $checkKeysByKey = [];
307 $checkKeysFlat = [];
308 foreach ( $checkKeys as $i => $checkKeyGroup ) {
309 $prefixed = self::prefixCacheKeys( (array)$checkKeyGroup, self::TIME_KEY_PREFIX );
310 $checkKeysFlat = array_merge( $checkKeysFlat, $prefixed );
311 // Is this check keys for a specific cache key, or for all keys being fetched?
312 if ( is_int( $i ) ) {
313 $checkKeysForAll = array_merge( $checkKeysForAll, $prefixed );
314 } else {
315 $checkKeysByKey[$i] = isset( $checkKeysByKey[$i] )
316 ? array_merge( $checkKeysByKey[$i], $prefixed )
317 : $prefixed;
318 }
319 }
320
321 // Fetch all of the raw values
322 $keysGet = array_merge( $valueKeys, $checkKeysFlat );
323 if ( $this->warmupCache ) {
324 $wrappedValues = array_intersect_key( $this->warmupCache, array_flip( $keysGet ) );
325 $keysGet = array_diff( $keysGet, array_keys( $wrappedValues ) ); // keys left to fetch
326 $this->warmupKeyMisses += count( $keysGet );
327 } else {
328 $wrappedValues = [];
329 }
330 if ( $keysGet ) {
331 $wrappedValues += $this->cache->getMulti( $keysGet );
332 }
333 // Time used to compare/init "check" keys (derived after getMulti() to be pessimistic)
334 $now = $this->getCurrentTime();
335
336 // Collect timestamps from all "check" keys
337 $purgeValuesForAll = $this->processCheckKeys( $checkKeysForAll, $wrappedValues, $now );
338 $purgeValuesByKey = [];
339 foreach ( $checkKeysByKey as $cacheKey => $checks ) {
340 $purgeValuesByKey[$cacheKey] =
341 $this->processCheckKeys( $checks, $wrappedValues, $now );
342 }
343
344 // Get the main cache value for each key and validate them
345 foreach ( $valueKeys as $vKey ) {
346 if ( !isset( $wrappedValues[$vKey] ) ) {
347 continue; // not found
348 }
349
350 $key = substr( $vKey, $vPrefixLen ); // unprefix
351
352 list( $value, $curTTL ) = $this->unwrap( $wrappedValues[$vKey], $now );
353 if ( $value !== false ) {
354 $result[$key] = $value;
355
356 // Force dependant keys to be invalid for a while after purging
357 // to reduce race conditions involving stale data getting cached
358 $purgeValues = $purgeValuesForAll;
359 if ( isset( $purgeValuesByKey[$key] ) ) {
360 $purgeValues = array_merge( $purgeValues, $purgeValuesByKey[$key] );
361 }
362 foreach ( $purgeValues as $purge ) {
363 $safeTimestamp = $purge[self::FLD_TIME] + $purge[self::FLD_HOLDOFF];
364 if ( $safeTimestamp >= $wrappedValues[$vKey][self::FLD_TIME] ) {
365 // How long ago this value was expired by *this* check key
366 $ago = min( $purge[self::FLD_TIME] - $now, self::TINY_NEGATIVE );
367 // How long ago this value was expired by *any* known check key
368 $curTTL = min( $curTTL, $ago );
369 }
370 }
371 }
372 $curTTLs[$key] = $curTTL;
373 $asOfs[$key] = ( $value !== false ) ? $wrappedValues[$vKey][self::FLD_TIME] : null;
374 }
375
376 return $result;
377 }
378
379 /**
380 * @since 1.27
381 * @param array $timeKeys List of prefixed time check keys
382 * @param array $wrappedValues
383 * @param float $now
384 * @return array List of purge value arrays
385 */
386 private function processCheckKeys( array $timeKeys, array $wrappedValues, $now ) {
387 $purgeValues = [];
388 foreach ( $timeKeys as $timeKey ) {
389 $purge = isset( $wrappedValues[$timeKey] )
390 ? $this->parsePurgeValue( $wrappedValues[$timeKey] )
391 : false;
392 if ( $purge === false ) {
393 // Key is not set or invalid; regenerate
394 $newVal = $this->makePurgeValue( $now, self::HOLDOFF_TTL );
395 $this->cache->add( $timeKey, $newVal, self::CHECK_KEY_TTL );
396 $purge = $this->parsePurgeValue( $newVal );
397 }
398 $purgeValues[] = $purge;
399 }
400 return $purgeValues;
401 }
402
403 /**
404 * Set the value of a key in cache
405 *
406 * Simply calling this method when source data changes is not valid because
407 * the changes do not replicate to the other WAN sites. In that case, delete()
408 * should be used instead. This method is intended for use on cache misses.
409 *
410 * If the data was read from a snapshot-isolated transactions (e.g. the default
411 * REPEATABLE-READ in innoDB), use 'since' to avoid the following race condition:
412 * - a) T1 starts
413 * - b) T2 updates a row, calls delete(), and commits
414 * - c) The HOLDOFF_TTL passes, expiring the delete() tombstone
415 * - d) T1 reads the row and calls set() due to a cache miss
416 * - e) Stale value is stuck in cache
417 *
418 * Setting 'lag' and 'since' help avoids keys getting stuck in stale states.
419 *
420 * Example usage:
421 * @code
422 * $dbr = wfGetDB( DB_REPLICA );
423 * $setOpts = Database::getCacheSetOptions( $dbr );
424 * // Fetch the row from the DB
425 * $row = $dbr->selectRow( ... );
426 * $key = $cache->makeKey( 'building', $buildingId );
427 * $cache->set( $key, $row, $cache::TTL_DAY, $setOpts );
428 * @endcode
429 *
430 * @param string $key Cache key
431 * @param mixed $value
432 * @param int $ttl Seconds to live. Special values are:
433 * - WANObjectCache::TTL_INDEFINITE: Cache forever
434 * @param array $opts Options map:
435 * - lag : Seconds of replica DB lag. Typically, this is either the replica DB lag
436 * before the data was read or, if applicable, the replica DB lag before
437 * the snapshot-isolated transaction the data was read from started.
438 * Use false to indicate that replication is not running.
439 * Default: 0 seconds
440 * - since : UNIX timestamp of the data in $value. Typically, this is either
441 * the current time the data was read or (if applicable) the time when
442 * the snapshot-isolated transaction the data was read from started.
443 * Default: 0 seconds
444 * - pending : Whether this data is possibly from an uncommitted write transaction.
445 * Generally, other threads should not see values from the future and
446 * they certainly should not see ones that ended up getting rolled back.
447 * Default: false
448 * - lockTSE : if excessive replication/snapshot lag is detected, then store the value
449 * with this TTL and flag it as stale. This is only useful if the reads for this key
450 * use getWithSetCallback() with "lockTSE" set. Note that if "staleTTL" is set
451 * then it will still add on to this TTL in the excessive lag scenario.
452 * Default: WANObjectCache::TSE_NONE
453 * - staleTTL : Seconds to keep the key around if it is stale. The get()/getMulti()
454 * methods return such stale values with a $curTTL of 0, and getWithSetCallback()
455 * will call the regeneration callback in such cases, passing in the old value
456 * and its as-of time to the callback. This is useful if adaptiveTTL() is used
457 * on the old value's as-of time when it is verified as still being correct.
458 * Default: WANObjectCache::STALE_TTL_NONE.
459 * @note Options added in 1.28: staleTTL
460 * @return bool Success
461 */
462 final public function set( $key, $value, $ttl = 0, array $opts = [] ) {
463 $now = $this->getCurrentTime();
464 $lockTSE = isset( $opts['lockTSE'] ) ? $opts['lockTSE'] : self::TSE_NONE;
465 $staleTTL = isset( $opts['staleTTL'] ) ? $opts['staleTTL'] : self::STALE_TTL_NONE;
466 $age = isset( $opts['since'] ) ? max( 0, $now - $opts['since'] ) : 0;
467 $lag = isset( $opts['lag'] ) ? $opts['lag'] : 0;
468
469 // Do not cache potentially uncommitted data as it might get rolled back
470 if ( !empty( $opts['pending'] ) ) {
471 $this->logger->info( 'Rejected set() for {cachekey} due to pending writes.',
472 [ 'cachekey' => $key ] );
473
474 return true; // no-op the write for being unsafe
475 }
476
477 $wrapExtra = []; // additional wrapped value fields
478 // Check if there's a risk of writing stale data after the purge tombstone expired
479 if ( $lag === false || ( $lag + $age ) > self::MAX_READ_LAG ) {
480 // Case A: read lag with "lockTSE"; save but record value as stale
481 if ( $lockTSE >= 0 ) {
482 $ttl = max( 1, (int)$lockTSE ); // set() expects seconds
483 $wrapExtra[self::FLD_FLAGS] = self::FLG_STALE; // mark as stale
484 // Case B: any long-running transaction; ignore this set()
485 } elseif ( $age > self::MAX_READ_LAG ) {
486 $this->logger->info( 'Rejected set() for {cachekey} due to snapshot lag.',
487 [ 'cachekey' => $key ] );
488
489 return true; // no-op the write for being unsafe
490 // Case C: high replication lag; lower TTL instead of ignoring all set()s
491 } elseif ( $lag === false || $lag > self::MAX_READ_LAG ) {
492 $ttl = $ttl ? min( $ttl, self::TTL_LAGGED ) : self::TTL_LAGGED;
493 $this->logger->warning( 'Lowered set() TTL for {cachekey} due to replication lag.',
494 [ 'cachekey' => $key ] );
495 // Case D: medium length request with medium replication lag; ignore this set()
496 } else {
497 $this->logger->info( 'Rejected set() for {cachekey} due to high read lag.',
498 [ 'cachekey' => $key ] );
499
500 return true; // no-op the write for being unsafe
501 }
502 }
503
504 // Wrap that value with time/TTL/version metadata
505 $wrapped = $this->wrap( $value, $ttl, $now ) + $wrapExtra;
506
507 $func = function ( $cache, $key, $cWrapped ) use ( $wrapped ) {
508 return ( is_string( $cWrapped ) )
509 ? false // key is tombstoned; do nothing
510 : $wrapped;
511 };
512
513 return $this->cache->merge( self::VALUE_KEY_PREFIX . $key, $func, $ttl + $staleTTL, 1 );
514 }
515
516 /**
517 * Purge a key from all datacenters
518 *
519 * This should only be called when the underlying data (being cached)
520 * changes in a significant way. This deletes the key and starts a hold-off
521 * period where the key cannot be written to for a few seconds (HOLDOFF_TTL).
522 * This is done to avoid the following race condition:
523 * - a) Some DB data changes and delete() is called on a corresponding key
524 * - b) A request refills the key with a stale value from a lagged DB
525 * - c) The stale value is stuck there until the key is expired/evicted
526 *
527 * This is implemented by storing a special "tombstone" value at the cache
528 * key that this class recognizes; get() calls will return false for the key
529 * and any set() calls will refuse to replace tombstone values at the key.
530 * For this to always avoid stale value writes, the following must hold:
531 * - a) Replication lag is bounded to being less than HOLDOFF_TTL; or
532 * - b) If lag is higher, the DB will have gone into read-only mode already
533 *
534 * Note that set() can also be lag-aware and lower the TTL if it's high.
535 *
536 * When using potentially long-running ACID transactions, a good pattern is
537 * to use a pre-commit hook to issue the delete. This means that immediately
538 * after commit, callers will see the tombstone in cache upon purge relay.
539 * It also avoids the following race condition:
540 * - a) T1 begins, changes a row, and calls delete()
541 * - b) The HOLDOFF_TTL passes, expiring the delete() tombstone
542 * - c) T2 starts, reads the row and calls set() due to a cache miss
543 * - d) T1 finally commits
544 * - e) Stale value is stuck in cache
545 *
546 * Example usage:
547 * @code
548 * $dbw->startAtomic( __METHOD__ ); // start of request
549 * ... <execute some stuff> ...
550 * // Update the row in the DB
551 * $dbw->update( ... );
552 * $key = $cache->makeKey( 'homes', $homeId );
553 * // Purge the corresponding cache entry just before committing
554 * $dbw->onTransactionPreCommitOrIdle( function() use ( $cache, $key ) {
555 * $cache->delete( $key );
556 * } );
557 * ... <execute some stuff> ...
558 * $dbw->endAtomic( __METHOD__ ); // end of request
559 * @endcode
560 *
561 * The $ttl parameter can be used when purging values that have not actually changed
562 * recently. For example, a cleanup script to purge cache entries does not really need
563 * a hold-off period, so it can use HOLDOFF_NONE. Likewise for user-requested purge.
564 * Note that $ttl limits the effective range of 'lockTSE' for getWithSetCallback().
565 *
566 * If called twice on the same key, then the last hold-off TTL takes precedence. For
567 * idempotence, the $ttl should not vary for different delete() calls on the same key.
568 *
569 * @param string $key Cache key
570 * @param int $ttl Tombstone TTL; Default: WANObjectCache::HOLDOFF_TTL
571 * @return bool True if the item was purged or not found, false on failure
572 */
573 final public function delete( $key, $ttl = self::HOLDOFF_TTL ) {
574 $key = self::VALUE_KEY_PREFIX . $key;
575
576 if ( $ttl <= 0 ) {
577 // Publish the purge to all datacenters
578 $ok = $this->relayDelete( $key );
579 } else {
580 // Publish the purge to all datacenters
581 $ok = $this->relayPurge( $key, $ttl, self::HOLDOFF_NONE );
582 }
583
584 return $ok;
585 }
586
587 /**
588 * Fetch the value of a timestamp "check" key
589 *
590 * The key will be *initialized* to the current time if not set,
591 * so only call this method if this behavior is actually desired
592 *
593 * The timestamp can be used to check whether a cached value is valid.
594 * Callers should not assume that this returns the same timestamp in
595 * all datacenters due to relay delays.
596 *
597 * The level of staleness can roughly be estimated from this key, but
598 * if the key was evicted from cache, such calculations may show the
599 * time since expiry as ~0 seconds.
600 *
601 * Note that "check" keys won't collide with other regular keys.
602 *
603 * @param string $key
604 * @return float UNIX timestamp
605 */
606 final public function getCheckKeyTime( $key ) {
607 return $this->getMultiCheckKeyTime( [ $key ] )[$key];
608 }
609
610 /**
611 * Fetch the values of each timestamp "check" key
612 *
613 * This works like getCheckKeyTime() except it takes a list of keys
614 * and returns a map of timestamps instead of just that of one key
615 *
616 * This might be useful if both:
617 * - a) a class of entities each depend on hundreds of other entities
618 * - b) these other entities are depended upon by millions of entities
619 *
620 * The later entities can each use a "check" key to invalidate their dependee entities.
621 * However, it is expensive for the former entities to verify against all of the relevant
622 * "check" keys during each getWithSetCallback() call. A less expensive approach is to do
623 * these verifications only after a "time-till-verify" (TTV) has passed. This is a middle
624 * ground between using blind TTLs and using constant verification. The adaptiveTTL() method
625 * can be used to dynamically adjust the TTV. Also, the initial TTV can make use of the
626 * last-modified times of the dependant entities (either from the DB or the "check" keys).
627 *
628 * Example usage:
629 * @code
630 * $value = $cache->getWithSetCallback(
631 * $cache->makeGlobalKey( 'wikibase-item', $id ),
632 * self::INITIAL_TTV, // initial time-till-verify
633 * function ( $oldValue, &$ttv, &$setOpts, $oldAsOf ) use ( $checkKeys, $cache ) {
634 * $now = microtime( true );
635 * // Use $oldValue if it passes max ultimate age and "check" key comparisons
636 * if ( $oldValue &&
637 * $oldAsOf > max( $cache->getMultiCheckKeyTime( $checkKeys ) ) &&
638 * ( $now - $oldValue['ctime'] ) <= self::MAX_CACHE_AGE
639 * ) {
640 * // Increase time-till-verify by 50% of last time to reduce overhead
641 * $ttv = $cache->adaptiveTTL( $oldAsOf, self::MAX_TTV, self::MIN_TTV, 1.5 );
642 * // Unlike $oldAsOf, "ctime" is the ultimate age of the cached data
643 * return $oldValue;
644 * }
645 *
646 * $mtimes = []; // dependency last-modified times; passed by reference
647 * $value = [ 'data' => $this->fetchEntityData( $mtimes ), 'ctime' => $now ];
648 * // Guess time-till-change among the dependencies, e.g. 1/(total change rate)
649 * $ttc = 1 / array_sum( array_map(
650 * function ( $mtime ) use ( $now ) {
651 * return 1 / ( $mtime ? ( $now - $mtime ) : 900 );
652 * },
653 * $mtimes
654 * ) );
655 * // The time-to-verify should not be overly pessimistic nor optimistic
656 * $ttv = min( max( $ttc, self::MIN_TTV ), self::MAX_TTV );
657 *
658 * return $value;
659 * },
660 * [ 'staleTTL' => $cache::TTL_DAY ] // keep around to verify and re-save
661 * );
662 * @endcode
663 *
664 * @see WANObjectCache::getCheckKeyTime()
665 * @see WANObjectCache::getWithSetCallback()
666 *
667 * @param array $keys
668 * @return float[] Map of (key => UNIX timestamp)
669 * @since 1.31
670 */
671 final public function getMultiCheckKeyTime( array $keys ) {
672 $rawKeys = [];
673 foreach ( $keys as $key ) {
674 $rawKeys[$key] = self::TIME_KEY_PREFIX . $key;
675 }
676
677 $rawValues = $this->cache->getMulti( $rawKeys );
678 $rawValues += array_fill_keys( $rawKeys, false );
679
680 $times = [];
681 foreach ( $rawKeys as $key => $rawKey ) {
682 $purge = $this->parsePurgeValue( $rawValues[$rawKey] );
683 if ( $purge !== false ) {
684 $time = $purge[self::FLD_TIME];
685 } else {
686 // Casting assures identical floats for the next getCheckKeyTime() calls
687 $now = (string)$this->getCurrentTime();
688 $this->cache->add(
689 $rawKey,
690 $this->makePurgeValue( $now, self::HOLDOFF_TTL ),
691 self::CHECK_KEY_TTL
692 );
693 $time = (float)$now;
694 }
695
696 $times[$key] = $time;
697 }
698
699 return $times;
700 }
701
702 /**
703 * Purge a "check" key from all datacenters, invalidating keys that use it
704 *
705 * This should only be called when the underlying data (being cached)
706 * changes in a significant way, and it is impractical to call delete()
707 * on all keys that should be changed. When get() is called on those
708 * keys, the relevant "check" keys must be supplied for this to work.
709 *
710 * The "check" key essentially represents a last-modified time of an entity.
711 * When the key is touched, the timestamp will be updated to the current time.
712 * Keys using the "check" key via get(), getMulti(), or getWithSetCallback() will
713 * be invalidated. This approach is useful if many keys depend on a single entity.
714 *
715 * The timestamp of the "check" key is treated as being HOLDOFF_TTL seconds in the
716 * future by get*() methods in order to avoid race conditions where keys are updated
717 * with stale values (e.g. from a lagged replica DB). A high TTL is set on the "check"
718 * key, making it possible to know the timestamp of the last change to the corresponding
719 * entities in most cases. This might use more cache space than resetCheckKey().
720 *
721 * When a few important keys get a large number of hits, a high cache time is usually
722 * desired as well as "lockTSE" logic. The resetCheckKey() method is less appropriate
723 * in such cases since the "time since expiry" cannot be inferred, causing any get()
724 * after the reset to treat the key as being "hot", resulting in more stale value usage.
725 *
726 * Note that "check" keys won't collide with other regular keys.
727 *
728 * @see WANObjectCache::get()
729 * @see WANObjectCache::getWithSetCallback()
730 * @see WANObjectCache::resetCheckKey()
731 *
732 * @param string $key Cache key
733 * @param int $holdoff HOLDOFF_TTL or HOLDOFF_NONE constant
734 * @return bool True if the item was purged or not found, false on failure
735 */
736 final public function touchCheckKey( $key, $holdoff = self::HOLDOFF_TTL ) {
737 // Publish the purge to all datacenters
738 return $this->relayPurge( self::TIME_KEY_PREFIX . $key, self::CHECK_KEY_TTL, $holdoff );
739 }
740
741 /**
742 * Delete a "check" key from all datacenters, invalidating keys that use it
743 *
744 * This is similar to touchCheckKey() in that keys using it via get(), getMulti(),
745 * or getWithSetCallback() will be invalidated. The differences are:
746 * - a) The "check" key will be deleted from all caches and lazily
747 * re-initialized when accessed (rather than set everywhere)
748 * - b) Thus, dependent keys will be known to be invalid, but not
749 * for how long (they are treated as "just" purged), which
750 * effects any lockTSE logic in getWithSetCallback()
751 * - c) Since "check" keys are initialized only on the server the key hashes
752 * to, any temporary ejection of that server will cause the value to be
753 * seen as purged as a new server will initialize the "check" key.
754 *
755 * The advantage here is that the "check" keys, which have high TTLs, will only
756 * be created when a get*() method actually uses that key. This is better when
757 * a large number of "check" keys are invalided in a short period of time.
758 *
759 * Note that "check" keys won't collide with other regular keys.
760 *
761 * @see WANObjectCache::get()
762 * @see WANObjectCache::getWithSetCallback()
763 * @see WANObjectCache::touchCheckKey()
764 *
765 * @param string $key Cache key
766 * @return bool True if the item was purged or not found, false on failure
767 */
768 final public function resetCheckKey( $key ) {
769 // Publish the purge to all datacenters
770 return $this->relayDelete( self::TIME_KEY_PREFIX . $key );
771 }
772
773 /**
774 * Method to fetch/regenerate cache keys
775 *
776 * On cache miss, the key will be set to the callback result via set()
777 * (unless the callback returns false) and that result will be returned.
778 * The arguments supplied to the callback are:
779 * - $oldValue : current cache value or false if not present
780 * - &$ttl : a reference to the TTL which can be altered
781 * - &$setOpts : a reference to options for set() which can be altered
782 * - $oldAsOf : generation UNIX timestamp of $oldValue or null if not present (since 1.28)
783 *
784 * It is strongly recommended to set the 'lag' and 'since' fields to avoid race conditions
785 * that can cause stale values to get stuck at keys. Usually, callbacks ignore the current
786 * value, but it can be used to maintain "most recent X" values that come from time or
787 * sequence based source data, provided that the "as of" id/time is tracked. Note that
788 * preemptive regeneration and $checkKeys can result in a non-false current value.
789 *
790 * Usage of $checkKeys is similar to get() and getMulti(). However, rather than the caller
791 * having to inspect a "current time left" variable (e.g. $curTTL, $curTTLs), a cache
792 * regeneration will automatically be triggered using the callback.
793 *
794 * The $ttl argument and "hotTTR" option (in $opts) use time-dependant randomization
795 * to avoid stampedes. Keys that are slow to regenerate and either heavily used
796 * or subject to explicit (unpredictable) purges, may need additional mechanisms.
797 * The simplest way to avoid stampedes for such keys is to use 'lockTSE' (in $opts).
798 * If explicit purges are needed, also:
799 * - a) Pass $key into $checkKeys
800 * - b) Use touchCheckKey( $key ) instead of delete( $key )
801 *
802 * Example usage (typical key):
803 * @code
804 * $catInfo = $cache->getWithSetCallback(
805 * // Key to store the cached value under
806 * $cache->makeKey( 'cat-attributes', $catId ),
807 * // Time-to-live (in seconds)
808 * $cache::TTL_MINUTE,
809 * // Function that derives the new key value
810 * function ( $oldValue, &$ttl, array &$setOpts ) {
811 * $dbr = wfGetDB( DB_REPLICA );
812 * // Account for any snapshot/replica DB lag
813 * $setOpts += Database::getCacheSetOptions( $dbr );
814 *
815 * return $dbr->selectRow( ... );
816 * }
817 * );
818 * @endcode
819 *
820 * Example usage (key that is expensive and hot):
821 * @code
822 * $catConfig = $cache->getWithSetCallback(
823 * // Key to store the cached value under
824 * $cache->makeKey( 'site-cat-config' ),
825 * // Time-to-live (in seconds)
826 * $cache::TTL_DAY,
827 * // Function that derives the new key value
828 * function ( $oldValue, &$ttl, array &$setOpts ) {
829 * $dbr = wfGetDB( DB_REPLICA );
830 * // Account for any snapshot/replica DB lag
831 * $setOpts += Database::getCacheSetOptions( $dbr );
832 *
833 * return CatConfig::newFromRow( $dbr->selectRow( ... ) );
834 * },
835 * [
836 * // Calling touchCheckKey() on this key invalidates the cache
837 * 'checkKeys' => [ $cache->makeKey( 'site-cat-config' ) ],
838 * // Try to only let one datacenter thread manage cache updates at a time
839 * 'lockTSE' => 30,
840 * // Avoid querying cache servers multiple times in a web request
841 * 'pcTTL' => $cache::TTL_PROC_LONG
842 * ]
843 * );
844 * @endcode
845 *
846 * Example usage (key with dynamic dependencies):
847 * @code
848 * $catState = $cache->getWithSetCallback(
849 * // Key to store the cached value under
850 * $cache->makeKey( 'cat-state', $cat->getId() ),
851 * // Time-to-live (seconds)
852 * $cache::TTL_HOUR,
853 * // Function that derives the new key value
854 * function ( $oldValue, &$ttl, array &$setOpts ) {
855 * // Determine new value from the DB
856 * $dbr = wfGetDB( DB_REPLICA );
857 * // Account for any snapshot/replica DB lag
858 * $setOpts += Database::getCacheSetOptions( $dbr );
859 *
860 * return CatState::newFromResults( $dbr->select( ... ) );
861 * },
862 * [
863 * // The "check" keys that represent things the value depends on;
864 * // Calling touchCheckKey() on any of them invalidates the cache
865 * 'checkKeys' => [
866 * $cache->makeKey( 'sustenance-bowls', $cat->getRoomId() ),
867 * $cache->makeKey( 'people-present', $cat->getHouseId() ),
868 * $cache->makeKey( 'cat-laws', $cat->getCityId() ),
869 * ]
870 * ]
871 * );
872 * @endcode
873 *
874 * Example usage (hot key holding most recent 100 events):
875 * @code
876 * $lastCatActions = $cache->getWithSetCallback(
877 * // Key to store the cached value under
878 * $cache->makeKey( 'cat-last-actions', 100 ),
879 * // Time-to-live (in seconds)
880 * 10,
881 * // Function that derives the new key value
882 * function ( $oldValue, &$ttl, array &$setOpts ) {
883 * $dbr = wfGetDB( DB_REPLICA );
884 * // Account for any snapshot/replica DB lag
885 * $setOpts += Database::getCacheSetOptions( $dbr );
886 *
887 * // Start off with the last cached list
888 * $list = $oldValue ?: [];
889 * // Fetch the last 100 relevant rows in descending order;
890 * // only fetch rows newer than $list[0] to reduce scanning
891 * $rows = iterator_to_array( $dbr->select( ... ) );
892 * // Merge them and get the new "last 100" rows
893 * return array_slice( array_merge( $new, $list ), 0, 100 );
894 * },
895 * [
896 * // Try to only let one datacenter thread manage cache updates at a time
897 * 'lockTSE' => 30,
898 * // Use a magic value when no cache value is ready rather than stampeding
899 * 'busyValue' => 'computing'
900 * ]
901 * );
902 * @endcode
903 *
904 * @see WANObjectCache::get()
905 * @see WANObjectCache::set()
906 *
907 * @param string $key Cache key made from makeKey() or makeGlobalKey()
908 * @param int $ttl Seconds to live for key updates. Special values are:
909 * - WANObjectCache::TTL_INDEFINITE: Cache forever (subject to LRU-style evictions)
910 * - WANObjectCache::TTL_UNCACHEABLE: Do not cache (if the key exists, it is not deleted)
911 * @param callable $callback Value generation function
912 * @param array $opts Options map:
913 * - checkKeys: List of "check" keys. The key at $key will be seen as invalid when either
914 * touchCheckKey() or resetCheckKey() is called on any of the keys in this list. This
915 * is useful if thousands or millions of keys depend on the same entity. The entity can
916 * simply have its "check" key updated whenever the entity is modified.
917 * Default: [].
918 * - graceTTL: Consider reusing expired values instead of refreshing them if they expired
919 * less than this many seconds ago. The odds of a refresh becomes more likely over time,
920 * becoming certain once the grace period is reached. This can reduce traffic spikes
921 * when millions of keys are compared to the same "check" key and touchCheckKey()
922 * or resetCheckKey() is called on that "check" key.
923 * Default: WANObjectCache::GRACE_TTL_NONE.
924 * - lockTSE: If the key is tombstoned or expired (by checkKeys) less than this many seconds
925 * ago, then try to have a single thread handle cache regeneration at any given time.
926 * Other threads will try to use stale values if possible. If, on miss, the time since
927 * expiration is low, the assumption is that the key is hot and that a stampede is worth
928 * avoiding. Setting this above WANObjectCache::HOLDOFF_TTL makes no difference. The
929 * higher this is set, the higher the worst-case staleness can be.
930 * Use WANObjectCache::TSE_NONE to disable this logic.
931 * Default: WANObjectCache::TSE_NONE.
932 * - busyValue: If no value exists and another thread is currently regenerating it, use this
933 * as a fallback value (or a callback to generate such a value). This assures that cache
934 * stampedes cannot happen if the value falls out of cache. This can be used as insurance
935 * against cache regeneration becoming very slow for some reason (greater than the TTL).
936 * Default: null.
937 * - pcTTL: Process cache the value in this PHP instance for this many seconds. This avoids
938 * network I/O when a key is read several times. This will not cache when the callback
939 * returns false, however. Note that any purges will not be seen while process cached;
940 * since the callback should use replica DBs and they may be lagged or have snapshot
941 * isolation anyway, this should not typically matter.
942 * Default: WANObjectCache::TTL_UNCACHEABLE.
943 * - pcGroup: Process cache group to use instead of the primary one. If set, this must be
944 * of the format ALPHANUMERIC_NAME:MAX_KEY_SIZE, e.g. "mydata:10". Use this for storing
945 * large values, small yet numerous values, or some values with a high cost of eviction.
946 * It is generally preferable to use a class constant when setting this value.
947 * This has no effect unless pcTTL is used.
948 * Default: WANObjectCache::PC_PRIMARY.
949 * - version: Integer version number. This allows for callers to make breaking changes to
950 * how values are stored while maintaining compatability and correct cache purges. New
951 * versions are stored alongside older versions concurrently. Avoid storing class objects
952 * however, as this reduces compatibility (due to serialization).
953 * Default: null.
954 * - minAsOf: Reject values if they were generated before this UNIX timestamp.
955 * This is useful if the source of a key is suspected of having possibly changed
956 * recently, and the caller wants any such changes to be reflected.
957 * Default: WANObjectCache::MIN_TIMESTAMP_NONE.
958 * - hotTTR: Expected time-till-refresh (TTR) in seconds for keys that average ~1 hit per
959 * second (e.g. 1Hz). Keys with a hit rate higher than 1Hz will refresh sooner than this
960 * TTR and vise versa. Such refreshes won't happen until keys are "ageNew" seconds old.
961 * This uses randomization to avoid triggering cache stampedes. The TTR is useful at
962 * reducing the impact of missed cache purges, since the effect of a heavily referenced
963 * key being stale is worse than that of a rarely referenced key. Unlike simply lowering
964 * $ttl, seldomly used keys are largely unaffected by this option, which makes it
965 * possible to have a high hit rate for the "long-tail" of less-used keys.
966 * Default: WANObjectCache::HOT_TTR.
967 * - lowTTL: Consider pre-emptive updates when the current TTL (seconds) of the key is less
968 * than this. It becomes more likely over time, becoming certain once the key is expired.
969 * This helps avoid cache stampedes that might be triggered due to the key expiring.
970 * Default: WANObjectCache::LOW_TTL.
971 * - ageNew: Consider popularity refreshes only once a key reaches this age in seconds.
972 * Default: WANObjectCache::AGE_NEW.
973 * - staleTTL: Seconds to keep the key around if it is stale. This means that on cache
974 * miss the callback may get $oldValue/$oldAsOf values for keys that have already been
975 * expired for this specified time. This is useful if adaptiveTTL() is used on the old
976 * value's as-of time when it is verified as still being correct.
977 * Default: WANObjectCache::STALE_TTL_NONE
978 * @return mixed Value found or written to the key
979 * @note Options added in 1.28: version, busyValue, hotTTR, ageNew, pcGroup, minAsOf
980 * @note Options added in 1.31: staleTTL, graceTTL
981 * @note Callable type hints are not used to avoid class-autoloading
982 */
983 final public function getWithSetCallback( $key, $ttl, $callback, array $opts = [] ) {
984 $pcTTL = isset( $opts['pcTTL'] ) ? $opts['pcTTL'] : self::TTL_UNCACHEABLE;
985
986 // Try the process cache if enabled and the cache callback is not within a cache callback.
987 // Process cache use in nested callbacks is not lag-safe with regard to HOLDOFF_TTL since
988 // the in-memory value is further lagged than the shared one since it uses a blind TTL.
989 if ( $pcTTL >= 0 && $this->callbackDepth == 0 ) {
990 $group = isset( $opts['pcGroup'] ) ? $opts['pcGroup'] : self::PC_PRIMARY;
991 $procCache = $this->getProcessCache( $group );
992 $value = $procCache->get( $key );
993 } else {
994 $procCache = false;
995 $value = false;
996 }
997
998 if ( $value === false ) {
999 // Fetch the value over the network
1000 if ( isset( $opts['version'] ) ) {
1001 $version = $opts['version'];
1002 $asOf = null;
1003 $cur = $this->doGetWithSetCallback(
1004 $key,
1005 $ttl,
1006 function ( $oldValue, &$ttl, &$setOpts, $oldAsOf )
1007 use ( $callback, $version ) {
1008 if ( is_array( $oldValue )
1009 && array_key_exists( self::VFLD_DATA, $oldValue )
1010 && array_key_exists( self::VFLD_VERSION, $oldValue )
1011 && $oldValue[self::VFLD_VERSION] === $version
1012 ) {
1013 $oldData = $oldValue[self::VFLD_DATA];
1014 } else {
1015 // VFLD_DATA is not set if an old, unversioned, key is present
1016 $oldData = false;
1017 $oldAsOf = null;
1018 }
1019
1020 return [
1021 self::VFLD_DATA => $callback( $oldData, $ttl, $setOpts, $oldAsOf ),
1022 self::VFLD_VERSION => $version
1023 ];
1024 },
1025 $opts,
1026 $asOf
1027 );
1028 if ( $cur[self::VFLD_VERSION] === $version ) {
1029 // Value created or existed before with version; use it
1030 $value = $cur[self::VFLD_DATA];
1031 } else {
1032 // Value existed before with a different version; use variant key.
1033 // Reflect purges to $key by requiring that this key value be newer.
1034 $value = $this->doGetWithSetCallback(
1035 $this->makeGlobalKey( 'WANCache-key-variant', md5( $key ), $version ),
1036 $ttl,
1037 $callback,
1038 // Regenerate value if not newer than $key
1039 [ 'version' => null, 'minAsOf' => $asOf ] + $opts
1040 );
1041 }
1042 } else {
1043 $value = $this->doGetWithSetCallback( $key, $ttl, $callback, $opts );
1044 }
1045
1046 // Update the process cache if enabled
1047 if ( $procCache && $value !== false ) {
1048 $procCache->set( $key, $value, $pcTTL );
1049 }
1050 }
1051
1052 return $value;
1053 }
1054
1055 /**
1056 * Do the actual I/O for getWithSetCallback() when needed
1057 *
1058 * @see WANObjectCache::getWithSetCallback()
1059 *
1060 * @param string $key
1061 * @param int $ttl
1062 * @param callback $callback
1063 * @param array $opts Options map for getWithSetCallback()
1064 * @param float &$asOf Cache generation timestamp of returned value [returned]
1065 * @return mixed
1066 * @note Callable type hints are not used to avoid class-autoloading
1067 */
1068 protected function doGetWithSetCallback( $key, $ttl, $callback, array $opts, &$asOf = null ) {
1069 $lowTTL = isset( $opts['lowTTL'] ) ? $opts['lowTTL'] : min( self::LOW_TTL, $ttl );
1070 $lockTSE = isset( $opts['lockTSE'] ) ? $opts['lockTSE'] : self::TSE_NONE;
1071 $staleTTL = isset( $opts['staleTTL'] ) ? $opts['staleTTL'] : self::STALE_TTL_NONE;
1072 $graceTTL = isset( $opts['graceTTL'] ) ? $opts['graceTTL'] : self::GRACE_TTL_NONE;
1073 $checkKeys = isset( $opts['checkKeys'] ) ? $opts['checkKeys'] : [];
1074 $busyValue = isset( $opts['busyValue'] ) ? $opts['busyValue'] : null;
1075 $popWindow = isset( $opts['hotTTR'] ) ? $opts['hotTTR'] : self::HOT_TTR;
1076 $ageNew = isset( $opts['ageNew'] ) ? $opts['ageNew'] : self::AGE_NEW;
1077 $minTime = isset( $opts['minAsOf'] ) ? $opts['minAsOf'] : self::MIN_TIMESTAMP_NONE;
1078 $versioned = isset( $opts['version'] );
1079
1080 // Get a collection name to describe this class of key
1081 $kClass = $this->determineKeyClass( $key );
1082
1083 // Get the current key value
1084 $curTTL = null;
1085 $cValue = $this->get( $key, $curTTL, $checkKeys, $asOf ); // current value
1086 $value = $cValue; // return value
1087
1088 $preCallbackTime = $this->getCurrentTime();
1089 // Determine if a cached value regeneration is needed or desired
1090 if ( $value !== false
1091 && $this->isAliveOrInGracePeriod( $curTTL, $graceTTL )
1092 && $this->isValid( $value, $versioned, $asOf, $minTime )
1093 ) {
1094 $preemptiveRefresh = (
1095 $this->worthRefreshExpiring( $curTTL, $lowTTL ) ||
1096 $this->worthRefreshPopular( $asOf, $ageNew, $popWindow, $preCallbackTime )
1097 );
1098
1099 if ( !$preemptiveRefresh ) {
1100 $this->stats->increment( "wanobjectcache.$kClass.hit.good" );
1101
1102 return $value;
1103 } elseif ( $this->asyncHandler ) {
1104 // Update the cache value later, such during post-send of an HTTP request
1105 $func = $this->asyncHandler;
1106 $func( function () use ( $key, $ttl, $callback, $opts, $asOf ) {
1107 $opts['minAsOf'] = INF; // force a refresh
1108 $this->doGetWithSetCallback( $key, $ttl, $callback, $opts, $asOf );
1109 } );
1110 $this->stats->increment( "wanobjectcache.$kClass.hit.refresh" );
1111
1112 return $value;
1113 }
1114 }
1115
1116 // A deleted key with a negative TTL left must be tombstoned
1117 $isTombstone = ( $curTTL !== null && $value === false );
1118 if ( $isTombstone && $lockTSE <= 0 ) {
1119 // Use the INTERIM value for tombstoned keys to reduce regeneration load
1120 $lockTSE = self::INTERIM_KEY_TTL;
1121 }
1122 // Assume a key is hot if requested soon after invalidation
1123 $isHot = ( $curTTL !== null && $curTTL <= 0 && abs( $curTTL ) <= $lockTSE );
1124 // Use the mutex if there is no value and a busy fallback is given
1125 $checkBusy = ( $busyValue !== null && $value === false );
1126 // Decide whether a single thread should handle regenerations.
1127 // This avoids stampedes when $checkKeys are bumped and when preemptive
1128 // renegerations take too long. It also reduces regenerations while $key
1129 // is tombstoned. This balances cache freshness with avoiding DB load.
1130 $useMutex = ( $isHot || ( $isTombstone && $lockTSE > 0 ) || $checkBusy );
1131
1132 $lockAcquired = false;
1133 if ( $useMutex ) {
1134 // Acquire a datacenter-local non-blocking lock
1135 if ( $this->cache->add( self::MUTEX_KEY_PREFIX . $key, 1, self::LOCK_TTL ) ) {
1136 // Lock acquired; this thread should update the key
1137 $lockAcquired = true;
1138 } elseif ( $value !== false && $this->isValid( $value, $versioned, $asOf, $minTime ) ) {
1139 $this->stats->increment( "wanobjectcache.$kClass.hit.stale" );
1140 // If it cannot be acquired; then the stale value can be used
1141 return $value;
1142 } else {
1143 // Use the INTERIM value for tombstoned keys to reduce regeneration load.
1144 // For hot keys, either another thread has the lock or the lock failed;
1145 // use the INTERIM value from the last thread that regenerated it.
1146 $value = $this->getInterimValue( $key, $versioned, $minTime, $asOf );
1147 if ( $value !== false ) {
1148 $this->stats->increment( "wanobjectcache.$kClass.hit.volatile" );
1149
1150 return $value;
1151 }
1152 // Use the busy fallback value if nothing else
1153 if ( $busyValue !== null ) {
1154 $this->stats->increment( "wanobjectcache.$kClass.miss.busy" );
1155
1156 return is_callable( $busyValue ) ? $busyValue() : $busyValue;
1157 }
1158 }
1159 }
1160
1161 if ( !is_callable( $callback ) ) {
1162 throw new InvalidArgumentException( "Invalid cache miss callback provided." );
1163 }
1164
1165 // Generate the new value from the callback...
1166 $setOpts = [];
1167 ++$this->callbackDepth;
1168 try {
1169 $value = call_user_func_array( $callback, [ $cValue, &$ttl, &$setOpts, $asOf ] );
1170 } finally {
1171 --$this->callbackDepth;
1172 }
1173 $valueIsCacheable = ( $value !== false && $ttl >= 0 );
1174
1175 // When delete() is called, writes are write-holed by the tombstone,
1176 // so use a special INTERIM key to pass the new value around threads.
1177 if ( ( $isTombstone && $lockTSE > 0 ) && $valueIsCacheable ) {
1178 $tempTTL = max( 1, (int)$lockTSE ); // set() expects seconds
1179 $newAsOf = $this->getCurrentTime();
1180 $wrapped = $this->wrap( $value, $tempTTL, $newAsOf );
1181 // Avoid using set() to avoid pointless mcrouter broadcasting
1182 $this->setInterimValue( $key, $wrapped, $tempTTL );
1183 }
1184
1185 if ( $valueIsCacheable ) {
1186 $setOpts['lockTSE'] = $lockTSE;
1187 $setOpts['staleTTL'] = $staleTTL;
1188 // Use best known "since" timestamp if not provided
1189 $setOpts += [ 'since' => $preCallbackTime ];
1190 // Update the cache; this will fail if the key is tombstoned
1191 $this->set( $key, $value, $ttl, $setOpts );
1192 }
1193
1194 if ( $lockAcquired ) {
1195 // Avoid using delete() to avoid pointless mcrouter broadcasting
1196 $this->cache->changeTTL( self::MUTEX_KEY_PREFIX . $key, (int)$preCallbackTime - 60 );
1197 }
1198
1199 $this->stats->increment( "wanobjectcache.$kClass.miss.compute" );
1200
1201 return $value;
1202 }
1203
1204 /**
1205 * @param string $key
1206 * @param bool $versioned
1207 * @param float $minTime
1208 * @param mixed $asOf
1209 * @return mixed
1210 */
1211 protected function getInterimValue( $key, $versioned, $minTime, &$asOf ) {
1212 if ( !$this->useInterimHoldOffCaching ) {
1213 return false; // disabled
1214 }
1215
1216 $wrapped = $this->cache->get( self::INTERIM_KEY_PREFIX . $key );
1217 list( $value ) = $this->unwrap( $wrapped, $this->getCurrentTime() );
1218 if ( $value !== false && $this->isValid( $value, $versioned, $asOf, $minTime ) ) {
1219 $asOf = $wrapped[self::FLD_TIME];
1220
1221 return $value;
1222 }
1223
1224 return false;
1225 }
1226
1227 /**
1228 * @param string $key
1229 * @param array $wrapped
1230 * @param int $tempTTL
1231 */
1232 protected function setInterimValue( $key, $wrapped, $tempTTL ) {
1233 $this->cache->merge(
1234 self::INTERIM_KEY_PREFIX . $key,
1235 function () use ( $wrapped ) {
1236 return $wrapped;
1237 },
1238 $tempTTL,
1239 1
1240 );
1241 }
1242
1243 /**
1244 * Method to fetch multiple cache keys at once with regeneration
1245 *
1246 * This works the same as getWithSetCallback() except:
1247 * - a) The $keys argument expects the result of WANObjectCache::makeMultiKeys()
1248 * - b) The $callback argument expects a callback taking the following arguments:
1249 * - $id: ID of an entity to query
1250 * - $oldValue : the prior cache value or false if none was present
1251 * - &$ttl : a reference to the new value TTL in seconds
1252 * - &$setOpts : a reference to options for set() which can be altered
1253 * - $oldAsOf : generation UNIX timestamp of $oldValue or null if not present
1254 * Aside from the additional $id argument, the other arguments function the same
1255 * way they do in getWithSetCallback().
1256 * - c) The return value is a map of (cache key => value) in the order of $keyedIds
1257 *
1258 * @see WANObjectCache::getWithSetCallback()
1259 * @see WANObjectCache::getMultiWithUnionSetCallback()
1260 *
1261 * Example usage:
1262 * @code
1263 * $rows = $cache->getMultiWithSetCallback(
1264 * // Map of cache keys to entity IDs
1265 * $cache->makeMultiKeys(
1266 * $this->fileVersionIds(),
1267 * function ( $id, WANObjectCache $cache ) {
1268 * return $cache->makeKey( 'file-version', $id );
1269 * }
1270 * ),
1271 * // Time-to-live (in seconds)
1272 * $cache::TTL_DAY,
1273 * // Function that derives the new key value
1274 * function ( $id, $oldValue, &$ttl, array &$setOpts ) {
1275 * $dbr = wfGetDB( DB_REPLICA );
1276 * // Account for any snapshot/replica DB lag
1277 * $setOpts += Database::getCacheSetOptions( $dbr );
1278 *
1279 * // Load the row for this file
1280 * $queryInfo = File::getQueryInfo();
1281 * $row = $dbr->selectRow(
1282 * $queryInfo['tables'],
1283 * $queryInfo['fields'],
1284 * [ 'id' => $id ],
1285 * __METHOD__,
1286 * [],
1287 * $queryInfo['joins']
1288 * );
1289 *
1290 * return $row ? (array)$row : false;
1291 * },
1292 * [
1293 * // Process cache for 30 seconds
1294 * 'pcTTL' => 30,
1295 * // Use a dedicated 500 item cache (initialized on-the-fly)
1296 * 'pcGroup' => 'file-versions:500'
1297 * ]
1298 * );
1299 * $files = array_map( [ __CLASS__, 'newFromRow' ], $rows );
1300 * @endcode
1301 *
1302 * @param ArrayIterator $keyedIds Result of WANObjectCache::makeMultiKeys()
1303 * @param int $ttl Seconds to live for key updates
1304 * @param callable $callback Callback the yields entity regeneration callbacks
1305 * @param array $opts Options map
1306 * @return array Map of (cache key => value) in the same order as $keyedIds
1307 * @since 1.28
1308 */
1309 final public function getMultiWithSetCallback(
1310 ArrayIterator $keyedIds, $ttl, callable $callback, array $opts = []
1311 ) {
1312 $valueKeys = array_keys( $keyedIds->getArrayCopy() );
1313 $checkKeys = isset( $opts['checkKeys'] ) ? $opts['checkKeys'] : [];
1314
1315 // Load required keys into process cache in one go
1316 $this->warmupCache = $this->getRawKeysForWarmup(
1317 $this->getNonProcessCachedKeys( $valueKeys, $opts ),
1318 $checkKeys
1319 );
1320 $this->warmupKeyMisses = 0;
1321
1322 // Wrap $callback to match the getWithSetCallback() format while passing $id to $callback
1323 $id = null; // current entity ID
1324 $func = function ( $oldValue, &$ttl, &$setOpts, $oldAsOf ) use ( $callback, &$id ) {
1325 return $callback( $id, $oldValue, $ttl, $setOpts, $oldAsOf );
1326 };
1327
1328 $values = [];
1329 foreach ( $keyedIds as $key => $id ) { // preserve order
1330 $values[$key] = $this->getWithSetCallback( $key, $ttl, $func, $opts );
1331 }
1332
1333 $this->warmupCache = [];
1334
1335 return $values;
1336 }
1337
1338 /**
1339 * Method to fetch/regenerate multiple cache keys at once
1340 *
1341 * This works the same as getWithSetCallback() except:
1342 * - a) The $keys argument expects the result of WANObjectCache::makeMultiKeys()
1343 * - b) The $callback argument expects a callback returning a map of (ID => new value)
1344 * for all entity IDs in $regenById and it takes the following arguments:
1345 * - $ids: a list of entity IDs to regenerate
1346 * - &$ttls: a reference to the (entity ID => new TTL) map
1347 * - &$setOpts: a reference to options for set() which can be altered
1348 * - c) The return value is a map of (cache key => value) in the order of $keyedIds
1349 * - d) The "lockTSE" and "busyValue" options are ignored
1350 *
1351 * @see WANObjectCache::getWithSetCallback()
1352 * @see WANObjectCache::getMultiWithSetCallback()
1353 *
1354 * Example usage:
1355 * @code
1356 * $rows = $cache->getMultiWithUnionSetCallback(
1357 * // Map of cache keys to entity IDs
1358 * $cache->makeMultiKeys(
1359 * $this->fileVersionIds(),
1360 * function ( $id, WANObjectCache $cache ) {
1361 * return $cache->makeKey( 'file-version', $id );
1362 * }
1363 * ),
1364 * // Time-to-live (in seconds)
1365 * $cache::TTL_DAY,
1366 * // Function that derives the new key value
1367 * function ( array $ids, array &$ttls, array &$setOpts ) {
1368 * $dbr = wfGetDB( DB_REPLICA );
1369 * // Account for any snapshot/replica DB lag
1370 * $setOpts += Database::getCacheSetOptions( $dbr );
1371 *
1372 * // Load the rows for these files
1373 * $rows = [];
1374 * $queryInfo = File::getQueryInfo();
1375 * $res = $dbr->select(
1376 * $queryInfo['tables'],
1377 * $queryInfo['fields'],
1378 * [ 'id' => $ids ],
1379 * __METHOD__,
1380 * [],
1381 * $queryInfo['joins']
1382 * );
1383 * foreach ( $res as $row ) {
1384 * $rows[$row->id] = $row;
1385 * $mtime = wfTimestamp( TS_UNIX, $row->timestamp );
1386 * $ttls[$row->id] = $this->adaptiveTTL( $mtime, $ttls[$row->id] );
1387 * }
1388 *
1389 * return $rows;
1390 * },
1391 * ]
1392 * );
1393 * $files = array_map( [ __CLASS__, 'newFromRow' ], $rows );
1394 * @endcode
1395 *
1396 * @param ArrayIterator $keyedIds Result of WANObjectCache::makeMultiKeys()
1397 * @param int $ttl Seconds to live for key updates
1398 * @param callable $callback Callback the yields entity regeneration callbacks
1399 * @param array $opts Options map
1400 * @return array Map of (cache key => value) in the same order as $keyedIds
1401 * @since 1.30
1402 */
1403 final public function getMultiWithUnionSetCallback(
1404 ArrayIterator $keyedIds, $ttl, callable $callback, array $opts = []
1405 ) {
1406 $idsByValueKey = $keyedIds->getArrayCopy();
1407 $valueKeys = array_keys( $idsByValueKey );
1408 $checkKeys = isset( $opts['checkKeys'] ) ? $opts['checkKeys'] : [];
1409 unset( $opts['lockTSE'] ); // incompatible
1410 unset( $opts['busyValue'] ); // incompatible
1411
1412 // Load required keys into process cache in one go
1413 $keysGet = $this->getNonProcessCachedKeys( $valueKeys, $opts );
1414 $this->warmupCache = $this->getRawKeysForWarmup( $keysGet, $checkKeys );
1415 $this->warmupKeyMisses = 0;
1416
1417 // IDs of entities known to be in need of regeneration
1418 $idsRegen = [];
1419
1420 // Find out which keys are missing/deleted/stale
1421 $curTTLs = [];
1422 $asOfs = [];
1423 $curByKey = $this->getMulti( $keysGet, $curTTLs, $checkKeys, $asOfs );
1424 foreach ( $keysGet as $key ) {
1425 if ( !array_key_exists( $key, $curByKey ) || $curTTLs[$key] < 0 ) {
1426 $idsRegen[] = $idsByValueKey[$key];
1427 }
1428 }
1429
1430 // Run the callback to populate the regeneration value map for all required IDs
1431 $newSetOpts = [];
1432 $newTTLsById = array_fill_keys( $idsRegen, $ttl );
1433 $newValsById = $idsRegen ? $callback( $idsRegen, $newTTLsById, $newSetOpts ) : [];
1434
1435 // Wrap $callback to match the getWithSetCallback() format while passing $id to $callback
1436 $id = null; // current entity ID
1437 $func = function ( $oldValue, &$ttl, &$setOpts, $oldAsOf )
1438 use ( $callback, &$id, $newValsById, $newTTLsById, $newSetOpts )
1439 {
1440 if ( array_key_exists( $id, $newValsById ) ) {
1441 // Value was already regerated as expected, so use the value in $newValsById
1442 $newValue = $newValsById[$id];
1443 $ttl = $newTTLsById[$id];
1444 $setOpts = $newSetOpts;
1445 } else {
1446 // Pre-emptive/popularity refresh and version mismatch cases are not detected
1447 // above and thus $newValsById has no entry. Run $callback on this single entity.
1448 $ttls = [ $id => $ttl ];
1449 $newValue = $callback( [ $id ], $ttls, $setOpts )[$id];
1450 $ttl = $ttls[$id];
1451 }
1452
1453 return $newValue;
1454 };
1455
1456 // Run the cache-aside logic using warmupCache instead of persistent cache queries
1457 $values = [];
1458 foreach ( $idsByValueKey as $key => $id ) { // preserve order
1459 $values[$key] = $this->getWithSetCallback( $key, $ttl, $func, $opts );
1460 }
1461
1462 $this->warmupCache = [];
1463
1464 return $values;
1465 }
1466
1467 /**
1468 * Locally set a key to expire soon if it is stale based on $purgeTimestamp
1469 *
1470 * This sets stale keys' time-to-live at HOLDOFF_TTL seconds, which both avoids
1471 * broadcasting in mcrouter setups and also avoids races with new tombstones.
1472 *
1473 * @param string $key Cache key
1474 * @param int $purgeTimestamp UNIX timestamp of purge
1475 * @param bool &$isStale Whether the key is stale
1476 * @return bool Success
1477 * @since 1.28
1478 */
1479 final public function reap( $key, $purgeTimestamp, &$isStale = false ) {
1480 $minAsOf = $purgeTimestamp + self::HOLDOFF_TTL;
1481 $wrapped = $this->cache->get( self::VALUE_KEY_PREFIX . $key );
1482 if ( is_array( $wrapped ) && $wrapped[self::FLD_TIME] < $minAsOf ) {
1483 $isStale = true;
1484 $this->logger->warning( "Reaping stale value key '$key'." );
1485 $ttlReap = self::HOLDOFF_TTL; // avoids races with tombstone creation
1486 $ok = $this->cache->changeTTL( self::VALUE_KEY_PREFIX . $key, $ttlReap );
1487 if ( !$ok ) {
1488 $this->logger->error( "Could not complete reap of key '$key'." );
1489 }
1490
1491 return $ok;
1492 }
1493
1494 $isStale = false;
1495
1496 return true;
1497 }
1498
1499 /**
1500 * Locally set a "check" key to expire soon if it is stale based on $purgeTimestamp
1501 *
1502 * @param string $key Cache key
1503 * @param int $purgeTimestamp UNIX timestamp of purge
1504 * @param bool &$isStale Whether the key is stale
1505 * @return bool Success
1506 * @since 1.28
1507 */
1508 final public function reapCheckKey( $key, $purgeTimestamp, &$isStale = false ) {
1509 $purge = $this->parsePurgeValue( $this->cache->get( self::TIME_KEY_PREFIX . $key ) );
1510 if ( $purge && $purge[self::FLD_TIME] < $purgeTimestamp ) {
1511 $isStale = true;
1512 $this->logger->warning( "Reaping stale check key '$key'." );
1513 $ok = $this->cache->changeTTL( self::TIME_KEY_PREFIX . $key, 1 );
1514 if ( !$ok ) {
1515 $this->logger->error( "Could not complete reap of check key '$key'." );
1516 }
1517
1518 return $ok;
1519 }
1520
1521 $isStale = false;
1522
1523 return false;
1524 }
1525
1526 /**
1527 * @see BagOStuff::makeKey()
1528 * @param string $class Key class
1529 * @param string $component [optional] Key component (starting with a key collection name)
1530 * @return string Colon-delimited list of $keyspace followed by escaped components of $args
1531 * @since 1.27
1532 */
1533 public function makeKey( $class, $component = null ) {
1534 return call_user_func_array( [ $this->cache, __FUNCTION__ ], func_get_args() );
1535 }
1536
1537 /**
1538 * @see BagOStuff::makeGlobalKey()
1539 * @param string $class Key class
1540 * @param string $component [optional] Key component (starting with a key collection name)
1541 * @return string Colon-delimited list of $keyspace followed by escaped components of $args
1542 * @since 1.27
1543 */
1544 public function makeGlobalKey( $class, $component = null ) {
1545 return call_user_func_array( [ $this->cache, __FUNCTION__ ], func_get_args() );
1546 }
1547
1548 /**
1549 * @param array $entities List of entity IDs
1550 * @param callable $keyFunc Callback yielding a key from (entity ID, this WANObjectCache)
1551 * @return ArrayIterator Iterator yielding (cache key => entity ID) in $entities order
1552 * @since 1.28
1553 */
1554 final public function makeMultiKeys( array $entities, callable $keyFunc ) {
1555 $map = [];
1556 foreach ( $entities as $entity ) {
1557 $map[$keyFunc( $entity, $this )] = $entity;
1558 }
1559
1560 return new ArrayIterator( $map );
1561 }
1562
1563 /**
1564 * Get the "last error" registered; clearLastError() should be called manually
1565 * @return int ERR_* class constant for the "last error" registry
1566 */
1567 final public function getLastError() {
1568 if ( $this->lastRelayError ) {
1569 // If the cache and the relayer failed, focus on the latter.
1570 // An update not making it to the relayer means it won't show up
1571 // in other DCs (nor will consistent re-hashing see up-to-date values).
1572 // On the other hand, if just the cache update failed, then it should
1573 // eventually be applied by the relayer.
1574 return $this->lastRelayError;
1575 }
1576
1577 $code = $this->cache->getLastError();
1578 switch ( $code ) {
1579 case BagOStuff::ERR_NONE:
1580 return self::ERR_NONE;
1581 case BagOStuff::ERR_NO_RESPONSE:
1582 return self::ERR_NO_RESPONSE;
1583 case BagOStuff::ERR_UNREACHABLE:
1584 return self::ERR_UNREACHABLE;
1585 default:
1586 return self::ERR_UNEXPECTED;
1587 }
1588 }
1589
1590 /**
1591 * Clear the "last error" registry
1592 */
1593 final public function clearLastError() {
1594 $this->cache->clearLastError();
1595 $this->lastRelayError = self::ERR_NONE;
1596 }
1597
1598 /**
1599 * Clear the in-process caches; useful for testing
1600 *
1601 * @since 1.27
1602 */
1603 public function clearProcessCache() {
1604 $this->processCaches = [];
1605 }
1606
1607 /**
1608 * Enable or disable the use of brief caching for tombstoned keys
1609 *
1610 * When a key is purged via delete(), there normally is a period where caching
1611 * is hold-off limited to an extremely short time. This method will disable that
1612 * caching, forcing the callback to run for any of:
1613 * - WANObjectCache::getWithSetCallback()
1614 * - WANObjectCache::getMultiWithSetCallback()
1615 * - WANObjectCache::getMultiWithUnionSetCallback()
1616 *
1617 * This is useful when both:
1618 * - a) the database used by the callback is known to be up-to-date enough
1619 * for some particular purpose (e.g. replica DB has applied transaction X)
1620 * - b) the caller needs to exploit that fact, and therefore needs to avoid the
1621 * use of inherently volatile and possibly stale interim keys
1622 *
1623 * @see WANObjectCache::delete()
1624 * @param bool $enabled Whether to enable interim caching
1625 * @since 1.31
1626 */
1627 final public function useInterimHoldOffCaching( $enabled ) {
1628 $this->useInterimHoldOffCaching = $enabled;
1629 }
1630
1631 /**
1632 * @param int $flag ATTR_* class constant
1633 * @return int QOS_* class constant
1634 * @since 1.28
1635 */
1636 public function getQoS( $flag ) {
1637 return $this->cache->getQoS( $flag );
1638 }
1639
1640 /**
1641 * Get a TTL that is higher for objects that have not changed recently
1642 *
1643 * This is useful for keys that get explicit purges and DB or purge relay
1644 * lag is a potential concern (especially how it interacts with CDN cache)
1645 *
1646 * Example usage:
1647 * @code
1648 * // Last-modified time of page
1649 * $mtime = wfTimestamp( TS_UNIX, $page->getTimestamp() );
1650 * // Get adjusted TTL. If $mtime is 3600 seconds ago and $minTTL/$factor left at
1651 * // defaults, then $ttl is 3600 * .2 = 720. If $minTTL was greater than 720, then
1652 * // $ttl would be $minTTL. If $maxTTL was smaller than 720, $ttl would be $maxTTL.
1653 * $ttl = $cache->adaptiveTTL( $mtime, $cache::TTL_DAY );
1654 * @endcode
1655 *
1656 * Another use case is when there are no applicable "last modified" fields in the DB,
1657 * and there are too many dependencies for explicit purges to be viable, and the rate of
1658 * change to relevant content is unstable, and it is highly valued to have the cached value
1659 * be as up-to-date as possible.
1660 *
1661 * Example usage:
1662 * @code
1663 * $query = "<some complex query>";
1664 * $idListFromComplexQuery = $cache->getWithSetCallback(
1665 * $cache->makeKey( 'complex-graph-query', $hashOfQuery ),
1666 * GraphQueryClass::STARTING_TTL,
1667 * function ( $oldValue, &$ttl, array &$setOpts, $oldAsOf ) use ( $query, $cache ) {
1668 * $gdb = $this->getReplicaGraphDbConnection();
1669 * // Account for any snapshot/replica DB lag
1670 * $setOpts += GraphDatabase::getCacheSetOptions( $gdb );
1671 *
1672 * $newList = iterator_to_array( $gdb->query( $query ) );
1673 * sort( $newList, SORT_NUMERIC ); // normalize
1674 *
1675 * $minTTL = GraphQueryClass::MIN_TTL;
1676 * $maxTTL = GraphQueryClass::MAX_TTL;
1677 * if ( $oldValue !== false ) {
1678 * // Note that $oldAsOf is the last time this callback ran
1679 * $ttl = ( $newList === $oldValue )
1680 * // No change: cache for 150% of the age of $oldValue
1681 * ? $cache->adaptiveTTL( $oldAsOf, $maxTTL, $minTTL, 1.5 )
1682 * // Changed: cache for %50 of the age of $oldValue
1683 * : $cache->adaptiveTTL( $oldAsOf, $maxTTL, $minTTL, .5 );
1684 * }
1685 *
1686 * return $newList;
1687 * },
1688 * [
1689 * // Keep stale values around for doing comparisons for TTL calculations.
1690 * // High values improve long-tail keys hit-rates, though might waste space.
1691 * 'staleTTL' => GraphQueryClass::GRACE_TTL
1692 * ]
1693 * );
1694 * @endcode
1695 *
1696 * @param int|float $mtime UNIX timestamp
1697 * @param int $maxTTL Maximum TTL (seconds)
1698 * @param int $minTTL Minimum TTL (seconds); Default: 30
1699 * @param float $factor Value in the range (0,1); Default: .2
1700 * @return int Adaptive TTL
1701 * @since 1.28
1702 */
1703 public function adaptiveTTL( $mtime, $maxTTL, $minTTL = 30, $factor = 0.2 ) {
1704 if ( is_float( $mtime ) || ctype_digit( $mtime ) ) {
1705 $mtime = (int)$mtime; // handle fractional seconds and string integers
1706 }
1707
1708 if ( !is_int( $mtime ) || $mtime <= 0 ) {
1709 return $minTTL; // no last-modified time provided
1710 }
1711
1712 $age = $this->getCurrentTime() - $mtime;
1713
1714 return (int)min( $maxTTL, max( $minTTL, $factor * $age ) );
1715 }
1716
1717 /**
1718 * @return int Number of warmup key cache misses last round
1719 * @since 1.30
1720 */
1721 final public function getWarmupKeyMisses() {
1722 return $this->warmupKeyMisses;
1723 }
1724
1725 /**
1726 * Do the actual async bus purge of a key
1727 *
1728 * This must set the key to "PURGED:<UNIX timestamp>:<holdoff>"
1729 *
1730 * @param string $key Cache key
1731 * @param int $ttl How long to keep the tombstone [seconds]
1732 * @param int $holdoff HOLDOFF_* constant controlling how long to ignore sets for this key
1733 * @return bool Success
1734 */
1735 protected function relayPurge( $key, $ttl, $holdoff ) {
1736 if ( $this->purgeRelayer instanceof EventRelayerNull ) {
1737 // This handles the mcrouter and the single-DC case
1738 $ok = $this->cache->set( $key,
1739 $this->makePurgeValue( $this->getCurrentTime(), self::HOLDOFF_NONE ),
1740 $ttl
1741 );
1742 } else {
1743 $event = $this->cache->modifySimpleRelayEvent( [
1744 'cmd' => 'set',
1745 'key' => $key,
1746 'val' => 'PURGED:$UNIXTIME$:' . (int)$holdoff,
1747 'ttl' => max( $ttl, 1 ),
1748 'sbt' => true, // substitute $UNIXTIME$ with actual microtime
1749 ] );
1750
1751 $ok = $this->purgeRelayer->notify( $this->purgeChannel, $event );
1752 if ( !$ok ) {
1753 $this->lastRelayError = self::ERR_RELAY;
1754 }
1755 }
1756
1757 return $ok;
1758 }
1759
1760 /**
1761 * Do the actual async bus delete of a key
1762 *
1763 * @param string $key Cache key
1764 * @return bool Success
1765 */
1766 protected function relayDelete( $key ) {
1767 if ( $this->purgeRelayer instanceof EventRelayerNull ) {
1768 // This handles the mcrouter and the single-DC case
1769 $ok = $this->cache->delete( $key );
1770 } else {
1771 $event = $this->cache->modifySimpleRelayEvent( [
1772 'cmd' => 'delete',
1773 'key' => $key,
1774 ] );
1775
1776 $ok = $this->purgeRelayer->notify( $this->purgeChannel, $event );
1777 if ( !$ok ) {
1778 $this->lastRelayError = self::ERR_RELAY;
1779 }
1780 }
1781
1782 return $ok;
1783 }
1784
1785 /**
1786 * Check if a key is fresh or in the grace window and thus due for randomized reuse
1787 *
1788 * If $curTTL > 0 (e.g. not expired) this returns true. Otherwise, the chance of returning
1789 * true decrease steadily from 100% to 0% as the |$curTTL| moves from 0 to $graceTTL seconds.
1790 * This handles widely varying levels of cache access traffic.
1791 *
1792 * If $curTTL <= -$graceTTL (e.g. already expired), then this returns false.
1793 *
1794 * @param float $curTTL Approximate TTL left on the key if present
1795 * @param int $graceTTL Consider using stale values if $curTTL is greater than this
1796 * @return bool
1797 */
1798 protected function isAliveOrInGracePeriod( $curTTL, $graceTTL ) {
1799 if ( $curTTL > 0 ) {
1800 return true;
1801 } elseif ( $graceTTL <= 0 ) {
1802 return false;
1803 }
1804
1805 $ageStale = abs( $curTTL ); // seconds of staleness
1806 $curGTTL = ( $graceTTL - $ageStale ); // current grace-time-to-live
1807 if ( $curGTTL <= 0 ) {
1808 return false; // already out of grace period
1809 }
1810
1811 // Chance of using a stale value is the complement of the chance of refreshing it
1812 return !$this->worthRefreshExpiring( $curGTTL, $graceTTL );
1813 }
1814
1815 /**
1816 * Check if a key is nearing expiration and thus due for randomized regeneration
1817 *
1818 * This returns false if $curTTL >= $lowTTL. Otherwise, the chance of returning true
1819 * increases steadily from 0% to 100% as the $curTTL moves from $lowTTL to 0 seconds.
1820 * This handles widely varying levels of cache access traffic.
1821 *
1822 * If $curTTL <= 0 (e.g. already expired), then this returns false.
1823 *
1824 * @param float $curTTL Approximate TTL left on the key if present
1825 * @param float $lowTTL Consider a refresh when $curTTL is less than this
1826 * @return bool
1827 */
1828 protected function worthRefreshExpiring( $curTTL, $lowTTL ) {
1829 if ( $lowTTL <= 0 ) {
1830 return false;
1831 } elseif ( $curTTL >= $lowTTL ) {
1832 return false;
1833 } elseif ( $curTTL <= 0 ) {
1834 return false;
1835 }
1836
1837 $chance = ( 1 - $curTTL / $lowTTL );
1838
1839 return mt_rand( 1, 1e9 ) <= 1e9 * $chance;
1840 }
1841
1842 /**
1843 * Check if a key is due for randomized regeneration due to its popularity
1844 *
1845 * This is used so that popular keys can preemptively refresh themselves for higher
1846 * consistency (especially in the case of purge loss/delay). Unpopular keys can remain
1847 * in cache with their high nominal TTL. This means popular keys keep good consistency,
1848 * whether the data changes frequently or not, and long-tail keys get to stay in cache
1849 * and get hits too. Similar to worthRefreshExpiring(), randomization is used.
1850 *
1851 * @param float $asOf UNIX timestamp of the value
1852 * @param int $ageNew Age of key when this might recommend refreshing (seconds)
1853 * @param int $timeTillRefresh Age of key when it should be refreshed if popular (seconds)
1854 * @param float $now The current UNIX timestamp
1855 * @return bool
1856 */
1857 protected function worthRefreshPopular( $asOf, $ageNew, $timeTillRefresh, $now ) {
1858 if ( $ageNew < 0 || $timeTillRefresh <= 0 ) {
1859 return false;
1860 }
1861
1862 $age = $now - $asOf;
1863 $timeOld = $age - $ageNew;
1864 if ( $timeOld <= 0 ) {
1865 return false;
1866 }
1867
1868 // Lifecycle is: new, ramp-up refresh chance, full refresh chance.
1869 // Note that the "expected # of refreshes" for the ramp-up time range is half of what it
1870 // would be if P(refresh) was at its full value during that time range.
1871 $refreshWindowSec = max( $timeTillRefresh - $ageNew - self::RAMPUP_TTL / 2, 1 );
1872 // P(refresh) * (# hits in $refreshWindowSec) = (expected # of refreshes)
1873 // P(refresh) * ($refreshWindowSec * $popularHitsPerSec) = 1
1874 // P(refresh) = 1/($refreshWindowSec * $popularHitsPerSec)
1875 $chance = 1 / ( self::HIT_RATE_HIGH * $refreshWindowSec );
1876
1877 // Ramp up $chance from 0 to its nominal value over RAMPUP_TTL seconds to avoid stampedes
1878 $chance *= ( $timeOld <= self::RAMPUP_TTL ) ? $timeOld / self::RAMPUP_TTL : 1;
1879
1880 return mt_rand( 1, 1e9 ) <= 1e9 * $chance;
1881 }
1882
1883 /**
1884 * Check whether $value is appropriately versioned and not older than $minTime (if set)
1885 *
1886 * @param array $value
1887 * @param bool $versioned
1888 * @param float $asOf The time $value was generated
1889 * @param float $minTime The last time the main value was generated (0.0 if unknown)
1890 * @return bool
1891 */
1892 protected function isValid( $value, $versioned, $asOf, $minTime ) {
1893 if ( $versioned && !isset( $value[self::VFLD_VERSION] ) ) {
1894 return false;
1895 } elseif ( $minTime > 0 && $asOf < $minTime ) {
1896 return false;
1897 }
1898
1899 return true;
1900 }
1901
1902 /**
1903 * Do not use this method outside WANObjectCache
1904 *
1905 * @param mixed $value
1906 * @param int $ttl [0=forever]
1907 * @param float $now Unix Current timestamp just before calling set()
1908 * @return array
1909 */
1910 protected function wrap( $value, $ttl, $now ) {
1911 return [
1912 self::FLD_VERSION => self::VERSION,
1913 self::FLD_VALUE => $value,
1914 self::FLD_TTL => $ttl,
1915 self::FLD_TIME => $now
1916 ];
1917 }
1918
1919 /**
1920 * Do not use this method outside WANObjectCache
1921 *
1922 * @param array|string|bool $wrapped
1923 * @param float $now Unix Current timestamp (preferrably pre-query)
1924 * @return array (mixed; false if absent/tombstoned/invalid, current time left)
1925 */
1926 protected function unwrap( $wrapped, $now ) {
1927 // Check if the value is a tombstone
1928 $purge = $this->parsePurgeValue( $wrapped );
1929 if ( $purge !== false ) {
1930 // Purged values should always have a negative current $ttl
1931 $curTTL = min( $purge[self::FLD_TIME] - $now, self::TINY_NEGATIVE );
1932 return [ false, $curTTL ];
1933 }
1934
1935 if ( !is_array( $wrapped ) // not found
1936 || !isset( $wrapped[self::FLD_VERSION] ) // wrong format
1937 || $wrapped[self::FLD_VERSION] !== self::VERSION // wrong version
1938 ) {
1939 return [ false, null ];
1940 }
1941
1942 $flags = isset( $wrapped[self::FLD_FLAGS] ) ? $wrapped[self::FLD_FLAGS] : 0;
1943 if ( ( $flags & self::FLG_STALE ) == self::FLG_STALE ) {
1944 // Treat as expired, with the cache time as the expiration
1945 $age = $now - $wrapped[self::FLD_TIME];
1946 $curTTL = min( -$age, self::TINY_NEGATIVE );
1947 } elseif ( $wrapped[self::FLD_TTL] > 0 ) {
1948 // Get the approximate time left on the key
1949 $age = $now - $wrapped[self::FLD_TIME];
1950 $curTTL = max( $wrapped[self::FLD_TTL] - $age, 0.0 );
1951 } else {
1952 // Key had no TTL, so the time left is unbounded
1953 $curTTL = INF;
1954 }
1955
1956 return [ $wrapped[self::FLD_VALUE], $curTTL ];
1957 }
1958
1959 /**
1960 * @param array $keys
1961 * @param string $prefix
1962 * @return string[]
1963 */
1964 protected static function prefixCacheKeys( array $keys, $prefix ) {
1965 $res = [];
1966 foreach ( $keys as $key ) {
1967 $res[] = $prefix . $key;
1968 }
1969
1970 return $res;
1971 }
1972
1973 /**
1974 * @param string $key String of the format <scope>:<class>[:<class or variable>]...
1975 * @return string
1976 */
1977 protected function determineKeyClass( $key ) {
1978 $parts = explode( ':', $key );
1979
1980 return isset( $parts[1] ) ? $parts[1] : $parts[0]; // sanity
1981 }
1982
1983 /**
1984 * @return float UNIX timestamp
1985 * @codeCoverageIgnore
1986 */
1987 protected function getCurrentTime() {
1988 return microtime( true );
1989 }
1990
1991 /**
1992 * @param string $value Wrapped value like "PURGED:<timestamp>:<holdoff>"
1993 * @return array|bool Array containing a UNIX timestamp (float) and holdoff period (integer),
1994 * or false if value isn't a valid purge value
1995 */
1996 protected function parsePurgeValue( $value ) {
1997 if ( !is_string( $value ) ) {
1998 return false;
1999 }
2000 $segments = explode( ':', $value, 3 );
2001 if ( !isset( $segments[0] ) || !isset( $segments[1] )
2002 || "{$segments[0]}:" !== self::PURGE_VAL_PREFIX
2003 ) {
2004 return false;
2005 }
2006 if ( !isset( $segments[2] ) ) {
2007 // Back-compat with old purge values without holdoff
2008 $segments[2] = self::HOLDOFF_TTL;
2009 }
2010 return [
2011 self::FLD_TIME => (float)$segments[1],
2012 self::FLD_HOLDOFF => (int)$segments[2],
2013 ];
2014 }
2015
2016 /**
2017 * @param float $timestamp
2018 * @param int $holdoff In seconds
2019 * @return string Wrapped purge value
2020 */
2021 protected function makePurgeValue( $timestamp, $holdoff ) {
2022 return self::PURGE_VAL_PREFIX . (float)$timestamp . ':' . (int)$holdoff;
2023 }
2024
2025 /**
2026 * @param string $group
2027 * @return HashBagOStuff
2028 */
2029 protected function getProcessCache( $group ) {
2030 if ( !isset( $this->processCaches[$group] ) ) {
2031 list( , $n ) = explode( ':', $group );
2032 $this->processCaches[$group] = new HashBagOStuff( [ 'maxKeys' => (int)$n ] );
2033 }
2034
2035 return $this->processCaches[$group];
2036 }
2037
2038 /**
2039 * @param array $keys
2040 * @param array $opts
2041 * @return array List of keys
2042 */
2043 private function getNonProcessCachedKeys( array $keys, array $opts ) {
2044 $keysFound = [];
2045 if ( isset( $opts['pcTTL'] ) && $opts['pcTTL'] > 0 && $this->callbackDepth == 0 ) {
2046 $pcGroup = isset( $opts['pcGroup'] ) ? $opts['pcGroup'] : self::PC_PRIMARY;
2047 $procCache = $this->getProcessCache( $pcGroup );
2048 foreach ( $keys as $key ) {
2049 if ( $procCache->get( $key ) !== false ) {
2050 $keysFound[] = $key;
2051 }
2052 }
2053 }
2054
2055 return array_diff( $keys, $keysFound );
2056 }
2057
2058 /**
2059 * @param array $keys
2060 * @param array $checkKeys
2061 * @return array Map of (cache key => mixed)
2062 */
2063 private function getRawKeysForWarmup( array $keys, array $checkKeys ) {
2064 if ( !$keys ) {
2065 return [];
2066 }
2067
2068 $keysWarmUp = [];
2069 // Get all the value keys to fetch...
2070 foreach ( $keys as $key ) {
2071 $keysWarmUp[] = self::VALUE_KEY_PREFIX . $key;
2072 }
2073 // Get all the check keys to fetch...
2074 foreach ( $checkKeys as $i => $checkKeyOrKeys ) {
2075 if ( is_int( $i ) ) {
2076 // Single check key that applies to all value keys
2077 $keysWarmUp[] = self::TIME_KEY_PREFIX . $checkKeyOrKeys;
2078 } else {
2079 // List of check keys that apply to value key $i
2080 $keysWarmUp = array_merge(
2081 $keysWarmUp,
2082 self::prefixCacheKeys( $checkKeyOrKeys, self::TIME_KEY_PREFIX )
2083 );
2084 }
2085 }
2086
2087 $warmupCache = $this->cache->getMulti( $keysWarmUp );
2088 $warmupCache += array_fill_keys( $keysWarmUp, false );
2089
2090 return $warmupCache;
2091 }
2092 }