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.
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.
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
22 use Liuggio\StatsdClient\Factory\StatsdDataFactoryInterface
;
23 use Psr\Log\LoggerAwareInterface
;
24 use Psr\Log\LoggerInterface
;
25 use Psr\Log\NullLogger
;
28 * Multi-datacenter aware caching interface
30 * ### Using WANObjectCache
32 * All operations go to the local datacenter cache, except for delete(),
33 * touchCheckKey(), and resetCheckKey(), which broadcast to all datacenters.
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.
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
46 * The validation strategy refers to the natural avoidance of stale data
47 * by one of the following means:
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 If outdated, the value is recomputed.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.
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().
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.
72 * The need for immediate updates should be avoided. If needed, solutions must be
73 * sought outside WANObjectCache.
75 * ### Deploying WANObjectCache
77 * There are two supported ways to set up broadcasted operations:
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.
97 * Broadcasted operations like delete() and touchCheckKey() are intended to run
98 * immediately in the local datacenter and asynchronously in remote datacenters.
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.
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
116 class WANObjectCache
implements IExpiringStore
, IStoreKeyEncoder
, LoggerAwareInterface
{
117 /** @var BagOStuff The local datacenter cache */
119 /** @var MapCacheLRU[] Map of group PHP instance caches */
120 protected $processCaches = [];
121 /** @var LoggerInterface */
123 /** @var StatsdDataFactoryInterface */
125 /** @var callable|null Function that takes a WAN cache callback and runs it later */
126 protected $asyncHandler;
128 /** @var bool Whether to use mcrouter key prefixing for routing */
129 protected $mcrouterAware;
130 /** @var string Physical region for mcrouter use */
132 /** @var string Cache cluster name for mcrouter use */
134 /** @var bool Whether to use "interim" caching while keys are tombstoned */
135 protected $useInterimHoldOffCaching = true;
136 /** @var float Unix timestamp of the oldest possible valid values */
138 /** @var string Stable secret used for hasing long strings into key components */
141 /** @var int Callback stack depth for getWithSetCallback() */
142 private $callbackDepth = 0;
143 /** @var mixed[] Temporary warm-up cache */
144 private $warmupCache = [];
145 /** @var int Key fetched */
146 private $warmupKeyMisses = 0;
148 /** @var float|null */
149 private $wallClockOverride;
151 /** @var int Max expected seconds to pass between delete() and DB commit finishing */
152 const MAX_COMMIT_DELAY
= 3;
153 /** @var int Max expected seconds of combined lag from replication and view snapshots */
154 const MAX_READ_LAG
= 7;
155 /** @var int Seconds to tombstone keys on delete() and treat as volatile after invalidation */
156 const HOLDOFF_TTL
= self
::MAX_COMMIT_DELAY + self
::MAX_READ_LAG +
1;
158 /** @var int Idiom for getWithSetCallback() meaning "do not store the callback result" */
159 const TTL_UNCACHEABLE
= -1;
161 /** @var int Consider regeneration if the key will expire within this many seconds */
163 /** @var int Max TTL, in seconds, to store keys when a data sourced is lagged */
164 const TTL_LAGGED
= 30;
166 /** @var int Expected time-till-refresh, in seconds, if the key is accessed once per second */
168 /** @var int Minimum key age, in seconds, for expected time-till-refresh to be considered */
171 /** @var int Idiom for getWithSetCallback() meaning "no cache stampede mutex required" */
174 /** @var int Idiom for set()/getWithSetCallback() meaning "no post-expiration persistence" */
175 const STALE_TTL_NONE
= 0;
176 /** @var int Idiom for set()/getWithSetCallback() meaning "no post-expiration grace period" */
177 const GRACE_TTL_NONE
= 0;
178 /** @var int Idiom for delete()/touchCheckKey() meaning "no hold-off period" */
179 const HOLDOFF_TTL_NONE
= 0;
180 /** @var int Alias for HOLDOFF_TTL_NONE (b/c) (deprecated since 1.34) */
181 const HOLDOFF_NONE
= self
::HOLDOFF_TTL_NONE
;
183 /** @var float Idiom for getWithSetCallback() meaning "no minimum required as-of timestamp" */
184 const MIN_TIMESTAMP_NONE
= 0.0;
186 /** @var string Default process cache name and max key count */
187 const PC_PRIMARY
= 'primary:1000';
189 /** @var int Idion for get()/getMulti() to return extra information by reference */
190 const PASS_BY_REF
= -1;
192 /** @var int Seconds to keep dependency purge keys around */
193 private static $CHECK_KEY_TTL = self
::TTL_YEAR
;
194 /** @var int Seconds to keep interim value keys for tombstoned keys around */
195 private static $INTERIM_KEY_TTL = 1;
197 /** @var int Seconds to keep lock keys around */
198 private static $LOCK_TTL = 10;
199 /** @var int Seconds to no-op key set() calls to avoid large blob I/O stampedes */
200 private static $COOLOFF_TTL = 1;
201 /** @var int Seconds to ramp up the chance of regeneration due to expected time-till-refresh */
202 private static $RAMPUP_TTL = 30;
204 /** @var float Tiny negative float to use when CTL comes up >= 0 due to clock skew */
205 private static $TINY_NEGATIVE = -0.000001;
206 /** @var float Tiny positive float to use when using "minTime" to assert an inequality */
207 private static $TINY_POSTIVE = 0.000001;
209 /** @var int Milliseconds of key fetch/validate/regenerate delay prone to set() stampedes */
210 private static $SET_DELAY_HIGH_MS = 50;
211 /** @var int Min millisecond set() backoff during hold-off (far less than INTERIM_KEY_TTL) */
212 private static $RECENT_SET_LOW_MS = 50;
213 /** @var int Max millisecond set() backoff during hold-off (far less than INTERIM_KEY_TTL) */
214 private static $RECENT_SET_HIGH_MS = 100;
216 /** @var int Consider value generation slow if it takes more than this many seconds */
217 private static $GENERATION_SLOW_SEC = 3;
219 /** @var int Key to the tombstone entry timestamp */
220 private static $PURGE_TIME = 0;
221 /** @var int Key to the tombstone entry hold-off TTL */
222 private static $PURGE_HOLDOFF = 1;
224 /** @var int Cache format version number */
225 private static $VERSION = 1;
227 /** @var int Key to WAN cache version number */
228 private static $FLD_FORMAT_VERSION = 0;
229 /** @var int Key to the cached value */
230 private static $FLD_VALUE = 1;
231 /** @var int Key to the original TTL */
232 private static $FLD_TTL = 2;
233 /** @var int Key to the cache timestamp */
234 private static $FLD_TIME = 3;
235 /** @var int Key to the flags bit field (reserved number) */
236 private static /** @noinspection PhpUnusedPrivateFieldInspection */ $FLD_FLAGS = 4;
237 /** @var int Key to collection cache version number */
238 private static $FLD_VALUE_VERSION = 5;
239 /** @var int Key to how long it took to generate the value */
240 private static $FLD_GENERATION_TIME = 6;
242 private static $VALUE_KEY_PREFIX = 'WANCache:v:';
243 private static $INTERIM_KEY_PREFIX = 'WANCache:i:';
244 private static $TIME_KEY_PREFIX = 'WANCache:t:';
245 private static $MUTEX_KEY_PREFIX = 'WANCache:m:';
246 private static $COOLOFF_KEY_PREFIX = 'WANCache:c:';
248 private static $PURGE_VAL_PREFIX = 'PURGED:';
251 * @param array $params
252 * - cache : BagOStuff object for a persistent cache
253 * - logger : LoggerInterface object
254 * - stats : StatsdDataFactoryInterface object
255 * - asyncHandler : A function that takes a callback and runs it later. If supplied,
256 * whenever a preemptive refresh would be triggered in getWithSetCallback(), the
257 * current cache value is still used instead. However, the async-handler function
258 * receives a WAN cache callback that, when run, will execute the value generation
259 * callback supplied by the getWithSetCallback() caller. The result will be saved
260 * as normal. The handler is expected to call the WAN cache callback at an opportune
261 * time (e.g. HTTP post-send), though generally within a few 100ms. [optional]
262 * - region: the current physical region. This is required when using mcrouter as the
263 * backing store proxy. [optional]
264 * - cluster: name of the cache cluster used by this WAN cache. The name must be the
265 * same in all datacenters; the ("region","cluster") tuple is what distinguishes
266 * the counterpart cache clusters among all the datacenter. The contents of
267 * https://github.com/facebook/mcrouter/wiki/Config-Files give background on this.
268 * This is required when using mcrouter as the backing store proxy. [optional]
269 * - mcrouterAware: set as true if mcrouter is the backing store proxy and mcrouter
270 * is configured to interpret /<region>/<cluster>/ key prefixes as routes. This
271 * requires that "region" and "cluster" are both set above. [optional]
272 * - epoch: lowest UNIX timestamp a value/tombstone must have to be valid. [optional]
273 * - secret: stable secret used for hashing long strings into key components. [optional]
275 public function __construct( array $params ) {
276 $this->cache
= $params['cache'];
277 $this->region
= $params['region'] ??
'main';
278 $this->cluster
= $params['cluster'] ??
'wan-main';
279 $this->mcrouterAware
= !empty( $params['mcrouterAware'] );
280 $this->epoch
= $params['epoch'] ??
0;
281 $this->secret
= $params['secret'] ??
(string)$this->epoch
;
283 $this->setLogger( $params['logger'] ??
new NullLogger() );
284 $this->stats
= $params['stats'] ??
new NullStatsdDataFactory();
285 $this->asyncHandler
= $params['asyncHandler'] ??
null;
289 * @param LoggerInterface $logger
291 public function setLogger( LoggerInterface
$logger ) {
292 $this->logger
= $logger;
296 * Get an instance that wraps EmptyBagOStuff
298 * @return WANObjectCache
300 public static function newEmpty() {
301 return new static( [ 'cache' => new EmptyBagOStuff() ] );
305 * Fetch the value of a key from cache
307 * If supplied, $curTTL is set to the remaining TTL (current time left):
308 * - a) INF; if $key exists, has no TTL, and is not invalidated by $checkKeys
309 * - b) float (>=0); if $key exists, has a TTL, and is not invalidated by $checkKeys
310 * - c) float (<0); if $key is tombstoned, stale, or existing but invalidated by $checkKeys
311 * - d) null; if $key does not exist and is not tombstoned
313 * If a key is tombstoned, $curTTL will reflect the time since delete().
315 * The timestamp of $key will be checked against the last-purge timestamp
316 * of each of $checkKeys. Those $checkKeys not in cache will have the last-purge
317 * initialized to the current timestamp. If any of $checkKeys have a timestamp
318 * greater than that of $key, then $curTTL will reflect how long ago $key
319 * became invalid. Callers can use $curTTL to know when the value is stale.
320 * The $checkKeys parameter allow mass invalidations by updating a single key:
321 * - a) Each "check" key represents "last purged" of some source data
322 * - b) Callers pass in relevant "check" keys as $checkKeys in get()
323 * - c) When the source data that "check" keys represent changes,
324 * the touchCheckKey() method is called on them
326 * Source data entities might exists in a DB that uses snapshot isolation
327 * (e.g. the default REPEATABLE-READ in innoDB). Even for mutable data, that
328 * isolation can largely be maintained by doing the following:
329 * - a) Calling delete() on entity change *and* creation, before DB commit
330 * - b) Keeping transaction duration shorter than the delete() hold-off TTL
331 * - c) Disabling interim key caching via useInterimHoldOffCaching() before get() calls
333 * However, pre-snapshot values might still be seen if an update was made
334 * in a remote datacenter but the purge from delete() didn't relay yet.
336 * Consider using getWithSetCallback() instead of get() and set() cycles.
337 * That method has cache slam avoiding features for hot/expensive keys.
339 * Pass $info as WANObjectCache::PASS_BY_REF to transform it into a cache key metadata map.
340 * This map includes the following metadata:
341 * - asOf: UNIX timestamp of the value or null if the key is nonexistant
342 * - tombAsOf: UNIX timestamp of the tombstone or null if the key is not tombstoned
343 * - lastCKPurge: UNIX timestamp of the highest check key or null if none provided
344 * - version: cached value version number or null if the key is nonexistant
346 * Otherwise, $info will transform into the cached value timestamp.
348 * @param string $key Cache key made from makeKey() or makeGlobalKey()
349 * @param mixed|null &$curTTL Approximate TTL left on the key if present/tombstoned [returned]
350 * @param string[] $checkKeys The "check" keys used to validate the value
351 * @param mixed|null &$info Key info if WANObjectCache::PASS_BY_REF [returned]
352 * @return mixed Value of cache key or false on failure
354 final public function get(
355 $key, &$curTTL = null, array $checkKeys = [], &$info = null
357 $curTTLs = self
::PASS_BY_REF
;
358 $infoByKey = self
::PASS_BY_REF
;
359 $values = $this->getMulti( [ $key ], $curTTLs, $checkKeys, $infoByKey );
360 $curTTL = $curTTLs[$key] ??
null;
361 if ( $info === self
::PASS_BY_REF
) {
363 'asOf' => $infoByKey[$key]['asOf'] ??
null,
364 'tombAsOf' => $infoByKey[$key]['tombAsOf'] ??
null,
365 'lastCKPurge' => $infoByKey[$key]['lastCKPurge'] ??
null,
366 'version' => $infoByKey[$key]['version'] ??
null
369 $info = $infoByKey[$key]['asOf'] ??
null; // b/c
372 return $values[$key] ??
false;
376 * Fetch the value of several keys from cache
378 * Pass $info as WANObjectCache::PASS_BY_REF to transform it into a map of cache keys
379 * to cache key metadata maps, each having the same style as those of WANObjectCache::get().
380 * All the cache keys listed in $keys will have an entry.
382 * Othwerwise, $info will transform into a map of (cache key => cached value timestamp).
383 * Only the cache keys listed in $keys that exists or are tombstoned will have an entry.
385 * $checkKeys holds the "check" keys used to validate values of applicable keys. The integer
386 * indexes hold "check" keys that apply to all of $keys while the string indexes hold "check"
387 * keys that only apply to the cache key with that name.
389 * @see WANObjectCache::get()
391 * @param string[] $keys List of cache keys made from makeKey() or makeGlobalKey()
392 * @param mixed|null &$curTTLs Map of (key => TTL left) for existing/tombstoned keys [returned]
393 * @param string[]|string[][] $checkKeys Map of (integer or cache key => "check" key(s))
394 * @param mixed|null &$info Map of (key => info) if WANObjectCache::PASS_BY_REF [returned]
395 * @return mixed[] Map of (key => value) for existing values; order of $keys is preserved
397 final public function getMulti(
400 array $checkKeys = [],
407 $vPrefixLen = strlen( self
::$VALUE_KEY_PREFIX );
408 $valueKeys = self
::prefixCacheKeys( $keys, self
::$VALUE_KEY_PREFIX );
410 $checkKeysForAll = [];
411 $checkKeysByKey = [];
413 foreach ( $checkKeys as $i => $checkKeyGroup ) {
414 $prefixed = self
::prefixCacheKeys( (array)$checkKeyGroup, self
::$TIME_KEY_PREFIX );
415 $checkKeysFlat = array_merge( $checkKeysFlat, $prefixed );
416 // Are these check keys for a specific cache key, or for all keys being fetched?
417 if ( is_int( $i ) ) {
418 $checkKeysForAll = array_merge( $checkKeysForAll, $prefixed );
420 $checkKeysByKey[$i] = $prefixed;
424 // Fetch all of the raw values
425 $keysGet = array_merge( $valueKeys, $checkKeysFlat );
426 if ( $this->warmupCache
) {
427 $wrappedValues = array_intersect_key( $this->warmupCache
, array_flip( $keysGet ) );
428 $keysGet = array_diff( $keysGet, array_keys( $wrappedValues ) ); // keys left to fetch
429 $this->warmupKeyMisses +
= count( $keysGet );
434 $wrappedValues +
= $this->cache
->getMulti( $keysGet );
436 // Time used to compare/init "check" keys (derived after getMulti() to be pessimistic)
437 $now = $this->getCurrentTime();
439 // Collect timestamps from all "check" keys
440 $purgeValuesForAll = $this->processCheckKeys( $checkKeysForAll, $wrappedValues, $now );
441 $purgeValuesByKey = [];
442 foreach ( $checkKeysByKey as $cacheKey => $checks ) {
443 $purgeValuesByKey[$cacheKey] =
444 $this->processCheckKeys( $checks, $wrappedValues, $now );
447 // Get the main cache value for each key and validate them
448 foreach ( $valueKeys as $vKey ) {
449 $key = substr( $vKey, $vPrefixLen ); // unprefix
450 list( $value, $keyInfo ) = $this->unwrap( $wrappedValues[$vKey] ??
false, $now );
451 // Force dependent keys to be seen as stale for a while after purging
452 // to reduce race conditions involving stale data getting cached
453 $purgeValues = $purgeValuesForAll;
454 if ( isset( $purgeValuesByKey[$key] ) ) {
455 $purgeValues = array_merge( $purgeValues, $purgeValuesByKey[$key] );
458 $lastCKPurge = null; // timestamp of the highest check key
459 foreach ( $purgeValues as $purge ) {
460 $lastCKPurge = max( $purge[self
::$PURGE_TIME], $lastCKPurge );
461 $safeTimestamp = $purge[self
::$PURGE_TIME] +
$purge[self
::$PURGE_HOLDOFF];
462 if ( $value !== false && $safeTimestamp >= $keyInfo['asOf'] ) {
463 // How long ago this value was invalidated by *this* check key
464 $ago = min( $purge[self
::$PURGE_TIME] - $now, self
::$TINY_NEGATIVE );
465 // How long ago this value was invalidated by *any* known check key
466 $keyInfo['curTTL'] = min( $keyInfo['curTTL'], $ago );
469 $keyInfo[ 'lastCKPurge'] = $lastCKPurge;
471 if ( $value !== false ) {
472 $result[$key] = $value;
474 if ( $keyInfo['curTTL'] !== null ) {
475 $curTTLs[$key] = $keyInfo['curTTL'];
478 $infoByKey[$key] = ( $info === self
::PASS_BY_REF
)
480 : $keyInfo['asOf']; // b/c
490 * @param string[] $timeKeys List of prefixed time check keys
491 * @param mixed[] $wrappedValues
493 * @return array[] List of purge value arrays
495 private function processCheckKeys( array $timeKeys, array $wrappedValues, $now ) {
497 foreach ( $timeKeys as $timeKey ) {
498 $purge = isset( $wrappedValues[$timeKey] )
499 ?
$this->parsePurgeValue( $wrappedValues[$timeKey] )
501 if ( $purge === false ) {
502 // Key is not set or malformed; regenerate
503 $newVal = $this->makePurgeValue( $now, self
::HOLDOFF_TTL
);
504 $this->cache
->add( $timeKey, $newVal, self
::$CHECK_KEY_TTL );
505 $purge = $this->parsePurgeValue( $newVal );
507 $purgeValues[] = $purge;
514 * Set the value of a key in cache
516 * Simply calling this method when source data changes is not valid because
517 * the changes do not replicate to the other WAN sites. In that case, delete()
518 * should be used instead. This method is intended for use on cache misses.
520 * If the data was read from a snapshot-isolated transactions (e.g. the default
521 * REPEATABLE-READ in innoDB), use 'since' to avoid the following race condition:
523 * - b) T2 updates a row, calls delete(), and commits
524 * - c) The HOLDOFF_TTL passes, expiring the delete() tombstone
525 * - d) T1 reads the row and calls set() due to a cache miss
526 * - e) Stale value is stuck in cache
528 * Setting 'lag' and 'since' help avoids keys getting stuck in stale states.
530 * Be aware that this does not update the process cache for getWithSetCallback()
531 * callers. Keys accessed via that method are not generally meant to also be set
532 * using this primitive method.
534 * Do not use this method on versioned keys accessed via getWithSetCallback().
538 * $dbr = wfGetDB( DB_REPLICA );
539 * $setOpts = Database::getCacheSetOptions( $dbr );
540 * // Fetch the row from the DB
541 * $row = $dbr->selectRow( ... );
542 * $key = $cache->makeKey( 'building', $buildingId );
543 * $cache->set( $key, $row, $cache::TTL_DAY, $setOpts );
546 * @param string $key Cache key
547 * @param mixed $value
548 * @param int $ttl Seconds to live. Special values are:
549 * - WANObjectCache::TTL_INDEFINITE: Cache forever (default)
550 * - WANObjectCache::TTL_UNCACHEABLE: Do not cache (if the key exists, it is not deleted)
551 * @param array $opts Options map:
552 * - lag: Seconds of replica DB lag. Typically, this is either the replica DB lag
553 * before the data was read or, if applicable, the replica DB lag before
554 * the snapshot-isolated transaction the data was read from started.
555 * Use false to indicate that replication is not running.
557 * - since: UNIX timestamp of the data in $value. Typically, this is either
558 * the current time the data was read or (if applicable) the time when
559 * the snapshot-isolated transaction the data was read from started.
561 * - pending: Whether this data is possibly from an uncommitted write transaction.
562 * Generally, other threads should not see values from the future and
563 * they certainly should not see ones that ended up getting rolled back.
565 * - lockTSE: If excessive replication/snapshot lag is detected, then store the value
566 * with this TTL and flag it as stale. This is only useful if the reads for this key
567 * use getWithSetCallback() with "lockTSE" set. Note that if "staleTTL" is set
568 * then it will still add on to this TTL in the excessive lag scenario.
569 * Default: WANObjectCache::TSE_NONE
570 * - staleTTL: Seconds to keep the key around if it is stale. The get()/getMulti()
571 * methods return such stale values with a $curTTL of 0, and getWithSetCallback()
572 * will call the regeneration callback in such cases, passing in the old value
573 * and its as-of time to the callback. This is useful if adaptiveTTL() is used
574 * on the old value's as-of time when it is verified as still being correct.
575 * Default: WANObjectCache::STALE_TTL_NONE
576 * - creating: Optimize for the case where the key does not already exist.
578 * - version: Integer version number signifiying the format of the value.
580 * - walltime: How long the value took to generate in seconds. Default: 0.0
581 * @codingStandardsIgnoreStart
582 * @phan-param array{lag?:int,since?:int,pending?:bool,lockTSE?:int,staleTTL?:int,creating?:bool,version?:?string,walltime?:int|float} $opts
583 * @codingStandardsIgnoreEnd
584 * @note Options added in 1.28: staleTTL
585 * @note Options added in 1.33: creating
586 * @note Options added in 1.34: version, walltime
587 * @return bool Success
589 final public function set( $key, $value, $ttl = self
::TTL_INDEFINITE
, array $opts = [] ) {
590 $now = $this->getCurrentTime();
591 $lag = $opts['lag'] ??
0;
592 $age = isset( $opts['since'] ) ?
max( 0, $now - $opts['since'] ) : 0;
593 $pending = $opts['pending'] ??
false;
594 $lockTSE = $opts['lockTSE'] ?? self
::TSE_NONE
;
595 $staleTTL = $opts['staleTTL'] ?? self
::STALE_TTL_NONE
;
596 $creating = $opts['creating'] ??
false;
597 $version = $opts['version'] ??
null;
598 $walltime = $opts['walltime'] ??
0.0;
604 // Do not cache potentially uncommitted data as it might get rolled back
607 'Rejected set() for {cachekey} due to pending writes.',
608 [ 'cachekey' => $key ]
611 return true; // no-op the write for being unsafe
614 $logicalTTL = null; // logical TTL override
615 // Check if there's a risk of writing stale data after the purge tombstone expired
616 if ( $lag === false ||
( $lag +
$age ) > self
::MAX_READ_LAG
) {
617 // Case A: any long-running transaction
618 if ( $age > self
::MAX_READ_LAG
) {
619 if ( $lockTSE >= 0 ) {
620 // Store value as *almost* stale to avoid cache and mutex stampedes
621 $logicalTTL = self
::TTL_SECOND
;
623 'Lowered set() TTL for {cachekey} due to snapshot lag.',
624 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ]
628 'Rejected set() for {cachekey} due to snapshot lag.',
629 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ]
632 return true; // no-op the write for being unsafe
634 // Case B: high replication lag; lower TTL instead of ignoring all set()s
635 } elseif ( $lag === false ||
$lag > self
::MAX_READ_LAG
) {
636 if ( $lockTSE >= 0 ) {
637 $logicalTTL = min( $ttl ?
: INF
, self
::TTL_LAGGED
);
639 $ttl = min( $ttl ?
: INF
, self
::TTL_LAGGED
);
641 $this->logger
->warning(
642 'Lowered set() TTL for {cachekey} due to replication lag.',
643 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ]
645 // Case C: medium length request with medium replication lag
646 } elseif ( $lockTSE >= 0 ) {
647 // Store value as *almost* stale to avoid cache and mutex stampedes
648 $logicalTTL = self
::TTL_SECOND
;
650 'Lowered set() TTL for {cachekey} due to high read lag.',
651 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ]
655 'Rejected set() for {cachekey} due to high read lag.',
656 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ]
659 return true; // no-op the write for being unsafe
663 // Wrap that value with time/TTL/version metadata
664 $wrapped = $this->wrap( $value, $logicalTTL ?
: $ttl, $version, $now, $walltime );
665 $storeTTL = $ttl +
$staleTTL;
668 $ok = $this->cache
->add( self
::$VALUE_KEY_PREFIX . $key, $wrapped, $storeTTL );
670 $ok = $this->cache
->merge(
671 self
::$VALUE_KEY_PREFIX . $key,
672 function ( $cache, $key, $cWrapped ) use ( $wrapped ) {
673 // A string value means that it is a tombstone; do nothing in that case
674 return ( is_string( $cWrapped ) ) ?
false : $wrapped;
685 * Purge a key from all datacenters
687 * This should only be called when the underlying data (being cached)
688 * changes in a significant way. This deletes the key and starts a hold-off
689 * period where the key cannot be written to for a few seconds (HOLDOFF_TTL).
690 * This is done to avoid the following race condition:
691 * - a) Some DB data changes and delete() is called on a corresponding key
692 * - b) A request refills the key with a stale value from a lagged DB
693 * - c) The stale value is stuck there until the key is expired/evicted
695 * This is implemented by storing a special "tombstone" value at the cache
696 * key that this class recognizes; get() calls will return false for the key
697 * and any set() calls will refuse to replace tombstone values at the key.
698 * For this to always avoid stale value writes, the following must hold:
699 * - a) Replication lag is bounded to being less than HOLDOFF_TTL; or
700 * - b) If lag is higher, the DB will have gone into read-only mode already
702 * Note that set() can also be lag-aware and lower the TTL if it's high.
704 * Be aware that this does not clear the process cache. Even if it did, callbacks
705 * used by getWithSetCallback() might still return stale data in the case of either
706 * uncommitted or not-yet-replicated changes (callback generally use replica DBs).
708 * When using potentially long-running ACID transactions, a good pattern is
709 * to use a pre-commit hook to issue the delete. This means that immediately
710 * after commit, callers will see the tombstone in cache upon purge relay.
711 * It also avoids the following race condition:
712 * - a) T1 begins, changes a row, and calls delete()
713 * - b) The HOLDOFF_TTL passes, expiring the delete() tombstone
714 * - c) T2 starts, reads the row and calls set() due to a cache miss
715 * - d) T1 finally commits
716 * - e) Stale value is stuck in cache
720 * $dbw->startAtomic( __METHOD__ ); // start of request
721 * ... <execute some stuff> ...
722 * // Update the row in the DB
723 * $dbw->update( ... );
724 * $key = $cache->makeKey( 'homes', $homeId );
725 * // Purge the corresponding cache entry just before committing
726 * $dbw->onTransactionPreCommitOrIdle( function() use ( $cache, $key ) {
727 * $cache->delete( $key );
729 * ... <execute some stuff> ...
730 * $dbw->endAtomic( __METHOD__ ); // end of request
733 * The $ttl parameter can be used when purging values that have not actually changed
734 * recently. For example, a cleanup script to purge cache entries does not really need
735 * a hold-off period, so it can use HOLDOFF_TTL_NONE. Likewise for user-requested purge.
736 * Note that $ttl limits the effective range of 'lockTSE' for getWithSetCallback().
738 * If called twice on the same key, then the last hold-off TTL takes precedence. For
739 * idempotence, the $ttl should not vary for different delete() calls on the same key.
741 * @param string $key Cache key
742 * @param int $ttl Tombstone TTL; Default: WANObjectCache::HOLDOFF_TTL
743 * @return bool True if the item was purged or not found, false on failure
745 final public function delete( $key, $ttl = self
::HOLDOFF_TTL
) {
747 // Publish the purge to all datacenters
748 $ok = $this->relayDelete( self
::$VALUE_KEY_PREFIX . $key );
750 // Publish the purge to all datacenters
751 $ok = $this->relayPurge( self
::$VALUE_KEY_PREFIX . $key, $ttl, self
::HOLDOFF_TTL_NONE
);
754 $kClass = $this->determineKeyClassForStats( $key );
755 $this->stats
->increment( "wanobjectcache.$kClass.delete." . ( $ok ?
'ok' : 'error' ) );
761 * Fetch the value of a timestamp "check" key
763 * The key will be *initialized* to the current time if not set,
764 * so only call this method if this behavior is actually desired
766 * The timestamp can be used to check whether a cached value is valid.
767 * Callers should not assume that this returns the same timestamp in
768 * all datacenters due to relay delays.
770 * The level of staleness can roughly be estimated from this key, but
771 * if the key was evicted from cache, such calculations may show the
772 * time since expiry as ~0 seconds.
774 * Note that "check" keys won't collide with other regular keys.
777 * @return float UNIX timestamp
779 final public function getCheckKeyTime( $key ) {
780 return $this->getMultiCheckKeyTime( [ $key ] )[$key];
784 * Fetch the values of each timestamp "check" key
786 * This works like getCheckKeyTime() except it takes a list of keys
787 * and returns a map of timestamps instead of just that of one key
789 * This might be useful if both:
790 * - a) a class of entities each depend on hundreds of other entities
791 * - b) these other entities are depended upon by millions of entities
793 * The later entities can each use a "check" key to invalidate their dependee entities.
794 * However, it is expensive for the former entities to verify against all of the relevant
795 * "check" keys during each getWithSetCallback() call. A less expensive approach is to do
796 * these verifications only after a "time-till-verify" (TTV) has passed. This is a middle
797 * ground between using blind TTLs and using constant verification. The adaptiveTTL() method
798 * can be used to dynamically adjust the TTV. Also, the initial TTV can make use of the
799 * last-modified times of the dependant entities (either from the DB or the "check" keys).
803 * $value = $cache->getWithSetCallback(
804 * $cache->makeGlobalKey( 'wikibase-item', $id ),
805 * self::INITIAL_TTV, // initial time-till-verify
806 * function ( $oldValue, &$ttv, &$setOpts, $oldAsOf ) use ( $checkKeys, $cache ) {
807 * $now = microtime( true );
808 * // Use $oldValue if it passes max ultimate age and "check" key comparisons
810 * $oldAsOf > max( $cache->getMultiCheckKeyTime( $checkKeys ) ) &&
811 * ( $now - $oldValue['ctime'] ) <= self::MAX_CACHE_AGE
813 * // Increase time-till-verify by 50% of last time to reduce overhead
814 * $ttv = $cache->adaptiveTTL( $oldAsOf, self::MAX_TTV, self::MIN_TTV, 1.5 );
815 * // Unlike $oldAsOf, "ctime" is the ultimate age of the cached data
819 * $mtimes = []; // dependency last-modified times; passed by reference
820 * $value = [ 'data' => $this->fetchEntityData( $mtimes ), 'ctime' => $now ];
821 * // Guess time-till-change among the dependencies, e.g. 1/(total change rate)
822 * $ttc = 1 / array_sum( array_map(
823 * function ( $mtime ) use ( $now ) {
824 * return 1 / ( $mtime ? ( $now - $mtime ) : 900 );
828 * // The time-to-verify should not be overly pessimistic nor optimistic
829 * $ttv = min( max( $ttc, self::MIN_TTV ), self::MAX_TTV );
833 * [ 'staleTTL' => $cache::TTL_DAY ] // keep around to verify and re-save
837 * @see WANObjectCache::getCheckKeyTime()
838 * @see WANObjectCache::getWithSetCallback()
840 * @param string[] $keys
841 * @return float[] Map of (key => UNIX timestamp)
844 final public function getMultiCheckKeyTime( array $keys ) {
846 foreach ( $keys as $key ) {
847 $rawKeys[$key] = self
::$TIME_KEY_PREFIX . $key;
850 $rawValues = $this->cache
->getMulti( $rawKeys );
851 $rawValues +
= array_fill_keys( $rawKeys, false );
854 foreach ( $rawKeys as $key => $rawKey ) {
855 $purge = $this->parsePurgeValue( $rawValues[$rawKey] );
856 if ( $purge !== false ) {
857 $time = $purge[self
::$PURGE_TIME];
859 // Casting assures identical floats for the next getCheckKeyTime() calls
860 $now = (string)$this->getCurrentTime();
863 $this->makePurgeValue( $now, self
::HOLDOFF_TTL
),
869 $times[$key] = $time;
876 * Purge a "check" key from all datacenters, invalidating keys that use it
878 * This should only be called when the underlying data (being cached)
879 * changes in a significant way, and it is impractical to call delete()
880 * on all keys that should be changed. When get() is called on those
881 * keys, the relevant "check" keys must be supplied for this to work.
883 * The "check" key essentially represents a last-modified time of an entity.
884 * When the key is touched, the timestamp will be updated to the current time.
885 * Keys using the "check" key via get(), getMulti(), or getWithSetCallback() will
886 * be invalidated. This approach is useful if many keys depend on a single entity.
888 * The timestamp of the "check" key is treated as being HOLDOFF_TTL seconds in the
889 * future by get*() methods in order to avoid race conditions where keys are updated
890 * with stale values (e.g. from a lagged replica DB). A high TTL is set on the "check"
891 * key, making it possible to know the timestamp of the last change to the corresponding
892 * entities in most cases. This might use more cache space than resetCheckKey().
894 * When a few important keys get a large number of hits, a high cache time is usually
895 * desired as well as "lockTSE" logic. The resetCheckKey() method is less appropriate
896 * in such cases since the "time since expiry" cannot be inferred, causing any get()
897 * after the reset to treat the key as being "hot", resulting in more stale value usage.
899 * Note that "check" keys won't collide with other regular keys.
901 * @see WANObjectCache::get()
902 * @see WANObjectCache::getWithSetCallback()
903 * @see WANObjectCache::resetCheckKey()
905 * @param string $key Cache key
906 * @param int $holdoff HOLDOFF_TTL or HOLDOFF_TTL_NONE constant
907 * @return bool True if the item was purged or not found, false on failure
909 final public function touchCheckKey( $key, $holdoff = self
::HOLDOFF_TTL
) {
910 // Publish the purge to all datacenters
911 $ok = $this->relayPurge( self
::$TIME_KEY_PREFIX . $key, self
::$CHECK_KEY_TTL, $holdoff );
913 $kClass = $this->determineKeyClassForStats( $key );
914 $this->stats
->increment( "wanobjectcache.$kClass.ck_touch." . ( $ok ?
'ok' : 'error' ) );
920 * Delete a "check" key from all datacenters, invalidating keys that use it
922 * This is similar to touchCheckKey() in that keys using it via get(), getMulti(),
923 * or getWithSetCallback() will be invalidated. The differences are:
924 * - a) The "check" key will be deleted from all caches and lazily
925 * re-initialized when accessed (rather than set everywhere)
926 * - b) Thus, dependent keys will be known to be stale, but not
927 * for how long (they are treated as "just" purged), which
928 * effects any lockTSE logic in getWithSetCallback()
929 * - c) Since "check" keys are initialized only on the server the key hashes
930 * to, any temporary ejection of that server will cause the value to be
931 * seen as purged as a new server will initialize the "check" key.
933 * The advantage here is that the "check" keys, which have high TTLs, will only
934 * be created when a get*() method actually uses that key. This is better when
935 * a large number of "check" keys are invalided in a short period of time.
937 * Note that "check" keys won't collide with other regular keys.
939 * @see WANObjectCache::get()
940 * @see WANObjectCache::getWithSetCallback()
941 * @see WANObjectCache::touchCheckKey()
943 * @param string $key Cache key
944 * @return bool True if the item was purged or not found, false on failure
946 final public function resetCheckKey( $key ) {
947 // Publish the purge to all datacenters
948 $ok = $this->relayDelete( self
::$TIME_KEY_PREFIX . $key );
950 $kClass = $this->determineKeyClassForStats( $key );
951 $this->stats
->increment( "wanobjectcache.$kClass.ck_reset." . ( $ok ?
'ok' : 'error' ) );
957 * Method to fetch/regenerate cache keys
959 * On cache miss, the key will be set to the callback result via set()
960 * (unless the callback returns false) and that result will be returned.
961 * The arguments supplied to the callback are:
962 * - $oldValue : current cache value or false if not present
963 * - &$ttl : a reference to the TTL which can be altered
964 * - &$setOpts : a reference to options for set() which can be altered
965 * - $oldAsOf : generation UNIX timestamp of $oldValue or null if not present (since 1.28)
967 * It is strongly recommended to set the 'lag' and 'since' fields to avoid race conditions
968 * that can cause stale values to get stuck at keys. Usually, callbacks ignore the current
969 * value, but it can be used to maintain "most recent X" values that come from time or
970 * sequence based source data, provided that the "as of" id/time is tracked. Note that
971 * preemptive regeneration and $checkKeys can result in a non-false current value.
973 * Usage of $checkKeys is similar to get() and getMulti(). However, rather than the caller
974 * having to inspect a "current time left" variable (e.g. $curTTL, $curTTLs), a cache
975 * regeneration will automatically be triggered using the callback.
977 * The $ttl argument and "hotTTR" option (in $opts) use time-dependant randomization
978 * to avoid stampedes. Keys that are slow to regenerate and either heavily used
979 * or subject to explicit (unpredictable) purges, may need additional mechanisms.
980 * The simplest way to avoid stampedes for such keys is to use 'lockTSE' (in $opts).
981 * If explicit purges are needed, also:
982 * - a) Pass $key into $checkKeys
983 * - b) Use touchCheckKey( $key ) instead of delete( $key )
985 * Example usage (typical key):
987 * $catInfo = $cache->getWithSetCallback(
988 * // Key to store the cached value under
989 * $cache->makeKey( 'cat-attributes', $catId ),
990 * // Time-to-live (in seconds)
991 * $cache::TTL_MINUTE,
992 * // Function that derives the new key value
993 * function ( $oldValue, &$ttl, array &$setOpts ) {
994 * $dbr = wfGetDB( DB_REPLICA );
995 * // Account for any snapshot/replica DB lag
996 * $setOpts += Database::getCacheSetOptions( $dbr );
998 * return $dbr->selectRow( ... );
1003 * Example usage (key that is expensive and hot):
1005 * $catConfig = $cache->getWithSetCallback(
1006 * // Key to store the cached value under
1007 * $cache->makeKey( 'site-cat-config' ),
1008 * // Time-to-live (in seconds)
1010 * // Function that derives the new key value
1011 * function ( $oldValue, &$ttl, array &$setOpts ) {
1012 * $dbr = wfGetDB( DB_REPLICA );
1013 * // Account for any snapshot/replica DB lag
1014 * $setOpts += Database::getCacheSetOptions( $dbr );
1016 * return CatConfig::newFromRow( $dbr->selectRow( ... ) );
1019 * // Calling touchCheckKey() on this key invalidates the cache
1020 * 'checkKeys' => [ $cache->makeKey( 'site-cat-config' ) ],
1021 * // Try to only let one datacenter thread manage cache updates at a time
1023 * // Avoid querying cache servers multiple times in a web request
1024 * 'pcTTL' => $cache::TTL_PROC_LONG
1029 * Example usage (key with dynamic dependencies):
1031 * $catState = $cache->getWithSetCallback(
1032 * // Key to store the cached value under
1033 * $cache->makeKey( 'cat-state', $cat->getId() ),
1034 * // Time-to-live (seconds)
1036 * // Function that derives the new key value
1037 * function ( $oldValue, &$ttl, array &$setOpts ) {
1038 * // Determine new value from the DB
1039 * $dbr = wfGetDB( DB_REPLICA );
1040 * // Account for any snapshot/replica DB lag
1041 * $setOpts += Database::getCacheSetOptions( $dbr );
1043 * return CatState::newFromResults( $dbr->select( ... ) );
1046 * // The "check" keys that represent things the value depends on;
1047 * // Calling touchCheckKey() on any of them invalidates the cache
1049 * $cache->makeKey( 'sustenance-bowls', $cat->getRoomId() ),
1050 * $cache->makeKey( 'people-present', $cat->getHouseId() ),
1051 * $cache->makeKey( 'cat-laws', $cat->getCityId() ),
1057 * Example usage (key that is expensive with too many DB dependencies for "check keys"):
1059 * $catToys = $cache->getWithSetCallback(
1060 * // Key to store the cached value under
1061 * $cache->makeKey( 'cat-toys', $catId ),
1062 * // Time-to-live (seconds)
1064 * // Function that derives the new key value
1065 * function ( $oldValue, &$ttl, array &$setOpts ) {
1066 * // Determine new value from the DB
1067 * $dbr = wfGetDB( DB_REPLICA );
1068 * // Account for any snapshot/replica DB lag
1069 * $setOpts += Database::getCacheSetOptions( $dbr );
1071 * return CatToys::newFromResults( $dbr->select( ... ) );
1074 * // Get the highest timestamp of any of the cat's toys
1075 * 'touchedCallback' => function ( $value ) use ( $catId ) {
1076 * $dbr = wfGetDB( DB_REPLICA );
1077 * $ts = $dbr->selectField( 'cat_toys', 'MAX(ct_touched)', ... );
1079 * return wfTimestampOrNull( TS_UNIX, $ts );
1081 * // Avoid DB queries for repeated access
1082 * 'pcTTL' => $cache::TTL_PROC_SHORT
1087 * Example usage (hot key holding most recent 100 events):
1089 * $lastCatActions = $cache->getWithSetCallback(
1090 * // Key to store the cached value under
1091 * $cache->makeKey( 'cat-last-actions', 100 ),
1092 * // Time-to-live (in seconds)
1094 * // Function that derives the new key value
1095 * function ( $oldValue, &$ttl, array &$setOpts ) {
1096 * $dbr = wfGetDB( DB_REPLICA );
1097 * // Account for any snapshot/replica DB lag
1098 * $setOpts += Database::getCacheSetOptions( $dbr );
1100 * // Start off with the last cached list
1101 * $list = $oldValue ?: [];
1102 * // Fetch the last 100 relevant rows in descending order;
1103 * // only fetch rows newer than $list[0] to reduce scanning
1104 * $rows = iterator_to_array( $dbr->select( ... ) );
1105 * // Merge them and get the new "last 100" rows
1106 * return array_slice( array_merge( $new, $list ), 0, 100 );
1109 * // Try to only let one datacenter thread manage cache updates at a time
1111 * // Use a magic value when no cache value is ready rather than stampeding
1112 * 'busyValue' => 'computing'
1117 * Example usage (key holding an LRU subkey:value map; this can avoid flooding cache with
1118 * keys for an unlimited set of (constraint,situation) pairs, thereby avoiding elevated
1119 * cache evictions and wasted memory):
1121 * $catSituationTolerabilityCache = $this->cache->getWithSetCallback(
1122 * // Group by constraint ID/hash, cat family ID/hash, or something else useful
1123 * $this->cache->makeKey( 'cat-situation-tolerability-checks', $groupKey ),
1124 * WANObjectCache::TTL_DAY, // rarely used groups should fade away
1125 * // The $scenarioKey format is $constraintId:<ID/hash of $situation>
1126 * function ( $cacheMap ) use ( $scenarioKey, $constraintId, $situation ) {
1127 * $lruCache = MapCacheLRU::newFromArray( $cacheMap ?: [], self::CACHE_SIZE );
1128 * $result = $lruCache->get( $scenarioKey ); // triggers LRU bump if present
1129 * if ( $result === null || $this->isScenarioResultExpired( $result ) ) {
1130 * $result = $this->checkScenarioTolerability( $constraintId, $situation );
1131 * $lruCache->set( $scenarioKey, $result, 3 / 8 );
1133 * // Save the new LRU cache map and reset the map's TTL
1134 * return $lruCache->toArray();
1137 * // Once map is > 1 sec old, consider refreshing
1139 * // Update within 5 seconds after "ageNew" given a 1hz cache check rate
1141 * // Avoid querying cache servers multiple times in a request; this also means
1142 * // that a request can only alter the value of any given constraint key once
1143 * 'pcTTL' => WANObjectCache::TTL_PROC_LONG
1146 * $tolerability = isset( $catSituationTolerabilityCache[$scenarioKey] )
1147 * ? $catSituationTolerabilityCache[$scenarioKey]
1148 * : $this->checkScenarioTolerability( $constraintId, $situation );
1151 * @see WANObjectCache::get()
1152 * @see WANObjectCache::set()
1154 * @param string $key Cache key made from makeKey() or makeGlobalKey()
1155 * @param int $ttl Seconds to live for key updates. Special values are:
1156 * - WANObjectCache::TTL_INDEFINITE: Cache forever (subject to LRU-style evictions)
1157 * - WANObjectCache::TTL_UNCACHEABLE: Do not cache (if the key exists, it is not deleted)
1158 * @param callable $callback Value generation function
1159 * @param array $opts Options map:
1160 * - checkKeys: List of "check" keys. The key at $key will be seen as stale when either
1161 * touchCheckKey() or resetCheckKey() is called on any of the keys in this list. This
1162 * is useful if thousands or millions of keys depend on the same entity. The entity can
1163 * simply have its "check" key updated whenever the entity is modified.
1165 * - graceTTL: If the key is invalidated (by "checkKeys"/"touchedCallback") less than this
1166 * many seconds ago, consider reusing the stale value. The odds of a refresh becomes
1167 * more likely over time, becoming certain once the grace period is reached. This can
1168 * reduce traffic spikes when millions of keys are compared to the same "check" key and
1169 * touchCheckKey() or resetCheckKey() is called on that "check" key. This option is not
1170 * useful for avoiding traffic spikes in the case of the key simply expiring on account
1171 * of its TTL (use "lowTTL" instead).
1172 * Default: WANObjectCache::GRACE_TTL_NONE.
1173 * - lockTSE: If the key is tombstoned or invalidated (by "checkKeys"/"touchedCallback")
1174 * less than this many seconds ago, try to have a single thread handle cache regeneration
1175 * at any given time. Other threads will use stale values if possible. If, on miss,
1176 * the time since expiration is low, the assumption is that the key is hot and that a
1177 * stampede is worth avoiding. Note that if the key falls out of cache then concurrent
1178 * threads will all run the callback on cache miss until the value is saved in cache.
1179 * The only stampede protection in that case is from duplicate cache sets when the
1180 * callback takes longer than WANObjectCache::SET_DELAY_HIGH_MS milliseconds; consider
1181 * using "busyValue" if such stampedes are a problem. Note that the higher "lockTSE" is
1182 * set, the higher the worst-case staleness of returned values can be. Also note that
1183 * this option does not by itself handle the case of the key simply expiring on account
1184 * of its TTL, so make sure that "lowTTL" is not disabled when using this option. Avoid
1185 * combining this option with delete() as it can always cause a stampede due to their
1186 * being no stale value available until after a thread completes the callback.
1187 * Use WANObjectCache::TSE_NONE to disable this logic.
1188 * Default: WANObjectCache::TSE_NONE.
1189 * - busyValue: Specify a placeholder value to use when no value exists and another thread
1190 * is currently regenerating it. This assures that cache stampedes cannot happen if the
1191 * value falls out of cache. This also mitigates stampedes when value regeneration
1192 * becomes very slow (greater than $ttl/"lowTTL"). If this is a closure, then it will
1193 * be invoked to get the placeholder when needed.
1195 * - pcTTL: Process cache the value in this PHP instance for this many seconds. This avoids
1196 * network I/O when a key is read several times. This will not cache when the callback
1197 * returns false, however. Note that any purges will not be seen while process cached;
1198 * since the callback should use replica DBs and they may be lagged or have snapshot
1199 * isolation anyway, this should not typically matter.
1200 * Default: WANObjectCache::TTL_UNCACHEABLE.
1201 * - pcGroup: Process cache group to use instead of the primary one. If set, this must be
1202 * of the format ALPHANUMERIC_NAME:MAX_KEY_SIZE, e.g. "mydata:10". Use this for storing
1203 * large values, small yet numerous values, or some values with a high cost of eviction.
1204 * It is generally preferable to use a class constant when setting this value.
1205 * This has no effect unless pcTTL is used.
1206 * Default: WANObjectCache::PC_PRIMARY.
1207 * - version: Integer version number. This lets callers make breaking changes to the format
1208 * of cached values without causing problems for sites that use non-instantaneous code
1209 * deployments. Old and new code will recognize incompatible versions and purges from
1210 * both old and new code will been seen by each other. When this method encounters an
1211 * incompatibly versioned value at the provided key, a "variant key" will be used for
1212 * reading from and saving to cache. The variant key is specific to the key and version
1213 * number provided to this method. If the variant key value is older than that of the
1214 * provided key, or the provided key is non-existant, then the variant key will be seen
1215 * as non-existant. Therefore, delete() calls invalidate the provided key's variant keys.
1216 * The "checkKeys" and "touchedCallback" options still apply to variant keys as usual.
1217 * Avoid storing class objects, as this reduces compatibility (due to serialization).
1219 * - minAsOf: Reject values if they were generated before this UNIX timestamp.
1220 * This is useful if the source of a key is suspected of having possibly changed
1221 * recently, and the caller wants any such changes to be reflected.
1222 * Default: WANObjectCache::MIN_TIMESTAMP_NONE.
1223 * - hotTTR: Expected time-till-refresh (TTR) in seconds for keys that average ~1 hit per
1224 * second (e.g. 1Hz). Keys with a hit rate higher than 1Hz will refresh sooner than this
1225 * TTR and vise versa. Such refreshes won't happen until keys are "ageNew" seconds old.
1226 * This uses randomization to avoid triggering cache stampedes. The TTR is useful at
1227 * reducing the impact of missed cache purges, since the effect of a heavily referenced
1228 * key being stale is worse than that of a rarely referenced key. Unlike simply lowering
1229 * $ttl, seldomly used keys are largely unaffected by this option, which makes it
1230 * possible to have a high hit rate for the "long-tail" of less-used keys.
1231 * Default: WANObjectCache::HOT_TTR.
1232 * - lowTTL: Consider pre-emptive updates when the current TTL (seconds) of the key is less
1233 * than this. It becomes more likely over time, becoming certain once the key is expired.
1234 * This helps avoid cache stampedes that might be triggered due to the key expiring.
1235 * Default: WANObjectCache::LOW_TTL.
1236 * - ageNew: Consider popularity refreshes only once a key reaches this age in seconds.
1237 * Default: WANObjectCache::AGE_NEW.
1238 * - staleTTL: Seconds to keep the key around if it is stale. This means that on cache
1239 * miss the callback may get $oldValue/$oldAsOf values for keys that have already been
1240 * expired for this specified time. This is useful if adaptiveTTL() is used on the old
1241 * value's as-of time when it is verified as still being correct.
1242 * Default: WANObjectCache::STALE_TTL_NONE
1243 * - touchedCallback: A callback that takes the current value and returns a UNIX timestamp
1244 * indicating the last time a dynamic dependency changed. Null can be returned if there
1245 * are no relevant dependency changes to check. This can be used to check against things
1246 * like last-modified times of files or DB timestamp fields. This should generally not be
1247 * used for small and easily queried values in a DB if the callback itself ends up doing
1248 * a similarly expensive DB query to check a timestamp. Usages of this option makes the
1249 * most sense for values that are moderately to highly expensive to regenerate and easy
1250 * to query for dependency timestamps. The use of "pcTTL" reduces timestamp queries.
1252 * @codingStandardsIgnoreStart
1253 * @phan-param array{checkKeys?:string[],graceTTL?:int,lockTSE?:int,busyValue?:mixed,pcTTL?:int,pcGroup?:string,version?:int,minAsOf?:int,hotTTR?:int,lowTTL?:int,ageNew?:int,staleTTL?:int,touchedCallback?:callable} $opts
1254 * @codingStandardsIgnoreEnd
1255 * @return mixed Value found or written to the key
1256 * @note Options added in 1.28: version, busyValue, hotTTR, ageNew, pcGroup, minAsOf
1257 * @note Options added in 1.31: staleTTL, graceTTL
1258 * @note Options added in 1.33: touchedCallback
1259 * @note Callable type hints are not used to avoid class-autoloading
1261 final public function getWithSetCallback( $key, $ttl, $callback, array $opts = [] ) {
1262 $version = $opts['version'] ??
null;
1263 $pcTTL = $opts['pcTTL'] ?? self
::TTL_UNCACHEABLE
;
1264 $pCache = ( $pcTTL >= 0 )
1265 ?
$this->getProcessCache( $opts['pcGroup'] ?? self
::PC_PRIMARY
)
1268 // Use the process cache if requested as long as no outer cache callback is running.
1269 // Nested callback process cache use is not lag-safe with regard to HOLDOFF_TTL since
1270 // process cached values are more lagged than persistent ones as they are not purged.
1271 if ( $pCache && $this->callbackDepth
== 0 ) {
1272 $cached = $pCache->get( $this->getProcessCacheKey( $key, $version ), INF
, false );
1273 if ( $cached !== false ) {
1278 $res = $this->fetchOrRegenerate( $key, $ttl, $callback, $opts );
1279 list( $value, $valueVersion, $curAsOf ) = $res;
1280 if ( $valueVersion !== $version ) {
1281 // Current value has a different version; use the variant key for this version.
1282 // Regenerate the variant value if it is not newer than the main value at $key
1283 // so that purges to the main key propagate to the variant value.
1284 list( $value ) = $this->fetchOrRegenerate(
1285 $this->makeGlobalKey( 'WANCache-key-variant', md5( $key ), $version ),
1288 [ 'version' => null, 'minAsOf' => $curAsOf ] +
$opts
1292 // Update the process cache if enabled
1293 if ( $pCache && $value !== false ) {
1294 $pCache->set( $this->getProcessCacheKey( $key, $version ), $value );
1301 * Do the actual I/O for getWithSetCallback() when needed
1303 * @see WANObjectCache::getWithSetCallback()
1305 * @param string $key
1307 * @param callable $callback
1308 * @param array $opts Options map for getWithSetCallback()
1309 * @return array Ordered list of the following:
1310 * - Cached or regenerated value
1311 * - Cached or regenerated value version number or null if not versioned
1312 * - Timestamp of the current cached value at the key or null if there is no value
1313 * @note Callable type hints are not used to avoid class-autoloading
1314 * @suppress PhanTypeArraySuspicious
1316 private function fetchOrRegenerate( $key, $ttl, $callback, array $opts ) {
1317 $checkKeys = $opts['checkKeys'] ??
[];
1318 $graceTTL = $opts['graceTTL'] ?? self
::GRACE_TTL_NONE
;
1319 $minAsOf = $opts['minAsOf'] ?? self
::MIN_TIMESTAMP_NONE
;
1320 $hotTTR = $opts['hotTTR'] ?? self
::HOT_TTR
;
1321 $lowTTL = $opts['lowTTL'] ??
min( self
::LOW_TTL
, $ttl );
1322 $ageNew = $opts['ageNew'] ?? self
::AGE_NEW
;
1323 $touchedCb = $opts['touchedCallback'] ??
null;
1324 $initialTime = $this->getCurrentTime();
1326 $kClass = $this->determineKeyClassForStats( $key );
1328 // Get the current key value and its metadata
1329 $curTTL = self
::PASS_BY_REF
;
1330 $curInfo = self
::PASS_BY_REF
; /** @var array $curInfo */
1331 $curValue = $this->get( $key, $curTTL, $checkKeys, $curInfo );
1332 // Apply any $touchedCb invalidation timestamp to get the "last purge timestamp"
1333 list( $curTTL, $LPT ) = $this->resolveCTL( $curValue, $curTTL, $curInfo, $touchedCb );
1334 // Use the cached value if it exists and is not due for synchronous regeneration
1336 $this->isValid( $curValue, $curInfo['asOf'], $minAsOf ) &&
1337 $this->isAliveOrInGracePeriod( $curTTL, $graceTTL )
1339 $preemptiveRefresh = (
1340 $this->worthRefreshExpiring( $curTTL, $lowTTL ) ||
1341 $this->worthRefreshPopular( $curInfo['asOf'], $ageNew, $hotTTR, $initialTime )
1343 if ( !$preemptiveRefresh ) {
1344 $this->stats
->increment( "wanobjectcache.$kClass.hit.good" );
1346 return [ $curValue, $curInfo['version'], $curInfo['asOf'] ];
1347 } elseif ( $this->scheduleAsyncRefresh( $key, $ttl, $callback, $opts ) ) {
1348 $this->stats
->increment( "wanobjectcache.$kClass.hit.refresh" );
1350 return [ $curValue, $curInfo['version'], $curInfo['asOf'] ];
1354 // Determine if there is stale or volatile cached value that is still usable
1355 $isKeyTombstoned = ( $curInfo['tombAsOf'] !== null );
1356 if ( $isKeyTombstoned ) {
1357 // Key is write-holed; use the (volatile) interim key as an alternative
1358 list( $possValue, $possInfo ) = $this->getInterimValue( $key, $minAsOf );
1359 // Update the "last purge time" since the $touchedCb timestamp depends on $value
1360 $LPT = $this->resolveTouched( $possValue, $LPT, $touchedCb );
1362 $possValue = $curValue;
1363 $possInfo = $curInfo;
1366 // Avoid overhead from callback runs, regeneration locks, and cache sets during
1367 // hold-off periods for the key by reusing very recently generated cached values
1369 $this->isValid( $possValue, $possInfo['asOf'], $minAsOf, $LPT ) &&
1370 $this->isVolatileValueAgeNegligible( $initialTime - $possInfo['asOf'] )
1372 $this->stats
->increment( "wanobjectcache.$kClass.hit.volatile" );
1374 return [ $possValue, $possInfo['version'], $curInfo['asOf'] ];
1377 $lockTSE = $opts['lockTSE'] ?? self
::TSE_NONE
;
1378 $busyValue = $opts['busyValue'] ??
null;
1379 $staleTTL = $opts['staleTTL'] ?? self
::STALE_TTL_NONE
;
1380 $version = $opts['version'] ??
null;
1382 // Determine whether one thread per datacenter should handle regeneration at a time
1383 $useRegenerationLock =
1384 // Note that since tombstones no-op set(), $lockTSE and $curTTL cannot be used to
1385 // deduce the key hotness because |$curTTL| will always keep increasing until the
1386 // tombstone expires or is overwritten by a new tombstone. Also, even if $lockTSE
1387 // is not set, constant regeneration of a key for the tombstone lifetime might be
1388 // very expensive. Assume tombstoned keys are possibly hot in order to reduce
1389 // the risk of high regeneration load after the delete() method is called.
1391 // Assume a key is hot if requested soon ($lockTSE seconds) after invalidation.
1392 // This avoids stampedes when timestamps from $checkKeys/$touchedCb bump.
1393 ( $curTTL !== null && $curTTL <= 0 && abs( $curTTL ) <= $lockTSE ) ||
1394 // Assume a key is hot if there is no value and a busy fallback is given.
1395 // This avoids stampedes on eviction or preemptive regeneration taking too long.
1396 ( $busyValue !== null && $possValue === false );
1398 // If a regeneration lock is required, threads that do not get the lock will try to use
1399 // the stale value, the interim value, or the $busyValue placeholder, in that order. If
1400 // none of those are set then all threads will bypass the lock and regenerate the value.
1401 $hasLock = $useRegenerationLock && $this->claimStampedeLock( $key );
1402 if ( $useRegenerationLock && !$hasLock ) {
1403 if ( $this->isValid( $possValue, $possInfo['asOf'], $minAsOf ) ) {
1404 $this->stats
->increment( "wanobjectcache.$kClass.hit.stale" );
1406 return [ $possValue, $possInfo['version'], $curInfo['asOf'] ];
1407 } elseif ( $busyValue !== null ) {
1408 $miss = is_infinite( $minAsOf ) ?
'renew' : 'miss';
1409 $this->stats
->increment( "wanobjectcache.$kClass.$miss.busy" );
1411 return [ $this->resolveBusyValue( $busyValue ), $version, $curInfo['asOf'] ];
1415 // Generate the new value given any prior value with a matching version
1417 $preCallbackTime = $this->getCurrentTime();
1418 ++
$this->callbackDepth
;
1421 ( $curInfo['version'] === $version ) ?
$curValue : false,
1424 ( $curInfo['version'] === $version ) ?
$curInfo['asOf'] : null
1427 --$this->callbackDepth
;
1429 $postCallbackTime = $this->getCurrentTime();
1431 // How long it took to fetch, validate, and generate the value
1432 $elapsed = max( $postCallbackTime - $initialTime, 0.0 );
1434 // Attempt to save the newly generated value if applicable
1436 // Callback yielded a cacheable value
1437 ( $value !== false && $ttl >= 0 ) &&
1438 // Current thread was not raced out of a regeneration lock or key is tombstoned
1439 ( !$useRegenerationLock ||
$hasLock ||
$isKeyTombstoned ) &&
1440 // Key does not appear to be undergoing a set() stampede
1441 $this->checkAndSetCooloff( $key, $kClass, $elapsed, $lockTSE, $hasLock )
1443 // How long it took to generate the value
1444 $walltime = max( $postCallbackTime - $preCallbackTime, 0.0 );
1445 $this->stats
->timing( "wanobjectcache.$kClass.regen_walltime", 1e3
* $walltime );
1446 // If the key is write-holed then use the (volatile) interim key as an alternative
1447 if ( $isKeyTombstoned ) {
1448 $this->setInterimValue( $key, $value, $lockTSE, $version, $walltime );
1451 'since' => $setOpts['since'] ??
$preCallbackTime,
1452 'version' => $version,
1453 'staleTTL' => $staleTTL,
1454 'lockTSE' => $lockTSE, // informs lag vs performance trade-offs
1455 'creating' => ( $curValue === false ), // optimization
1456 'walltime' => $walltime
1458 $this->set( $key, $value, $ttl, $finalSetOpts );
1462 $this->yieldStampedeLock( $key, $hasLock );
1464 $miss = is_infinite( $minAsOf ) ?
'renew' : 'miss';
1465 $this->stats
->increment( "wanobjectcache.$kClass.$miss.compute" );
1467 return [ $value, $version, $curInfo['asOf'] ];
1471 * @param string $key
1472 * @return bool Success
1474 private function claimStampedeLock( $key ) {
1475 // Note that locking is not bypassed due to I/O errors; this avoids stampedes
1476 return $this->cache
->add( self
::$MUTEX_KEY_PREFIX . $key, 1, self
::$LOCK_TTL );
1480 * @param string $key
1481 * @param bool $hasLock
1483 private function yieldStampedeLock( $key, $hasLock ) {
1485 // The backend might be a mcrouter proxy set to broadcast DELETE to *all* the local
1486 // datacenter cache servers via OperationSelectorRoute (for increased consistency).
1487 // Since that would be excessive for these locks, use TOUCH to expire the key.
1488 $this->cache
->changeTTL( self
::$MUTEX_KEY_PREFIX . $key, $this->getCurrentTime() - 60 );
1493 * @param float $age Age of volatile/interim key in seconds
1494 * @return bool Whether the age of a volatile value is negligible
1496 private function isVolatileValueAgeNegligible( $age ) {
1497 return ( $age < mt_rand( self
::$RECENT_SET_LOW_MS, self
::$RECENT_SET_HIGH_MS ) / 1e3
);
1501 * @param string $key
1502 * @param string $kClass
1503 * @param float $elapsed Seconds spent regenerating the value
1504 * @param float $lockTSE
1505 * @param bool $hasLock
1506 * @return bool Whether it is OK to proceed with a key set operation
1508 private function checkAndSetCooloff( $key, $kClass, $elapsed, $lockTSE, $hasLock ) {
1509 $this->stats
->timing( "wanobjectcache.$kClass.regen_set_delay", 1e3
* $elapsed );
1511 // If $lockTSE is set, the lock was bypassed because there was no stale/interim value,
1512 // and $elapsed indicates that regeration is slow, then there is a risk of set()
1513 // stampedes with large blobs. With a typical scale-out infrastructure, CPU and query
1514 // load from $callback invocations is distributed among appservers and replica DBs,
1515 // but cache operations for a given key route to a single cache server (e.g. striped
1516 // consistent hashing).
1517 if ( $lockTSE < 0 ||
$hasLock ) {
1518 return true; // either not a priori hot or thread has the lock
1519 } elseif ( $elapsed <= self
::$SET_DELAY_HIGH_MS * 1e3
) {
1520 return true; // not enough time for threads to pile up
1523 $this->cache
->clearLastError();
1525 !$this->cache
->add( self
::$COOLOFF_KEY_PREFIX . $key, 1, self
::$COOLOFF_TTL ) &&
1526 // Don't treat failures due to I/O errors as the key being in cooloff
1527 $this->cache
->getLastError() === BagOStuff
::ERR_NONE
1529 $this->stats
->increment( "wanobjectcache.$kClass.cooloff_bounce" );
1538 * @param mixed $value
1539 * @param float|null $curTTL
1540 * @param array $curInfo
1541 * @param callable|null $touchedCallback
1542 * @return array (current time left or null, UNIX timestamp of last purge or null)
1543 * @note Callable type hints are not used to avoid class-autoloading
1545 private function resolveCTL( $value, $curTTL, $curInfo, $touchedCallback ) {
1546 if ( $touchedCallback === null ||
$value === false ) {
1547 return [ $curTTL, max( $curInfo['tombAsOf'], $curInfo['lastCKPurge'] ) ];
1550 $touched = $touchedCallback( $value );
1551 if ( $touched !== null && $touched >= $curInfo['asOf'] ) {
1552 $curTTL = min( $curTTL, self
::$TINY_NEGATIVE, $curInfo['asOf'] - $touched );
1555 return [ $curTTL, max( $curInfo['tombAsOf'], $curInfo['lastCKPurge'], $touched ) ];
1559 * @param mixed $value
1560 * @param float|null $lastPurge
1561 * @param callable|null $touchedCallback
1562 * @return float|null UNIX timestamp of last purge or null
1563 * @note Callable type hints are not used to avoid class-autoloading
1565 private function resolveTouched( $value, $lastPurge, $touchedCallback ) {
1566 return ( $touchedCallback === null ||
$value === false )
1567 ?
$lastPurge // nothing to derive the "touched timestamp" from
1568 : max( $touchedCallback( $value ), $lastPurge );
1572 * @param string $key
1573 * @param float $minAsOf Minimum acceptable "as of" timestamp
1574 * @return array (cached value or false, cache key metadata map)
1576 private function getInterimValue( $key, $minAsOf ) {
1577 $now = $this->getCurrentTime();
1579 if ( $this->useInterimHoldOffCaching
) {
1580 $wrapped = $this->cache
->get( self
::$INTERIM_KEY_PREFIX . $key );
1582 list( $value, $keyInfo ) = $this->unwrap( $wrapped, $now );
1583 if ( $this->isValid( $value, $keyInfo['asOf'], $minAsOf ) ) {
1584 return [ $value, $keyInfo ];
1588 return $this->unwrap( false, $now );
1592 * @param string $key
1593 * @param mixed $value
1595 * @param int|null $version Value version number
1596 * @param float $walltime How long it took to generate the value in seconds
1598 private function setInterimValue( $key, $value, $ttl, $version, $walltime ) {
1599 $ttl = max( self
::$INTERIM_KEY_TTL, (int)$ttl );
1601 $wrapped = $this->wrap( $value, $ttl, $version, $this->getCurrentTime(), $walltime );
1602 $this->cache
->merge(
1603 self
::$INTERIM_KEY_PREFIX . $key,
1604 function () use ( $wrapped ) {
1613 * @param mixed $busyValue
1616 private function resolveBusyValue( $busyValue ) {
1617 return ( $busyValue instanceof Closure
) ?
$busyValue() : $busyValue;
1621 * Method to fetch multiple cache keys at once with regeneration
1623 * This works the same as getWithSetCallback() except:
1624 * - a) The $keys argument expects the result of WANObjectCache::makeMultiKeys()
1625 * - b) The $callback argument expects a callback taking the following arguments:
1626 * - $id: ID of an entity to query
1627 * - $oldValue : the prior cache value or false if none was present
1628 * - &$ttl : a reference to the new value TTL in seconds
1629 * - &$setOpts : a reference to options for set() which can be altered
1630 * - $oldAsOf : generation UNIX timestamp of $oldValue or null if not present
1631 * Aside from the additional $id argument, the other arguments function the same
1632 * way they do in getWithSetCallback().
1633 * - c) The return value is a map of (cache key => value) in the order of $keyedIds
1635 * @see WANObjectCache::getWithSetCallback()
1636 * @see WANObjectCache::getMultiWithUnionSetCallback()
1640 * $rows = $cache->getMultiWithSetCallback(
1641 * // Map of cache keys to entity IDs
1642 * $cache->makeMultiKeys(
1643 * $this->fileVersionIds(),
1644 * function ( $id ) use ( $cache ) {
1645 * return $cache->makeKey( 'file-version', $id );
1648 * // Time-to-live (in seconds)
1650 * // Function that derives the new key value
1651 * function ( $id, $oldValue, &$ttl, array &$setOpts ) {
1652 * $dbr = wfGetDB( DB_REPLICA );
1653 * // Account for any snapshot/replica DB lag
1654 * $setOpts += Database::getCacheSetOptions( $dbr );
1656 * // Load the row for this file
1657 * $queryInfo = File::getQueryInfo();
1658 * $row = $dbr->selectRow(
1659 * $queryInfo['tables'],
1660 * $queryInfo['fields'],
1664 * $queryInfo['joins']
1667 * return $row ? (array)$row : false;
1670 * // Process cache for 30 seconds
1672 * // Use a dedicated 500 item cache (initialized on-the-fly)
1673 * 'pcGroup' => 'file-versions:500'
1676 * $files = array_map( [ __CLASS__, 'newFromRow' ], $rows );
1679 * @param ArrayIterator $keyedIds Result of WANObjectCache::makeMultiKeys()
1680 * @param int $ttl Seconds to live for key updates
1681 * @param callable $callback Callback the yields entity regeneration callbacks
1682 * @param array $opts Options map
1683 * @return mixed[] Map of (cache key => value) in the same order as $keyedIds
1686 final public function getMultiWithSetCallback(
1687 ArrayIterator
$keyedIds, $ttl, callable
$callback, array $opts = []
1689 // Load required keys into process cache in one go
1690 $this->warmupCache
= $this->getRawKeysForWarmup(
1691 $this->getNonProcessCachedMultiKeys( $keyedIds, $opts ),
1692 $opts['checkKeys'] ??
[]
1694 $this->warmupKeyMisses
= 0;
1696 // Wrap $callback to match the getWithSetCallback() format while passing $id to $callback
1697 $id = null; // current entity ID
1698 $func = function ( $oldValue, &$ttl, &$setOpts, $oldAsOf ) use ( $callback, &$id ) {
1699 return $callback( $id, $oldValue, $ttl, $setOpts, $oldAsOf );
1703 foreach ( $keyedIds as $key => $id ) { // preserve order
1704 $values[$key] = $this->getWithSetCallback( $key, $ttl, $func, $opts );
1707 $this->warmupCache
= [];
1713 * Method to fetch/regenerate multiple cache keys at once
1715 * This works the same as getWithSetCallback() except:
1716 * - a) The $keys argument expects the result of WANObjectCache::makeMultiKeys()
1717 * - b) The $callback argument expects a callback returning a map of (ID => new value)
1718 * for all entity IDs in $ids and it takes the following arguments:
1719 * - $ids: a list of entity IDs that require cache regeneration
1720 * - &$ttls: a reference to the (entity ID => new TTL) map
1721 * - &$setOpts: a reference to options for set() which can be altered
1722 * - c) The return value is a map of (cache key => value) in the order of $keyedIds
1723 * - d) The "lockTSE" and "busyValue" options are ignored
1725 * @see WANObjectCache::getWithSetCallback()
1726 * @see WANObjectCache::getMultiWithSetCallback()
1730 * $rows = $cache->getMultiWithUnionSetCallback(
1731 * // Map of cache keys to entity IDs
1732 * $cache->makeMultiKeys(
1733 * $this->fileVersionIds(),
1734 * function ( $id ) use ( $cache ) {
1735 * return $cache->makeKey( 'file-version', $id );
1738 * // Time-to-live (in seconds)
1740 * // Function that derives the new key value
1741 * function ( array $ids, array &$ttls, array &$setOpts ) {
1742 * $dbr = wfGetDB( DB_REPLICA );
1743 * // Account for any snapshot/replica DB lag
1744 * $setOpts += Database::getCacheSetOptions( $dbr );
1746 * // Load the rows for these files
1748 * $queryInfo = File::getQueryInfo();
1749 * $res = $dbr->select(
1750 * $queryInfo['tables'],
1751 * $queryInfo['fields'],
1755 * $queryInfo['joins']
1757 * foreach ( $res as $row ) {
1758 * $rows[$row->id] = $row;
1759 * $mtime = wfTimestamp( TS_UNIX, $row->timestamp );
1760 * $ttls[$row->id] = $this->adaptiveTTL( $mtime, $ttls[$row->id] );
1767 * $files = array_map( [ __CLASS__, 'newFromRow' ], $rows );
1770 * @param ArrayIterator $keyedIds Result of WANObjectCache::makeMultiKeys()
1771 * @param int $ttl Seconds to live for key updates
1772 * @param callable $callback Callback the yields entity regeneration callbacks
1773 * @param array $opts Options map
1774 * @return mixed[] Map of (cache key => value) in the same order as $keyedIds
1777 final public function getMultiWithUnionSetCallback(
1778 ArrayIterator
$keyedIds, $ttl, callable
$callback, array $opts = []
1780 $checkKeys = $opts['checkKeys'] ??
[];
1781 unset( $opts['lockTSE'] ); // incompatible
1782 unset( $opts['busyValue'] ); // incompatible
1784 // Load required keys into process cache in one go
1785 $keysByIdGet = $this->getNonProcessCachedMultiKeys( $keyedIds, $opts );
1786 $this->warmupCache
= $this->getRawKeysForWarmup( $keysByIdGet, $checkKeys );
1787 $this->warmupKeyMisses
= 0;
1789 // IDs of entities known to be in need of regeneration
1792 // Find out which keys are missing/deleted/stale
1795 $curByKey = $this->getMulti( $keysByIdGet, $curTTLs, $checkKeys, $asOfs );
1796 foreach ( $keysByIdGet as $id => $key ) {
1797 if ( !array_key_exists( $key, $curByKey ) ||
$curTTLs[$key] < 0 ) {
1802 // Run the callback to populate the regeneration value map for all required IDs
1804 $newTTLsById = array_fill_keys( $idsRegen, $ttl );
1805 $newValsById = $idsRegen ?
$callback( $idsRegen, $newTTLsById, $newSetOpts ) : [];
1807 // Wrap $callback to match the getWithSetCallback() format while passing $id to $callback
1808 $id = null; // current entity ID
1809 $func = function ( $oldValue, &$ttl, &$setOpts, $oldAsOf )
1810 use ( $callback, &$id, $newValsById, $newTTLsById, $newSetOpts )
1812 if ( array_key_exists( $id, $newValsById ) ) {
1813 // Value was already regerated as expected, so use the value in $newValsById
1814 $newValue = $newValsById[$id];
1815 $ttl = $newTTLsById[$id];
1816 $setOpts = $newSetOpts;
1818 // Pre-emptive/popularity refresh and version mismatch cases are not detected
1819 // above and thus $newValsById has no entry. Run $callback on this single entity.
1820 $ttls = [ $id => $ttl ];
1821 $newValue = $callback( [ $id ], $ttls, $setOpts )[$id];
1828 // Run the cache-aside logic using warmupCache instead of persistent cache queries
1830 foreach ( $keyedIds as $key => $id ) { // preserve order
1831 $values[$key] = $this->getWithSetCallback( $key, $ttl, $func, $opts );
1834 $this->warmupCache
= [];
1840 * Set a key to soon expire in the local cluster if it pre-dates $purgeTimestamp
1842 * This sets stale keys' time-to-live at HOLDOFF_TTL seconds, which both avoids
1843 * broadcasting in mcrouter setups and also avoids races with new tombstones.
1845 * @param string $key Cache key
1846 * @param int $purgeTimestamp UNIX timestamp of purge
1847 * @param bool &$isStale Whether the key is stale
1848 * @return bool Success
1851 final public function reap( $key, $purgeTimestamp, &$isStale = false ) {
1852 $minAsOf = $purgeTimestamp + self
::HOLDOFF_TTL
;
1853 $wrapped = $this->cache
->get( self
::$VALUE_KEY_PREFIX . $key );
1854 if ( is_array( $wrapped ) && $wrapped[self
::$FLD_TIME] < $minAsOf ) {
1856 $this->logger
->warning( "Reaping stale value key '$key'." );
1857 $ttlReap = self
::HOLDOFF_TTL
; // avoids races with tombstone creation
1858 $ok = $this->cache
->changeTTL( self
::$VALUE_KEY_PREFIX . $key, $ttlReap );
1860 $this->logger
->error( "Could not complete reap of key '$key'." );
1872 * Set a "check" key to soon expire in the local cluster if it pre-dates $purgeTimestamp
1874 * @param string $key Cache key
1875 * @param int $purgeTimestamp UNIX timestamp of purge
1876 * @param bool &$isStale Whether the key is stale
1877 * @return bool Success
1880 final public function reapCheckKey( $key, $purgeTimestamp, &$isStale = false ) {
1881 $purge = $this->parsePurgeValue( $this->cache
->get( self
::$TIME_KEY_PREFIX . $key ) );
1882 if ( $purge && $purge[self
::$PURGE_TIME] < $purgeTimestamp ) {
1884 $this->logger
->warning( "Reaping stale check key '$key'." );
1885 $ok = $this->cache
->changeTTL( self
::$TIME_KEY_PREFIX . $key, self
::TTL_SECOND
);
1887 $this->logger
->error( "Could not complete reap of check key '$key'." );
1899 * @see BagOStuff::makeKey()
1900 * @param string $class Key class
1901 * @param string ...$components Key components (starting with a key collection name)
1902 * @return string Colon-delimited list of $keyspace followed by escaped components
1905 public function makeKey( $class, ...$components ) {
1906 return $this->cache
->makeKey( ...func_get_args() );
1910 * @see BagOStuff::makeGlobalKey()
1911 * @param string $class Key class
1912 * @param string ...$components Key components (starting with a key collection name)
1913 * @return string Colon-delimited list of $keyspace followed by escaped components
1916 public function makeGlobalKey( $class, ...$components ) {
1917 return $this->cache
->makeGlobalKey( ...func_get_args() );
1921 * Hash a possibly long string into a suitable component for makeKey()/makeGlobalKey()
1923 * @param string $component A raw component used in building a cache key
1924 * @return string 64 character HMAC using a stable secret for public collision resistance
1927 public function hash256( $component ) {
1928 return hash_hmac( 'sha256', $component, $this->secret
);
1932 * Get an iterator of (cache key => entity ID) for a list of entity IDs
1934 * The callback takes an ID string and returns a key via makeKey()/makeGlobalKey().
1935 * There should be no network nor filesystem I/O used in the callback. The entity
1936 * ID/key mapping must be 1:1 or an exception will be thrown. If hashing is needed,
1937 * then use the hash256() method.
1939 * Example usage for the default keyspace:
1941 * $keyedIds = $cache->makeMultiKeys(
1943 * function ( $module ) use ( $cache ) {
1944 * return $cache->makeKey( 'module-info', $module );
1949 * Example usage for mixed default and global keyspace:
1951 * $keyedIds = $cache->makeMultiKeys(
1953 * function ( $filter ) use ( $cache ) {
1954 * return ( strpos( $filter, 'central:' ) === 0 )
1955 * ? $cache->makeGlobalKey( 'regex-filter', $filter )
1956 * : $cache->makeKey( 'regex-filter', $filter )
1961 * Example usage with hashing:
1963 * $keyedIds = $cache->makeMultiKeys(
1965 * function ( $url ) use ( $cache ) {
1966 * return $cache->makeKey( 'url-info', $cache->hash256( $url ) );
1971 * @see WANObjectCache::makeKey()
1972 * @see WANObjectCache::makeGlobalKey()
1973 * @see WANObjectCache::hash256()
1975 * @param string[]|int[] $ids List of entity IDs
1976 * @param callable $keyCallback Function returning makeKey()/makeGlobalKey() on the input ID
1977 * @return ArrayIterator Iterator of (cache key => ID); order of $ids is preserved
1978 * @throws UnexpectedValueException
1981 final public function makeMultiKeys( array $ids, $keyCallback ) {
1983 foreach ( $ids as $id ) {
1984 // Discourage triggering of automatic makeKey() hashing in some backends
1985 if ( strlen( $id ) > 64 ) {
1986 $this->logger
->warning( __METHOD__
. ": long ID '$id'; use hash256()" );
1988 $key = $keyCallback( $id, $this );
1989 // Edge case: ignore key collisions due to duplicate $ids like "42" and 42
1990 if ( !isset( $idByKey[$key] ) ) {
1991 $idByKey[$key] = $id;
1992 } elseif ( (string)$id !== (string)$idByKey[$key] ) {
1993 throw new UnexpectedValueException(
1994 "Cache key collision; IDs ('$id','{$idByKey[$key]}') map to '$key'"
1999 return new ArrayIterator( $idByKey );
2003 * Get an (ID => value) map from (i) a non-unique list of entity IDs, and (ii) the list
2004 * of corresponding entity values by first appearance of each ID in the entity ID list
2006 * For use with getMultiWithSetCallback() and getMultiWithUnionSetCallback().
2008 * *Only* use this method if the entity ID/key mapping is trivially 1:1 without exception.
2009 * Key generation method must utitilize the *full* entity ID in the key (not a hash of it).
2013 * $poems = $cache->getMultiWithSetCallback(
2014 * $cache->makeMultiKeys(
2016 * function ( $uuid ) use ( $cache ) {
2017 * return $cache->makeKey( 'poem', $uuid );
2021 * function ( $uuid ) use ( $url ) {
2022 * return $this->http->run( [ 'method' => 'GET', 'url' => "$url/$uuid" ] );
2025 * $poemsByUUID = $cache->multiRemap( $uuids, $poems );
2028 * @see WANObjectCache::makeMultiKeys()
2029 * @see WANObjectCache::getMultiWithSetCallback()
2030 * @see WANObjectCache::getMultiWithUnionSetCallback()
2032 * @param string[]|int[] $ids Entity ID list makeMultiKeys()
2033 * @param mixed[] $res Result of getMultiWithSetCallback()/getMultiWithUnionSetCallback()
2034 * @return mixed[] Map of (ID => value); order of $ids is preserved
2037 final public function multiRemap( array $ids, array $res ) {
2038 if ( count( $ids ) !== count( $res ) ) {
2039 // If makeMultiKeys() is called on a list of non-unique IDs, then the resulting
2040 // ArrayIterator will have less entries due to "first appearance" de-duplication
2041 $ids = array_keys( array_flip( $ids ) );
2042 if ( count( $ids ) !== count( $res ) ) {
2043 throw new UnexpectedValueException( "Multi-key result does not match ID list" );
2047 return array_combine( $ids, $res );
2051 * Get the "last error" registered; clearLastError() should be called manually
2052 * @return int ERR_* class constant for the "last error" registry
2054 final public function getLastError() {
2055 $code = $this->cache
->getLastError();
2057 case BagOStuff
::ERR_NONE
:
2058 return self
::ERR_NONE
;
2059 case BagOStuff
::ERR_NO_RESPONSE
:
2060 return self
::ERR_NO_RESPONSE
;
2061 case BagOStuff
::ERR_UNREACHABLE
:
2062 return self
::ERR_UNREACHABLE
;
2064 return self
::ERR_UNEXPECTED
;
2069 * Clear the "last error" registry
2071 final public function clearLastError() {
2072 $this->cache
->clearLastError();
2076 * Clear the in-process caches; useful for testing
2080 public function clearProcessCache() {
2081 $this->processCaches
= [];
2085 * Enable or disable the use of brief caching for tombstoned keys
2087 * When a key is purged via delete(), there normally is a period where caching
2088 * is hold-off limited to an extremely short time. This method will disable that
2089 * caching, forcing the callback to run for any of:
2090 * - WANObjectCache::getWithSetCallback()
2091 * - WANObjectCache::getMultiWithSetCallback()
2092 * - WANObjectCache::getMultiWithUnionSetCallback()
2094 * This is useful when both:
2095 * - a) the database used by the callback is known to be up-to-date enough
2096 * for some particular purpose (e.g. replica DB has applied transaction X)
2097 * - b) the caller needs to exploit that fact, and therefore needs to avoid the
2098 * use of inherently volatile and possibly stale interim keys
2100 * @see WANObjectCache::delete()
2101 * @param bool $enabled Whether to enable interim caching
2104 final public function useInterimHoldOffCaching( $enabled ) {
2105 $this->useInterimHoldOffCaching
= $enabled;
2109 * @param int $flag ATTR_* class constant
2110 * @return int QOS_* class constant
2113 public function getQoS( $flag ) {
2114 return $this->cache
->getQoS( $flag );
2118 * Get a TTL that is higher for objects that have not changed recently
2120 * This is useful for keys that get explicit purges and DB or purge relay
2121 * lag is a potential concern (especially how it interacts with CDN cache)
2125 * // Last-modified time of page
2126 * $mtime = wfTimestamp( TS_UNIX, $page->getTimestamp() );
2127 * // Get adjusted TTL. If $mtime is 3600 seconds ago and $minTTL/$factor left at
2128 * // defaults, then $ttl is 3600 * .2 = 720. If $minTTL was greater than 720, then
2129 * // $ttl would be $minTTL. If $maxTTL was smaller than 720, $ttl would be $maxTTL.
2130 * $ttl = $cache->adaptiveTTL( $mtime, $cache::TTL_DAY );
2133 * Another use case is when there are no applicable "last modified" fields in the DB,
2134 * and there are too many dependencies for explicit purges to be viable, and the rate of
2135 * change to relevant content is unstable, and it is highly valued to have the cached value
2136 * be as up-to-date as possible.
2140 * $query = "<some complex query>";
2141 * $idListFromComplexQuery = $cache->getWithSetCallback(
2142 * $cache->makeKey( 'complex-graph-query', $hashOfQuery ),
2143 * GraphQueryClass::STARTING_TTL,
2144 * function ( $oldValue, &$ttl, array &$setOpts, $oldAsOf ) use ( $query, $cache ) {
2145 * $gdb = $this->getReplicaGraphDbConnection();
2146 * // Account for any snapshot/replica DB lag
2147 * $setOpts += GraphDatabase::getCacheSetOptions( $gdb );
2149 * $newList = iterator_to_array( $gdb->query( $query ) );
2150 * sort( $newList, SORT_NUMERIC ); // normalize
2152 * $minTTL = GraphQueryClass::MIN_TTL;
2153 * $maxTTL = GraphQueryClass::MAX_TTL;
2154 * if ( $oldValue !== false ) {
2155 * // Note that $oldAsOf is the last time this callback ran
2156 * $ttl = ( $newList === $oldValue )
2157 * // No change: cache for 150% of the age of $oldValue
2158 * ? $cache->adaptiveTTL( $oldAsOf, $maxTTL, $minTTL, 1.5 )
2159 * // Changed: cache for 50% of the age of $oldValue
2160 * : $cache->adaptiveTTL( $oldAsOf, $maxTTL, $minTTL, .5 );
2166 * // Keep stale values around for doing comparisons for TTL calculations.
2167 * // High values improve long-tail keys hit-rates, though might waste space.
2168 * 'staleTTL' => GraphQueryClass::GRACE_TTL
2173 * @param int|float $mtime UNIX timestamp
2174 * @param int $maxTTL Maximum TTL (seconds)
2175 * @param int $minTTL Minimum TTL (seconds); Default: 30
2176 * @param float $factor Value in the range (0,1); Default: .2
2177 * @return int Adaptive TTL
2180 public function adaptiveTTL( $mtime, $maxTTL, $minTTL = 30, $factor = 0.2 ) {
2181 if ( is_float( $mtime ) ||
ctype_digit( $mtime ) ) {
2182 $mtime = (int)$mtime; // handle fractional seconds and string integers
2185 if ( !is_int( $mtime ) ||
$mtime <= 0 ) {
2186 return $minTTL; // no last-modified time provided
2189 $age = $this->getCurrentTime() - $mtime;
2191 return (int)min( $maxTTL, max( $minTTL, $factor * $age ) );
2195 * @return int Number of warmup key cache misses last round
2198 final public function getWarmupKeyMisses() {
2199 return $this->warmupKeyMisses
;
2203 * Do the actual async bus purge of a key
2205 * This must set the key to "PURGED:<UNIX timestamp>:<holdoff>"
2207 * @param string $key Cache key
2208 * @param int $ttl Seconds to keep the tombstone around
2209 * @param int $holdoff HOLDOFF_* constant controlling how long to ignore sets for this key
2210 * @return bool Success
2212 protected function relayPurge( $key, $ttl, $holdoff ) {
2213 if ( $this->mcrouterAware
) {
2214 // See https://github.com/facebook/mcrouter/wiki/Multi-cluster-broadcast-setup
2215 // Wildcards select all matching routes, e.g. the WAN cluster on all DCs
2216 $ok = $this->cache
->set(
2217 "/*/{$this->cluster}/{$key}",
2218 $this->makePurgeValue( $this->getCurrentTime(), $holdoff ),
2222 // Some other proxy handles broadcasting or there is only one datacenter
2223 $ok = $this->cache
->set(
2225 $this->makePurgeValue( $this->getCurrentTime(), $holdoff ),
2234 * Do the actual async bus delete of a key
2236 * @param string $key Cache key
2237 * @return bool Success
2239 protected function relayDelete( $key ) {
2240 if ( $this->mcrouterAware
) {
2241 // See https://github.com/facebook/mcrouter/wiki/Multi-cluster-broadcast-setup
2242 // Wildcards select all matching routes, e.g. the WAN cluster on all DCs
2243 $ok = $this->cache
->delete( "/*/{$this->cluster}/{$key}" );
2245 // Some other proxy handles broadcasting or there is only one datacenter
2246 $ok = $this->cache
->delete( $key );
2253 * @param string $key
2254 * @param int $ttl Seconds to live
2255 * @param callable $callback
2256 * @param array $opts
2257 * @return bool Success
2258 * @note Callable type hints are not used to avoid class-autoloading
2260 private function scheduleAsyncRefresh( $key, $ttl, $callback, $opts ) {
2261 if ( !$this->asyncHandler
) {
2264 // Update the cache value later, such during post-send of an HTTP request
2265 $func = $this->asyncHandler
;
2266 $func( function () use ( $key, $ttl, $callback, $opts ) {
2267 $opts['minAsOf'] = INF
; // force a refresh
2268 $this->fetchOrRegenerate( $key, $ttl, $callback, $opts );
2275 * Check if a key is fresh or in the grace window and thus due for randomized reuse
2277 * If $curTTL > 0 (e.g. not expired) this returns true. Otherwise, the chance of returning
2278 * true decrease steadily from 100% to 0% as the |$curTTL| moves from 0 to $graceTTL seconds.
2279 * This handles widely varying levels of cache access traffic.
2281 * If $curTTL <= -$graceTTL (e.g. already expired), then this returns false.
2283 * @param float $curTTL Approximate TTL left on the key if present
2284 * @param int $graceTTL Consider using stale values if $curTTL is greater than this
2287 private function isAliveOrInGracePeriod( $curTTL, $graceTTL ) {
2288 if ( $curTTL > 0 ) {
2290 } elseif ( $graceTTL <= 0 ) {
2294 $ageStale = abs( $curTTL ); // seconds of staleness
2295 $curGTTL = ( $graceTTL - $ageStale ); // current grace-time-to-live
2296 if ( $curGTTL <= 0 ) {
2297 return false; // already out of grace period
2300 // Chance of using a stale value is the complement of the chance of refreshing it
2301 return !$this->worthRefreshExpiring( $curGTTL, $graceTTL );
2305 * Check if a key is nearing expiration and thus due for randomized regeneration
2307 * This returns false if $curTTL >= $lowTTL. Otherwise, the chance of returning true
2308 * increases steadily from 0% to 100% as the $curTTL moves from $lowTTL to 0 seconds.
2309 * This handles widely varying levels of cache access traffic.
2311 * If $curTTL <= 0 (e.g. already expired), then this returns false.
2313 * @param float $curTTL Approximate TTL left on the key if present
2314 * @param float $lowTTL Consider a refresh when $curTTL is less than this
2317 protected function worthRefreshExpiring( $curTTL, $lowTTL ) {
2318 if ( $lowTTL <= 0 ) {
2320 } elseif ( $curTTL >= $lowTTL ) {
2322 } elseif ( $curTTL <= 0 ) {
2326 $chance = ( 1 - $curTTL / $lowTTL );
2328 // @phan-suppress-next-line PhanTypeMismatchArgumentInternal
2329 return mt_rand( 1, 1e9
) <= 1e9
* $chance;
2333 * Check if a key is due for randomized regeneration due to its popularity
2335 * This is used so that popular keys can preemptively refresh themselves for higher
2336 * consistency (especially in the case of purge loss/delay). Unpopular keys can remain
2337 * in cache with their high nominal TTL. This means popular keys keep good consistency,
2338 * whether the data changes frequently or not, and long-tail keys get to stay in cache
2339 * and get hits too. Similar to worthRefreshExpiring(), randomization is used.
2341 * @param float $asOf UNIX timestamp of the value
2342 * @param int $ageNew Age of key when this might recommend refreshing (seconds)
2343 * @param int $timeTillRefresh Age of key when it should be refreshed if popular (seconds)
2344 * @param float $now The current UNIX timestamp
2347 protected function worthRefreshPopular( $asOf, $ageNew, $timeTillRefresh, $now ) {
2348 if ( $ageNew < 0 ||
$timeTillRefresh <= 0 ) {
2352 $age = $now - $asOf;
2353 $timeOld = $age - $ageNew;
2354 if ( $timeOld <= 0 ) {
2358 $popularHitsPerSec = 1;
2359 // Lifecycle is: new, ramp-up refresh chance, full refresh chance.
2360 // Note that the "expected # of refreshes" for the ramp-up time range is half
2361 // of what it would be if P(refresh) was at its full value during that time range.
2362 $refreshWindowSec = max( $timeTillRefresh - $ageNew - self
::$RAMPUP_TTL / 2, 1 );
2363 // P(refresh) * (# hits in $refreshWindowSec) = (expected # of refreshes)
2364 // P(refresh) * ($refreshWindowSec * $popularHitsPerSec) = 1 (by definition)
2365 // P(refresh) = 1/($refreshWindowSec * $popularHitsPerSec)
2366 $chance = 1 / ( $popularHitsPerSec * $refreshWindowSec );
2368 // Ramp up $chance from 0 to its nominal value over RAMPUP_TTL seconds to avoid stampedes
2369 $chance *= ( $timeOld <= self
::$RAMPUP_TTL ) ?
$timeOld / self
::$RAMPUP_TTL : 1;
2371 // @phan-suppress-next-line PhanTypeMismatchArgumentInternal
2372 return mt_rand( 1, 1e9
) <= 1e9
* $chance;
2376 * Check if $value is not false, versioned (if needed), and not older than $minTime (if set)
2378 * @param array|bool $value
2379 * @param float $asOf The time $value was generated
2380 * @param float $minAsOf Minimum acceptable "as of" timestamp
2381 * @param float|null $purgeTime The last time the value was invalidated
2384 protected function isValid( $value, $asOf, $minAsOf, $purgeTime = null ) {
2385 // Avoid reading any key not generated after the latest delete() or touch
2386 $safeMinAsOf = max( $minAsOf, $purgeTime + self
::$TINY_POSTIVE );
2388 if ( $value === false ) {
2390 } elseif ( $safeMinAsOf > 0 && $asOf < $minAsOf ) {
2398 * @param mixed $value
2399 * @param int $ttl Seconds to live or zero for "indefinite"
2400 * @param int|null $version Value version number or null if not versioned
2401 * @param float $now Unix Current timestamp just before calling set()
2402 * @param float $walltime How long it took to generate the value in seconds
2405 private function wrap( $value, $ttl, $version, $now, $walltime ) {
2406 // Returns keys in ascending integer order for PHP7 array packing:
2407 // https://nikic.github.io/2014/12/22/PHPs-new-hashtable-implementation.html
2409 self
::$FLD_FORMAT_VERSION => self
::$VERSION,
2410 self
::$FLD_VALUE => $value,
2411 self
::$FLD_TTL => $ttl,
2412 self
::$FLD_TIME => $now
2414 if ( $version !== null ) {
2415 $wrapped[self
::$FLD_VALUE_VERSION] = $version;
2417 if ( $walltime >= self
::$GENERATION_SLOW_SEC ) {
2418 $wrapped[self
::$FLD_GENERATION_TIME] = $walltime;
2425 * @param array|string|bool $wrapped The entry at a cache key
2426 * @param float $now Unix Current timestamp (preferrably pre-query)
2427 * @return array (value or false if absent/tombstoned/malformed, value metadata map).
2428 * The cache key metadata includes the following metadata:
2429 * - asOf: UNIX timestamp of the value or null if there is no value
2430 * - curTTL: remaining time-to-live (negative if tombstoned) or null if there is no value
2431 * - version: value version number or null if the if there is no value
2432 * - tombAsOf: UNIX timestamp of the tombstone or null if there is no tombstone
2433 * @phan-return array{0:mixed,1:array{asOf:?mixed,curTTL:?int|float,version:?mixed,tombAsOf:?mixed}}
2435 private function unwrap( $wrapped, $now ) {
2437 $info = [ 'asOf' => null, 'curTTL' => null, 'version' => null, 'tombAsOf' => null ];
2439 if ( is_array( $wrapped ) ) {
2440 // Entry expected to be a cached value; validate it
2442 ( $wrapped[self
::$FLD_FORMAT_VERSION] ??
null ) === self
::$VERSION &&
2443 $wrapped[self
::$FLD_TIME] >= $this->epoch
2445 if ( $wrapped[self
::$FLD_TTL] > 0 ) {
2446 // Get the approximate time left on the key
2447 $age = $now - $wrapped[self
::$FLD_TIME];
2448 $curTTL = max( $wrapped[self
::$FLD_TTL] - $age, 0.0 );
2450 // Key had no TTL, so the time left is unbounded
2453 $value = $wrapped[self
::$FLD_VALUE];
2454 $info['version'] = $wrapped[self
::$FLD_VALUE_VERSION] ??
null;
2455 $info['asOf'] = $wrapped[self
::$FLD_TIME];
2456 $info['curTTL'] = $curTTL;
2459 // Entry expected to be a tombstone; parse it
2460 $purge = $this->parsePurgeValue( $wrapped );
2461 if ( $purge !== false ) {
2462 // Tombstoned keys should always have a negative current $ttl
2463 $info['curTTL'] = min( $purge[self
::$PURGE_TIME] - $now, self
::$TINY_NEGATIVE );
2464 $info['tombAsOf'] = $purge[self
::$PURGE_TIME];
2468 return [ $value, $info ];
2472 * @param string[] $keys
2473 * @param string $prefix
2474 * @return string[] Prefix keys; the order of $keys is preserved
2476 protected static function prefixCacheKeys( array $keys, $prefix ) {
2478 foreach ( $keys as $key ) {
2479 $res[] = $prefix . $key;
2486 * @param string $key String of the format <scope>:<class>[:<class or variable>]...
2487 * @return string A collection name to describe this class of key
2489 private function determineKeyClassForStats( $key ) {
2490 $parts = explode( ':', $key, 3 );
2492 return $parts[1] ??
$parts[0]; // sanity
2496 * @param string|array|bool $value Possible string of the form "PURGED:<timestamp>:<holdoff>"
2497 * @return array|bool Array containing a UNIX timestamp (float) and holdoff period (integer),
2498 * or false if value isn't a valid purge value
2500 private function parsePurgeValue( $value ) {
2501 if ( !is_string( $value ) ) {
2505 $segments = explode( ':', $value, 3 );
2507 !isset( $segments[0] ) ||
2508 !isset( $segments[1] ) ||
2509 "{$segments[0]}:" !== self
::$PURGE_VAL_PREFIX
2514 if ( !isset( $segments[2] ) ) {
2515 // Back-compat with old purge values without holdoff
2516 $segments[2] = self
::HOLDOFF_TTL
;
2519 if ( $segments[1] < $this->epoch
) {
2520 // Values this old are ignored
2525 self
::$PURGE_TIME => (float)$segments[1],
2526 self
::$PURGE_HOLDOFF => (int)$segments[2],
2531 * @param float $timestamp
2532 * @param int $holdoff In seconds
2533 * @return string Wrapped purge value
2535 private function makePurgeValue( $timestamp, $holdoff ) {
2536 return self
::$PURGE_VAL_PREFIX . (float)$timestamp . ':' . (int)$holdoff;
2540 * @param string $group
2541 * @return MapCacheLRU
2543 private function getProcessCache( $group ) {
2544 if ( !isset( $this->processCaches
[$group] ) ) {
2545 list( , $size ) = explode( ':', $group );
2546 $this->processCaches
[$group] = new MapCacheLRU( (int)$size );
2549 return $this->processCaches
[$group];
2553 * @param string $key
2554 * @param int $version
2557 private function getProcessCacheKey( $key, $version ) {
2558 return $key . ' ' . (int)$version;
2562 * @param ArrayIterator $keys
2563 * @param array $opts
2564 * @return string[] Map of (ID => cache key)
2566 private function getNonProcessCachedMultiKeys( ArrayIterator
$keys, array $opts ) {
2567 $pcTTL = $opts['pcTTL'] ?? self
::TTL_UNCACHEABLE
;
2570 if ( $pcTTL > 0 && $this->callbackDepth
== 0 ) {
2571 $version = $opts['version'] ??
null;
2572 $pCache = $this->getProcessCache( $opts['pcGroup'] ?? self
::PC_PRIMARY
);
2573 foreach ( $keys as $key => $id ) {
2574 if ( !$pCache->has( $this->getProcessCacheKey( $key, $version ), $pcTTL ) ) {
2575 $keysMissing[$id] = $key;
2580 return $keysMissing;
2584 * @param string[] $keys
2585 * @param string[]|string[][] $checkKeys
2586 * @return string[] List of cache keys
2588 private function getRawKeysForWarmup( array $keys, array $checkKeys ) {
2594 // Get all the value keys to fetch...
2595 foreach ( $keys as $key ) {
2596 $keysWarmUp[] = self
::$VALUE_KEY_PREFIX . $key;
2598 // Get all the check keys to fetch...
2599 foreach ( $checkKeys as $i => $checkKeyOrKeys ) {
2600 if ( is_int( $i ) ) {
2601 // Single check key that applies to all value keys
2602 $keysWarmUp[] = self
::$TIME_KEY_PREFIX . $checkKeyOrKeys;
2604 // List of check keys that apply to value key $i
2605 $keysWarmUp = array_merge(
2607 self
::prefixCacheKeys( $checkKeyOrKeys, self
::$TIME_KEY_PREFIX )
2612 $warmupCache = $this->cache
->getMulti( $keysWarmUp );
2613 $warmupCache +
= array_fill_keys( $keysWarmUp, false );
2615 return $warmupCache;
2619 * @return float UNIX timestamp
2620 * @codeCoverageIgnore
2622 protected function getCurrentTime() {
2623 if ( $this->wallClockOverride
) {
2624 return $this->wallClockOverride
;
2627 $clockTime = (float)time(); // call this first
2628 // microtime() uses an initial gettimeofday() call added to usage clocks.
2629 // This can severely drift from time() and the microtime() value of other threads
2630 // due to undercounting of the amount of time elapsed. Instead of seeing the current
2631 // time as being in the past, use the value of time(). This avoids setting cache values
2632 // that will immediately be seen as expired and possibly cause stampedes.
2633 return max( microtime( true ), $clockTime );
2637 * @param float|null &$time Mock UNIX timestamp for testing
2638 * @codeCoverageIgnore
2640 public function setMockTime( &$time ) {
2641 $this->wallClockOverride
=& $time;
2642 $this->cache
->setMockTime( $time );