Add meta=userinfo&uiprop=latestcontrib
[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 * Otherwise, $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 (default)
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 = self::TTL_INDEFINITE, 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 if ( $ttl <= 0 ) {
694 // Publish the purge to all datacenters
695 $ok = $this->relayDelete( self::VALUE_KEY_PREFIX . $key );
696 } else {
697 // Publish the purge to all datacenters
698 $ok = $this->relayPurge( self::VALUE_KEY_PREFIX . $key, $ttl, self::HOLDOFF_NONE );
699 }
700
701 $kClass = $this->determineKeyClassForStats( $key );
702 $this->stats->increment( "wanobjectcache.$kClass.delete." . ( $ok ? 'ok' : 'error' ) );
703
704 return $ok;
705 }
706
707 /**
708 * Fetch the value of a timestamp "check" key
709 *
710 * The key will be *initialized* to the current time if not set,
711 * so only call this method if this behavior is actually desired
712 *
713 * The timestamp can be used to check whether a cached value is valid.
714 * Callers should not assume that this returns the same timestamp in
715 * all datacenters due to relay delays.
716 *
717 * The level of staleness can roughly be estimated from this key, but
718 * if the key was evicted from cache, such calculations may show the
719 * time since expiry as ~0 seconds.
720 *
721 * Note that "check" keys won't collide with other regular keys.
722 *
723 * @param string $key
724 * @return float UNIX timestamp
725 */
726 final public function getCheckKeyTime( $key ) {
727 return $this->getMultiCheckKeyTime( [ $key ] )[$key];
728 }
729
730 /**
731 * Fetch the values of each timestamp "check" key
732 *
733 * This works like getCheckKeyTime() except it takes a list of keys
734 * and returns a map of timestamps instead of just that of one key
735 *
736 * This might be useful if both:
737 * - a) a class of entities each depend on hundreds of other entities
738 * - b) these other entities are depended upon by millions of entities
739 *
740 * The later entities can each use a "check" key to invalidate their dependee entities.
741 * However, it is expensive for the former entities to verify against all of the relevant
742 * "check" keys during each getWithSetCallback() call. A less expensive approach is to do
743 * these verifications only after a "time-till-verify" (TTV) has passed. This is a middle
744 * ground between using blind TTLs and using constant verification. The adaptiveTTL() method
745 * can be used to dynamically adjust the TTV. Also, the initial TTV can make use of the
746 * last-modified times of the dependant entities (either from the DB or the "check" keys).
747 *
748 * Example usage:
749 * @code
750 * $value = $cache->getWithSetCallback(
751 * $cache->makeGlobalKey( 'wikibase-item', $id ),
752 * self::INITIAL_TTV, // initial time-till-verify
753 * function ( $oldValue, &$ttv, &$setOpts, $oldAsOf ) use ( $checkKeys, $cache ) {
754 * $now = microtime( true );
755 * // Use $oldValue if it passes max ultimate age and "check" key comparisons
756 * if ( $oldValue &&
757 * $oldAsOf > max( $cache->getMultiCheckKeyTime( $checkKeys ) ) &&
758 * ( $now - $oldValue['ctime'] ) <= self::MAX_CACHE_AGE
759 * ) {
760 * // Increase time-till-verify by 50% of last time to reduce overhead
761 * $ttv = $cache->adaptiveTTL( $oldAsOf, self::MAX_TTV, self::MIN_TTV, 1.5 );
762 * // Unlike $oldAsOf, "ctime" is the ultimate age of the cached data
763 * return $oldValue;
764 * }
765 *
766 * $mtimes = []; // dependency last-modified times; passed by reference
767 * $value = [ 'data' => $this->fetchEntityData( $mtimes ), 'ctime' => $now ];
768 * // Guess time-till-change among the dependencies, e.g. 1/(total change rate)
769 * $ttc = 1 / array_sum( array_map(
770 * function ( $mtime ) use ( $now ) {
771 * return 1 / ( $mtime ? ( $now - $mtime ) : 900 );
772 * },
773 * $mtimes
774 * ) );
775 * // The time-to-verify should not be overly pessimistic nor optimistic
776 * $ttv = min( max( $ttc, self::MIN_TTV ), self::MAX_TTV );
777 *
778 * return $value;
779 * },
780 * [ 'staleTTL' => $cache::TTL_DAY ] // keep around to verify and re-save
781 * );
782 * @endcode
783 *
784 * @see WANObjectCache::getCheckKeyTime()
785 * @see WANObjectCache::getWithSetCallback()
786 *
787 * @param array $keys
788 * @return float[] Map of (key => UNIX timestamp)
789 * @since 1.31
790 */
791 final public function getMultiCheckKeyTime( array $keys ) {
792 $rawKeys = [];
793 foreach ( $keys as $key ) {
794 $rawKeys[$key] = self::TIME_KEY_PREFIX . $key;
795 }
796
797 $rawValues = $this->cache->getMulti( $rawKeys );
798 $rawValues += array_fill_keys( $rawKeys, false );
799
800 $times = [];
801 foreach ( $rawKeys as $key => $rawKey ) {
802 $purge = $this->parsePurgeValue( $rawValues[$rawKey] );
803 if ( $purge !== false ) {
804 $time = $purge[self::FLD_TIME];
805 } else {
806 // Casting assures identical floats for the next getCheckKeyTime() calls
807 $now = (string)$this->getCurrentTime();
808 $this->cache->add(
809 $rawKey,
810 $this->makePurgeValue( $now, self::HOLDOFF_TTL ),
811 self::CHECK_KEY_TTL
812 );
813 $time = (float)$now;
814 }
815
816 $times[$key] = $time;
817 }
818
819 return $times;
820 }
821
822 /**
823 * Purge a "check" key from all datacenters, invalidating keys that use it
824 *
825 * This should only be called when the underlying data (being cached)
826 * changes in a significant way, and it is impractical to call delete()
827 * on all keys that should be changed. When get() is called on those
828 * keys, the relevant "check" keys must be supplied for this to work.
829 *
830 * The "check" key essentially represents a last-modified time of an entity.
831 * When the key is touched, the timestamp will be updated to the current time.
832 * Keys using the "check" key via get(), getMulti(), or getWithSetCallback() will
833 * be invalidated. This approach is useful if many keys depend on a single entity.
834 *
835 * The timestamp of the "check" key is treated as being HOLDOFF_TTL seconds in the
836 * future by get*() methods in order to avoid race conditions where keys are updated
837 * with stale values (e.g. from a lagged replica DB). A high TTL is set on the "check"
838 * key, making it possible to know the timestamp of the last change to the corresponding
839 * entities in most cases. This might use more cache space than resetCheckKey().
840 *
841 * When a few important keys get a large number of hits, a high cache time is usually
842 * desired as well as "lockTSE" logic. The resetCheckKey() method is less appropriate
843 * in such cases since the "time since expiry" cannot be inferred, causing any get()
844 * after the reset to treat the key as being "hot", resulting in more stale value usage.
845 *
846 * Note that "check" keys won't collide with other regular keys.
847 *
848 * @see WANObjectCache::get()
849 * @see WANObjectCache::getWithSetCallback()
850 * @see WANObjectCache::resetCheckKey()
851 *
852 * @param string $key Cache key
853 * @param int $holdoff HOLDOFF_TTL or HOLDOFF_NONE constant
854 * @return bool True if the item was purged or not found, false on failure
855 */
856 final public function touchCheckKey( $key, $holdoff = self::HOLDOFF_TTL ) {
857 // Publish the purge to all datacenters
858 $ok = $this->relayPurge( self::TIME_KEY_PREFIX . $key, self::CHECK_KEY_TTL, $holdoff );
859
860 $kClass = $this->determineKeyClassForStats( $key );
861 $this->stats->increment( "wanobjectcache.$kClass.ck_touch." . ( $ok ? 'ok' : 'error' ) );
862
863 return $ok;
864 }
865
866 /**
867 * Delete a "check" key from all datacenters, invalidating keys that use it
868 *
869 * This is similar to touchCheckKey() in that keys using it via get(), getMulti(),
870 * or getWithSetCallback() will be invalidated. The differences are:
871 * - a) The "check" key will be deleted from all caches and lazily
872 * re-initialized when accessed (rather than set everywhere)
873 * - b) Thus, dependent keys will be known to be stale, but not
874 * for how long (they are treated as "just" purged), which
875 * effects any lockTSE logic in getWithSetCallback()
876 * - c) Since "check" keys are initialized only on the server the key hashes
877 * to, any temporary ejection of that server will cause the value to be
878 * seen as purged as a new server will initialize the "check" key.
879 *
880 * The advantage here is that the "check" keys, which have high TTLs, will only
881 * be created when a get*() method actually uses that key. This is better when
882 * a large number of "check" keys are invalided in a short period of time.
883 *
884 * Note that "check" keys won't collide with other regular keys.
885 *
886 * @see WANObjectCache::get()
887 * @see WANObjectCache::getWithSetCallback()
888 * @see WANObjectCache::touchCheckKey()
889 *
890 * @param string $key Cache key
891 * @return bool True if the item was purged or not found, false on failure
892 */
893 final public function resetCheckKey( $key ) {
894 // Publish the purge to all datacenters
895 $ok = $this->relayDelete( self::TIME_KEY_PREFIX . $key );
896
897 $kClass = $this->determineKeyClassForStats( $key );
898 $this->stats->increment( "wanobjectcache.$kClass.ck_reset." . ( $ok ? 'ok' : 'error' ) );
899
900 return $ok;
901 }
902
903 /**
904 * Method to fetch/regenerate cache keys
905 *
906 * On cache miss, the key will be set to the callback result via set()
907 * (unless the callback returns false) and that result will be returned.
908 * The arguments supplied to the callback are:
909 * - $oldValue : current cache value or false if not present
910 * - &$ttl : a reference to the TTL which can be altered
911 * - &$setOpts : a reference to options for set() which can be altered
912 * - $oldAsOf : generation UNIX timestamp of $oldValue or null if not present (since 1.28)
913 *
914 * It is strongly recommended to set the 'lag' and 'since' fields to avoid race conditions
915 * that can cause stale values to get stuck at keys. Usually, callbacks ignore the current
916 * value, but it can be used to maintain "most recent X" values that come from time or
917 * sequence based source data, provided that the "as of" id/time is tracked. Note that
918 * preemptive regeneration and $checkKeys can result in a non-false current value.
919 *
920 * Usage of $checkKeys is similar to get() and getMulti(). However, rather than the caller
921 * having to inspect a "current time left" variable (e.g. $curTTL, $curTTLs), a cache
922 * regeneration will automatically be triggered using the callback.
923 *
924 * The $ttl argument and "hotTTR" option (in $opts) use time-dependant randomization
925 * to avoid stampedes. Keys that are slow to regenerate and either heavily used
926 * or subject to explicit (unpredictable) purges, may need additional mechanisms.
927 * The simplest way to avoid stampedes for such keys is to use 'lockTSE' (in $opts).
928 * If explicit purges are needed, also:
929 * - a) Pass $key into $checkKeys
930 * - b) Use touchCheckKey( $key ) instead of delete( $key )
931 *
932 * Example usage (typical key):
933 * @code
934 * $catInfo = $cache->getWithSetCallback(
935 * // Key to store the cached value under
936 * $cache->makeKey( 'cat-attributes', $catId ),
937 * // Time-to-live (in seconds)
938 * $cache::TTL_MINUTE,
939 * // Function that derives the new key value
940 * function ( $oldValue, &$ttl, array &$setOpts ) {
941 * $dbr = wfGetDB( DB_REPLICA );
942 * // Account for any snapshot/replica DB lag
943 * $setOpts += Database::getCacheSetOptions( $dbr );
944 *
945 * return $dbr->selectRow( ... );
946 * }
947 * );
948 * @endcode
949 *
950 * Example usage (key that is expensive and hot):
951 * @code
952 * $catConfig = $cache->getWithSetCallback(
953 * // Key to store the cached value under
954 * $cache->makeKey( 'site-cat-config' ),
955 * // Time-to-live (in seconds)
956 * $cache::TTL_DAY,
957 * // Function that derives the new key value
958 * function ( $oldValue, &$ttl, array &$setOpts ) {
959 * $dbr = wfGetDB( DB_REPLICA );
960 * // Account for any snapshot/replica DB lag
961 * $setOpts += Database::getCacheSetOptions( $dbr );
962 *
963 * return CatConfig::newFromRow( $dbr->selectRow( ... ) );
964 * },
965 * [
966 * // Calling touchCheckKey() on this key invalidates the cache
967 * 'checkKeys' => [ $cache->makeKey( 'site-cat-config' ) ],
968 * // Try to only let one datacenter thread manage cache updates at a time
969 * 'lockTSE' => 30,
970 * // Avoid querying cache servers multiple times in a web request
971 * 'pcTTL' => $cache::TTL_PROC_LONG
972 * ]
973 * );
974 * @endcode
975 *
976 * Example usage (key with dynamic dependencies):
977 * @code
978 * $catState = $cache->getWithSetCallback(
979 * // Key to store the cached value under
980 * $cache->makeKey( 'cat-state', $cat->getId() ),
981 * // Time-to-live (seconds)
982 * $cache::TTL_HOUR,
983 * // Function that derives the new key value
984 * function ( $oldValue, &$ttl, array &$setOpts ) {
985 * // Determine new value from the DB
986 * $dbr = wfGetDB( DB_REPLICA );
987 * // Account for any snapshot/replica DB lag
988 * $setOpts += Database::getCacheSetOptions( $dbr );
989 *
990 * return CatState::newFromResults( $dbr->select( ... ) );
991 * },
992 * [
993 * // The "check" keys that represent things the value depends on;
994 * // Calling touchCheckKey() on any of them invalidates the cache
995 * 'checkKeys' => [
996 * $cache->makeKey( 'sustenance-bowls', $cat->getRoomId() ),
997 * $cache->makeKey( 'people-present', $cat->getHouseId() ),
998 * $cache->makeKey( 'cat-laws', $cat->getCityId() ),
999 * ]
1000 * ]
1001 * );
1002 * @endcode
1003 *
1004 * Example usage (key that is expensive with too many DB dependencies for "check keys"):
1005 * @code
1006 * $catToys = $cache->getWithSetCallback(
1007 * // Key to store the cached value under
1008 * $cache->makeKey( 'cat-toys', $catId ),
1009 * // Time-to-live (seconds)
1010 * $cache::TTL_HOUR,
1011 * // Function that derives the new key value
1012 * function ( $oldValue, &$ttl, array &$setOpts ) {
1013 * // Determine new value from the DB
1014 * $dbr = wfGetDB( DB_REPLICA );
1015 * // Account for any snapshot/replica DB lag
1016 * $setOpts += Database::getCacheSetOptions( $dbr );
1017 *
1018 * return CatToys::newFromResults( $dbr->select( ... ) );
1019 * },
1020 * [
1021 * // Get the highest timestamp of any of the cat's toys
1022 * 'touchedCallback' => function ( $value ) use ( $catId ) {
1023 * $dbr = wfGetDB( DB_REPLICA );
1024 * $ts = $dbr->selectField( 'cat_toys', 'MAX(ct_touched)', ... );
1025 *
1026 * return wfTimestampOrNull( TS_UNIX, $ts );
1027 * },
1028 * // Avoid DB queries for repeated access
1029 * 'pcTTL' => $cache::TTL_PROC_SHORT
1030 * ]
1031 * );
1032 * @endcode
1033 *
1034 * Example usage (hot key holding most recent 100 events):
1035 * @code
1036 * $lastCatActions = $cache->getWithSetCallback(
1037 * // Key to store the cached value under
1038 * $cache->makeKey( 'cat-last-actions', 100 ),
1039 * // Time-to-live (in seconds)
1040 * 10,
1041 * // Function that derives the new key value
1042 * function ( $oldValue, &$ttl, array &$setOpts ) {
1043 * $dbr = wfGetDB( DB_REPLICA );
1044 * // Account for any snapshot/replica DB lag
1045 * $setOpts += Database::getCacheSetOptions( $dbr );
1046 *
1047 * // Start off with the last cached list
1048 * $list = $oldValue ?: [];
1049 * // Fetch the last 100 relevant rows in descending order;
1050 * // only fetch rows newer than $list[0] to reduce scanning
1051 * $rows = iterator_to_array( $dbr->select( ... ) );
1052 * // Merge them and get the new "last 100" rows
1053 * return array_slice( array_merge( $new, $list ), 0, 100 );
1054 * },
1055 * [
1056 * // Try to only let one datacenter thread manage cache updates at a time
1057 * 'lockTSE' => 30,
1058 * // Use a magic value when no cache value is ready rather than stampeding
1059 * 'busyValue' => 'computing'
1060 * ]
1061 * );
1062 * @endcode
1063 *
1064 * Example usage (key holding an LRU subkey:value map; this can avoid flooding cache with
1065 * keys for an unlimited set of (constraint,situation) pairs, thereby avoiding elevated
1066 * cache evictions and wasted memory):
1067 * @code
1068 * $catSituationTolerabilityCache = $this->cache->getWithSetCallback(
1069 * // Group by constraint ID/hash, cat family ID/hash, or something else useful
1070 * $this->cache->makeKey( 'cat-situation-tolerability-checks', $groupKey ),
1071 * WANObjectCache::TTL_DAY, // rarely used groups should fade away
1072 * // The $scenarioKey format is $constraintId:<ID/hash of $situation>
1073 * function ( $cacheMap ) use ( $scenarioKey, $constraintId, $situation ) {
1074 * $lruCache = MapCacheLRU::newFromArray( $cacheMap ?: [], self::CACHE_SIZE );
1075 * $result = $lruCache->get( $scenarioKey ); // triggers LRU bump if present
1076 * if ( $result === null || $this->isScenarioResultExpired( $result ) ) {
1077 * $result = $this->checkScenarioTolerability( $constraintId, $situation );
1078 * $lruCache->set( $scenarioKey, $result, 3 / 8 );
1079 * }
1080 * // Save the new LRU cache map and reset the map's TTL
1081 * return $lruCache->toArray();
1082 * },
1083 * [
1084 * // Once map is > 1 sec old, consider refreshing
1085 * 'ageNew' => 1,
1086 * // Update within 5 seconds after "ageNew" given a 1hz cache check rate
1087 * 'hotTTR' => 5,
1088 * // Avoid querying cache servers multiple times in a request; this also means
1089 * // that a request can only alter the value of any given constraint key once
1090 * 'pcTTL' => WANObjectCache::TTL_PROC_LONG
1091 * ]
1092 * );
1093 * $tolerability = isset( $catSituationTolerabilityCache[$scenarioKey] )
1094 * ? $catSituationTolerabilityCache[$scenarioKey]
1095 * : $this->checkScenarioTolerability( $constraintId, $situation );
1096 * @endcode
1097 *
1098 * @see WANObjectCache::get()
1099 * @see WANObjectCache::set()
1100 *
1101 * @param string $key Cache key made from makeKey() or makeGlobalKey()
1102 * @param int $ttl Seconds to live for key updates. Special values are:
1103 * - WANObjectCache::TTL_INDEFINITE: Cache forever (subject to LRU-style evictions)
1104 * - WANObjectCache::TTL_UNCACHEABLE: Do not cache (if the key exists, it is not deleted)
1105 * @param callable $callback Value generation function
1106 * @param array $opts Options map:
1107 * - checkKeys: List of "check" keys. The key at $key will be seen as stale when either
1108 * touchCheckKey() or resetCheckKey() is called on any of the keys in this list. This
1109 * is useful if thousands or millions of keys depend on the same entity. The entity can
1110 * simply have its "check" key updated whenever the entity is modified.
1111 * Default: [].
1112 * - graceTTL: If the key is invalidated (by "checkKeys"/"touchedCallback") less than this
1113 * many seconds ago, consider reusing the stale value. The odds of a refresh becomes
1114 * more likely over time, becoming certain once the grace period is reached. This can
1115 * reduce traffic spikes when millions of keys are compared to the same "check" key and
1116 * touchCheckKey() or resetCheckKey() is called on that "check" key. This option is not
1117 * useful for avoiding traffic spikes in the case of the key simply expiring on account
1118 * of its TTL (use "lowTTL" instead).
1119 * Default: WANObjectCache::GRACE_TTL_NONE.
1120 * - lockTSE: If the key is tombstoned or invalidated (by "checkKeys"/"touchedCallback")
1121 * less than this many seconds ago, try to have a single thread handle cache regeneration
1122 * at any given time. Other threads will use stale values if possible. If, on miss,
1123 * the time since expiration is low, the assumption is that the key is hot and that a
1124 * stampede is worth avoiding. Note that if the key falls out of cache then concurrent
1125 * threads will all run the callback on cache miss until the value is saved in cache.
1126 * The only stampede protection in that case is from duplicate cache sets when the
1127 * callback takes longer than WANObjectCache::SET_DELAY_HIGH_SEC seconds; consider
1128 * using "busyValue" if such stampedes are a problem. Note that the higher "lockTSE" is
1129 * set, the higher the worst-case staleness of returned values can be. Also note that
1130 * this option does not by itself handle the case of the key simply expiring on account
1131 * of its TTL, so make sure that "lowTTL" is not disabled when using this option. Avoid
1132 * combining this option with delete() as it can always cause a stampede due to their
1133 * being no stale value available until after a thread completes the callback.
1134 * Use WANObjectCache::TSE_NONE to disable this logic.
1135 * Default: WANObjectCache::TSE_NONE.
1136 * - busyValue: If no value exists and another thread is currently regenerating it, use this
1137 * as a fallback value (or a callback to generate such a value). This assures that cache
1138 * stampedes cannot happen if the value falls out of cache. This can be used as insurance
1139 * against cache regeneration becoming very slow for some reason (greater than the TTL).
1140 * Default: null.
1141 * - pcTTL: Process cache the value in this PHP instance for this many seconds. This avoids
1142 * network I/O when a key is read several times. This will not cache when the callback
1143 * returns false, however. Note that any purges will not be seen while process cached;
1144 * since the callback should use replica DBs and they may be lagged or have snapshot
1145 * isolation anyway, this should not typically matter.
1146 * Default: WANObjectCache::TTL_UNCACHEABLE.
1147 * - pcGroup: Process cache group to use instead of the primary one. If set, this must be
1148 * of the format ALPHANUMERIC_NAME:MAX_KEY_SIZE, e.g. "mydata:10". Use this for storing
1149 * large values, small yet numerous values, or some values with a high cost of eviction.
1150 * It is generally preferable to use a class constant when setting this value.
1151 * This has no effect unless pcTTL is used.
1152 * Default: WANObjectCache::PC_PRIMARY.
1153 * - version: Integer version number. This allows for callers to make breaking changes to
1154 * how values are stored while maintaining compatability and correct cache purges. New
1155 * versions are stored alongside older versions concurrently. Avoid storing class objects
1156 * however, as this reduces compatibility (due to serialization).
1157 * Default: null.
1158 * - minAsOf: Reject values if they were generated before this UNIX timestamp.
1159 * This is useful if the source of a key is suspected of having possibly changed
1160 * recently, and the caller wants any such changes to be reflected.
1161 * Default: WANObjectCache::MIN_TIMESTAMP_NONE.
1162 * - hotTTR: Expected time-till-refresh (TTR) in seconds for keys that average ~1 hit per
1163 * second (e.g. 1Hz). Keys with a hit rate higher than 1Hz will refresh sooner than this
1164 * TTR and vise versa. Such refreshes won't happen until keys are "ageNew" seconds old.
1165 * This uses randomization to avoid triggering cache stampedes. The TTR is useful at
1166 * reducing the impact of missed cache purges, since the effect of a heavily referenced
1167 * key being stale is worse than that of a rarely referenced key. Unlike simply lowering
1168 * $ttl, seldomly used keys are largely unaffected by this option, which makes it
1169 * possible to have a high hit rate for the "long-tail" of less-used keys.
1170 * Default: WANObjectCache::HOT_TTR.
1171 * - lowTTL: Consider pre-emptive updates when the current TTL (seconds) of the key is less
1172 * than this. It becomes more likely over time, becoming certain once the key is expired.
1173 * This helps avoid cache stampedes that might be triggered due to the key expiring.
1174 * Default: WANObjectCache::LOW_TTL.
1175 * - ageNew: Consider popularity refreshes only once a key reaches this age in seconds.
1176 * Default: WANObjectCache::AGE_NEW.
1177 * - staleTTL: Seconds to keep the key around if it is stale. This means that on cache
1178 * miss the callback may get $oldValue/$oldAsOf values for keys that have already been
1179 * expired for this specified time. This is useful if adaptiveTTL() is used on the old
1180 * value's as-of time when it is verified as still being correct.
1181 * Default: WANObjectCache::STALE_TTL_NONE
1182 * - touchedCallback: A callback that takes the current value and returns a UNIX timestamp
1183 * indicating the last time a dynamic dependency changed. Null can be returned if there
1184 * are no relevant dependency changes to check. This can be used to check against things
1185 * like last-modified times of files or DB timestamp fields. This should generally not be
1186 * used for small and easily queried values in a DB if the callback itself ends up doing
1187 * a similarly expensive DB query to check a timestamp. Usages of this option makes the
1188 * most sense for values that are moderately to highly expensive to regenerate and easy
1189 * to query for dependency timestamps. The use of "pcTTL" reduces timestamp queries.
1190 * Default: null.
1191 * @return mixed Value found or written to the key
1192 * @note Options added in 1.28: version, busyValue, hotTTR, ageNew, pcGroup, minAsOf
1193 * @note Options added in 1.31: staleTTL, graceTTL
1194 * @note Options added in 1.33: touchedCallback
1195 * @note Callable type hints are not used to avoid class-autoloading
1196 */
1197 final public function getWithSetCallback( $key, $ttl, $callback, array $opts = [] ) {
1198 $pcTTL = $opts['pcTTL'] ?? self::TTL_UNCACHEABLE;
1199
1200 // Try the process cache if enabled and the cache callback is not within a cache callback.
1201 // Process cache use in nested callbacks is not lag-safe with regard to HOLDOFF_TTL since
1202 // the in-memory value is further lagged than the shared one since it uses a blind TTL.
1203 if ( $pcTTL >= 0 && $this->callbackDepth == 0 ) {
1204 $group = $opts['pcGroup'] ?? self::PC_PRIMARY;
1205 $procCache = $this->getProcessCache( $group );
1206 $value = $procCache->has( $key, $pcTTL ) ? $procCache->get( $key ) : false;
1207 } else {
1208 $procCache = false;
1209 $value = false;
1210 }
1211
1212 if ( $value === false ) {
1213 // Fetch the value over the network
1214 if ( isset( $opts['version'] ) ) {
1215 $version = $opts['version'];
1216 $asOf = null;
1217 $cur = $this->doGetWithSetCallback(
1218 $key,
1219 $ttl,
1220 function ( $oldValue, &$ttl, &$setOpts, $oldAsOf )
1221 use ( $callback, $version ) {
1222 if ( is_array( $oldValue )
1223 && array_key_exists( self::VFLD_DATA, $oldValue )
1224 && array_key_exists( self::VFLD_VERSION, $oldValue )
1225 && $oldValue[self::VFLD_VERSION] === $version
1226 ) {
1227 $oldData = $oldValue[self::VFLD_DATA];
1228 } else {
1229 // VFLD_DATA is not set if an old, unversioned, key is present
1230 $oldData = false;
1231 $oldAsOf = null;
1232 }
1233
1234 return [
1235 self::VFLD_DATA => $callback( $oldData, $ttl, $setOpts, $oldAsOf ),
1236 self::VFLD_VERSION => $version
1237 ];
1238 },
1239 $opts,
1240 $asOf
1241 );
1242 if ( $cur[self::VFLD_VERSION] === $version ) {
1243 // Value created or existed before with version; use it
1244 $value = $cur[self::VFLD_DATA];
1245 } else {
1246 // Value existed before with a different version; use variant key.
1247 // Reflect purges to $key by requiring that this key value be newer.
1248 $value = $this->doGetWithSetCallback(
1249 $this->makeGlobalKey( 'WANCache-key-variant', md5( $key ), $version ),
1250 $ttl,
1251 $callback,
1252 // Regenerate value if not newer than $key
1253 [ 'version' => null, 'minAsOf' => $asOf ] + $opts
1254 );
1255 }
1256 } else {
1257 $value = $this->doGetWithSetCallback( $key, $ttl, $callback, $opts );
1258 }
1259
1260 // Update the process cache if enabled
1261 if ( $procCache && $value !== false ) {
1262 $procCache->set( $key, $value );
1263 }
1264 }
1265
1266 return $value;
1267 }
1268
1269 /**
1270 * Do the actual I/O for getWithSetCallback() when needed
1271 *
1272 * @see WANObjectCache::getWithSetCallback()
1273 *
1274 * @param string $key
1275 * @param int $ttl
1276 * @param callable $callback
1277 * @param array $opts Options map for getWithSetCallback()
1278 * @param float|null &$asOf Cache generation timestamp of returned value [returned]
1279 * @return mixed
1280 * @note Callable type hints are not used to avoid class-autoloading
1281 */
1282 protected function doGetWithSetCallback( $key, $ttl, $callback, array $opts, &$asOf = null ) {
1283 $lowTTL = $opts['lowTTL'] ?? min( self::LOW_TTL, $ttl );
1284 $lockTSE = $opts['lockTSE'] ?? self::TSE_NONE;
1285 $staleTTL = $opts['staleTTL'] ?? self::STALE_TTL_NONE;
1286 $graceTTL = $opts['graceTTL'] ?? self::GRACE_TTL_NONE;
1287 $checkKeys = $opts['checkKeys'] ?? [];
1288 $busyValue = $opts['busyValue'] ?? null;
1289 $popWindow = $opts['hotTTR'] ?? self::HOT_TTR;
1290 $ageNew = $opts['ageNew'] ?? self::AGE_NEW;
1291 $minTime = $opts['minAsOf'] ?? self::MIN_TIMESTAMP_NONE;
1292 $needsVersion = isset( $opts['version'] );
1293 $touchedCb = $opts['touchedCallback'] ?? null;
1294 $initialTime = $this->getCurrentTime();
1295
1296 $kClass = $this->determineKeyClassForStats( $key );
1297
1298 // Get the current key value
1299 $curTTL = self::PASS_BY_REF;
1300 $curInfo = self::PASS_BY_REF; /** @var array $curInfo */
1301 $curValue = $this->get( $key, $curTTL, $checkKeys, $curInfo );
1302 // Apply any $touchedCb invalidation timestamp to get the "last purge timestamp"
1303 list( $curTTL, $LPT ) = $this->resolveCTL( $curValue, $curTTL, $curInfo, $touchedCb );
1304 // Keep track of the best candidate value and its timestamp
1305 $value = $curValue; // return value
1306 $asOf = $curInfo['asOf']; // return value timestamp
1307
1308 // Determine if a cached value regeneration is needed or desired
1309 if (
1310 $this->isValid( $value, $needsVersion, $asOf, $minTime ) &&
1311 $this->isAliveOrInGracePeriod( $curTTL, $graceTTL )
1312 ) {
1313 $preemptiveRefresh = (
1314 $this->worthRefreshExpiring( $curTTL, $lowTTL ) ||
1315 $this->worthRefreshPopular( $asOf, $ageNew, $popWindow, $initialTime )
1316 );
1317
1318 if ( !$preemptiveRefresh ) {
1319 $this->stats->increment( "wanobjectcache.$kClass.hit.good" );
1320
1321 return $value;
1322 } elseif ( $this->scheduleAsyncRefresh( $key, $ttl, $callback, $opts ) ) {
1323 $this->stats->increment( "wanobjectcache.$kClass.hit.refresh" );
1324
1325 return $value;
1326 }
1327 }
1328
1329 $isKeyTombstoned = ( $curInfo['tombAsOf'] !== null );
1330 if ( $isKeyTombstoned ) {
1331 // Get the interim key value since the key is tombstoned (write-holed)
1332 list( $value, $asOf ) = $this->getInterimValue( $key, $needsVersion, $minTime );
1333 // Update the "last purge time" since the $touchedCb timestamp depends on $value
1334 $LPT = $this->resolveTouched( $value, $LPT, $touchedCb );
1335 }
1336
1337 // Reduce mutex and cache set spam while keys are in the tombstone/holdoff period by
1338 // checking if $value was genereated by a recent thread much less than a second ago.
1339 if (
1340 $this->isValid( $value, $needsVersion, $asOf, $minTime, $LPT ) &&
1341 $this->isVolatileValueAgeNegligible( $initialTime - $asOf )
1342 ) {
1343 $this->stats->increment( "wanobjectcache.$kClass.hit.volatile" );
1344
1345 return $value;
1346 }
1347
1348 // Decide if only one thread should handle regeneration at a time
1349 $useMutex =
1350 // Note that since tombstones no-op set(), $lockTSE and $curTTL cannot be used to
1351 // deduce the key hotness because |$curTTL| will always keep increasing until the
1352 // tombstone expires or is overwritten by a new tombstone. Also, even if $lockTSE
1353 // is not set, constant regeneration of a key for the tombstone lifetime might be
1354 // very expensive. Assume tombstoned keys are possibly hot in order to reduce
1355 // the risk of high regeneration load after the delete() method is called.
1356 $isKeyTombstoned ||
1357 // Assume a key is hot if requested soon ($lockTSE seconds) after invalidation.
1358 // This avoids stampedes when timestamps from $checkKeys/$touchedCb bump.
1359 ( $curTTL !== null && $curTTL <= 0 && abs( $curTTL ) <= $lockTSE ) ||
1360 // Assume a key is hot if there is no value and a busy fallback is given.
1361 // This avoids stampedes on eviction or preemptive regeneration taking too long.
1362 ( $busyValue !== null && $value === false );
1363
1364 $hasLock = false;
1365 if ( $useMutex ) {
1366 // Acquire a datacenter-local non-blocking lock
1367 if ( $this->cache->add( self::MUTEX_KEY_PREFIX . $key, 1, self::LOCK_TTL ) ) {
1368 // Lock acquired; this thread will recompute the value and update cache
1369 $hasLock = true;
1370 } elseif ( $this->isValid( $value, $needsVersion, $asOf, $minTime ) ) {
1371 // Lock not acquired and a stale value exists; use the stale value
1372 $this->stats->increment( "wanobjectcache.$kClass.hit.stale" );
1373
1374 return $value;
1375 } else {
1376 // Lock not acquired and no stale value exists
1377 if ( $busyValue !== null ) {
1378 // Use the busy fallback value if nothing else
1379 $miss = is_infinite( $minTime ) ? 'renew' : 'miss';
1380 $this->stats->increment( "wanobjectcache.$kClass.$miss.busy" );
1381
1382 return is_callable( $busyValue ) ? $busyValue() : $busyValue;
1383 }
1384 }
1385 }
1386
1387 if ( !is_callable( $callback ) ) {
1388 throw new InvalidArgumentException( "Invalid cache miss callback provided." );
1389 }
1390
1391 $preCallbackTime = $this->getCurrentTime();
1392 // Generate the new value from the callback...
1393 $setOpts = [];
1394 ++$this->callbackDepth;
1395 try {
1396 $value = call_user_func_array( $callback, [ $curValue, &$ttl, &$setOpts, $asOf ] );
1397 } finally {
1398 --$this->callbackDepth;
1399 }
1400 $valueIsCacheable = ( $value !== false && $ttl >= 0 );
1401
1402 if ( $valueIsCacheable ) {
1403 $ago = max( $this->getCurrentTime() - $initialTime, 0.0 );
1404 $this->stats->timing( "wanobjectcache.$kClass.regen_set_delay", 1000 * $ago );
1405
1406 if ( $isKeyTombstoned ) {
1407 if ( $this->checkAndSetCooloff( $key, $kClass, $ago, $lockTSE, $hasLock ) ) {
1408 // Use the interim key value since the key is tombstoned (write-holed)
1409 $tempTTL = max( self::INTERIM_KEY_TTL, (int)$lockTSE );
1410 $this->setInterimValue( $key, $value, $tempTTL, $this->getCurrentTime() );
1411 }
1412 } elseif ( !$useMutex || $hasLock ) {
1413 if ( $this->checkAndSetCooloff( $key, $kClass, $ago, $lockTSE, $hasLock ) ) {
1414 // Save the value unless a lock-winning thread is already expected to do that
1415 $setOpts['lockTSE'] = $lockTSE;
1416 $setOpts['staleTTL'] = $staleTTL;
1417 // Use best known "since" timestamp if not provided
1418 $setOpts += [ 'since' => $preCallbackTime ];
1419 // Update the cache; this will fail if the key is tombstoned
1420 $this->set( $key, $value, $ttl, $setOpts );
1421 }
1422 }
1423 }
1424
1425 if ( $hasLock ) {
1426 $this->cache->changeTTL( self::MUTEX_KEY_PREFIX . $key, (int)$initialTime - 60 );
1427 }
1428
1429 $miss = is_infinite( $minTime ) ? 'renew' : 'miss';
1430 $this->stats->increment( "wanobjectcache.$kClass.$miss.compute" );
1431
1432 return $value;
1433 }
1434
1435 /**
1436 * @param float $age Age of volatile/interim key in seconds
1437 * @return bool Whether the age of a volatile value is negligible
1438 */
1439 private function isVolatileValueAgeNegligible( $age ) {
1440 return ( $age < mt_rand( self::RECENT_SET_LOW_MS, self::RECENT_SET_HIGH_MS ) / 1e3 );
1441 }
1442
1443 /**
1444 * @param string $key
1445 * @param string $kClass
1446 * @param float $elapsed Seconds spent regenerating the value
1447 * @param float $lockTSE
1448 * @param $hasLock bool
1449 * @return bool Whether it is OK to proceed with a key set operation
1450 */
1451 private function checkAndSetCooloff( $key, $kClass, $elapsed, $lockTSE, $hasLock ) {
1452 // If $lockTSE is set, the lock was bypassed because there was no stale/interim value,
1453 // and $elapsed indicates that regeration is slow, then there is a risk of set()
1454 // stampedes with large blobs. With a typical scale-out infrastructure, CPU and query
1455 // load from $callback invocations is distributed among appservers and replica DBs,
1456 // but cache operations for a given key route to a single cache server (e.g. striped
1457 // consistent hashing).
1458 if ( $lockTSE < 0 || $hasLock ) {
1459 return true; // either not a priori hot or thread has the lock
1460 } elseif ( $elapsed <= self::SET_DELAY_HIGH_SEC ) {
1461 return true; // not enough time for threads to pile up
1462 }
1463
1464 $this->cache->clearLastError();
1465 if (
1466 !$this->cache->add( self::COOLOFF_KEY_PREFIX . $key, 1, self::COOLOFF_TTL ) &&
1467 // Don't treat failures due to I/O errors as the key being in cooloff
1468 $this->cache->getLastError() === BagOStuff::ERR_NONE
1469 ) {
1470 $this->stats->increment( "wanobjectcache.$kClass.cooloff_bounce" );
1471
1472 return false;
1473 }
1474
1475 return true;
1476 }
1477
1478 /**
1479 * @param mixed $value
1480 * @param float|null $curTTL
1481 * @param array $curInfo
1482 * @param callable|null $touchedCallback
1483 * @return array (current time left or null, UNIX timestamp of last purge or null)
1484 * @note Callable type hints are not used to avoid class-autoloading
1485 */
1486 protected function resolveCTL( $value, $curTTL, $curInfo, $touchedCallback ) {
1487 if ( $touchedCallback === null || $value === false ) {
1488 return [ $curTTL, max( $curInfo['tombAsOf'], $curInfo['lastCKPurge'] ) ];
1489 }
1490
1491 if ( !is_callable( $touchedCallback ) ) {
1492 throw new InvalidArgumentException( "Invalid expiration callback provided." );
1493 }
1494
1495 $touched = $touchedCallback( $value );
1496 if ( $touched !== null && $touched >= $curInfo['asOf'] ) {
1497 $curTTL = min( $curTTL, self::TINY_NEGATIVE, $curInfo['asOf'] - $touched );
1498 }
1499
1500 return [ $curTTL, max( $curInfo['tombAsOf'], $curInfo['lastCKPurge'], $touched ) ];
1501 }
1502
1503 /**
1504 * @param mixed $value
1505 * @param float|null $lastPurge
1506 * @param callable|null $touchedCallback
1507 * @return float|null UNIX timestamp of last purge or null
1508 * @note Callable type hints are not used to avoid class-autoloading
1509 */
1510 protected function resolveTouched( $value, $lastPurge, $touchedCallback ) {
1511 if ( $touchedCallback === null || $value === false ) {
1512 return $lastPurge;
1513 }
1514
1515 if ( !is_callable( $touchedCallback ) ) {
1516 throw new InvalidArgumentException( "Invalid expiration callback provided." );
1517 }
1518
1519 return max( $touchedCallback( $value ), $lastPurge );
1520 }
1521
1522 /**
1523 * @param string $key
1524 * @param bool $versioned
1525 * @param float $minTime
1526 * @return array (cached value or false, cached value timestamp or null)
1527 */
1528 protected function getInterimValue( $key, $versioned, $minTime ) {
1529 if ( !$this->useInterimHoldOffCaching ) {
1530 return [ false, null ]; // disabled
1531 }
1532
1533 $wrapped = $this->cache->get( self::INTERIM_KEY_PREFIX . $key );
1534 list( $value ) = $this->unwrap( $wrapped, $this->getCurrentTime() );
1535 $valueAsOf = $wrapped[self::FLD_TIME] ?? null;
1536 if ( $this->isValid( $value, $versioned, $valueAsOf, $minTime ) ) {
1537 return [ $value, $valueAsOf ];
1538 }
1539
1540 return [ false, null ];
1541 }
1542
1543 /**
1544 * @param string $key
1545 * @param mixed $value
1546 * @param int $tempTTL
1547 * @param float $newAsOf
1548 */
1549 protected function setInterimValue( $key, $value, $tempTTL, $newAsOf ) {
1550 $wrapped = $this->wrap( $value, $tempTTL, $newAsOf );
1551
1552 $this->cache->merge(
1553 self::INTERIM_KEY_PREFIX . $key,
1554 function () use ( $wrapped ) {
1555 return $wrapped;
1556 },
1557 $tempTTL,
1558 1
1559 );
1560 }
1561
1562 /**
1563 * Method to fetch multiple cache keys at once with regeneration
1564 *
1565 * This works the same as getWithSetCallback() except:
1566 * - a) The $keys argument expects the result of WANObjectCache::makeMultiKeys()
1567 * - b) The $callback argument expects a callback taking the following arguments:
1568 * - $id: ID of an entity to query
1569 * - $oldValue : the prior cache value or false if none was present
1570 * - &$ttl : a reference to the new value TTL in seconds
1571 * - &$setOpts : a reference to options for set() which can be altered
1572 * - $oldAsOf : generation UNIX timestamp of $oldValue or null if not present
1573 * Aside from the additional $id argument, the other arguments function the same
1574 * way they do in getWithSetCallback().
1575 * - c) The return value is a map of (cache key => value) in the order of $keyedIds
1576 *
1577 * @see WANObjectCache::getWithSetCallback()
1578 * @see WANObjectCache::getMultiWithUnionSetCallback()
1579 *
1580 * Example usage:
1581 * @code
1582 * $rows = $cache->getMultiWithSetCallback(
1583 * // Map of cache keys to entity IDs
1584 * $cache->makeMultiKeys(
1585 * $this->fileVersionIds(),
1586 * function ( $id, WANObjectCache $cache ) {
1587 * return $cache->makeKey( 'file-version', $id );
1588 * }
1589 * ),
1590 * // Time-to-live (in seconds)
1591 * $cache::TTL_DAY,
1592 * // Function that derives the new key value
1593 * function ( $id, $oldValue, &$ttl, array &$setOpts ) {
1594 * $dbr = wfGetDB( DB_REPLICA );
1595 * // Account for any snapshot/replica DB lag
1596 * $setOpts += Database::getCacheSetOptions( $dbr );
1597 *
1598 * // Load the row for this file
1599 * $queryInfo = File::getQueryInfo();
1600 * $row = $dbr->selectRow(
1601 * $queryInfo['tables'],
1602 * $queryInfo['fields'],
1603 * [ 'id' => $id ],
1604 * __METHOD__,
1605 * [],
1606 * $queryInfo['joins']
1607 * );
1608 *
1609 * return $row ? (array)$row : false;
1610 * },
1611 * [
1612 * // Process cache for 30 seconds
1613 * 'pcTTL' => 30,
1614 * // Use a dedicated 500 item cache (initialized on-the-fly)
1615 * 'pcGroup' => 'file-versions:500'
1616 * ]
1617 * );
1618 * $files = array_map( [ __CLASS__, 'newFromRow' ], $rows );
1619 * @endcode
1620 *
1621 * @param ArrayIterator $keyedIds Result of WANObjectCache::makeMultiKeys()
1622 * @param int $ttl Seconds to live for key updates
1623 * @param callable $callback Callback the yields entity regeneration callbacks
1624 * @param array $opts Options map
1625 * @return array Map of (cache key => value) in the same order as $keyedIds
1626 * @since 1.28
1627 */
1628 final public function getMultiWithSetCallback(
1629 ArrayIterator $keyedIds, $ttl, callable $callback, array $opts = []
1630 ) {
1631 $valueKeys = array_keys( $keyedIds->getArrayCopy() );
1632 $checkKeys = $opts['checkKeys'] ?? [];
1633 $pcTTL = $opts['pcTTL'] ?? self::TTL_UNCACHEABLE;
1634
1635 // Load required keys into process cache in one go
1636 $this->warmupCache = $this->getRawKeysForWarmup(
1637 $this->getNonProcessCachedKeys( $valueKeys, $opts, $pcTTL ),
1638 $checkKeys
1639 );
1640 $this->warmupKeyMisses = 0;
1641
1642 // Wrap $callback to match the getWithSetCallback() format while passing $id to $callback
1643 $id = null; // current entity ID
1644 $func = function ( $oldValue, &$ttl, &$setOpts, $oldAsOf ) use ( $callback, &$id ) {
1645 return $callback( $id, $oldValue, $ttl, $setOpts, $oldAsOf );
1646 };
1647
1648 $values = [];
1649 foreach ( $keyedIds as $key => $id ) { // preserve order
1650 $values[$key] = $this->getWithSetCallback( $key, $ttl, $func, $opts );
1651 }
1652
1653 $this->warmupCache = [];
1654
1655 return $values;
1656 }
1657
1658 /**
1659 * Method to fetch/regenerate multiple cache keys at once
1660 *
1661 * This works the same as getWithSetCallback() except:
1662 * - a) The $keys argument expects the result of WANObjectCache::makeMultiKeys()
1663 * - b) The $callback argument expects a callback returning a map of (ID => new value)
1664 * for all entity IDs in $ids and it takes the following arguments:
1665 * - $ids: a list of entity IDs that require cache regeneration
1666 * - &$ttls: a reference to the (entity ID => new TTL) map
1667 * - &$setOpts: a reference to options for set() which can be altered
1668 * - c) The return value is a map of (cache key => value) in the order of $keyedIds
1669 * - d) The "lockTSE" and "busyValue" options are ignored
1670 *
1671 * @see WANObjectCache::getWithSetCallback()
1672 * @see WANObjectCache::getMultiWithSetCallback()
1673 *
1674 * Example usage:
1675 * @code
1676 * $rows = $cache->getMultiWithUnionSetCallback(
1677 * // Map of cache keys to entity IDs
1678 * $cache->makeMultiKeys(
1679 * $this->fileVersionIds(),
1680 * function ( $id, WANObjectCache $cache ) {
1681 * return $cache->makeKey( 'file-version', $id );
1682 * }
1683 * ),
1684 * // Time-to-live (in seconds)
1685 * $cache::TTL_DAY,
1686 * // Function that derives the new key value
1687 * function ( array $ids, array &$ttls, array &$setOpts ) {
1688 * $dbr = wfGetDB( DB_REPLICA );
1689 * // Account for any snapshot/replica DB lag
1690 * $setOpts += Database::getCacheSetOptions( $dbr );
1691 *
1692 * // Load the rows for these files
1693 * $rows = [];
1694 * $queryInfo = File::getQueryInfo();
1695 * $res = $dbr->select(
1696 * $queryInfo['tables'],
1697 * $queryInfo['fields'],
1698 * [ 'id' => $ids ],
1699 * __METHOD__,
1700 * [],
1701 * $queryInfo['joins']
1702 * );
1703 * foreach ( $res as $row ) {
1704 * $rows[$row->id] = $row;
1705 * $mtime = wfTimestamp( TS_UNIX, $row->timestamp );
1706 * $ttls[$row->id] = $this->adaptiveTTL( $mtime, $ttls[$row->id] );
1707 * }
1708 *
1709 * return $rows;
1710 * },
1711 * ]
1712 * );
1713 * $files = array_map( [ __CLASS__, 'newFromRow' ], $rows );
1714 * @endcode
1715 *
1716 * @param ArrayIterator $keyedIds Result of WANObjectCache::makeMultiKeys()
1717 * @param int $ttl Seconds to live for key updates
1718 * @param callable $callback Callback the yields entity regeneration callbacks
1719 * @param array $opts Options map
1720 * @return array Map of (cache key => value) in the same order as $keyedIds
1721 * @since 1.30
1722 */
1723 final public function getMultiWithUnionSetCallback(
1724 ArrayIterator $keyedIds, $ttl, callable $callback, array $opts = []
1725 ) {
1726 $idsByValueKey = $keyedIds->getArrayCopy();
1727 $valueKeys = array_keys( $idsByValueKey );
1728 $checkKeys = $opts['checkKeys'] ?? [];
1729 $pcTTL = $opts['pcTTL'] ?? self::TTL_UNCACHEABLE;
1730 unset( $opts['lockTSE'] ); // incompatible
1731 unset( $opts['busyValue'] ); // incompatible
1732
1733 // Load required keys into process cache in one go
1734 $keysGet = $this->getNonProcessCachedKeys( $valueKeys, $opts, $pcTTL );
1735 $this->warmupCache = $this->getRawKeysForWarmup( $keysGet, $checkKeys );
1736 $this->warmupKeyMisses = 0;
1737
1738 // IDs of entities known to be in need of regeneration
1739 $idsRegen = [];
1740
1741 // Find out which keys are missing/deleted/stale
1742 $curTTLs = [];
1743 $asOfs = [];
1744 $curByKey = $this->getMulti( $keysGet, $curTTLs, $checkKeys, $asOfs );
1745 foreach ( $keysGet as $key ) {
1746 if ( !array_key_exists( $key, $curByKey ) || $curTTLs[$key] < 0 ) {
1747 $idsRegen[] = $idsByValueKey[$key];
1748 }
1749 }
1750
1751 // Run the callback to populate the regeneration value map for all required IDs
1752 $newSetOpts = [];
1753 $newTTLsById = array_fill_keys( $idsRegen, $ttl );
1754 $newValsById = $idsRegen ? $callback( $idsRegen, $newTTLsById, $newSetOpts ) : [];
1755
1756 // Wrap $callback to match the getWithSetCallback() format while passing $id to $callback
1757 $id = null; // current entity ID
1758 $func = function ( $oldValue, &$ttl, &$setOpts, $oldAsOf )
1759 use ( $callback, &$id, $newValsById, $newTTLsById, $newSetOpts )
1760 {
1761 if ( array_key_exists( $id, $newValsById ) ) {
1762 // Value was already regerated as expected, so use the value in $newValsById
1763 $newValue = $newValsById[$id];
1764 $ttl = $newTTLsById[$id];
1765 $setOpts = $newSetOpts;
1766 } else {
1767 // Pre-emptive/popularity refresh and version mismatch cases are not detected
1768 // above and thus $newValsById has no entry. Run $callback on this single entity.
1769 $ttls = [ $id => $ttl ];
1770 $newValue = $callback( [ $id ], $ttls, $setOpts )[$id];
1771 $ttl = $ttls[$id];
1772 }
1773
1774 return $newValue;
1775 };
1776
1777 // Run the cache-aside logic using warmupCache instead of persistent cache queries
1778 $values = [];
1779 foreach ( $idsByValueKey as $key => $id ) { // preserve order
1780 $values[$key] = $this->getWithSetCallback( $key, $ttl, $func, $opts );
1781 }
1782
1783 $this->warmupCache = [];
1784
1785 return $values;
1786 }
1787
1788 /**
1789 * Set a key to soon expire in the local cluster if it pre-dates $purgeTimestamp
1790 *
1791 * This sets stale keys' time-to-live at HOLDOFF_TTL seconds, which both avoids
1792 * broadcasting in mcrouter setups and also avoids races with new tombstones.
1793 *
1794 * @param string $key Cache key
1795 * @param int $purgeTimestamp UNIX timestamp of purge
1796 * @param bool &$isStale Whether the key is stale
1797 * @return bool Success
1798 * @since 1.28
1799 */
1800 final public function reap( $key, $purgeTimestamp, &$isStale = false ) {
1801 $minAsOf = $purgeTimestamp + self::HOLDOFF_TTL;
1802 $wrapped = $this->cache->get( self::VALUE_KEY_PREFIX . $key );
1803 if ( is_array( $wrapped ) && $wrapped[self::FLD_TIME] < $minAsOf ) {
1804 $isStale = true;
1805 $this->logger->warning( "Reaping stale value key '$key'." );
1806 $ttlReap = self::HOLDOFF_TTL; // avoids races with tombstone creation
1807 $ok = $this->cache->changeTTL( self::VALUE_KEY_PREFIX . $key, $ttlReap );
1808 if ( !$ok ) {
1809 $this->logger->error( "Could not complete reap of key '$key'." );
1810 }
1811
1812 return $ok;
1813 }
1814
1815 $isStale = false;
1816
1817 return true;
1818 }
1819
1820 /**
1821 * Set a "check" key to soon expire in the local cluster if it pre-dates $purgeTimestamp
1822 *
1823 * @param string $key Cache key
1824 * @param int $purgeTimestamp UNIX timestamp of purge
1825 * @param bool &$isStale Whether the key is stale
1826 * @return bool Success
1827 * @since 1.28
1828 */
1829 final public function reapCheckKey( $key, $purgeTimestamp, &$isStale = false ) {
1830 $purge = $this->parsePurgeValue( $this->cache->get( self::TIME_KEY_PREFIX . $key ) );
1831 if ( $purge && $purge[self::FLD_TIME] < $purgeTimestamp ) {
1832 $isStale = true;
1833 $this->logger->warning( "Reaping stale check key '$key'." );
1834 $ok = $this->cache->changeTTL( self::TIME_KEY_PREFIX . $key, self::TTL_SECOND );
1835 if ( !$ok ) {
1836 $this->logger->error( "Could not complete reap of check key '$key'." );
1837 }
1838
1839 return $ok;
1840 }
1841
1842 $isStale = false;
1843
1844 return false;
1845 }
1846
1847 /**
1848 * @see BagOStuff::makeKey()
1849 * @param string $class Key class
1850 * @param string|null $component [optional] Key component (starting with a key collection name)
1851 * @return string Colon-delimited list of $keyspace followed by escaped components of $args
1852 * @since 1.27
1853 */
1854 public function makeKey( $class, $component = null ) {
1855 return $this->cache->makeKey( ...func_get_args() );
1856 }
1857
1858 /**
1859 * @see BagOStuff::makeGlobalKey()
1860 * @param string $class Key class
1861 * @param string|null $component [optional] Key component (starting with a key collection name)
1862 * @return string Colon-delimited list of $keyspace followed by escaped components of $args
1863 * @since 1.27
1864 */
1865 public function makeGlobalKey( $class, $component = null ) {
1866 return $this->cache->makeGlobalKey( ...func_get_args() );
1867 }
1868
1869 /**
1870 * @param array $entities List of entity IDs
1871 * @param callable $keyFunc Callback yielding a key from (entity ID, this WANObjectCache)
1872 * @return ArrayIterator Iterator yielding (cache key => entity ID) in $entities order
1873 * @since 1.28
1874 */
1875 final public function makeMultiKeys( array $entities, callable $keyFunc ) {
1876 $map = [];
1877 foreach ( $entities as $entity ) {
1878 $map[$keyFunc( $entity, $this )] = $entity;
1879 }
1880
1881 return new ArrayIterator( $map );
1882 }
1883
1884 /**
1885 * Get the "last error" registered; clearLastError() should be called manually
1886 * @return int ERR_* class constant for the "last error" registry
1887 */
1888 final public function getLastError() {
1889 $code = $this->cache->getLastError();
1890 switch ( $code ) {
1891 case BagOStuff::ERR_NONE:
1892 return self::ERR_NONE;
1893 case BagOStuff::ERR_NO_RESPONSE:
1894 return self::ERR_NO_RESPONSE;
1895 case BagOStuff::ERR_UNREACHABLE:
1896 return self::ERR_UNREACHABLE;
1897 default:
1898 return self::ERR_UNEXPECTED;
1899 }
1900 }
1901
1902 /**
1903 * Clear the "last error" registry
1904 */
1905 final public function clearLastError() {
1906 $this->cache->clearLastError();
1907 }
1908
1909 /**
1910 * Clear the in-process caches; useful for testing
1911 *
1912 * @since 1.27
1913 */
1914 public function clearProcessCache() {
1915 $this->processCaches = [];
1916 }
1917
1918 /**
1919 * Enable or disable the use of brief caching for tombstoned keys
1920 *
1921 * When a key is purged via delete(), there normally is a period where caching
1922 * is hold-off limited to an extremely short time. This method will disable that
1923 * caching, forcing the callback to run for any of:
1924 * - WANObjectCache::getWithSetCallback()
1925 * - WANObjectCache::getMultiWithSetCallback()
1926 * - WANObjectCache::getMultiWithUnionSetCallback()
1927 *
1928 * This is useful when both:
1929 * - a) the database used by the callback is known to be up-to-date enough
1930 * for some particular purpose (e.g. replica DB has applied transaction X)
1931 * - b) the caller needs to exploit that fact, and therefore needs to avoid the
1932 * use of inherently volatile and possibly stale interim keys
1933 *
1934 * @see WANObjectCache::delete()
1935 * @param bool $enabled Whether to enable interim caching
1936 * @since 1.31
1937 */
1938 final public function useInterimHoldOffCaching( $enabled ) {
1939 $this->useInterimHoldOffCaching = $enabled;
1940 }
1941
1942 /**
1943 * @param int $flag ATTR_* class constant
1944 * @return int QOS_* class constant
1945 * @since 1.28
1946 */
1947 public function getQoS( $flag ) {
1948 return $this->cache->getQoS( $flag );
1949 }
1950
1951 /**
1952 * Get a TTL that is higher for objects that have not changed recently
1953 *
1954 * This is useful for keys that get explicit purges and DB or purge relay
1955 * lag is a potential concern (especially how it interacts with CDN cache)
1956 *
1957 * Example usage:
1958 * @code
1959 * // Last-modified time of page
1960 * $mtime = wfTimestamp( TS_UNIX, $page->getTimestamp() );
1961 * // Get adjusted TTL. If $mtime is 3600 seconds ago and $minTTL/$factor left at
1962 * // defaults, then $ttl is 3600 * .2 = 720. If $minTTL was greater than 720, then
1963 * // $ttl would be $minTTL. If $maxTTL was smaller than 720, $ttl would be $maxTTL.
1964 * $ttl = $cache->adaptiveTTL( $mtime, $cache::TTL_DAY );
1965 * @endcode
1966 *
1967 * Another use case is when there are no applicable "last modified" fields in the DB,
1968 * and there are too many dependencies for explicit purges to be viable, and the rate of
1969 * change to relevant content is unstable, and it is highly valued to have the cached value
1970 * be as up-to-date as possible.
1971 *
1972 * Example usage:
1973 * @code
1974 * $query = "<some complex query>";
1975 * $idListFromComplexQuery = $cache->getWithSetCallback(
1976 * $cache->makeKey( 'complex-graph-query', $hashOfQuery ),
1977 * GraphQueryClass::STARTING_TTL,
1978 * function ( $oldValue, &$ttl, array &$setOpts, $oldAsOf ) use ( $query, $cache ) {
1979 * $gdb = $this->getReplicaGraphDbConnection();
1980 * // Account for any snapshot/replica DB lag
1981 * $setOpts += GraphDatabase::getCacheSetOptions( $gdb );
1982 *
1983 * $newList = iterator_to_array( $gdb->query( $query ) );
1984 * sort( $newList, SORT_NUMERIC ); // normalize
1985 *
1986 * $minTTL = GraphQueryClass::MIN_TTL;
1987 * $maxTTL = GraphQueryClass::MAX_TTL;
1988 * if ( $oldValue !== false ) {
1989 * // Note that $oldAsOf is the last time this callback ran
1990 * $ttl = ( $newList === $oldValue )
1991 * // No change: cache for 150% of the age of $oldValue
1992 * ? $cache->adaptiveTTL( $oldAsOf, $maxTTL, $minTTL, 1.5 )
1993 * // Changed: cache for 50% of the age of $oldValue
1994 * : $cache->adaptiveTTL( $oldAsOf, $maxTTL, $minTTL, .5 );
1995 * }
1996 *
1997 * return $newList;
1998 * },
1999 * [
2000 * // Keep stale values around for doing comparisons for TTL calculations.
2001 * // High values improve long-tail keys hit-rates, though might waste space.
2002 * 'staleTTL' => GraphQueryClass::GRACE_TTL
2003 * ]
2004 * );
2005 * @endcode
2006 *
2007 * @param int|float $mtime UNIX timestamp
2008 * @param int $maxTTL Maximum TTL (seconds)
2009 * @param int $minTTL Minimum TTL (seconds); Default: 30
2010 * @param float $factor Value in the range (0,1); Default: .2
2011 * @return int Adaptive TTL
2012 * @since 1.28
2013 */
2014 public function adaptiveTTL( $mtime, $maxTTL, $minTTL = 30, $factor = 0.2 ) {
2015 if ( is_float( $mtime ) || ctype_digit( $mtime ) ) {
2016 $mtime = (int)$mtime; // handle fractional seconds and string integers
2017 }
2018
2019 if ( !is_int( $mtime ) || $mtime <= 0 ) {
2020 return $minTTL; // no last-modified time provided
2021 }
2022
2023 $age = $this->getCurrentTime() - $mtime;
2024
2025 return (int)min( $maxTTL, max( $minTTL, $factor * $age ) );
2026 }
2027
2028 /**
2029 * @return int Number of warmup key cache misses last round
2030 * @since 1.30
2031 */
2032 final public function getWarmupKeyMisses() {
2033 return $this->warmupKeyMisses;
2034 }
2035
2036 /**
2037 * Do the actual async bus purge of a key
2038 *
2039 * This must set the key to "PURGED:<UNIX timestamp>:<holdoff>"
2040 *
2041 * @param string $key Cache key
2042 * @param int $ttl How long to keep the tombstone [seconds]
2043 * @param int $holdoff HOLDOFF_* constant controlling how long to ignore sets for this key
2044 * @return bool Success
2045 */
2046 protected function relayPurge( $key, $ttl, $holdoff ) {
2047 if ( $this->mcrouterAware ) {
2048 // See https://github.com/facebook/mcrouter/wiki/Multi-cluster-broadcast-setup
2049 // Wildcards select all matching routes, e.g. the WAN cluster on all DCs
2050 $ok = $this->cache->set(
2051 "/*/{$this->cluster}/{$key}",
2052 $this->makePurgeValue( $this->getCurrentTime(), self::HOLDOFF_NONE ),
2053 $ttl
2054 );
2055 } else {
2056 // This handles the mcrouter and the single-DC case
2057 $ok = $this->cache->set(
2058 $key,
2059 $this->makePurgeValue( $this->getCurrentTime(), self::HOLDOFF_NONE ),
2060 $ttl
2061 );
2062 }
2063
2064 return $ok;
2065 }
2066
2067 /**
2068 * Do the actual async bus delete of a key
2069 *
2070 * @param string $key Cache key
2071 * @return bool Success
2072 */
2073 protected function relayDelete( $key ) {
2074 if ( $this->mcrouterAware ) {
2075 // See https://github.com/facebook/mcrouter/wiki/Multi-cluster-broadcast-setup
2076 // Wildcards select all matching routes, e.g. the WAN cluster on all DCs
2077 $ok = $this->cache->delete( "/*/{$this->cluster}/{$key}" );
2078 } else {
2079 // Some other proxy handles broadcasting or there is only one datacenter
2080 $ok = $this->cache->delete( $key );
2081 }
2082
2083 return $ok;
2084 }
2085
2086 /**
2087 * @param string $key
2088 * @param int $ttl
2089 * @param callable $callback
2090 * @param array $opts
2091 * @return bool Success
2092 */
2093 private function scheduleAsyncRefresh( $key, $ttl, $callback, $opts ) {
2094 if ( !$this->asyncHandler ) {
2095 return false;
2096 }
2097 // Update the cache value later, such during post-send of an HTTP request
2098 $func = $this->asyncHandler;
2099 $func( function () use ( $key, $ttl, $callback, $opts ) {
2100 $asOf = null; // unused
2101 $opts['minAsOf'] = INF; // force a refresh
2102 $this->doGetWithSetCallback( $key, $ttl, $callback, $opts, $asOf );
2103 } );
2104
2105 return true;
2106 }
2107
2108 /**
2109 * Check if a key is fresh or in the grace window and thus due for randomized reuse
2110 *
2111 * If $curTTL > 0 (e.g. not expired) this returns true. Otherwise, the chance of returning
2112 * true decrease steadily from 100% to 0% as the |$curTTL| moves from 0 to $graceTTL seconds.
2113 * This handles widely varying levels of cache access traffic.
2114 *
2115 * If $curTTL <= -$graceTTL (e.g. already expired), then this returns false.
2116 *
2117 * @param float $curTTL Approximate TTL left on the key if present
2118 * @param int $graceTTL Consider using stale values if $curTTL is greater than this
2119 * @return bool
2120 */
2121 protected function isAliveOrInGracePeriod( $curTTL, $graceTTL ) {
2122 if ( $curTTL > 0 ) {
2123 return true;
2124 } elseif ( $graceTTL <= 0 ) {
2125 return false;
2126 }
2127
2128 $ageStale = abs( $curTTL ); // seconds of staleness
2129 $curGTTL = ( $graceTTL - $ageStale ); // current grace-time-to-live
2130 if ( $curGTTL <= 0 ) {
2131 return false; // already out of grace period
2132 }
2133
2134 // Chance of using a stale value is the complement of the chance of refreshing it
2135 return !$this->worthRefreshExpiring( $curGTTL, $graceTTL );
2136 }
2137
2138 /**
2139 * Check if a key is nearing expiration and thus due for randomized regeneration
2140 *
2141 * This returns false if $curTTL >= $lowTTL. Otherwise, the chance of returning true
2142 * increases steadily from 0% to 100% as the $curTTL moves from $lowTTL to 0 seconds.
2143 * This handles widely varying levels of cache access traffic.
2144 *
2145 * If $curTTL <= 0 (e.g. already expired), then this returns false.
2146 *
2147 * @param float $curTTL Approximate TTL left on the key if present
2148 * @param float $lowTTL Consider a refresh when $curTTL is less than this
2149 * @return bool
2150 */
2151 protected function worthRefreshExpiring( $curTTL, $lowTTL ) {
2152 if ( $lowTTL <= 0 ) {
2153 return false;
2154 } elseif ( $curTTL >= $lowTTL ) {
2155 return false;
2156 } elseif ( $curTTL <= 0 ) {
2157 return false;
2158 }
2159
2160 $chance = ( 1 - $curTTL / $lowTTL );
2161
2162 return mt_rand( 1, 1e9 ) <= 1e9 * $chance;
2163 }
2164
2165 /**
2166 * Check if a key is due for randomized regeneration due to its popularity
2167 *
2168 * This is used so that popular keys can preemptively refresh themselves for higher
2169 * consistency (especially in the case of purge loss/delay). Unpopular keys can remain
2170 * in cache with their high nominal TTL. This means popular keys keep good consistency,
2171 * whether the data changes frequently or not, and long-tail keys get to stay in cache
2172 * and get hits too. Similar to worthRefreshExpiring(), randomization is used.
2173 *
2174 * @param float $asOf UNIX timestamp of the value
2175 * @param int $ageNew Age of key when this might recommend refreshing (seconds)
2176 * @param int $timeTillRefresh Age of key when it should be refreshed if popular (seconds)
2177 * @param float $now The current UNIX timestamp
2178 * @return bool
2179 */
2180 protected function worthRefreshPopular( $asOf, $ageNew, $timeTillRefresh, $now ) {
2181 if ( $ageNew < 0 || $timeTillRefresh <= 0 ) {
2182 return false;
2183 }
2184
2185 $age = $now - $asOf;
2186 $timeOld = $age - $ageNew;
2187 if ( $timeOld <= 0 ) {
2188 return false;
2189 }
2190
2191 // Lifecycle is: new, ramp-up refresh chance, full refresh chance.
2192 // Note that the "expected # of refreshes" for the ramp-up time range is half of what it
2193 // would be if P(refresh) was at its full value during that time range.
2194 $refreshWindowSec = max( $timeTillRefresh - $ageNew - self::RAMPUP_TTL / 2, 1 );
2195 // P(refresh) * (# hits in $refreshWindowSec) = (expected # of refreshes)
2196 // P(refresh) * ($refreshWindowSec * $popularHitsPerSec) = 1
2197 // P(refresh) = 1/($refreshWindowSec * $popularHitsPerSec)
2198 $chance = 1 / ( self::HIT_RATE_HIGH * $refreshWindowSec );
2199
2200 // Ramp up $chance from 0 to its nominal value over RAMPUP_TTL seconds to avoid stampedes
2201 $chance *= ( $timeOld <= self::RAMPUP_TTL ) ? $timeOld / self::RAMPUP_TTL : 1;
2202
2203 return mt_rand( 1, 1e9 ) <= 1e9 * $chance;
2204 }
2205
2206 /**
2207 * Check if $value is not false, versioned (if needed), and not older than $minTime (if set)
2208 *
2209 * @param array|bool $value
2210 * @param bool $versioned
2211 * @param float $asOf The time $value was generated
2212 * @param float $minTime The last time the main value was generated (0.0 if unknown)
2213 * @param float|null $purgeTime The last time the value was invalidated
2214 * @return bool
2215 */
2216 protected function isValid( $value, $versioned, $asOf, $minTime, $purgeTime = null ) {
2217 // Avoid reading any key not generated after the latest delete() or touch
2218 $safeMinTime = max( $minTime, $purgeTime + self::TINY_POSTIVE );
2219
2220 if ( $value === false ) {
2221 return false;
2222 } elseif ( $versioned && !isset( $value[self::VFLD_VERSION] ) ) {
2223 return false;
2224 } elseif ( $safeMinTime > 0 && $asOf < $minTime ) {
2225 return false;
2226 }
2227
2228 return true;
2229 }
2230
2231 /**
2232 * Do not use this method outside WANObjectCache
2233 *
2234 * @param mixed $value
2235 * @param int $ttl [0=forever]
2236 * @param float $now Unix Current timestamp just before calling set()
2237 * @return array
2238 */
2239 protected function wrap( $value, $ttl, $now ) {
2240 return [
2241 self::FLD_VERSION => self::VERSION,
2242 self::FLD_VALUE => $value,
2243 self::FLD_TTL => $ttl,
2244 self::FLD_TIME => $now
2245 ];
2246 }
2247
2248 /**
2249 * Do not use this method outside WANObjectCache
2250 *
2251 * The cached value will be false if absent/tombstoned/malformed
2252 *
2253 * @param array|string|bool $wrapped
2254 * @param float $now Unix Current timestamp (preferrably pre-query)
2255 * @return array (cached value or false, current TTL, value timestamp, tombstone timestamp)
2256 */
2257 protected function unwrap( $wrapped, $now ) {
2258 // Check if the value is a tombstone
2259 $purge = $this->parsePurgeValue( $wrapped );
2260 if ( $purge !== false ) {
2261 // Purged values should always have a negative current $ttl
2262 $curTTL = min( $purge[self::FLD_TIME] - $now, self::TINY_NEGATIVE );
2263 return [ false, $curTTL, null, $purge[self::FLD_TIME] ];
2264 }
2265
2266 if ( !is_array( $wrapped ) // not found
2267 || !isset( $wrapped[self::FLD_VERSION] ) // wrong format
2268 || $wrapped[self::FLD_VERSION] !== self::VERSION // wrong version
2269 ) {
2270 return [ false, null, null, null ];
2271 }
2272
2273 if ( $wrapped[self::FLD_TTL] > 0 ) {
2274 // Get the approximate time left on the key
2275 $age = $now - $wrapped[self::FLD_TIME];
2276 $curTTL = max( $wrapped[self::FLD_TTL] - $age, 0.0 );
2277 } else {
2278 // Key had no TTL, so the time left is unbounded
2279 $curTTL = INF;
2280 }
2281
2282 if ( $wrapped[self::FLD_TIME] < $this->epoch ) {
2283 // Values this old are ignored
2284 return [ false, null, null, null ];
2285 }
2286
2287 return [ $wrapped[self::FLD_VALUE], $curTTL, $wrapped[self::FLD_TIME], null ];
2288 }
2289
2290 /**
2291 * @param array $keys
2292 * @param string $prefix
2293 * @return string[]
2294 */
2295 protected static function prefixCacheKeys( array $keys, $prefix ) {
2296 $res = [];
2297 foreach ( $keys as $key ) {
2298 $res[] = $prefix . $key;
2299 }
2300
2301 return $res;
2302 }
2303
2304 /**
2305 * @param string $key String of the format <scope>:<class>[:<class or variable>]...
2306 * @return string A collection name to describe this class of key
2307 */
2308 protected function determineKeyClassForStats( $key ) {
2309 $parts = explode( ':', $key );
2310
2311 return $parts[1] ?? $parts[0]; // sanity
2312 }
2313
2314 /**
2315 * @param string|array|bool $value Possible string of the form "PURGED:<timestamp>:<holdoff>"
2316 * @return array|bool Array containing a UNIX timestamp (float) and holdoff period (integer),
2317 * or false if value isn't a valid purge value
2318 */
2319 protected function parsePurgeValue( $value ) {
2320 if ( !is_string( $value ) ) {
2321 return false;
2322 }
2323
2324 $segments = explode( ':', $value, 3 );
2325 if ( !isset( $segments[0] ) || !isset( $segments[1] )
2326 || "{$segments[0]}:" !== self::PURGE_VAL_PREFIX
2327 ) {
2328 return false;
2329 }
2330
2331 if ( !isset( $segments[2] ) ) {
2332 // Back-compat with old purge values without holdoff
2333 $segments[2] = self::HOLDOFF_TTL;
2334 }
2335
2336 if ( $segments[1] < $this->epoch ) {
2337 // Values this old are ignored
2338 return false;
2339 }
2340
2341 return [
2342 self::FLD_TIME => (float)$segments[1],
2343 self::FLD_HOLDOFF => (int)$segments[2],
2344 ];
2345 }
2346
2347 /**
2348 * @param float $timestamp
2349 * @param int $holdoff In seconds
2350 * @return string Wrapped purge value
2351 */
2352 protected function makePurgeValue( $timestamp, $holdoff ) {
2353 return self::PURGE_VAL_PREFIX . (float)$timestamp . ':' . (int)$holdoff;
2354 }
2355
2356 /**
2357 * @param string $group
2358 * @return MapCacheLRU
2359 */
2360 protected function getProcessCache( $group ) {
2361 if ( !isset( $this->processCaches[$group] ) ) {
2362 list( , $n ) = explode( ':', $group );
2363 $this->processCaches[$group] = new MapCacheLRU( (int)$n );
2364 }
2365
2366 return $this->processCaches[$group];
2367 }
2368
2369 /**
2370 * @param array $keys
2371 * @param array $opts
2372 * @param int $pcTTL
2373 * @return array List of keys
2374 */
2375 private function getNonProcessCachedKeys( array $keys, array $opts, $pcTTL ) {
2376 $keysFound = [];
2377 if ( isset( $opts['pcTTL'] ) && $opts['pcTTL'] > 0 && $this->callbackDepth == 0 ) {
2378 $pcGroup = $opts['pcGroup'] ?? self::PC_PRIMARY;
2379 $procCache = $this->getProcessCache( $pcGroup );
2380 foreach ( $keys as $key ) {
2381 if ( $procCache->has( $key, $pcTTL ) ) {
2382 $keysFound[] = $key;
2383 }
2384 }
2385 }
2386
2387 return array_diff( $keys, $keysFound );
2388 }
2389
2390 /**
2391 * @param array $keys
2392 * @param array $checkKeys
2393 * @return array Map of (cache key => mixed)
2394 */
2395 private function getRawKeysForWarmup( array $keys, array $checkKeys ) {
2396 if ( !$keys ) {
2397 return [];
2398 }
2399
2400 $keysWarmUp = [];
2401 // Get all the value keys to fetch...
2402 foreach ( $keys as $key ) {
2403 $keysWarmUp[] = self::VALUE_KEY_PREFIX . $key;
2404 }
2405 // Get all the check keys to fetch...
2406 foreach ( $checkKeys as $i => $checkKeyOrKeys ) {
2407 if ( is_int( $i ) ) {
2408 // Single check key that applies to all value keys
2409 $keysWarmUp[] = self::TIME_KEY_PREFIX . $checkKeyOrKeys;
2410 } else {
2411 // List of check keys that apply to value key $i
2412 $keysWarmUp = array_merge(
2413 $keysWarmUp,
2414 self::prefixCacheKeys( $checkKeyOrKeys, self::TIME_KEY_PREFIX )
2415 );
2416 }
2417 }
2418
2419 $warmupCache = $this->cache->getMulti( $keysWarmUp );
2420 $warmupCache += array_fill_keys( $keysWarmUp, false );
2421
2422 return $warmupCache;
2423 }
2424
2425 /**
2426 * @return float UNIX timestamp
2427 * @codeCoverageIgnore
2428 */
2429 protected function getCurrentTime() {
2430 if ( $this->wallClockOverride ) {
2431 return $this->wallClockOverride;
2432 }
2433
2434 $clockTime = (float)time(); // call this first
2435 // microtime() uses an initial gettimeofday() call added to usage clocks.
2436 // This can severely drift from time() and the microtime() value of other threads
2437 // due to undercounting of the amount of time elapsed. Instead of seeing the current
2438 // time as being in the past, use the value of time(). This avoids setting cache values
2439 // that will immediately be seen as expired and possibly cause stampedes.
2440 return max( microtime( true ), $clockTime );
2441 }
2442
2443 /**
2444 * @param float|null &$time Mock UNIX timestamp for testing
2445 * @codeCoverageIgnore
2446 */
2447 public function setMockTime( &$time ) {
2448 $this->wallClockOverride =& $time;
2449 $this->cache->setMockTime( $time );
2450 }
2451 }