objectcache: move ERR_* constants to IExpiringStore
[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 * ### Using WANObjectCache
31 *
32 * All operations go to the local datacenter cache, except for delete(),
33 * touchCheckKey(), and resetCheckKey(), which broadcast to all datacenters.
34 *
35 * This class is intended for caching data from primary stores.
36 * If the get() method does not return a value, then the caller
37 * should query the new value and backfill the cache using set().
38 * The preferred way to do this logic is through getWithSetCallback().
39 * When querying the store on cache miss, the closest DB replica
40 * should be used. Try to avoid heavyweight DB master or quorum reads.
41 *
42 * To ensure consumers of the cache see new values in a timely manner,
43 * you either need to follow either the validation strategy, or the
44 * purge strategy.
45 *
46 * The validation strategy refers to the natural avoidance of stale data
47 * by one of the following means:
48 *
49 * - A) The cached value is immutable.
50 * If the consumer has access to an identifier that uniquely describes a value,
51 * cached value need not change. Instead, the key can change. This also allows
52 * all servers to access their perceived current version. This is important
53 * in context of multiple deployed versions of your application and/or cross-dc
54 * database replication, to ensure deterministic values without oscillation.
55 * - B) Validity is checked against the source after get().
56 * This is the inverse of A. The unique identifier is embedded inside the value
57 * and validated after on retreival. If outdated, the value is recomputed.
58 * - C) The value is cached with a modest TTL (without validation).
59 * If value recomputation is reasonably performant, and the value is allowed to
60 * be stale, one should consider using TTL only – using the value's age as
61 * method of validation.
62 *
63 * The purge strategy refers to the the approach whereby your application knows that
64 * source data has changed and can react by purging the relevant cache keys.
65 * As purges are expensive, this strategy should be avoided if possible.
66 * The simplest purge method is delete().
67 *
68 * No matter which strategy you choose, callers must not rely on updates or purges
69 * being immediately visible to other servers. It should be treated similarly as
70 * one would a database replica.
71 *
72 * The need for immediate updates should be avoided. If needed, solutions must be
73 * sought outside WANObjectCache.
74 *
75 * ### Deploying WANObjectCache
76 *
77 * There are two supported ways to set up broadcasted operations:
78 *
79 * - A) Set up mcrouter as the underlying cache backend, using a memcached BagOStuff class
80 * for the 'cache' parameter. The 'region' and 'cluster' parameters must be provided
81 * and 'mcrouterAware' must be set to `true`.
82 * Configure mcrouter as follows:
83 * - 1) Use Route Prefixing based on region (datacenter) and cache cluster.
84 * See https://github.com/facebook/mcrouter/wiki/Routing-Prefix and
85 * https://github.com/facebook/mcrouter/wiki/Multi-cluster-broadcast-setup.
86 * - 2) To increase the consistency of delete() and touchCheckKey() during cache
87 * server membership changes, you can use the OperationSelectorRoute to
88 * configure 'set' and 'delete' operations to go to all servers in the cache
89 * cluster, instead of just one server determined by hashing.
90 * See https://github.com/facebook/mcrouter/wiki/List-of-Route-Handles.
91 * - B) Set up dynomite as a cache middleware between the web servers and either memcached
92 * or redis and use it as the underlying cache backend, using a memcached BagOStuff
93 * class for the 'cache' parameter. This will broadcast all key setting operations,
94 * not just purges, which can be useful for cache warming. Writes are eventually
95 * consistent via the Dynamo replication model. See https://github.com/Netflix/dynomite.
96 *
97 * Broadcasted operations like delete() and touchCheckKey() are intended to run
98 * immediately in the local datacenter and asynchronously in remote datacenters.
99 *
100 * This means that callers in all datacenters may see older values for however many
101 * milliseconds that the purge took to reach that datacenter. As with any cache, this
102 * should not be relied on for cases where reads are used to determine writes to source
103 * (e.g. non-cache) data stores, except when reading immutable data.
104 *
105 * All values are wrapped in metadata arrays. Keys use a "WANCache:" prefix
106 * to avoid collisions with keys that are not wrapped as metadata arrays. The
107 * prefixes are as follows:
108 * - a) "WANCache:v" : used for regular value keys
109 * - b) "WANCache:i" : used for temporarily storing values of tombstoned keys
110 * - c) "WANCache:t" : used for storing timestamp "check" keys
111 * - d) "WANCache:m" : used for temporary mutex keys to avoid cache stampedes
112 *
113 * @ingroup Cache
114 * @since 1.26
115 */
116 class WANObjectCache implements IExpiringStore, LoggerAwareInterface {
117 /** @var BagOStuff The local datacenter cache */
118 protected $cache;
119 /** @var MapCacheLRU[] Map of group PHP instance caches */
120 protected $processCaches = [];
121 /** @bar bool Whether to use mcrouter key prefixing for routing */
122 protected $mcrouterAware;
123 /** @var string Physical region for mcrouter use */
124 protected $region;
125 /** @var string Cache cluster name for mcrouter use */
126 protected $cluster;
127 /** @var LoggerInterface */
128 protected $logger;
129 /** @var StatsdDataFactoryInterface */
130 protected $stats;
131 /** @var bool Whether to use "interim" caching while keys are tombstoned */
132 protected $useInterimHoldOffCaching = true;
133 /** @var callable|null Function that takes a WAN cache callback and runs it later */
134 protected $asyncHandler;
135 /** @var float Unix timestamp of the oldest possible valid values */
136 protected $epoch;
137
138 /** @var int Callback stack depth for getWithSetCallback() */
139 private $callbackDepth = 0;
140 /** @var mixed[] Temporary warm-up cache */
141 private $warmupCache = [];
142 /** @var int Key fetched */
143 private $warmupKeyMisses = 0;
144
145 /** @var float|null */
146 private $wallClockOverride;
147
148 /** Max time expected to pass between delete() and DB commit finishing */
149 const MAX_COMMIT_DELAY = 3;
150 /** Max replication+snapshot lag before applying TTL_LAGGED or disallowing set() */
151 const MAX_READ_LAG = 7;
152 /** Seconds to tombstone keys on delete() */
153 const HOLDOFF_TTL = 11; // MAX_COMMIT_DELAY + MAX_READ_LAG + 1
154
155 /** Seconds to keep dependency purge keys around */
156 const CHECK_KEY_TTL = self::TTL_YEAR;
157 /** Seconds to keep interim value keys for tombstoned keys around */
158 const INTERIM_KEY_TTL = 1;
159
160 /** Seconds to keep lock keys around */
161 const LOCK_TTL = 10;
162 /** Seconds to no-op key set() calls to avoid large blob I/O stampedes */
163 const COOLOFF_TTL = 1;
164 /** Default remaining TTL at which to consider pre-emptive regeneration */
165 const LOW_TTL = 30;
166
167 /** Never consider performing "popularity" refreshes until a key reaches this age */
168 const AGE_NEW = 60;
169 /** The time length of the "popularity" refresh window for hot keys */
170 const HOT_TTR = 900;
171 /** Hits/second for a refresh to be expected within the "popularity" window */
172 const HIT_RATE_HIGH = 1;
173 /** Seconds to ramp up to the "popularity" refresh chance after a key is no longer new */
174 const RAMPUP_TTL = 30;
175
176 /** Idiom for getWithSetCallback() callbacks to avoid calling set() */
177 const TTL_UNCACHEABLE = -1;
178 /** Idiom for getWithSetCallback() callbacks to 'lockTSE' logic */
179 const TSE_NONE = -1;
180 /** Max TTL to store keys when a data sourced is lagged */
181 const TTL_LAGGED = 30;
182 /** Idiom for delete() for "no hold-off" */
183 const HOLDOFF_NONE = 0;
184 /** Idiom for set()/getWithSetCallback() for "do not augment the storage medium TTL" */
185 const STALE_TTL_NONE = 0;
186 /** Idiom for set()/getWithSetCallback() for "no post-expired grace period" */
187 const GRACE_TTL_NONE = 0;
188
189 /** Idiom for getWithSetCallback() for "no minimum required as-of timestamp" */
190 const MIN_TIMESTAMP_NONE = 0.0;
191
192 /** Tiny negative float to use when CTL comes up >= 0 due to clock skew */
193 const TINY_NEGATIVE = -0.000001;
194 /** Tiny positive float to use when using "minTime" to assert an inequality */
195 const TINY_POSTIVE = 0.000001;
196
197 /** Seconds of delay after get() where set() storms are a consideration with 'lockTSE' */
198 const SET_DELAY_HIGH_SEC = 0.1;
199 /** Min millisecond set() backoff for keys in hold-off (far less than INTERIM_KEY_TTL) */
200 const RECENT_SET_LOW_MS = 50;
201 /** Max millisecond set() backoff for keys in hold-off (far less than INTERIM_KEY_TTL) */
202 const RECENT_SET_HIGH_MS = 100;
203
204 /** Parameter to get()/getMulti() to return extra information by reference */
205 const PASS_BY_REF = -1;
206
207 /** Cache format version number */
208 const VERSION = 1;
209
210 const FLD_VERSION = 0; // key to cache version number
211 const FLD_VALUE = 1; // key to the cached value
212 const FLD_TTL = 2; // key to the original TTL
213 const FLD_TIME = 3; // key to the cache time
214 const FLD_FLAGS = 4; // key to the flags bitfield (reserved number)
215 const FLD_HOLDOFF = 5; // key to any hold-off TTL
216
217 const VALUE_KEY_PREFIX = 'WANCache:v:';
218 const INTERIM_KEY_PREFIX = 'WANCache:i:';
219 const TIME_KEY_PREFIX = 'WANCache:t:';
220 const MUTEX_KEY_PREFIX = 'WANCache:m:';
221 const COOLOFF_KEY_PREFIX = 'WANCache:c:';
222
223 const PURGE_VAL_PREFIX = 'PURGED:';
224
225 const VFLD_DATA = 'WOC:d'; // key to the value of versioned data
226 const VFLD_VERSION = 'WOC:v'; // key to the version of the value present
227
228 const PC_PRIMARY = 'primary:1000'; // process cache name and max key count
229
230 /**
231 * @param array $params
232 * - cache : BagOStuff object for a persistent cache
233 * - logger : LoggerInterface object
234 * - stats : StatsdDataFactoryInterface object
235 * - asyncHandler : A function that takes a callback and runs it later. If supplied,
236 * whenever a preemptive refresh would be triggered in getWithSetCallback(), the
237 * current cache value is still used instead. However, the async-handler function
238 * receives a WAN cache callback that, when run, will execute the value generation
239 * callback supplied by the getWithSetCallback() caller. The result will be saved
240 * as normal. The handler is expected to call the WAN cache callback at an opportune
241 * time (e.g. HTTP post-send), though generally within a few 100ms. [optional]
242 * - region: the current physical region. This is required when using mcrouter as the
243 * backing store proxy. [optional]
244 * - cluster: name of the cache cluster used by this WAN cache. The name must be the
245 * same in all datacenters; the ("region","cluster") tuple is what distinguishes
246 * the counterpart cache clusters among all the datacenter. The contents of
247 * https://github.com/facebook/mcrouter/wiki/Config-Files give background on this.
248 * This is required when using mcrouter as the backing store proxy. [optional]
249 * - mcrouterAware: set as true if mcrouter is the backing store proxy and mcrouter
250 * is configured to interpret /<region>/<cluster>/ key prefixes as routes. This
251 * requires that "region" and "cluster" are both set above. [optional]
252 * - epoch: lowest UNIX timestamp a value/tombstone must have to be valid. [optional]
253 */
254 public function __construct( array $params ) {
255 $this->cache = $params['cache'];
256 $this->region = $params['region'] ?? 'main';
257 $this->cluster = $params['cluster'] ?? 'wan-main';
258 $this->mcrouterAware = !empty( $params['mcrouterAware'] );
259 $this->epoch = $params['epoch'] ?? 1.0;
260
261 $this->setLogger( $params['logger'] ?? new NullLogger() );
262 $this->stats = $params['stats'] ?? new NullStatsdDataFactory();
263 $this->asyncHandler = $params['asyncHandler'] ?? null;
264 }
265
266 /**
267 * @param LoggerInterface $logger
268 */
269 public function setLogger( LoggerInterface $logger ) {
270 $this->logger = $logger;
271 }
272
273 /**
274 * Get an instance that wraps EmptyBagOStuff
275 *
276 * @return WANObjectCache
277 */
278 public static function newEmpty() {
279 return new static( [ 'cache' => new EmptyBagOStuff() ] );
280 }
281
282 /**
283 * Fetch the value of a key from cache
284 *
285 * If supplied, $curTTL is set to the remaining TTL (current time left):
286 * - a) INF; if $key exists, has no TTL, and is not invalidated by $checkKeys
287 * - b) float (>=0); if $key exists, has a TTL, and is not invalidated by $checkKeys
288 * - c) float (<0); if $key is tombstoned, stale, or existing but invalidated by $checkKeys
289 * - d) null; if $key does not exist and is not tombstoned
290 *
291 * If a key is tombstoned, $curTTL will reflect the time since delete().
292 *
293 * The timestamp of $key will be checked against the last-purge timestamp
294 * of each of $checkKeys. Those $checkKeys not in cache will have the last-purge
295 * initialized to the current timestamp. If any of $checkKeys have a timestamp
296 * greater than that of $key, then $curTTL will reflect how long ago $key
297 * became invalid. Callers can use $curTTL to know when the value is stale.
298 * The $checkKeys parameter allow mass invalidations by updating a single key:
299 * - a) Each "check" key represents "last purged" of some source data
300 * - b) Callers pass in relevant "check" keys as $checkKeys in get()
301 * - c) When the source data that "check" keys represent changes,
302 * the touchCheckKey() method is called on them
303 *
304 * Source data entities might exists in a DB that uses snapshot isolation
305 * (e.g. the default REPEATABLE-READ in innoDB). Even for mutable data, that
306 * isolation can largely be maintained by doing the following:
307 * - a) Calling delete() on entity change *and* creation, before DB commit
308 * - b) Keeping transaction duration shorter than the delete() hold-off TTL
309 * - c) Disabling interim key caching via useInterimHoldOffCaching() before get() calls
310 *
311 * However, pre-snapshot values might still be seen if an update was made
312 * in a remote datacenter but the purge from delete() didn't relay yet.
313 *
314 * Consider using getWithSetCallback() instead of get() and set() cycles.
315 * That method has cache slam avoiding features for hot/expensive keys.
316 *
317 * Pass $info as WANObjectCache::PASS_BY_REF to transform it into a cache key info map.
318 * This map includes the following metadata:
319 * - asOf: UNIX timestamp of the value or null if the key is nonexistant
320 * - tombAsOf: UNIX timestamp of the tombstone or null if the key is not tombstoned
321 * - lastCKPurge: UNIX timestamp of the highest check key or null if none provided
322 *
323 * Othwerwise, $info will transform into the cached value timestamp.
324 *
325 * @param string $key Cache key made from makeKey() or makeGlobalKey()
326 * @param mixed|null &$curTTL Approximate TTL left on the key if present/tombstoned [returned]
327 * @param array $checkKeys List of "check" keys
328 * @param mixed|null &$info Key info if WANObjectCache::PASS_BY_REF [returned]
329 * @return mixed Value of cache key or false on failure
330 */
331 final public function get(
332 $key, &$curTTL = null, array $checkKeys = [], &$info = null
333 ) {
334 $curTTLs = self::PASS_BY_REF;
335 $infoByKey = self::PASS_BY_REF;
336 $values = $this->getMulti( [ $key ], $curTTLs, $checkKeys, $infoByKey );
337 $curTTL = $curTTLs[$key] ?? null;
338 if ( $info === self::PASS_BY_REF ) {
339 $info = [
340 'asOf' => $infoByKey[$key]['asOf'] ?? null,
341 'tombAsOf' => $infoByKey[$key]['tombAsOf'] ?? null,
342 'lastCKPurge' => $infoByKey[$key]['lastCKPurge'] ?? null
343 ];
344 } else {
345 $info = $infoByKey[$key]['asOf'] ?? null; // b/c
346 }
347
348 return $values[$key] ?? false;
349 }
350
351 /**
352 * Fetch the value of several keys from cache
353 *
354 * Pass $info as WANObjectCache::PASS_BY_REF to transform it into a map of cache keys
355 * to cache key info maps, each having the same style as those of WANObjectCache::get().
356 * All the cache keys listed in $keys will have an entry.
357 *
358 * Othwerwise, $info will transform into a map of (cache key => cached value timestamp).
359 * Only the cache keys listed in $keys that exists or are tombstoned will have an entry.
360 *
361 * @see WANObjectCache::get()
362 *
363 * @param array $keys List of cache keys made from makeKey() or makeGlobalKey()
364 * @param mixed|null &$curTTLs Map of (key => TTL left) for existing/tombstoned keys [returned]
365 * @param array $checkKeys List of check keys to apply to all $keys. May also apply "check"
366 * keys to specific cache keys only by using cache keys as keys in the $checkKeys array.
367 * @param mixed|null &$info Map of (key => info) if WANObjectCache::PASS_BY_REF [returned]
368 * @return array Map of (key => value) for keys that exist and are not tombstoned
369 */
370 final public function getMulti(
371 array $keys,
372 &$curTTLs = [],
373 array $checkKeys = [],
374 &$info = null
375 ) {
376 $result = [];
377 $curTTLs = [];
378 $infoByKey = [];
379
380 $vPrefixLen = strlen( self::VALUE_KEY_PREFIX );
381 $valueKeys = self::prefixCacheKeys( $keys, self::VALUE_KEY_PREFIX );
382
383 $checkKeysForAll = [];
384 $checkKeysByKey = [];
385 $checkKeysFlat = [];
386 foreach ( $checkKeys as $i => $checkKeyGroup ) {
387 $prefixed = self::prefixCacheKeys( (array)$checkKeyGroup, self::TIME_KEY_PREFIX );
388 $checkKeysFlat = array_merge( $checkKeysFlat, $prefixed );
389 // Are these check keys for a specific cache key, or for all keys being fetched?
390 if ( is_int( $i ) ) {
391 $checkKeysForAll = array_merge( $checkKeysForAll, $prefixed );
392 } else {
393 $checkKeysByKey[$i] = $prefixed;
394 }
395 }
396
397 // Fetch all of the raw values
398 $keysGet = array_merge( $valueKeys, $checkKeysFlat );
399 if ( $this->warmupCache ) {
400 $wrappedValues = array_intersect_key( $this->warmupCache, array_flip( $keysGet ) );
401 $keysGet = array_diff( $keysGet, array_keys( $wrappedValues ) ); // keys left to fetch
402 $this->warmupKeyMisses += count( $keysGet );
403 } else {
404 $wrappedValues = [];
405 }
406 if ( $keysGet ) {
407 $wrappedValues += $this->cache->getMulti( $keysGet );
408 }
409 // Time used to compare/init "check" keys (derived after getMulti() to be pessimistic)
410 $now = $this->getCurrentTime();
411
412 // Collect timestamps from all "check" keys
413 $purgeValuesForAll = $this->processCheckKeys( $checkKeysForAll, $wrappedValues, $now );
414 $purgeValuesByKey = [];
415 foreach ( $checkKeysByKey as $cacheKey => $checks ) {
416 $purgeValuesByKey[$cacheKey] =
417 $this->processCheckKeys( $checks, $wrappedValues, $now );
418 }
419
420 // Get the main cache value for each key and validate them
421 foreach ( $valueKeys as $vKey ) {
422 $key = substr( $vKey, $vPrefixLen ); // unprefix
423 list( $value, $curTTL, $asOf, $tombAsOf ) = isset( $wrappedValues[$vKey] )
424 ? $this->unwrap( $wrappedValues[$vKey], $now )
425 : [ false, null, null, null ]; // not found
426 // Force dependent keys to be seen as stale for a while after purging
427 // to reduce race conditions involving stale data getting cached
428 $purgeValues = $purgeValuesForAll;
429 if ( isset( $purgeValuesByKey[$key] ) ) {
430 $purgeValues = array_merge( $purgeValues, $purgeValuesByKey[$key] );
431 }
432
433 $lastCKPurge = null; // timestamp of the highest check key
434 foreach ( $purgeValues as $purge ) {
435 $lastCKPurge = max( $purge[self::FLD_TIME], $lastCKPurge );
436 $safeTimestamp = $purge[self::FLD_TIME] + $purge[self::FLD_HOLDOFF];
437 if ( $value !== false && $safeTimestamp >= $asOf ) {
438 // How long ago this value was invalidated by *this* check key
439 $ago = min( $purge[self::FLD_TIME] - $now, self::TINY_NEGATIVE );
440 // How long ago this value was invalidated by *any* known check key
441 $curTTL = min( $curTTL, $ago );
442 }
443 }
444
445 if ( $value !== false ) {
446 $result[$key] = $value;
447 }
448 if ( $curTTL !== null ) {
449 $curTTLs[$key] = $curTTL;
450 }
451
452 $infoByKey[$key] = ( $info === self::PASS_BY_REF )
453 ? [ 'asOf' => $asOf, 'tombAsOf' => $tombAsOf, 'lastCKPurge' => $lastCKPurge ]
454 : $asOf; // b/c
455 }
456
457 $info = $infoByKey;
458
459 return $result;
460 }
461
462 /**
463 * @since 1.27
464 * @param array $timeKeys List of prefixed time check keys
465 * @param array $wrappedValues
466 * @param float $now
467 * @return array List of purge value arrays
468 */
469 private function processCheckKeys( array $timeKeys, array $wrappedValues, $now ) {
470 $purgeValues = [];
471 foreach ( $timeKeys as $timeKey ) {
472 $purge = isset( $wrappedValues[$timeKey] )
473 ? $this->parsePurgeValue( $wrappedValues[$timeKey] )
474 : false;
475 if ( $purge === false ) {
476 // Key is not set or malformed; regenerate
477 $newVal = $this->makePurgeValue( $now, self::HOLDOFF_TTL );
478 $this->cache->add( $timeKey, $newVal, self::CHECK_KEY_TTL );
479 $purge = $this->parsePurgeValue( $newVal );
480 }
481 $purgeValues[] = $purge;
482 }
483 return $purgeValues;
484 }
485
486 /**
487 * Set the value of a key in cache
488 *
489 * Simply calling this method when source data changes is not valid because
490 * the changes do not replicate to the other WAN sites. In that case, delete()
491 * should be used instead. This method is intended for use on cache misses.
492 *
493 * If the data was read from a snapshot-isolated transactions (e.g. the default
494 * REPEATABLE-READ in innoDB), use 'since' to avoid the following race condition:
495 * - a) T1 starts
496 * - b) T2 updates a row, calls delete(), and commits
497 * - c) The HOLDOFF_TTL passes, expiring the delete() tombstone
498 * - d) T1 reads the row and calls set() due to a cache miss
499 * - e) Stale value is stuck in cache
500 *
501 * Setting 'lag' and 'since' help avoids keys getting stuck in stale states.
502 *
503 * Be aware that this does not update the process cache for getWithSetCallback()
504 * callers. Keys accessed via that method are not generally meant to also be set
505 * using this primitive method.
506 *
507 * Do not use this method on versioned keys accessed via getWithSetCallback().
508 *
509 * Example usage:
510 * @code
511 * $dbr = wfGetDB( DB_REPLICA );
512 * $setOpts = Database::getCacheSetOptions( $dbr );
513 * // Fetch the row from the DB
514 * $row = $dbr->selectRow( ... );
515 * $key = $cache->makeKey( 'building', $buildingId );
516 * $cache->set( $key, $row, $cache::TTL_DAY, $setOpts );
517 * @endcode
518 *
519 * @param string $key Cache key
520 * @param mixed $value
521 * @param int $ttl Seconds to live. Special values are:
522 * - WANObjectCache::TTL_INDEFINITE: Cache forever
523 * @param array $opts Options map:
524 * - lag : Seconds of replica DB lag. Typically, this is either the replica DB lag
525 * before the data was read or, if applicable, the replica DB lag before
526 * the snapshot-isolated transaction the data was read from started.
527 * Use false to indicate that replication is not running.
528 * Default: 0 seconds
529 * - since : UNIX timestamp of the data in $value. Typically, this is either
530 * the current time the data was read or (if applicable) the time when
531 * the snapshot-isolated transaction the data was read from started.
532 * Default: 0 seconds
533 * - pending : Whether this data is possibly from an uncommitted write transaction.
534 * Generally, other threads should not see values from the future and
535 * they certainly should not see ones that ended up getting rolled back.
536 * Default: false
537 * - lockTSE : if excessive replication/snapshot lag is detected, then store the value
538 * with this TTL and flag it as stale. This is only useful if the reads for this key
539 * use getWithSetCallback() with "lockTSE" set. Note that if "staleTTL" is set
540 * then it will still add on to this TTL in the excessive lag scenario.
541 * Default: WANObjectCache::TSE_NONE
542 * - staleTTL : Seconds to keep the key around if it is stale. The get()/getMulti()
543 * methods return such stale values with a $curTTL of 0, and getWithSetCallback()
544 * will call the regeneration callback in such cases, passing in the old value
545 * and its as-of time to the callback. This is useful if adaptiveTTL() is used
546 * on the old value's as-of time when it is verified as still being correct.
547 * Default: WANObjectCache::STALE_TTL_NONE.
548 * @note Options added in 1.28: staleTTL
549 * @return bool Success
550 */
551 final public function set( $key, $value, $ttl = 0, array $opts = [] ) {
552 $now = $this->getCurrentTime();
553 $lockTSE = $opts['lockTSE'] ?? self::TSE_NONE;
554 $staleTTL = $opts['staleTTL'] ?? self::STALE_TTL_NONE;
555 $age = isset( $opts['since'] ) ? max( 0, $now - $opts['since'] ) : 0;
556 $lag = $opts['lag'] ?? 0;
557
558 // Do not cache potentially uncommitted data as it might get rolled back
559 if ( !empty( $opts['pending'] ) ) {
560 $this->logger->info(
561 'Rejected set() for {cachekey} due to pending writes.',
562 [ 'cachekey' => $key ]
563 );
564
565 return true; // no-op the write for being unsafe
566 }
567
568 $logicalTTL = null; // logical TTL override
569 // Check if there's a risk of writing stale data after the purge tombstone expired
570 if ( $lag === false || ( $lag + $age ) > self::MAX_READ_LAG ) {
571 // Case A: any long-running transaction
572 if ( $age > self::MAX_READ_LAG ) {
573 if ( $lockTSE >= 0 ) {
574 // Store value as *almost* stale to avoid cache and mutex stampedes
575 $logicalTTL = self::TTL_SECOND;
576 $this->logger->info(
577 'Lowered set() TTL for {cachekey} due to snapshot lag.',
578 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ]
579 );
580 } else {
581 $this->logger->info(
582 'Rejected set() for {cachekey} due to snapshot lag.',
583 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ]
584 );
585
586 return true; // no-op the write for being unsafe
587 }
588 // Case B: high replication lag; lower TTL instead of ignoring all set()s
589 } elseif ( $lag === false || $lag > self::MAX_READ_LAG ) {
590 if ( $lockTSE >= 0 ) {
591 $logicalTTL = min( $ttl ?: INF, self::TTL_LAGGED );
592 } else {
593 $ttl = min( $ttl ?: INF, self::TTL_LAGGED );
594 }
595 $this->logger->warning(
596 'Lowered set() TTL for {cachekey} due to replication lag.',
597 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ]
598 );
599 // Case C: medium length request with medium replication lag
600 } else {
601 if ( $lockTSE >= 0 ) {
602 // Store value as *almost* stale to avoid cache and mutex stampedes
603 $logicalTTL = self::TTL_SECOND;
604 $this->logger->info(
605 'Lowered set() TTL for {cachekey} due to high read lag.',
606 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ]
607 );
608 } else {
609 $this->logger->info(
610 'Rejected set() for {cachekey} due to high read lag.',
611 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ]
612 );
613
614 return true; // no-op the write for being unsafe
615 }
616 }
617 }
618
619 // Wrap that value with time/TTL/version metadata
620 $wrapped = $this->wrap( $value, $logicalTTL ?: $ttl, $now );
621
622 $func = function ( $cache, $key, $cWrapped ) use ( $wrapped ) {
623 return ( is_string( $cWrapped ) )
624 ? false // key is tombstoned; do nothing
625 : $wrapped;
626 };
627
628 return $this->cache->merge( self::VALUE_KEY_PREFIX . $key, $func, $ttl + $staleTTL, 1 );
629 }
630
631 /**
632 * Purge a key from all datacenters
633 *
634 * This should only be called when the underlying data (being cached)
635 * changes in a significant way. This deletes the key and starts a hold-off
636 * period where the key cannot be written to for a few seconds (HOLDOFF_TTL).
637 * This is done to avoid the following race condition:
638 * - a) Some DB data changes and delete() is called on a corresponding key
639 * - b) A request refills the key with a stale value from a lagged DB
640 * - c) The stale value is stuck there until the key is expired/evicted
641 *
642 * This is implemented by storing a special "tombstone" value at the cache
643 * key that this class recognizes; get() calls will return false for the key
644 * and any set() calls will refuse to replace tombstone values at the key.
645 * For this to always avoid stale value writes, the following must hold:
646 * - a) Replication lag is bounded to being less than HOLDOFF_TTL; or
647 * - b) If lag is higher, the DB will have gone into read-only mode already
648 *
649 * Note that set() can also be lag-aware and lower the TTL if it's high.
650 *
651 * Be aware that this does not clear the process cache. Even if it did, callbacks
652 * used by getWithSetCallback() might still return stale data in the case of either
653 * uncommitted or not-yet-replicated changes (callback generally use replica DBs).
654 *
655 * When using potentially long-running ACID transactions, a good pattern is
656 * to use a pre-commit hook to issue the delete. This means that immediately
657 * after commit, callers will see the tombstone in cache upon purge relay.
658 * It also avoids the following race condition:
659 * - a) T1 begins, changes a row, and calls delete()
660 * - b) The HOLDOFF_TTL passes, expiring the delete() tombstone
661 * - c) T2 starts, reads the row and calls set() due to a cache miss
662 * - d) T1 finally commits
663 * - e) Stale value is stuck in cache
664 *
665 * Example usage:
666 * @code
667 * $dbw->startAtomic( __METHOD__ ); // start of request
668 * ... <execute some stuff> ...
669 * // Update the row in the DB
670 * $dbw->update( ... );
671 * $key = $cache->makeKey( 'homes', $homeId );
672 * // Purge the corresponding cache entry just before committing
673 * $dbw->onTransactionPreCommitOrIdle( function() use ( $cache, $key ) {
674 * $cache->delete( $key );
675 * } );
676 * ... <execute some stuff> ...
677 * $dbw->endAtomic( __METHOD__ ); // end of request
678 * @endcode
679 *
680 * The $ttl parameter can be used when purging values that have not actually changed
681 * recently. For example, a cleanup script to purge cache entries does not really need
682 * a hold-off period, so it can use HOLDOFF_NONE. Likewise for user-requested purge.
683 * Note that $ttl limits the effective range of 'lockTSE' for getWithSetCallback().
684 *
685 * If called twice on the same key, then the last hold-off TTL takes precedence. For
686 * idempotence, the $ttl should not vary for different delete() calls on the same key.
687 *
688 * @param string $key Cache key
689 * @param int $ttl Tombstone TTL; Default: WANObjectCache::HOLDOFF_TTL
690 * @return bool True if the item was purged or not found, false on failure
691 */
692 final public function delete( $key, $ttl = self::HOLDOFF_TTL ) {
693 $key = self::VALUE_KEY_PREFIX . $key;
694
695 if ( $ttl <= 0 ) {
696 // Publish the purge to all datacenters
697 $ok = $this->relayDelete( $key );
698 } else {
699 // Publish the purge to all datacenters
700 $ok = $this->relayPurge( $key, $ttl, self::HOLDOFF_NONE );
701 }
702
703 return $ok;
704 }
705
706 /**
707 * Fetch the value of a timestamp "check" key
708 *
709 * The key will be *initialized* to the current time if not set,
710 * so only call this method if this behavior is actually desired
711 *
712 * The timestamp can be used to check whether a cached value is valid.
713 * Callers should not assume that this returns the same timestamp in
714 * all datacenters due to relay delays.
715 *
716 * The level of staleness can roughly be estimated from this key, but
717 * if the key was evicted from cache, such calculations may show the
718 * time since expiry as ~0 seconds.
719 *
720 * Note that "check" keys won't collide with other regular keys.
721 *
722 * @param string $key
723 * @return float UNIX timestamp
724 */
725 final public function getCheckKeyTime( $key ) {
726 return $this->getMultiCheckKeyTime( [ $key ] )[$key];
727 }
728
729 /**
730 * Fetch the values of each timestamp "check" key
731 *
732 * This works like getCheckKeyTime() except it takes a list of keys
733 * and returns a map of timestamps instead of just that of one key
734 *
735 * This might be useful if both:
736 * - a) a class of entities each depend on hundreds of other entities
737 * - b) these other entities are depended upon by millions of entities
738 *
739 * The later entities can each use a "check" key to invalidate their dependee entities.
740 * However, it is expensive for the former entities to verify against all of the relevant
741 * "check" keys during each getWithSetCallback() call. A less expensive approach is to do
742 * these verifications only after a "time-till-verify" (TTV) has passed. This is a middle
743 * ground between using blind TTLs and using constant verification. The adaptiveTTL() method
744 * can be used to dynamically adjust the TTV. Also, the initial TTV can make use of the
745 * last-modified times of the dependant entities (either from the DB or the "check" keys).
746 *
747 * Example usage:
748 * @code
749 * $value = $cache->getWithSetCallback(
750 * $cache->makeGlobalKey( 'wikibase-item', $id ),
751 * self::INITIAL_TTV, // initial time-till-verify
752 * function ( $oldValue, &$ttv, &$setOpts, $oldAsOf ) use ( $checkKeys, $cache ) {
753 * $now = microtime( true );
754 * // Use $oldValue if it passes max ultimate age and "check" key comparisons
755 * if ( $oldValue &&
756 * $oldAsOf > max( $cache->getMultiCheckKeyTime( $checkKeys ) ) &&
757 * ( $now - $oldValue['ctime'] ) <= self::MAX_CACHE_AGE
758 * ) {
759 * // Increase time-till-verify by 50% of last time to reduce overhead
760 * $ttv = $cache->adaptiveTTL( $oldAsOf, self::MAX_TTV, self::MIN_TTV, 1.5 );
761 * // Unlike $oldAsOf, "ctime" is the ultimate age of the cached data
762 * return $oldValue;
763 * }
764 *
765 * $mtimes = []; // dependency last-modified times; passed by reference
766 * $value = [ 'data' => $this->fetchEntityData( $mtimes ), 'ctime' => $now ];
767 * // Guess time-till-change among the dependencies, e.g. 1/(total change rate)
768 * $ttc = 1 / array_sum( array_map(
769 * function ( $mtime ) use ( $now ) {
770 * return 1 / ( $mtime ? ( $now - $mtime ) : 900 );
771 * },
772 * $mtimes
773 * ) );
774 * // The time-to-verify should not be overly pessimistic nor optimistic
775 * $ttv = min( max( $ttc, self::MIN_TTV ), self::MAX_TTV );
776 *
777 * return $value;
778 * },
779 * [ 'staleTTL' => $cache::TTL_DAY ] // keep around to verify and re-save
780 * );
781 * @endcode
782 *
783 * @see WANObjectCache::getCheckKeyTime()
784 * @see WANObjectCache::getWithSetCallback()
785 *
786 * @param array $keys
787 * @return float[] Map of (key => UNIX timestamp)
788 * @since 1.31
789 */
790 final public function getMultiCheckKeyTime( array $keys ) {
791 $rawKeys = [];
792 foreach ( $keys as $key ) {
793 $rawKeys[$key] = self::TIME_KEY_PREFIX . $key;
794 }
795
796 $rawValues = $this->cache->getMulti( $rawKeys );
797 $rawValues += array_fill_keys( $rawKeys, false );
798
799 $times = [];
800 foreach ( $rawKeys as $key => $rawKey ) {
801 $purge = $this->parsePurgeValue( $rawValues[$rawKey] );
802 if ( $purge !== false ) {
803 $time = $purge[self::FLD_TIME];
804 } else {
805 // Casting assures identical floats for the next getCheckKeyTime() calls
806 $now = (string)$this->getCurrentTime();
807 $this->cache->add(
808 $rawKey,
809 $this->makePurgeValue( $now, self::HOLDOFF_TTL ),
810 self::CHECK_KEY_TTL
811 );
812 $time = (float)$now;
813 }
814
815 $times[$key] = $time;
816 }
817
818 return $times;
819 }
820
821 /**
822 * Purge a "check" key from all datacenters, invalidating keys that use it
823 *
824 * This should only be called when the underlying data (being cached)
825 * changes in a significant way, and it is impractical to call delete()
826 * on all keys that should be changed. When get() is called on those
827 * keys, the relevant "check" keys must be supplied for this to work.
828 *
829 * The "check" key essentially represents a last-modified time of an entity.
830 * When the key is touched, the timestamp will be updated to the current time.
831 * Keys using the "check" key via get(), getMulti(), or getWithSetCallback() will
832 * be invalidated. This approach is useful if many keys depend on a single entity.
833 *
834 * The timestamp of the "check" key is treated as being HOLDOFF_TTL seconds in the
835 * future by get*() methods in order to avoid race conditions where keys are updated
836 * with stale values (e.g. from a lagged replica DB). A high TTL is set on the "check"
837 * key, making it possible to know the timestamp of the last change to the corresponding
838 * entities in most cases. This might use more cache space than resetCheckKey().
839 *
840 * When a few important keys get a large number of hits, a high cache time is usually
841 * desired as well as "lockTSE" logic. The resetCheckKey() method is less appropriate
842 * in such cases since the "time since expiry" cannot be inferred, causing any get()
843 * after the reset to treat the key as being "hot", resulting in more stale value usage.
844 *
845 * Note that "check" keys won't collide with other regular keys.
846 *
847 * @see WANObjectCache::get()
848 * @see WANObjectCache::getWithSetCallback()
849 * @see WANObjectCache::resetCheckKey()
850 *
851 * @param string $key Cache key
852 * @param int $holdoff HOLDOFF_TTL or HOLDOFF_NONE constant
853 * @return bool True if the item was purged or not found, false on failure
854 */
855 final public function touchCheckKey( $key, $holdoff = self::HOLDOFF_TTL ) {
856 // Publish the purge to all datacenters
857 return $this->relayPurge( self::TIME_KEY_PREFIX . $key, self::CHECK_KEY_TTL, $holdoff );
858 }
859
860 /**
861 * Delete a "check" key from all datacenters, invalidating keys that use it
862 *
863 * This is similar to touchCheckKey() in that keys using it via get(), getMulti(),
864 * or getWithSetCallback() will be invalidated. The differences are:
865 * - a) The "check" key will be deleted from all caches and lazily
866 * re-initialized when accessed (rather than set everywhere)
867 * - b) Thus, dependent keys will be known to be stale, but not
868 * for how long (they are treated as "just" purged), which
869 * effects any lockTSE logic in getWithSetCallback()
870 * - c) Since "check" keys are initialized only on the server the key hashes
871 * to, any temporary ejection of that server will cause the value to be
872 * seen as purged as a new server will initialize the "check" key.
873 *
874 * The advantage here is that the "check" keys, which have high TTLs, will only
875 * be created when a get*() method actually uses that key. This is better when
876 * a large number of "check" keys are invalided in a short period of time.
877 *
878 * Note that "check" keys won't collide with other regular keys.
879 *
880 * @see WANObjectCache::get()
881 * @see WANObjectCache::getWithSetCallback()
882 * @see WANObjectCache::touchCheckKey()
883 *
884 * @param string $key Cache key
885 * @return bool True if the item was purged or not found, false on failure
886 */
887 final public function resetCheckKey( $key ) {
888 // Publish the purge to all datacenters
889 return $this->relayDelete( self::TIME_KEY_PREFIX . $key );
890 }
891
892 /**
893 * Method to fetch/regenerate cache keys
894 *
895 * On cache miss, the key will be set to the callback result via set()
896 * (unless the callback returns false) and that result will be returned.
897 * The arguments supplied to the callback are:
898 * - $oldValue : current cache value or false if not present
899 * - &$ttl : a reference to the TTL which can be altered
900 * - &$setOpts : a reference to options for set() which can be altered
901 * - $oldAsOf : generation UNIX timestamp of $oldValue or null if not present (since 1.28)
902 *
903 * It is strongly recommended to set the 'lag' and 'since' fields to avoid race conditions
904 * that can cause stale values to get stuck at keys. Usually, callbacks ignore the current
905 * value, but it can be used to maintain "most recent X" values that come from time or
906 * sequence based source data, provided that the "as of" id/time is tracked. Note that
907 * preemptive regeneration and $checkKeys can result in a non-false current value.
908 *
909 * Usage of $checkKeys is similar to get() and getMulti(). However, rather than the caller
910 * having to inspect a "current time left" variable (e.g. $curTTL, $curTTLs), a cache
911 * regeneration will automatically be triggered using the callback.
912 *
913 * The $ttl argument and "hotTTR" option (in $opts) use time-dependant randomization
914 * to avoid stampedes. Keys that are slow to regenerate and either heavily used
915 * or subject to explicit (unpredictable) purges, may need additional mechanisms.
916 * The simplest way to avoid stampedes for such keys is to use 'lockTSE' (in $opts).
917 * If explicit purges are needed, also:
918 * - a) Pass $key into $checkKeys
919 * - b) Use touchCheckKey( $key ) instead of delete( $key )
920 *
921 * Example usage (typical key):
922 * @code
923 * $catInfo = $cache->getWithSetCallback(
924 * // Key to store the cached value under
925 * $cache->makeKey( 'cat-attributes', $catId ),
926 * // Time-to-live (in seconds)
927 * $cache::TTL_MINUTE,
928 * // Function that derives the new key value
929 * function ( $oldValue, &$ttl, array &$setOpts ) {
930 * $dbr = wfGetDB( DB_REPLICA );
931 * // Account for any snapshot/replica DB lag
932 * $setOpts += Database::getCacheSetOptions( $dbr );
933 *
934 * return $dbr->selectRow( ... );
935 * }
936 * );
937 * @endcode
938 *
939 * Example usage (key that is expensive and hot):
940 * @code
941 * $catConfig = $cache->getWithSetCallback(
942 * // Key to store the cached value under
943 * $cache->makeKey( 'site-cat-config' ),
944 * // Time-to-live (in seconds)
945 * $cache::TTL_DAY,
946 * // Function that derives the new key value
947 * function ( $oldValue, &$ttl, array &$setOpts ) {
948 * $dbr = wfGetDB( DB_REPLICA );
949 * // Account for any snapshot/replica DB lag
950 * $setOpts += Database::getCacheSetOptions( $dbr );
951 *
952 * return CatConfig::newFromRow( $dbr->selectRow( ... ) );
953 * },
954 * [
955 * // Calling touchCheckKey() on this key invalidates the cache
956 * 'checkKeys' => [ $cache->makeKey( 'site-cat-config' ) ],
957 * // Try to only let one datacenter thread manage cache updates at a time
958 * 'lockTSE' => 30,
959 * // Avoid querying cache servers multiple times in a web request
960 * 'pcTTL' => $cache::TTL_PROC_LONG
961 * ]
962 * );
963 * @endcode
964 *
965 * Example usage (key with dynamic dependencies):
966 * @code
967 * $catState = $cache->getWithSetCallback(
968 * // Key to store the cached value under
969 * $cache->makeKey( 'cat-state', $cat->getId() ),
970 * // Time-to-live (seconds)
971 * $cache::TTL_HOUR,
972 * // Function that derives the new key value
973 * function ( $oldValue, &$ttl, array &$setOpts ) {
974 * // Determine new value from the DB
975 * $dbr = wfGetDB( DB_REPLICA );
976 * // Account for any snapshot/replica DB lag
977 * $setOpts += Database::getCacheSetOptions( $dbr );
978 *
979 * return CatState::newFromResults( $dbr->select( ... ) );
980 * },
981 * [
982 * // The "check" keys that represent things the value depends on;
983 * // Calling touchCheckKey() on any of them invalidates the cache
984 * 'checkKeys' => [
985 * $cache->makeKey( 'sustenance-bowls', $cat->getRoomId() ),
986 * $cache->makeKey( 'people-present', $cat->getHouseId() ),
987 * $cache->makeKey( 'cat-laws', $cat->getCityId() ),
988 * ]
989 * ]
990 * );
991 * @endcode
992 *
993 * Example usage (key that is expensive with too many DB dependencies for "check keys"):
994 * @code
995 * $catToys = $cache->getWithSetCallback(
996 * // Key to store the cached value under
997 * $cache->makeKey( 'cat-toys', $catId ),
998 * // Time-to-live (seconds)
999 * $cache::TTL_HOUR,
1000 * // Function that derives the new key value
1001 * function ( $oldValue, &$ttl, array &$setOpts ) {
1002 * // Determine new value from the DB
1003 * $dbr = wfGetDB( DB_REPLICA );
1004 * // Account for any snapshot/replica DB lag
1005 * $setOpts += Database::getCacheSetOptions( $dbr );
1006 *
1007 * return CatToys::newFromResults( $dbr->select( ... ) );
1008 * },
1009 * [
1010 * // Get the highest timestamp of any of the cat's toys
1011 * 'touchedCallback' => function ( $value ) use ( $catId ) {
1012 * $dbr = wfGetDB( DB_REPLICA );
1013 * $ts = $dbr->selectField( 'cat_toys', 'MAX(ct_touched)', ... );
1014 *
1015 * return wfTimestampOrNull( TS_UNIX, $ts );
1016 * },
1017 * // Avoid DB queries for repeated access
1018 * 'pcTTL' => $cache::TTL_PROC_SHORT
1019 * ]
1020 * );
1021 * @endcode
1022 *
1023 * Example usage (hot key holding most recent 100 events):
1024 * @code
1025 * $lastCatActions = $cache->getWithSetCallback(
1026 * // Key to store the cached value under
1027 * $cache->makeKey( 'cat-last-actions', 100 ),
1028 * // Time-to-live (in seconds)
1029 * 10,
1030 * // Function that derives the new key value
1031 * function ( $oldValue, &$ttl, array &$setOpts ) {
1032 * $dbr = wfGetDB( DB_REPLICA );
1033 * // Account for any snapshot/replica DB lag
1034 * $setOpts += Database::getCacheSetOptions( $dbr );
1035 *
1036 * // Start off with the last cached list
1037 * $list = $oldValue ?: [];
1038 * // Fetch the last 100 relevant rows in descending order;
1039 * // only fetch rows newer than $list[0] to reduce scanning
1040 * $rows = iterator_to_array( $dbr->select( ... ) );
1041 * // Merge them and get the new "last 100" rows
1042 * return array_slice( array_merge( $new, $list ), 0, 100 );
1043 * },
1044 * [
1045 * // Try to only let one datacenter thread manage cache updates at a time
1046 * 'lockTSE' => 30,
1047 * // Use a magic value when no cache value is ready rather than stampeding
1048 * 'busyValue' => 'computing'
1049 * ]
1050 * );
1051 * @endcode
1052 *
1053 * Example usage (key holding an LRU subkey:value map; this can avoid flooding cache with
1054 * keys for an unlimited set of (constraint,situation) pairs, thereby avoiding elevated
1055 * cache evictions and wasted memory):
1056 * @code
1057 * $catSituationTolerabilityCache = $this->cache->getWithSetCallback(
1058 * // Group by constraint ID/hash, cat family ID/hash, or something else useful
1059 * $this->cache->makeKey( 'cat-situation-tolerability-checks', $groupKey ),
1060 * WANObjectCache::TTL_DAY, // rarely used groups should fade away
1061 * // The $scenarioKey format is $constraintId:<ID/hash of $situation>
1062 * function ( $cacheMap ) use ( $scenarioKey, $constraintId, $situation ) {
1063 * $lruCache = MapCacheLRU::newFromArray( $cacheMap ?: [], self::CACHE_SIZE );
1064 * $result = $lruCache->get( $scenarioKey ); // triggers LRU bump if present
1065 * if ( $result === null || $this->isScenarioResultExpired( $result ) ) {
1066 * $result = $this->checkScenarioTolerability( $constraintId, $situation );
1067 * $lruCache->set( $scenarioKey, $result, 3 / 8 );
1068 * }
1069 * // Save the new LRU cache map and reset the map's TTL
1070 * return $lruCache->toArray();
1071 * },
1072 * [
1073 * // Once map is > 1 sec old, consider refreshing
1074 * 'ageNew' => 1,
1075 * // Update within 5 seconds after "ageNew" given a 1hz cache check rate
1076 * 'hotTTR' => 5,
1077 * // Avoid querying cache servers multiple times in a request; this also means
1078 * // that a request can only alter the value of any given constraint key once
1079 * 'pcTTL' => WANObjectCache::TTL_PROC_LONG
1080 * ]
1081 * );
1082 * $tolerability = isset( $catSituationTolerabilityCache[$scenarioKey] )
1083 * ? $catSituationTolerabilityCache[$scenarioKey]
1084 * : $this->checkScenarioTolerability( $constraintId, $situation );
1085 * @endcode
1086 *
1087 * @see WANObjectCache::get()
1088 * @see WANObjectCache::set()
1089 *
1090 * @param string $key Cache key made from makeKey() or makeGlobalKey()
1091 * @param int $ttl Seconds to live for key updates. Special values are:
1092 * - WANObjectCache::TTL_INDEFINITE: Cache forever (subject to LRU-style evictions)
1093 * - WANObjectCache::TTL_UNCACHEABLE: Do not cache (if the key exists, it is not deleted)
1094 * @param callable $callback Value generation function
1095 * @param array $opts Options map:
1096 * - checkKeys: List of "check" keys. The key at $key will be seen as stale when either
1097 * touchCheckKey() or resetCheckKey() is called on any of the keys in this list. This
1098 * is useful if thousands or millions of keys depend on the same entity. The entity can
1099 * simply have its "check" key updated whenever the entity is modified.
1100 * Default: [].
1101 * - graceTTL: If the key is invalidated (by "checkKeys"/"touchedCallback") less than this
1102 * many seconds ago, consider reusing the stale value. The odds of a refresh becomes
1103 * more likely over time, becoming certain once the grace period is reached. This can
1104 * reduce traffic spikes when millions of keys are compared to the same "check" key and
1105 * touchCheckKey() or resetCheckKey() is called on that "check" key. This option is not
1106 * useful for avoiding traffic spikes in the case of the key simply expiring on account
1107 * of its TTL (use "lowTTL" instead).
1108 * Default: WANObjectCache::GRACE_TTL_NONE.
1109 * - lockTSE: If the key is tombstoned or invalidated (by "checkKeys"/"touchedCallback")
1110 * less than this many seconds ago, try to have a single thread handle cache regeneration
1111 * at any given time. Other threads will use stale values if possible. If, on miss,
1112 * the time since expiration is low, the assumption is that the key is hot and that a
1113 * stampede is worth avoiding. Note that if the key falls out of cache then concurrent
1114 * threads will all run the callback on cache miss until the value is saved in cache.
1115 * The only stampede protection in that case is from duplicate cache sets when the
1116 * callback takes longer than WANObjectCache::SET_DELAY_HIGH_SEC seconds; consider
1117 * using "busyValue" if such stampedes are a problem. Note that the higher "lockTSE" is
1118 * set, the higher the worst-case staleness of returned values can be. Also note that
1119 * this option does not by itself handle the case of the key simply expiring on account
1120 * of its TTL, so make sure that "lowTTL" is not disabled when using this option. Avoid
1121 * combining this option with delete() as it can always cause a stampede due to their
1122 * being no stale value available until after a thread completes the callback.
1123 * Use WANObjectCache::TSE_NONE to disable this logic.
1124 * Default: WANObjectCache::TSE_NONE.
1125 * - busyValue: If no value exists and another thread is currently regenerating it, use this
1126 * as a fallback value (or a callback to generate such a value). This assures that cache
1127 * stampedes cannot happen if the value falls out of cache. This can be used as insurance
1128 * against cache regeneration becoming very slow for some reason (greater than the TTL).
1129 * Default: null.
1130 * - pcTTL: Process cache the value in this PHP instance for this many seconds. This avoids
1131 * network I/O when a key is read several times. This will not cache when the callback
1132 * returns false, however. Note that any purges will not be seen while process cached;
1133 * since the callback should use replica DBs and they may be lagged or have snapshot
1134 * isolation anyway, this should not typically matter.
1135 * Default: WANObjectCache::TTL_UNCACHEABLE.
1136 * - pcGroup: Process cache group to use instead of the primary one. If set, this must be
1137 * of the format ALPHANUMERIC_NAME:MAX_KEY_SIZE, e.g. "mydata:10". Use this for storing
1138 * large values, small yet numerous values, or some values with a high cost of eviction.
1139 * It is generally preferable to use a class constant when setting this value.
1140 * This has no effect unless pcTTL is used.
1141 * Default: WANObjectCache::PC_PRIMARY.
1142 * - version: Integer version number. This allows for callers to make breaking changes to
1143 * how values are stored while maintaining compatability and correct cache purges. New
1144 * versions are stored alongside older versions concurrently. Avoid storing class objects
1145 * however, as this reduces compatibility (due to serialization).
1146 * Default: null.
1147 * - minAsOf: Reject values if they were generated before this UNIX timestamp.
1148 * This is useful if the source of a key is suspected of having possibly changed
1149 * recently, and the caller wants any such changes to be reflected.
1150 * Default: WANObjectCache::MIN_TIMESTAMP_NONE.
1151 * - hotTTR: Expected time-till-refresh (TTR) in seconds for keys that average ~1 hit per
1152 * second (e.g. 1Hz). Keys with a hit rate higher than 1Hz will refresh sooner than this
1153 * TTR and vise versa. Such refreshes won't happen until keys are "ageNew" seconds old.
1154 * This uses randomization to avoid triggering cache stampedes. The TTR is useful at
1155 * reducing the impact of missed cache purges, since the effect of a heavily referenced
1156 * key being stale is worse than that of a rarely referenced key. Unlike simply lowering
1157 * $ttl, seldomly used keys are largely unaffected by this option, which makes it
1158 * possible to have a high hit rate for the "long-tail" of less-used keys.
1159 * Default: WANObjectCache::HOT_TTR.
1160 * - lowTTL: Consider pre-emptive updates when the current TTL (seconds) of the key is less
1161 * than this. It becomes more likely over time, becoming certain once the key is expired.
1162 * This helps avoid cache stampedes that might be triggered due to the key expiring.
1163 * Default: WANObjectCache::LOW_TTL.
1164 * - ageNew: Consider popularity refreshes only once a key reaches this age in seconds.
1165 * Default: WANObjectCache::AGE_NEW.
1166 * - staleTTL: Seconds to keep the key around if it is stale. This means that on cache
1167 * miss the callback may get $oldValue/$oldAsOf values for keys that have already been
1168 * expired for this specified time. This is useful if adaptiveTTL() is used on the old
1169 * value's as-of time when it is verified as still being correct.
1170 * Default: WANObjectCache::STALE_TTL_NONE
1171 * - touchedCallback: A callback that takes the current value and returns a UNIX timestamp
1172 * indicating the last time a dynamic dependency changed. Null can be returned if there
1173 * are no relevant dependency changes to check. This can be used to check against things
1174 * like last-modified times of files or DB timestamp fields. This should generally not be
1175 * used for small and easily queried values in a DB if the callback itself ends up doing
1176 * a similarly expensive DB query to check a timestamp. Usages of this option makes the
1177 * most sense for values that are moderately to highly expensive to regenerate and easy
1178 * to query for dependency timestamps. The use of "pcTTL" reduces timestamp queries.
1179 * Default: null.
1180 * @return mixed Value found or written to the key
1181 * @note Options added in 1.28: version, busyValue, hotTTR, ageNew, pcGroup, minAsOf
1182 * @note Options added in 1.31: staleTTL, graceTTL
1183 * @note Options added in 1.33: touchedCallback
1184 * @note Callable type hints are not used to avoid class-autoloading
1185 */
1186 final public function getWithSetCallback( $key, $ttl, $callback, array $opts = [] ) {
1187 $pcTTL = $opts['pcTTL'] ?? self::TTL_UNCACHEABLE;
1188
1189 // Try the process cache if enabled and the cache callback is not within a cache callback.
1190 // Process cache use in nested callbacks is not lag-safe with regard to HOLDOFF_TTL since
1191 // the in-memory value is further lagged than the shared one since it uses a blind TTL.
1192 if ( $pcTTL >= 0 && $this->callbackDepth == 0 ) {
1193 $group = $opts['pcGroup'] ?? self::PC_PRIMARY;
1194 $procCache = $this->getProcessCache( $group );
1195 $value = $procCache->has( $key, $pcTTL ) ? $procCache->get( $key ) : false;
1196 } else {
1197 $procCache = false;
1198 $value = false;
1199 }
1200
1201 if ( $value === false ) {
1202 // Fetch the value over the network
1203 if ( isset( $opts['version'] ) ) {
1204 $version = $opts['version'];
1205 $asOf = null;
1206 $cur = $this->doGetWithSetCallback(
1207 $key,
1208 $ttl,
1209 function ( $oldValue, &$ttl, &$setOpts, $oldAsOf )
1210 use ( $callback, $version ) {
1211 if ( is_array( $oldValue )
1212 && array_key_exists( self::VFLD_DATA, $oldValue )
1213 && array_key_exists( self::VFLD_VERSION, $oldValue )
1214 && $oldValue[self::VFLD_VERSION] === $version
1215 ) {
1216 $oldData = $oldValue[self::VFLD_DATA];
1217 } else {
1218 // VFLD_DATA is not set if an old, unversioned, key is present
1219 $oldData = false;
1220 $oldAsOf = null;
1221 }
1222
1223 return [
1224 self::VFLD_DATA => $callback( $oldData, $ttl, $setOpts, $oldAsOf ),
1225 self::VFLD_VERSION => $version
1226 ];
1227 },
1228 $opts,
1229 $asOf
1230 );
1231 if ( $cur[self::VFLD_VERSION] === $version ) {
1232 // Value created or existed before with version; use it
1233 $value = $cur[self::VFLD_DATA];
1234 } else {
1235 // Value existed before with a different version; use variant key.
1236 // Reflect purges to $key by requiring that this key value be newer.
1237 $value = $this->doGetWithSetCallback(
1238 $this->makeGlobalKey( 'WANCache-key-variant', md5( $key ), $version ),
1239 $ttl,
1240 $callback,
1241 // Regenerate value if not newer than $key
1242 [ 'version' => null, 'minAsOf' => $asOf ] + $opts
1243 );
1244 }
1245 } else {
1246 $value = $this->doGetWithSetCallback( $key, $ttl, $callback, $opts );
1247 }
1248
1249 // Update the process cache if enabled
1250 if ( $procCache && $value !== false ) {
1251 $procCache->set( $key, $value );
1252 }
1253 }
1254
1255 return $value;
1256 }
1257
1258 /**
1259 * Do the actual I/O for getWithSetCallback() when needed
1260 *
1261 * @see WANObjectCache::getWithSetCallback()
1262 *
1263 * @param string $key
1264 * @param int $ttl
1265 * @param callable $callback
1266 * @param array $opts Options map for getWithSetCallback()
1267 * @param float|null &$asOf Cache generation timestamp of returned value [returned]
1268 * @return mixed
1269 * @note Callable type hints are not used to avoid class-autoloading
1270 */
1271 protected function doGetWithSetCallback( $key, $ttl, $callback, array $opts, &$asOf = null ) {
1272 $lowTTL = $opts['lowTTL'] ?? min( self::LOW_TTL, $ttl );
1273 $lockTSE = $opts['lockTSE'] ?? self::TSE_NONE;
1274 $staleTTL = $opts['staleTTL'] ?? self::STALE_TTL_NONE;
1275 $graceTTL = $opts['graceTTL'] ?? self::GRACE_TTL_NONE;
1276 $checkKeys = $opts['checkKeys'] ?? [];
1277 $busyValue = $opts['busyValue'] ?? null;
1278 $popWindow = $opts['hotTTR'] ?? self::HOT_TTR;
1279 $ageNew = $opts['ageNew'] ?? self::AGE_NEW;
1280 $minTime = $opts['minAsOf'] ?? self::MIN_TIMESTAMP_NONE;
1281 $needsVersion = isset( $opts['version'] );
1282 $touchedCb = $opts['touchedCallback'] ?? null;
1283 $initialTime = $this->getCurrentTime();
1284
1285 // Get a collection name to describe this class of key
1286 $kClass = $this->determineKeyClass( $key );
1287
1288 // Get the current key value
1289 $curTTL = self::PASS_BY_REF;
1290 $curInfo = self::PASS_BY_REF; /** @var array $curInfo */
1291 $curValue = $this->get( $key, $curTTL, $checkKeys, $curInfo );
1292 // Apply any $touchedCb invalidation timestamp to get the "last purge timestamp"
1293 list( $curTTL, $LPT ) = $this->resolveCTL( $curValue, $curTTL, $curInfo, $touchedCb );
1294 // Keep track of the best candidate value and its timestamp
1295 $value = $curValue; // return value
1296 $asOf = $curInfo['asOf']; // return value timestamp
1297
1298 // Determine if a cached value regeneration is needed or desired
1299 if (
1300 $this->isValid( $value, $needsVersion, $asOf, $minTime ) &&
1301 $this->isAliveOrInGracePeriod( $curTTL, $graceTTL )
1302 ) {
1303 $preemptiveRefresh = (
1304 $this->worthRefreshExpiring( $curTTL, $lowTTL ) ||
1305 $this->worthRefreshPopular( $asOf, $ageNew, $popWindow, $initialTime )
1306 );
1307
1308 if ( !$preemptiveRefresh ) {
1309 $this->stats->increment( "wanobjectcache.$kClass.hit.good" );
1310
1311 return $value;
1312 } elseif ( $this->scheduleAsyncRefresh( $key, $ttl, $callback, $opts ) ) {
1313 $this->stats->increment( "wanobjectcache.$kClass.hit.refresh" );
1314
1315 return $value;
1316 }
1317 }
1318
1319 $isKeyTombstoned = ( $curInfo['tombAsOf'] !== null );
1320 if ( $isKeyTombstoned ) {
1321 // Get the interim key value since the key is tombstoned (write-holed)
1322 list( $value, $asOf ) = $this->getInterimValue( $key, $needsVersion, $minTime );
1323 // Update the "last purge time" since the $touchedCb timestamp depends on $value
1324 $LPT = $this->resolveTouched( $value, $LPT, $touchedCb );
1325 }
1326
1327 // Reduce mutex and cache set spam while keys are in the tombstone/holdoff period by
1328 // checking if $value was genereated by a recent thread much less than a second ago.
1329 if (
1330 $this->isValid( $value, $needsVersion, $asOf, $minTime, $LPT ) &&
1331 $this->isVolatileValueAgeNegligible( $initialTime - $asOf )
1332 ) {
1333 $this->stats->increment( "wanobjectcache.$kClass.hit.volatile" );
1334
1335 return $value;
1336 }
1337
1338 // Decide if only one thread should handle regeneration at a time
1339 $useMutex =
1340 // Note that since tombstones no-op set(), $lockTSE and $curTTL cannot be used to
1341 // deduce the key hotness because |$curTTL| will always keep increasing until the
1342 // tombstone expires or is overwritten by a new tombstone. Also, even if $lockTSE
1343 // is not set, constant regeneration of a key for the tombstone lifetime might be
1344 // very expensive. Assume tombstoned keys are possibly hot in order to reduce
1345 // the risk of high regeneration load after the delete() method is called.
1346 $isKeyTombstoned ||
1347 // Assume a key is hot if requested soon ($lockTSE seconds) after invalidation.
1348 // This avoids stampedes when timestamps from $checkKeys/$touchedCb bump.
1349 ( $curTTL !== null && $curTTL <= 0 && abs( $curTTL ) <= $lockTSE ) ||
1350 // Assume a key is hot if there is no value and a busy fallback is given.
1351 // This avoids stampedes on eviction or preemptive regeneration taking too long.
1352 ( $busyValue !== null && $value === false );
1353
1354 $hasLock = false;
1355 if ( $useMutex ) {
1356 // Acquire a datacenter-local non-blocking lock
1357 if ( $this->cache->add( self::MUTEX_KEY_PREFIX . $key, 1, self::LOCK_TTL ) ) {
1358 // Lock acquired; this thread will recompute the value and update cache
1359 $hasLock = true;
1360 } elseif ( $this->isValid( $value, $needsVersion, $asOf, $minTime ) ) {
1361 // Lock not acquired and a stale value exists; use the stale value
1362 $this->stats->increment( "wanobjectcache.$kClass.hit.stale" );
1363
1364 return $value;
1365 } else {
1366 // Lock not acquired and no stale value exists
1367 if ( $busyValue !== null ) {
1368 // Use the busy fallback value if nothing else
1369 $miss = is_infinite( $minTime ) ? 'renew' : 'miss';
1370 $this->stats->increment( "wanobjectcache.$kClass.$miss.busy" );
1371
1372 return is_callable( $busyValue ) ? $busyValue() : $busyValue;
1373 }
1374 }
1375 }
1376
1377 if ( !is_callable( $callback ) ) {
1378 throw new InvalidArgumentException( "Invalid cache miss callback provided." );
1379 }
1380
1381 $preCallbackTime = $this->getCurrentTime();
1382 // Generate the new value from the callback...
1383 $setOpts = [];
1384 ++$this->callbackDepth;
1385 try {
1386 $value = call_user_func_array( $callback, [ $curValue, &$ttl, &$setOpts, $asOf ] );
1387 } finally {
1388 --$this->callbackDepth;
1389 }
1390 $valueIsCacheable = ( $value !== false && $ttl >= 0 );
1391
1392 if ( $valueIsCacheable ) {
1393 $ago = max( $this->getCurrentTime() - $initialTime, 0.0 );
1394 $this->stats->timing( "wanobjectcache.$kClass.regen_set_delay", 1000 * $ago );
1395
1396 if ( $isKeyTombstoned ) {
1397 if ( $this->checkAndSetCooloff( $key, $kClass, $ago, $lockTSE, $hasLock ) ) {
1398 // Use the interim key value since the key is tombstoned (write-holed)
1399 $tempTTL = max( self::INTERIM_KEY_TTL, (int)$lockTSE );
1400 $this->setInterimValue( $key, $value, $tempTTL, $this->getCurrentTime() );
1401 }
1402 } elseif ( !$useMutex || $hasLock ) {
1403 if ( $this->checkAndSetCooloff( $key, $kClass, $ago, $lockTSE, $hasLock ) ) {
1404 // Save the value unless a lock-winning thread is already expected to do that
1405 $setOpts['lockTSE'] = $lockTSE;
1406 $setOpts['staleTTL'] = $staleTTL;
1407 // Use best known "since" timestamp if not provided
1408 $setOpts += [ 'since' => $preCallbackTime ];
1409 // Update the cache; this will fail if the key is tombstoned
1410 $this->set( $key, $value, $ttl, $setOpts );
1411 }
1412 }
1413 }
1414
1415 if ( $hasLock ) {
1416 $this->cache->changeTTL( self::MUTEX_KEY_PREFIX . $key, (int)$initialTime - 60 );
1417 }
1418
1419 $miss = is_infinite( $minTime ) ? 'renew' : 'miss';
1420 $this->stats->increment( "wanobjectcache.$kClass.$miss.compute" );
1421
1422 return $value;
1423 }
1424
1425 /**
1426 * @param float $age Age of volatile/interim key in seconds
1427 * @return bool Whether the age of a volatile value is negligible
1428 */
1429 private function isVolatileValueAgeNegligible( $age ) {
1430 return ( $age < mt_rand( self::RECENT_SET_LOW_MS, self::RECENT_SET_HIGH_MS ) / 1e3 );
1431 }
1432
1433 /**
1434 * @param string $key
1435 * @param string $kClass
1436 * @param float $elapsed Seconds spent regenerating the value
1437 * @param float $lockTSE
1438 * @param $hasLock bool
1439 * @return bool Whether it is OK to proceed with a key set operation
1440 */
1441 private function checkAndSetCooloff( $key, $kClass, $elapsed, $lockTSE, $hasLock ) {
1442 // If $lockTSE is set, the lock was bypassed because there was no stale/interim value,
1443 // and $elapsed indicates that regeration is slow, then there is a risk of set()
1444 // stampedes with large blobs. With a typical scale-out infrastructure, CPU and query
1445 // load from $callback invocations is distributed among appservers and replica DBs,
1446 // but cache operations for a given key route to a single cache server (e.g. striped
1447 // consistent hashing).
1448 if ( $lockTSE < 0 || $hasLock ) {
1449 return true; // either not a priori hot or thread has the lock
1450 } elseif ( $elapsed <= self::SET_DELAY_HIGH_SEC ) {
1451 return true; // not enough time for threads to pile up
1452 }
1453
1454 $this->cache->clearLastError();
1455 if (
1456 !$this->cache->add( self::COOLOFF_KEY_PREFIX . $key, 1, self::COOLOFF_TTL ) &&
1457 // Don't treat failures due to I/O errors as the key being in cooloff
1458 $this->cache->getLastError() === BagOStuff::ERR_NONE
1459 ) {
1460 $this->stats->increment( "wanobjectcache.$kClass.cooloff_bounce" );
1461
1462 return false;
1463 }
1464
1465 return true;
1466 }
1467
1468 /**
1469 * @param mixed $value
1470 * @param float|null $curTTL
1471 * @param array $curInfo
1472 * @param callable|null $touchedCallback
1473 * @return array (current time left or null, UNIX timestamp of last purge or null)
1474 * @note Callable type hints are not used to avoid class-autoloading
1475 */
1476 protected function resolveCTL( $value, $curTTL, $curInfo, $touchedCallback ) {
1477 if ( $touchedCallback === null || $value === false ) {
1478 return [ $curTTL, max( $curInfo['tombAsOf'], $curInfo['lastCKPurge'] ) ];
1479 }
1480
1481 if ( !is_callable( $touchedCallback ) ) {
1482 throw new InvalidArgumentException( "Invalid expiration callback provided." );
1483 }
1484
1485 $touched = $touchedCallback( $value );
1486 if ( $touched !== null && $touched >= $curInfo['asOf'] ) {
1487 $curTTL = min( $curTTL, self::TINY_NEGATIVE, $curInfo['asOf'] - $touched );
1488 }
1489
1490 return [ $curTTL, max( $curInfo['tombAsOf'], $curInfo['lastCKPurge'], $touched ) ];
1491 }
1492
1493 /**
1494 * @param mixed $value
1495 * @param float|null $lastPurge
1496 * @param callable|null $touchedCallback
1497 * @return float|null UNIX timestamp of last purge or null
1498 * @note Callable type hints are not used to avoid class-autoloading
1499 */
1500 protected function resolveTouched( $value, $lastPurge, $touchedCallback ) {
1501 if ( $touchedCallback === null || $value === false ) {
1502 return $lastPurge;
1503 }
1504
1505 if ( !is_callable( $touchedCallback ) ) {
1506 throw new InvalidArgumentException( "Invalid expiration callback provided." );
1507 }
1508
1509 return max( $touchedCallback( $value ), $lastPurge );
1510 }
1511
1512 /**
1513 * @param string $key
1514 * @param bool $versioned
1515 * @param float $minTime
1516 * @return array (cached value or false, cached value timestamp or null)
1517 */
1518 protected function getInterimValue( $key, $versioned, $minTime ) {
1519 if ( !$this->useInterimHoldOffCaching ) {
1520 return [ false, null ]; // disabled
1521 }
1522
1523 $wrapped = $this->cache->get( self::INTERIM_KEY_PREFIX . $key );
1524 list( $value ) = $this->unwrap( $wrapped, $this->getCurrentTime() );
1525 $valueAsOf = $wrapped[self::FLD_TIME] ?? null;
1526 if ( $this->isValid( $value, $versioned, $valueAsOf, $minTime ) ) {
1527 return [ $value, $valueAsOf ];
1528 }
1529
1530 return [ false, null ];
1531 }
1532
1533 /**
1534 * @param string $key
1535 * @param mixed $value
1536 * @param int $tempTTL
1537 * @param float $newAsOf
1538 */
1539 protected function setInterimValue( $key, $value, $tempTTL, $newAsOf ) {
1540 $wrapped = $this->wrap( $value, $tempTTL, $newAsOf );
1541
1542 $this->cache->merge(
1543 self::INTERIM_KEY_PREFIX . $key,
1544 function () use ( $wrapped ) {
1545 return $wrapped;
1546 },
1547 $tempTTL,
1548 1
1549 );
1550 }
1551
1552 /**
1553 * Method to fetch multiple cache keys at once with regeneration
1554 *
1555 * This works the same as getWithSetCallback() except:
1556 * - a) The $keys argument expects the result of WANObjectCache::makeMultiKeys()
1557 * - b) The $callback argument expects a callback taking the following arguments:
1558 * - $id: ID of an entity to query
1559 * - $oldValue : the prior cache value or false if none was present
1560 * - &$ttl : a reference to the new value TTL in seconds
1561 * - &$setOpts : a reference to options for set() which can be altered
1562 * - $oldAsOf : generation UNIX timestamp of $oldValue or null if not present
1563 * Aside from the additional $id argument, the other arguments function the same
1564 * way they do in getWithSetCallback().
1565 * - c) The return value is a map of (cache key => value) in the order of $keyedIds
1566 *
1567 * @see WANObjectCache::getWithSetCallback()
1568 * @see WANObjectCache::getMultiWithUnionSetCallback()
1569 *
1570 * Example usage:
1571 * @code
1572 * $rows = $cache->getMultiWithSetCallback(
1573 * // Map of cache keys to entity IDs
1574 * $cache->makeMultiKeys(
1575 * $this->fileVersionIds(),
1576 * function ( $id, WANObjectCache $cache ) {
1577 * return $cache->makeKey( 'file-version', $id );
1578 * }
1579 * ),
1580 * // Time-to-live (in seconds)
1581 * $cache::TTL_DAY,
1582 * // Function that derives the new key value
1583 * function ( $id, $oldValue, &$ttl, array &$setOpts ) {
1584 * $dbr = wfGetDB( DB_REPLICA );
1585 * // Account for any snapshot/replica DB lag
1586 * $setOpts += Database::getCacheSetOptions( $dbr );
1587 *
1588 * // Load the row for this file
1589 * $queryInfo = File::getQueryInfo();
1590 * $row = $dbr->selectRow(
1591 * $queryInfo['tables'],
1592 * $queryInfo['fields'],
1593 * [ 'id' => $id ],
1594 * __METHOD__,
1595 * [],
1596 * $queryInfo['joins']
1597 * );
1598 *
1599 * return $row ? (array)$row : false;
1600 * },
1601 * [
1602 * // Process cache for 30 seconds
1603 * 'pcTTL' => 30,
1604 * // Use a dedicated 500 item cache (initialized on-the-fly)
1605 * 'pcGroup' => 'file-versions:500'
1606 * ]
1607 * );
1608 * $files = array_map( [ __CLASS__, 'newFromRow' ], $rows );
1609 * @endcode
1610 *
1611 * @param ArrayIterator $keyedIds Result of WANObjectCache::makeMultiKeys()
1612 * @param int $ttl Seconds to live for key updates
1613 * @param callable $callback Callback the yields entity regeneration callbacks
1614 * @param array $opts Options map
1615 * @return array Map of (cache key => value) in the same order as $keyedIds
1616 * @since 1.28
1617 */
1618 final public function getMultiWithSetCallback(
1619 ArrayIterator $keyedIds, $ttl, callable $callback, array $opts = []
1620 ) {
1621 $valueKeys = array_keys( $keyedIds->getArrayCopy() );
1622 $checkKeys = $opts['checkKeys'] ?? [];
1623 $pcTTL = $opts['pcTTL'] ?? self::TTL_UNCACHEABLE;
1624
1625 // Load required keys into process cache in one go
1626 $this->warmupCache = $this->getRawKeysForWarmup(
1627 $this->getNonProcessCachedKeys( $valueKeys, $opts, $pcTTL ),
1628 $checkKeys
1629 );
1630 $this->warmupKeyMisses = 0;
1631
1632 // Wrap $callback to match the getWithSetCallback() format while passing $id to $callback
1633 $id = null; // current entity ID
1634 $func = function ( $oldValue, &$ttl, &$setOpts, $oldAsOf ) use ( $callback, &$id ) {
1635 return $callback( $id, $oldValue, $ttl, $setOpts, $oldAsOf );
1636 };
1637
1638 $values = [];
1639 foreach ( $keyedIds as $key => $id ) { // preserve order
1640 $values[$key] = $this->getWithSetCallback( $key, $ttl, $func, $opts );
1641 }
1642
1643 $this->warmupCache = [];
1644
1645 return $values;
1646 }
1647
1648 /**
1649 * Method to fetch/regenerate multiple cache keys at once
1650 *
1651 * This works the same as getWithSetCallback() except:
1652 * - a) The $keys argument expects the result of WANObjectCache::makeMultiKeys()
1653 * - b) The $callback argument expects a callback returning a map of (ID => new value)
1654 * for all entity IDs in $ids and it takes the following arguments:
1655 * - $ids: a list of entity IDs that require cache regeneration
1656 * - &$ttls: a reference to the (entity ID => new TTL) map
1657 * - &$setOpts: a reference to options for set() which can be altered
1658 * - c) The return value is a map of (cache key => value) in the order of $keyedIds
1659 * - d) The "lockTSE" and "busyValue" options are ignored
1660 *
1661 * @see WANObjectCache::getWithSetCallback()
1662 * @see WANObjectCache::getMultiWithSetCallback()
1663 *
1664 * Example usage:
1665 * @code
1666 * $rows = $cache->getMultiWithUnionSetCallback(
1667 * // Map of cache keys to entity IDs
1668 * $cache->makeMultiKeys(
1669 * $this->fileVersionIds(),
1670 * function ( $id, WANObjectCache $cache ) {
1671 * return $cache->makeKey( 'file-version', $id );
1672 * }
1673 * ),
1674 * // Time-to-live (in seconds)
1675 * $cache::TTL_DAY,
1676 * // Function that derives the new key value
1677 * function ( array $ids, array &$ttls, array &$setOpts ) {
1678 * $dbr = wfGetDB( DB_REPLICA );
1679 * // Account for any snapshot/replica DB lag
1680 * $setOpts += Database::getCacheSetOptions( $dbr );
1681 *
1682 * // Load the rows for these files
1683 * $rows = [];
1684 * $queryInfo = File::getQueryInfo();
1685 * $res = $dbr->select(
1686 * $queryInfo['tables'],
1687 * $queryInfo['fields'],
1688 * [ 'id' => $ids ],
1689 * __METHOD__,
1690 * [],
1691 * $queryInfo['joins']
1692 * );
1693 * foreach ( $res as $row ) {
1694 * $rows[$row->id] = $row;
1695 * $mtime = wfTimestamp( TS_UNIX, $row->timestamp );
1696 * $ttls[$row->id] = $this->adaptiveTTL( $mtime, $ttls[$row->id] );
1697 * }
1698 *
1699 * return $rows;
1700 * },
1701 * ]
1702 * );
1703 * $files = array_map( [ __CLASS__, 'newFromRow' ], $rows );
1704 * @endcode
1705 *
1706 * @param ArrayIterator $keyedIds Result of WANObjectCache::makeMultiKeys()
1707 * @param int $ttl Seconds to live for key updates
1708 * @param callable $callback Callback the yields entity regeneration callbacks
1709 * @param array $opts Options map
1710 * @return array Map of (cache key => value) in the same order as $keyedIds
1711 * @since 1.30
1712 */
1713 final public function getMultiWithUnionSetCallback(
1714 ArrayIterator $keyedIds, $ttl, callable $callback, array $opts = []
1715 ) {
1716 $idsByValueKey = $keyedIds->getArrayCopy();
1717 $valueKeys = array_keys( $idsByValueKey );
1718 $checkKeys = $opts['checkKeys'] ?? [];
1719 $pcTTL = $opts['pcTTL'] ?? self::TTL_UNCACHEABLE;
1720 unset( $opts['lockTSE'] ); // incompatible
1721 unset( $opts['busyValue'] ); // incompatible
1722
1723 // Load required keys into process cache in one go
1724 $keysGet = $this->getNonProcessCachedKeys( $valueKeys, $opts, $pcTTL );
1725 $this->warmupCache = $this->getRawKeysForWarmup( $keysGet, $checkKeys );
1726 $this->warmupKeyMisses = 0;
1727
1728 // IDs of entities known to be in need of regeneration
1729 $idsRegen = [];
1730
1731 // Find out which keys are missing/deleted/stale
1732 $curTTLs = [];
1733 $asOfs = [];
1734 $curByKey = $this->getMulti( $keysGet, $curTTLs, $checkKeys, $asOfs );
1735 foreach ( $keysGet as $key ) {
1736 if ( !array_key_exists( $key, $curByKey ) || $curTTLs[$key] < 0 ) {
1737 $idsRegen[] = $idsByValueKey[$key];
1738 }
1739 }
1740
1741 // Run the callback to populate the regeneration value map for all required IDs
1742 $newSetOpts = [];
1743 $newTTLsById = array_fill_keys( $idsRegen, $ttl );
1744 $newValsById = $idsRegen ? $callback( $idsRegen, $newTTLsById, $newSetOpts ) : [];
1745
1746 // Wrap $callback to match the getWithSetCallback() format while passing $id to $callback
1747 $id = null; // current entity ID
1748 $func = function ( $oldValue, &$ttl, &$setOpts, $oldAsOf )
1749 use ( $callback, &$id, $newValsById, $newTTLsById, $newSetOpts )
1750 {
1751 if ( array_key_exists( $id, $newValsById ) ) {
1752 // Value was already regerated as expected, so use the value in $newValsById
1753 $newValue = $newValsById[$id];
1754 $ttl = $newTTLsById[$id];
1755 $setOpts = $newSetOpts;
1756 } else {
1757 // Pre-emptive/popularity refresh and version mismatch cases are not detected
1758 // above and thus $newValsById has no entry. Run $callback on this single entity.
1759 $ttls = [ $id => $ttl ];
1760 $newValue = $callback( [ $id ], $ttls, $setOpts )[$id];
1761 $ttl = $ttls[$id];
1762 }
1763
1764 return $newValue;
1765 };
1766
1767 // Run the cache-aside logic using warmupCache instead of persistent cache queries
1768 $values = [];
1769 foreach ( $idsByValueKey as $key => $id ) { // preserve order
1770 $values[$key] = $this->getWithSetCallback( $key, $ttl, $func, $opts );
1771 }
1772
1773 $this->warmupCache = [];
1774
1775 return $values;
1776 }
1777
1778 /**
1779 * Set a key to soon expire in the local cluster if it pre-dates $purgeTimestamp
1780 *
1781 * This sets stale keys' time-to-live at HOLDOFF_TTL seconds, which both avoids
1782 * broadcasting in mcrouter setups and also avoids races with new tombstones.
1783 *
1784 * @param string $key Cache key
1785 * @param int $purgeTimestamp UNIX timestamp of purge
1786 * @param bool &$isStale Whether the key is stale
1787 * @return bool Success
1788 * @since 1.28
1789 */
1790 final public function reap( $key, $purgeTimestamp, &$isStale = false ) {
1791 $minAsOf = $purgeTimestamp + self::HOLDOFF_TTL;
1792 $wrapped = $this->cache->get( self::VALUE_KEY_PREFIX . $key );
1793 if ( is_array( $wrapped ) && $wrapped[self::FLD_TIME] < $minAsOf ) {
1794 $isStale = true;
1795 $this->logger->warning( "Reaping stale value key '$key'." );
1796 $ttlReap = self::HOLDOFF_TTL; // avoids races with tombstone creation
1797 $ok = $this->cache->changeTTL( self::VALUE_KEY_PREFIX . $key, $ttlReap );
1798 if ( !$ok ) {
1799 $this->logger->error( "Could not complete reap of key '$key'." );
1800 }
1801
1802 return $ok;
1803 }
1804
1805 $isStale = false;
1806
1807 return true;
1808 }
1809
1810 /**
1811 * Set a "check" key to soon expire in the local cluster if it pre-dates $purgeTimestamp
1812 *
1813 * @param string $key Cache key
1814 * @param int $purgeTimestamp UNIX timestamp of purge
1815 * @param bool &$isStale Whether the key is stale
1816 * @return bool Success
1817 * @since 1.28
1818 */
1819 final public function reapCheckKey( $key, $purgeTimestamp, &$isStale = false ) {
1820 $purge = $this->parsePurgeValue( $this->cache->get( self::TIME_KEY_PREFIX . $key ) );
1821 if ( $purge && $purge[self::FLD_TIME] < $purgeTimestamp ) {
1822 $isStale = true;
1823 $this->logger->warning( "Reaping stale check key '$key'." );
1824 $ok = $this->cache->changeTTL( self::TIME_KEY_PREFIX . $key, self::TTL_SECOND );
1825 if ( !$ok ) {
1826 $this->logger->error( "Could not complete reap of check key '$key'." );
1827 }
1828
1829 return $ok;
1830 }
1831
1832 $isStale = false;
1833
1834 return false;
1835 }
1836
1837 /**
1838 * @see BagOStuff::makeKey()
1839 * @param string $class Key class
1840 * @param string|null $component [optional] Key component (starting with a key collection name)
1841 * @return string Colon-delimited list of $keyspace followed by escaped components of $args
1842 * @since 1.27
1843 */
1844 public function makeKey( $class, $component = null ) {
1845 return $this->cache->makeKey( ...func_get_args() );
1846 }
1847
1848 /**
1849 * @see BagOStuff::makeGlobalKey()
1850 * @param string $class Key class
1851 * @param string|null $component [optional] Key component (starting with a key collection name)
1852 * @return string Colon-delimited list of $keyspace followed by escaped components of $args
1853 * @since 1.27
1854 */
1855 public function makeGlobalKey( $class, $component = null ) {
1856 return $this->cache->makeGlobalKey( ...func_get_args() );
1857 }
1858
1859 /**
1860 * @param array $entities List of entity IDs
1861 * @param callable $keyFunc Callback yielding a key from (entity ID, this WANObjectCache)
1862 * @return ArrayIterator Iterator yielding (cache key => entity ID) in $entities order
1863 * @since 1.28
1864 */
1865 final public function makeMultiKeys( array $entities, callable $keyFunc ) {
1866 $map = [];
1867 foreach ( $entities as $entity ) {
1868 $map[$keyFunc( $entity, $this )] = $entity;
1869 }
1870
1871 return new ArrayIterator( $map );
1872 }
1873
1874 /**
1875 * Get the "last error" registered; clearLastError() should be called manually
1876 * @return int ERR_* class constant for the "last error" registry
1877 */
1878 final public function getLastError() {
1879 $code = $this->cache->getLastError();
1880 switch ( $code ) {
1881 case BagOStuff::ERR_NONE:
1882 return self::ERR_NONE;
1883 case BagOStuff::ERR_NO_RESPONSE:
1884 return self::ERR_NO_RESPONSE;
1885 case BagOStuff::ERR_UNREACHABLE:
1886 return self::ERR_UNREACHABLE;
1887 default:
1888 return self::ERR_UNEXPECTED;
1889 }
1890 }
1891
1892 /**
1893 * Clear the "last error" registry
1894 */
1895 final public function clearLastError() {
1896 $this->cache->clearLastError();
1897 }
1898
1899 /**
1900 * Clear the in-process caches; useful for testing
1901 *
1902 * @since 1.27
1903 */
1904 public function clearProcessCache() {
1905 $this->processCaches = [];
1906 }
1907
1908 /**
1909 * Enable or disable the use of brief caching for tombstoned keys
1910 *
1911 * When a key is purged via delete(), there normally is a period where caching
1912 * is hold-off limited to an extremely short time. This method will disable that
1913 * caching, forcing the callback to run for any of:
1914 * - WANObjectCache::getWithSetCallback()
1915 * - WANObjectCache::getMultiWithSetCallback()
1916 * - WANObjectCache::getMultiWithUnionSetCallback()
1917 *
1918 * This is useful when both:
1919 * - a) the database used by the callback is known to be up-to-date enough
1920 * for some particular purpose (e.g. replica DB has applied transaction X)
1921 * - b) the caller needs to exploit that fact, and therefore needs to avoid the
1922 * use of inherently volatile and possibly stale interim keys
1923 *
1924 * @see WANObjectCache::delete()
1925 * @param bool $enabled Whether to enable interim caching
1926 * @since 1.31
1927 */
1928 final public function useInterimHoldOffCaching( $enabled ) {
1929 $this->useInterimHoldOffCaching = $enabled;
1930 }
1931
1932 /**
1933 * @param int $flag ATTR_* class constant
1934 * @return int QOS_* class constant
1935 * @since 1.28
1936 */
1937 public function getQoS( $flag ) {
1938 return $this->cache->getQoS( $flag );
1939 }
1940
1941 /**
1942 * Get a TTL that is higher for objects that have not changed recently
1943 *
1944 * This is useful for keys that get explicit purges and DB or purge relay
1945 * lag is a potential concern (especially how it interacts with CDN cache)
1946 *
1947 * Example usage:
1948 * @code
1949 * // Last-modified time of page
1950 * $mtime = wfTimestamp( TS_UNIX, $page->getTimestamp() );
1951 * // Get adjusted TTL. If $mtime is 3600 seconds ago and $minTTL/$factor left at
1952 * // defaults, then $ttl is 3600 * .2 = 720. If $minTTL was greater than 720, then
1953 * // $ttl would be $minTTL. If $maxTTL was smaller than 720, $ttl would be $maxTTL.
1954 * $ttl = $cache->adaptiveTTL( $mtime, $cache::TTL_DAY );
1955 * @endcode
1956 *
1957 * Another use case is when there are no applicable "last modified" fields in the DB,
1958 * and there are too many dependencies for explicit purges to be viable, and the rate of
1959 * change to relevant content is unstable, and it is highly valued to have the cached value
1960 * be as up-to-date as possible.
1961 *
1962 * Example usage:
1963 * @code
1964 * $query = "<some complex query>";
1965 * $idListFromComplexQuery = $cache->getWithSetCallback(
1966 * $cache->makeKey( 'complex-graph-query', $hashOfQuery ),
1967 * GraphQueryClass::STARTING_TTL,
1968 * function ( $oldValue, &$ttl, array &$setOpts, $oldAsOf ) use ( $query, $cache ) {
1969 * $gdb = $this->getReplicaGraphDbConnection();
1970 * // Account for any snapshot/replica DB lag
1971 * $setOpts += GraphDatabase::getCacheSetOptions( $gdb );
1972 *
1973 * $newList = iterator_to_array( $gdb->query( $query ) );
1974 * sort( $newList, SORT_NUMERIC ); // normalize
1975 *
1976 * $minTTL = GraphQueryClass::MIN_TTL;
1977 * $maxTTL = GraphQueryClass::MAX_TTL;
1978 * if ( $oldValue !== false ) {
1979 * // Note that $oldAsOf is the last time this callback ran
1980 * $ttl = ( $newList === $oldValue )
1981 * // No change: cache for 150% of the age of $oldValue
1982 * ? $cache->adaptiveTTL( $oldAsOf, $maxTTL, $minTTL, 1.5 )
1983 * // Changed: cache for 50% of the age of $oldValue
1984 * : $cache->adaptiveTTL( $oldAsOf, $maxTTL, $minTTL, .5 );
1985 * }
1986 *
1987 * return $newList;
1988 * },
1989 * [
1990 * // Keep stale values around for doing comparisons for TTL calculations.
1991 * // High values improve long-tail keys hit-rates, though might waste space.
1992 * 'staleTTL' => GraphQueryClass::GRACE_TTL
1993 * ]
1994 * );
1995 * @endcode
1996 *
1997 * @param int|float $mtime UNIX timestamp
1998 * @param int $maxTTL Maximum TTL (seconds)
1999 * @param int $minTTL Minimum TTL (seconds); Default: 30
2000 * @param float $factor Value in the range (0,1); Default: .2
2001 * @return int Adaptive TTL
2002 * @since 1.28
2003 */
2004 public function adaptiveTTL( $mtime, $maxTTL, $minTTL = 30, $factor = 0.2 ) {
2005 if ( is_float( $mtime ) || ctype_digit( $mtime ) ) {
2006 $mtime = (int)$mtime; // handle fractional seconds and string integers
2007 }
2008
2009 if ( !is_int( $mtime ) || $mtime <= 0 ) {
2010 return $minTTL; // no last-modified time provided
2011 }
2012
2013 $age = $this->getCurrentTime() - $mtime;
2014
2015 return (int)min( $maxTTL, max( $minTTL, $factor * $age ) );
2016 }
2017
2018 /**
2019 * @return int Number of warmup key cache misses last round
2020 * @since 1.30
2021 */
2022 final public function getWarmupKeyMisses() {
2023 return $this->warmupKeyMisses;
2024 }
2025
2026 /**
2027 * Do the actual async bus purge of a key
2028 *
2029 * This must set the key to "PURGED:<UNIX timestamp>:<holdoff>"
2030 *
2031 * @param string $key Cache key
2032 * @param int $ttl How long to keep the tombstone [seconds]
2033 * @param int $holdoff HOLDOFF_* constant controlling how long to ignore sets for this key
2034 * @return bool Success
2035 */
2036 protected function relayPurge( $key, $ttl, $holdoff ) {
2037 if ( $this->mcrouterAware ) {
2038 // See https://github.com/facebook/mcrouter/wiki/Multi-cluster-broadcast-setup
2039 // Wildcards select all matching routes, e.g. the WAN cluster on all DCs
2040 $ok = $this->cache->set(
2041 "/*/{$this->cluster}/{$key}",
2042 $this->makePurgeValue( $this->getCurrentTime(), self::HOLDOFF_NONE ),
2043 $ttl
2044 );
2045 } else {
2046 // This handles the mcrouter and the single-DC case
2047 $ok = $this->cache->set(
2048 $key,
2049 $this->makePurgeValue( $this->getCurrentTime(), self::HOLDOFF_NONE ),
2050 $ttl
2051 );
2052 }
2053
2054 return $ok;
2055 }
2056
2057 /**
2058 * Do the actual async bus delete of a key
2059 *
2060 * @param string $key Cache key
2061 * @return bool Success
2062 */
2063 protected function relayDelete( $key ) {
2064 if ( $this->mcrouterAware ) {
2065 // See https://github.com/facebook/mcrouter/wiki/Multi-cluster-broadcast-setup
2066 // Wildcards select all matching routes, e.g. the WAN cluster on all DCs
2067 $ok = $this->cache->delete( "/*/{$this->cluster}/{$key}" );
2068 } else {
2069 // Some other proxy handles broadcasting or there is only one datacenter
2070 $ok = $this->cache->delete( $key );
2071 }
2072
2073 return $ok;
2074 }
2075
2076 /**
2077 * @param string $key
2078 * @param int $ttl
2079 * @param callable $callback
2080 * @param array $opts
2081 * @return bool Success
2082 */
2083 private function scheduleAsyncRefresh( $key, $ttl, $callback, $opts ) {
2084 if ( !$this->asyncHandler ) {
2085 return false;
2086 }
2087 // Update the cache value later, such during post-send of an HTTP request
2088 $func = $this->asyncHandler;
2089 $func( function () use ( $key, $ttl, $callback, $opts ) {
2090 $asOf = null; // unused
2091 $opts['minAsOf'] = INF; // force a refresh
2092 $this->doGetWithSetCallback( $key, $ttl, $callback, $opts, $asOf );
2093 } );
2094
2095 return true;
2096 }
2097
2098 /**
2099 * Check if a key is fresh or in the grace window and thus due for randomized reuse
2100 *
2101 * If $curTTL > 0 (e.g. not expired) this returns true. Otherwise, the chance of returning
2102 * true decrease steadily from 100% to 0% as the |$curTTL| moves from 0 to $graceTTL seconds.
2103 * This handles widely varying levels of cache access traffic.
2104 *
2105 * If $curTTL <= -$graceTTL (e.g. already expired), then this returns false.
2106 *
2107 * @param float $curTTL Approximate TTL left on the key if present
2108 * @param int $graceTTL Consider using stale values if $curTTL is greater than this
2109 * @return bool
2110 */
2111 protected function isAliveOrInGracePeriod( $curTTL, $graceTTL ) {
2112 if ( $curTTL > 0 ) {
2113 return true;
2114 } elseif ( $graceTTL <= 0 ) {
2115 return false;
2116 }
2117
2118 $ageStale = abs( $curTTL ); // seconds of staleness
2119 $curGTTL = ( $graceTTL - $ageStale ); // current grace-time-to-live
2120 if ( $curGTTL <= 0 ) {
2121 return false; // already out of grace period
2122 }
2123
2124 // Chance of using a stale value is the complement of the chance of refreshing it
2125 return !$this->worthRefreshExpiring( $curGTTL, $graceTTL );
2126 }
2127
2128 /**
2129 * Check if a key is nearing expiration and thus due for randomized regeneration
2130 *
2131 * This returns false if $curTTL >= $lowTTL. Otherwise, the chance of returning true
2132 * increases steadily from 0% to 100% as the $curTTL moves from $lowTTL to 0 seconds.
2133 * This handles widely varying levels of cache access traffic.
2134 *
2135 * If $curTTL <= 0 (e.g. already expired), then this returns false.
2136 *
2137 * @param float $curTTL Approximate TTL left on the key if present
2138 * @param float $lowTTL Consider a refresh when $curTTL is less than this
2139 * @return bool
2140 */
2141 protected function worthRefreshExpiring( $curTTL, $lowTTL ) {
2142 if ( $lowTTL <= 0 ) {
2143 return false;
2144 } elseif ( $curTTL >= $lowTTL ) {
2145 return false;
2146 } elseif ( $curTTL <= 0 ) {
2147 return false;
2148 }
2149
2150 $chance = ( 1 - $curTTL / $lowTTL );
2151
2152 return mt_rand( 1, 1e9 ) <= 1e9 * $chance;
2153 }
2154
2155 /**
2156 * Check if a key is due for randomized regeneration due to its popularity
2157 *
2158 * This is used so that popular keys can preemptively refresh themselves for higher
2159 * consistency (especially in the case of purge loss/delay). Unpopular keys can remain
2160 * in cache with their high nominal TTL. This means popular keys keep good consistency,
2161 * whether the data changes frequently or not, and long-tail keys get to stay in cache
2162 * and get hits too. Similar to worthRefreshExpiring(), randomization is used.
2163 *
2164 * @param float $asOf UNIX timestamp of the value
2165 * @param int $ageNew Age of key when this might recommend refreshing (seconds)
2166 * @param int $timeTillRefresh Age of key when it should be refreshed if popular (seconds)
2167 * @param float $now The current UNIX timestamp
2168 * @return bool
2169 */
2170 protected function worthRefreshPopular( $asOf, $ageNew, $timeTillRefresh, $now ) {
2171 if ( $ageNew < 0 || $timeTillRefresh <= 0 ) {
2172 return false;
2173 }
2174
2175 $age = $now - $asOf;
2176 $timeOld = $age - $ageNew;
2177 if ( $timeOld <= 0 ) {
2178 return false;
2179 }
2180
2181 // Lifecycle is: new, ramp-up refresh chance, full refresh chance.
2182 // Note that the "expected # of refreshes" for the ramp-up time range is half of what it
2183 // would be if P(refresh) was at its full value during that time range.
2184 $refreshWindowSec = max( $timeTillRefresh - $ageNew - self::RAMPUP_TTL / 2, 1 );
2185 // P(refresh) * (# hits in $refreshWindowSec) = (expected # of refreshes)
2186 // P(refresh) * ($refreshWindowSec * $popularHitsPerSec) = 1
2187 // P(refresh) = 1/($refreshWindowSec * $popularHitsPerSec)
2188 $chance = 1 / ( self::HIT_RATE_HIGH * $refreshWindowSec );
2189
2190 // Ramp up $chance from 0 to its nominal value over RAMPUP_TTL seconds to avoid stampedes
2191 $chance *= ( $timeOld <= self::RAMPUP_TTL ) ? $timeOld / self::RAMPUP_TTL : 1;
2192
2193 return mt_rand( 1, 1e9 ) <= 1e9 * $chance;
2194 }
2195
2196 /**
2197 * Check if $value is not false, versioned (if needed), and not older than $minTime (if set)
2198 *
2199 * @param array|bool $value
2200 * @param bool $versioned
2201 * @param float $asOf The time $value was generated
2202 * @param float $minTime The last time the main value was generated (0.0 if unknown)
2203 * @param float|null $purgeTime The last time the value was invalidated
2204 * @return bool
2205 */
2206 protected function isValid( $value, $versioned, $asOf, $minTime, $purgeTime = null ) {
2207 // Avoid reading any key not generated after the latest delete() or touch
2208 $safeMinTime = max( $minTime, $purgeTime + self::TINY_POSTIVE );
2209
2210 if ( $value === false ) {
2211 return false;
2212 } elseif ( $versioned && !isset( $value[self::VFLD_VERSION] ) ) {
2213 return false;
2214 } elseif ( $safeMinTime > 0 && $asOf < $minTime ) {
2215 return false;
2216 }
2217
2218 return true;
2219 }
2220
2221 /**
2222 * Do not use this method outside WANObjectCache
2223 *
2224 * @param mixed $value
2225 * @param int $ttl [0=forever]
2226 * @param float $now Unix Current timestamp just before calling set()
2227 * @return array
2228 */
2229 protected function wrap( $value, $ttl, $now ) {
2230 return [
2231 self::FLD_VERSION => self::VERSION,
2232 self::FLD_VALUE => $value,
2233 self::FLD_TTL => $ttl,
2234 self::FLD_TIME => $now
2235 ];
2236 }
2237
2238 /**
2239 * Do not use this method outside WANObjectCache
2240 *
2241 * The cached value will be false if absent/tombstoned/malformed
2242 *
2243 * @param array|string|bool $wrapped
2244 * @param float $now Unix Current timestamp (preferrably pre-query)
2245 * @return array (cached value or false, current TTL, value timestamp, tombstone timestamp)
2246 */
2247 protected function unwrap( $wrapped, $now ) {
2248 // Check if the value is a tombstone
2249 $purge = $this->parsePurgeValue( $wrapped );
2250 if ( $purge !== false ) {
2251 // Purged values should always have a negative current $ttl
2252 $curTTL = min( $purge[self::FLD_TIME] - $now, self::TINY_NEGATIVE );
2253 return [ false, $curTTL, null, $purge[self::FLD_TIME] ];
2254 }
2255
2256 if ( !is_array( $wrapped ) // not found
2257 || !isset( $wrapped[self::FLD_VERSION] ) // wrong format
2258 || $wrapped[self::FLD_VERSION] !== self::VERSION // wrong version
2259 ) {
2260 return [ false, null, null, null ];
2261 }
2262
2263 if ( $wrapped[self::FLD_TTL] > 0 ) {
2264 // Get the approximate time left on the key
2265 $age = $now - $wrapped[self::FLD_TIME];
2266 $curTTL = max( $wrapped[self::FLD_TTL] - $age, 0.0 );
2267 } else {
2268 // Key had no TTL, so the time left is unbounded
2269 $curTTL = INF;
2270 }
2271
2272 if ( $wrapped[self::FLD_TIME] < $this->epoch ) {
2273 // Values this old are ignored
2274 return [ false, null, null, null ];
2275 }
2276
2277 return [ $wrapped[self::FLD_VALUE], $curTTL, $wrapped[self::FLD_TIME], null ];
2278 }
2279
2280 /**
2281 * @param array $keys
2282 * @param string $prefix
2283 * @return string[]
2284 */
2285 protected static function prefixCacheKeys( array $keys, $prefix ) {
2286 $res = [];
2287 foreach ( $keys as $key ) {
2288 $res[] = $prefix . $key;
2289 }
2290
2291 return $res;
2292 }
2293
2294 /**
2295 * @param string $key String of the format <scope>:<class>[:<class or variable>]...
2296 * @return string
2297 */
2298 protected function determineKeyClass( $key ) {
2299 $parts = explode( ':', $key );
2300
2301 return $parts[1] ?? $parts[0]; // sanity
2302 }
2303
2304 /**
2305 * @param string|array|bool $value Possible string of the form "PURGED:<timestamp>:<holdoff>"
2306 * @return array|bool Array containing a UNIX timestamp (float) and holdoff period (integer),
2307 * or false if value isn't a valid purge value
2308 */
2309 protected function parsePurgeValue( $value ) {
2310 if ( !is_string( $value ) ) {
2311 return false;
2312 }
2313
2314 $segments = explode( ':', $value, 3 );
2315 if ( !isset( $segments[0] ) || !isset( $segments[1] )
2316 || "{$segments[0]}:" !== self::PURGE_VAL_PREFIX
2317 ) {
2318 return false;
2319 }
2320
2321 if ( !isset( $segments[2] ) ) {
2322 // Back-compat with old purge values without holdoff
2323 $segments[2] = self::HOLDOFF_TTL;
2324 }
2325
2326 if ( $segments[1] < $this->epoch ) {
2327 // Values this old are ignored
2328 return false;
2329 }
2330
2331 return [
2332 self::FLD_TIME => (float)$segments[1],
2333 self::FLD_HOLDOFF => (int)$segments[2],
2334 ];
2335 }
2336
2337 /**
2338 * @param float $timestamp
2339 * @param int $holdoff In seconds
2340 * @return string Wrapped purge value
2341 */
2342 protected function makePurgeValue( $timestamp, $holdoff ) {
2343 return self::PURGE_VAL_PREFIX . (float)$timestamp . ':' . (int)$holdoff;
2344 }
2345
2346 /**
2347 * @param string $group
2348 * @return MapCacheLRU
2349 */
2350 protected function getProcessCache( $group ) {
2351 if ( !isset( $this->processCaches[$group] ) ) {
2352 list( , $n ) = explode( ':', $group );
2353 $this->processCaches[$group] = new MapCacheLRU( (int)$n );
2354 }
2355
2356 return $this->processCaches[$group];
2357 }
2358
2359 /**
2360 * @param array $keys
2361 * @param array $opts
2362 * @param int $pcTTL
2363 * @return array List of keys
2364 */
2365 private function getNonProcessCachedKeys( array $keys, array $opts, $pcTTL ) {
2366 $keysFound = [];
2367 if ( isset( $opts['pcTTL'] ) && $opts['pcTTL'] > 0 && $this->callbackDepth == 0 ) {
2368 $pcGroup = $opts['pcGroup'] ?? self::PC_PRIMARY;
2369 $procCache = $this->getProcessCache( $pcGroup );
2370 foreach ( $keys as $key ) {
2371 if ( $procCache->has( $key, $pcTTL ) ) {
2372 $keysFound[] = $key;
2373 }
2374 }
2375 }
2376
2377 return array_diff( $keys, $keysFound );
2378 }
2379
2380 /**
2381 * @param array $keys
2382 * @param array $checkKeys
2383 * @return array Map of (cache key => mixed)
2384 */
2385 private function getRawKeysForWarmup( array $keys, array $checkKeys ) {
2386 if ( !$keys ) {
2387 return [];
2388 }
2389
2390 $keysWarmUp = [];
2391 // Get all the value keys to fetch...
2392 foreach ( $keys as $key ) {
2393 $keysWarmUp[] = self::VALUE_KEY_PREFIX . $key;
2394 }
2395 // Get all the check keys to fetch...
2396 foreach ( $checkKeys as $i => $checkKeyOrKeys ) {
2397 if ( is_int( $i ) ) {
2398 // Single check key that applies to all value keys
2399 $keysWarmUp[] = self::TIME_KEY_PREFIX . $checkKeyOrKeys;
2400 } else {
2401 // List of check keys that apply to value key $i
2402 $keysWarmUp = array_merge(
2403 $keysWarmUp,
2404 self::prefixCacheKeys( $checkKeyOrKeys, self::TIME_KEY_PREFIX )
2405 );
2406 }
2407 }
2408
2409 $warmupCache = $this->cache->getMulti( $keysWarmUp );
2410 $warmupCache += array_fill_keys( $keysWarmUp, false );
2411
2412 return $warmupCache;
2413 }
2414
2415 /**
2416 * @return float UNIX timestamp
2417 * @codeCoverageIgnore
2418 */
2419 protected function getCurrentTime() {
2420 if ( $this->wallClockOverride ) {
2421 return $this->wallClockOverride;
2422 }
2423
2424 $clockTime = (float)time(); // call this first
2425 // microtime() uses an initial gettimeofday() call added to usage clocks.
2426 // This can severely drift from time() and the microtime() value of other threads
2427 // due to undercounting of the amount of time elapsed. Instead of seeing the current
2428 // time as being in the past, use the value of time(). This avoids setting cache values
2429 // that will immediately be seen as expired and possibly cause stampedes.
2430 return max( microtime( true ), $clockTime );
2431 }
2432
2433 /**
2434 * @param float|null &$time Mock UNIX timestamp for testing
2435 * @codeCoverageIgnore
2436 */
2437 public function setMockTime( &$time ) {
2438 $this->wallClockOverride =& $time;
2439 $this->cache->setMockTime( $time );
2440 }
2441 }