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 If outdated, the value is recomputed.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 three supported ways to set up broadcasted operations:
79 * - A) Configure the 'purge' EventRelayer to point to a valid PubSub endpoint
80 * that has subscribed listeners on the cache servers applying the cache updates.
81 * - B) Omit the 'purge' EventRelayer parameter and set up mcrouter as the underlying cache
82 * backend, using a memcached BagOStuff class for the 'cache' parameter. The 'region'
83 * and 'cluster' parameters must be provided and 'mcrouterAware' must be set to `true`.
84 * Configure mcrouter as follows:
85 * - 1) Use Route Prefixing based on region (datacenter) and cache cluster.
86 * See https://github.com/facebook/mcrouter/wiki/Routing-Prefix and
87 * https://github.com/facebook/mcrouter/wiki/Multi-cluster-broadcast-setup.
88 * - 2) To increase the consistency of delete() and touchCheckKey() during cache
89 * server membership changes, you can use the OperationSelectorRoute to
90 * configure 'set' and 'delete' operations to go to all servers in the cache
91 * cluster, instead of just one server determined by hashing.
92 * See https://github.com/facebook/mcrouter/wiki/List-of-Route-Handles.
93 * - C) Omit the 'purge' EventRelayer parameter and set up dynomite as cache middleware
94 * between the web servers and either memcached or redis. This will broadcast all
95 * key setting operations, not just purges, which can be useful for cache warming.
96 * Writes are eventually consistent via the Dynamo replication model.
97 * See https://github.com/Netflix/dynomite.
99 * Broadcasted operations like delete() and touchCheckKey() are done asynchronously
100 * in all datacenters this way, though the local one should likely be near immediate.
102 * This means that callers in all datacenters may see older values for however many
103 * milliseconds that the purge took to reach that datacenter. As with any cache, this
104 * should not be relied on for cases where reads are used to determine writes to source
105 * (e.g. non-cache) data stores, except when reading immutable data.
107 * All values are wrapped in metadata arrays. Keys use a "WANCache:" prefix
108 * to avoid collisions with keys that are not wrapped as metadata arrays. The
109 * prefixes are as follows:
110 * - a) "WANCache:v" : used for regular value keys
111 * - b) "WANCache:i" : used for temporarily storing values of tombstoned keys
112 * - c) "WANCache:t" : used for storing timestamp "check" keys
113 * - d) "WANCache:m" : used for temporary mutex keys to avoid cache stampedes
118 class WANObjectCache
implements IExpiringStore
, LoggerAwareInterface
{
119 /** @var BagOStuff The local datacenter cache */
121 /** @var MapCacheLRU[] Map of group PHP instance caches */
122 protected $processCaches = [];
123 /** @var string Purge channel name */
124 protected $purgeChannel;
125 /** @var EventRelayer Bus that handles purge broadcasts */
126 protected $purgeRelayer;
127 /** @bar bool Whether to use mcrouter key prefixing for routing */
128 protected $mcrouterAware;
129 /** @var string Physical region for mcrouter use */
131 /** @var string Cache cluster name for mcrouter use */
133 /** @var LoggerInterface */
135 /** @var StatsdDataFactoryInterface */
137 /** @var bool Whether to use "interim" caching while keys are tombstoned */
138 protected $useInterimHoldOffCaching = true;
139 /** @var callable|null Function that takes a WAN cache callback and runs it later */
140 protected $asyncHandler;
141 /** @var float Unix timestamp of the oldest possible valid values */
144 /** @var int ERR_* constant for the "last error" registry */
145 protected $lastRelayError = self
::ERR_NONE
;
147 /** @var int Callback stack depth for getWithSetCallback() */
148 private $callbackDepth = 0;
149 /** @var mixed[] Temporary warm-up cache */
150 private $warmupCache = [];
151 /** @var int Key fetched */
152 private $warmupKeyMisses = 0;
154 /** @var float|null */
155 private $wallClockOverride;
157 /** Max time expected to pass between delete() and DB commit finishing */
158 const MAX_COMMIT_DELAY
= 3;
159 /** Max replication+snapshot lag before applying TTL_LAGGED or disallowing set() */
160 const MAX_READ_LAG
= 7;
161 /** Seconds to tombstone keys on delete() */
162 const HOLDOFF_TTL
= 11; // MAX_COMMIT_DELAY + MAX_READ_LAG + 1
164 /** Seconds to keep dependency purge keys around */
165 const CHECK_KEY_TTL
= self
::TTL_YEAR
;
166 /** Seconds to keep interim value keys for tombstoned keys around */
167 const INTERIM_KEY_TTL
= 1;
169 /** Seconds to keep lock keys around */
171 /** Seconds to no-op key set() calls to avoid large blob I/O stampedes */
172 const COOLOFF_TTL
= 1;
173 /** Default remaining TTL at which to consider pre-emptive regeneration */
176 /** Never consider performing "popularity" refreshes until a key reaches this age */
178 /** The time length of the "popularity" refresh window for hot keys */
180 /** Hits/second for a refresh to be expected within the "popularity" window */
181 const HIT_RATE_HIGH
= 1;
182 /** Seconds to ramp up to the "popularity" refresh chance after a key is no longer new */
183 const RAMPUP_TTL
= 30;
185 /** Idiom for getWithSetCallback() callbacks to avoid calling set() */
186 const TTL_UNCACHEABLE
= -1;
187 /** Idiom for getWithSetCallback() callbacks to 'lockTSE' logic */
189 /** Max TTL to store keys when a data sourced is lagged */
190 const TTL_LAGGED
= 30;
191 /** Idiom for delete() for "no hold-off" */
192 const HOLDOFF_NONE
= 0;
193 /** Idiom for set()/getWithSetCallback() for "do not augment the storage medium TTL" */
194 const STALE_TTL_NONE
= 0;
195 /** Idiom for set()/getWithSetCallback() for "no post-expired grace period" */
196 const GRACE_TTL_NONE
= 0;
198 /** Idiom for getWithSetCallback() for "no minimum required as-of timestamp" */
199 const MIN_TIMESTAMP_NONE
= 0.0;
201 /** Tiny negative float to use when CTL comes up >= 0 due to clock skew */
202 const TINY_NEGATIVE
= -0.000001;
204 /** Seconds of delay after get() where set() storms are a consideration with 'lockTSE' */
205 const SET_DELAY_HIGH_SEC
= 0.1;
207 /** Cache format version number */
210 const FLD_VERSION
= 0; // key to cache version number
211 const FLD_VALUE
= 1; // key to the cached value
212 const FLD_TTL
= 2; // key to the original TTL
213 const FLD_TIME
= 3; // key to the cache time
214 const FLD_FLAGS
= 4; // key to the flags bitfield
215 const FLD_HOLDOFF
= 5; // key to any hold-off TTL
217 /** @var int Treat this value as expired-on-arrival */
220 const ERR_NONE
= 0; // no error
221 const ERR_NO_RESPONSE
= 1; // no response
222 const ERR_UNREACHABLE
= 2; // can't connect
223 const ERR_UNEXPECTED
= 3; // response gave some error
224 const ERR_RELAY
= 4; // relay broadcast failed
226 const VALUE_KEY_PREFIX
= 'WANCache:v:';
227 const INTERIM_KEY_PREFIX
= 'WANCache:i:';
228 const TIME_KEY_PREFIX
= 'WANCache:t:';
229 const MUTEX_KEY_PREFIX
= 'WANCache:m:';
230 const COOLOFF_KEY_PREFIX
= 'WANCache:c:';
232 const PURGE_VAL_PREFIX
= 'PURGED:';
234 const VFLD_DATA
= 'WOC:d'; // key to the value of versioned data
235 const VFLD_VERSION
= 'WOC:v'; // key to the version of the value present
237 const PC_PRIMARY
= 'primary:1000'; // process cache name and max key count
239 const DEFAULT_PURGE_CHANNEL
= 'wancache-purge';
242 * @param array $params
243 * - cache : BagOStuff object for a persistent cache
244 * - channels : Map of (action => channel string). Actions include "purge".
245 * - relayers : Map of (action => EventRelayer object). Actions include "purge".
246 * - logger : LoggerInterface object
247 * - stats : LoggerInterface object
248 * - asyncHandler : A function that takes a callback and runs it later. If supplied,
249 * whenever a preemptive refresh would be triggered in getWithSetCallback(), the
250 * current cache value is still used instead. However, the async-handler function
251 * receives a WAN cache callback that, when run, will execute the value generation
252 * callback supplied by the getWithSetCallback() caller. The result will be saved
253 * as normal. The handler is expected to call the WAN cache callback at an opportune
254 * time (e.g. HTTP post-send), though generally within a few 100ms. [optional]
255 * - region: the current physical region. This is required when using mcrouter as the
256 * backing store proxy. [optional]
257 * - cluster: name of the cache cluster used by this WAN cache. The name must be the
258 * same in all datacenters; the ("region","cluster") tuple is what distinguishes
259 * the counterpart cache clusters among all the datacenter. The contents of
260 * https://github.com/facebook/mcrouter/wiki/Config-Files give background on this.
261 * This is required when using mcrouter as the backing store proxy. [optional]
262 * - mcrouterAware: set as true if mcrouter is the backing store proxy and mcrouter
263 * is configured to interpret /<region>/<cluster>/ key prefixes as routes. This
264 * requires that "region" and "cluster" are both set above. [optional]
265 * - epoch: lowest UNIX timestamp a value/tombstone must have to be valid. [optional]
267 public function __construct( array $params ) {
268 $this->cache
= $params['cache'];
269 $this->purgeChannel
= $params['channels']['purge'] ?? self
::DEFAULT_PURGE_CHANNEL
;
270 $this->purgeRelayer
= $params['relayers']['purge'] ??
new EventRelayerNull( [] );
271 $this->region
= $params['region'] ??
'main';
272 $this->cluster
= $params['cluster'] ??
'wan-main';
273 $this->mcrouterAware
= !empty( $params['mcrouterAware'] );
274 $this->epoch
= $params['epoch'] ??
1.0;
276 $this->setLogger( $params['logger'] ??
new NullLogger() );
277 $this->stats
= $params['stats'] ??
new NullStatsdDataFactory();
278 $this->asyncHandler
= $params['asyncHandler'] ??
null;
282 * @param LoggerInterface $logger
284 public function setLogger( LoggerInterface
$logger ) {
285 $this->logger
= $logger;
289 * Get an instance that wraps EmptyBagOStuff
291 * @return WANObjectCache
293 public static function newEmpty() {
295 'cache' => new EmptyBagOStuff()
300 * Fetch the value of a key from cache
302 * If supplied, $curTTL is set to the remaining TTL (current time left):
303 * - a) INF; if $key exists, has no TTL, and is not invalidated by $checkKeys
304 * - b) float (>=0); if $key exists, has a TTL, and is not invalidated by $checkKeys
305 * - c) float (<0); if $key is tombstoned, stale, or existing but invalidated by $checkKeys
306 * - d) null; if $key does not exist and is not tombstoned
308 * If a key is tombstoned, $curTTL will reflect the time since delete().
310 * The timestamp of $key will be checked against the last-purge timestamp
311 * of each of $checkKeys. Those $checkKeys not in cache will have the last-purge
312 * initialized to the current timestamp. If any of $checkKeys have a timestamp
313 * greater than that of $key, then $curTTL will reflect how long ago $key
314 * became invalid. Callers can use $curTTL to know when the value is stale.
315 * The $checkKeys parameter allow mass invalidations by updating a single key:
316 * - a) Each "check" key represents "last purged" of some source data
317 * - b) Callers pass in relevant "check" keys as $checkKeys in get()
318 * - c) When the source data that "check" keys represent changes,
319 * the touchCheckKey() method is called on them
321 * Source data entities might exists in a DB that uses snapshot isolation
322 * (e.g. the default REPEATABLE-READ in innoDB). Even for mutable data, that
323 * isolation can largely be maintained by doing the following:
324 * - a) Calling delete() on entity change *and* creation, before DB commit
325 * - b) Keeping transaction duration shorter than the delete() hold-off TTL
326 * - c) Disabling interim key caching via useInterimHoldOffCaching() before get() calls
328 * However, pre-snapshot values might still be seen if an update was made
329 * in a remote datacenter but the purge from delete() didn't relay yet.
331 * Consider using getWithSetCallback() instead of get() and set() cycles.
332 * That method has cache slam avoiding features for hot/expensive keys.
334 * @param string $key Cache key made from makeKey() or makeGlobalKey()
335 * @param mixed|null &$curTTL Approximate TTL left on the key if present/tombstoned [returned]
336 * @param array $checkKeys List of "check" keys
337 * @param float|null &$asOf UNIX timestamp of cached value; null on failure [returned]
338 * @return mixed Value of cache key or false on failure
340 final public function get( $key, &$curTTL = null, array $checkKeys = [], &$asOf = null ) {
343 $values = $this->getMulti( [ $key ], $curTTLs, $checkKeys, $asOfs );
344 $curTTL = $curTTLs[$key] ??
null;
345 $asOf = $asOfs[$key] ??
null;
347 return $values[$key] ??
false;
351 * Fetch the value of several keys from cache
353 * @see WANObjectCache::get()
355 * @param array $keys List of cache keys made from makeKey() or makeGlobalKey()
356 * @param array &$curTTLs Map of (key => approximate TTL left) for existing keys [returned]
357 * @param array $checkKeys List of check keys to apply to all $keys. May also apply "check"
358 * keys to specific cache keys only by using cache keys as keys in the $checkKeys array.
359 * @param float[] &$asOfs Map of (key => UNIX timestamp of cached value; null on failure)
360 * @return array Map of (key => value) for keys that exist and are not tombstoned
362 final public function getMulti(
363 array $keys, &$curTTLs = [], array $checkKeys = [], array &$asOfs = []
369 $vPrefixLen = strlen( self
::VALUE_KEY_PREFIX
);
370 $valueKeys = self
::prefixCacheKeys( $keys, self
::VALUE_KEY_PREFIX
);
372 $checkKeysForAll = [];
373 $checkKeysByKey = [];
375 foreach ( $checkKeys as $i => $checkKeyGroup ) {
376 $prefixed = self
::prefixCacheKeys( (array)$checkKeyGroup, self
::TIME_KEY_PREFIX
);
377 $checkKeysFlat = array_merge( $checkKeysFlat, $prefixed );
378 // Is this check keys for a specific cache key, or for all keys being fetched?
379 if ( is_int( $i ) ) {
380 $checkKeysForAll = array_merge( $checkKeysForAll, $prefixed );
382 $checkKeysByKey[$i] = isset( $checkKeysByKey[$i] )
383 ?
array_merge( $checkKeysByKey[$i], $prefixed )
388 // Fetch all of the raw values
389 $keysGet = array_merge( $valueKeys, $checkKeysFlat );
390 if ( $this->warmupCache
) {
391 $wrappedValues = array_intersect_key( $this->warmupCache
, array_flip( $keysGet ) );
392 $keysGet = array_diff( $keysGet, array_keys( $wrappedValues ) ); // keys left to fetch
393 $this->warmupKeyMisses +
= count( $keysGet );
398 $wrappedValues +
= $this->cache
->getMulti( $keysGet );
400 // Time used to compare/init "check" keys (derived after getMulti() to be pessimistic)
401 $now = $this->getCurrentTime();
403 // Collect timestamps from all "check" keys
404 $purgeValuesForAll = $this->processCheckKeys( $checkKeysForAll, $wrappedValues, $now );
405 $purgeValuesByKey = [];
406 foreach ( $checkKeysByKey as $cacheKey => $checks ) {
407 $purgeValuesByKey[$cacheKey] =
408 $this->processCheckKeys( $checks, $wrappedValues, $now );
411 // Get the main cache value for each key and validate them
412 foreach ( $valueKeys as $vKey ) {
413 if ( !isset( $wrappedValues[$vKey] ) ) {
414 continue; // not found
417 $key = substr( $vKey, $vPrefixLen ); // unprefix
419 list( $value, $curTTL ) = $this->unwrap( $wrappedValues[$vKey], $now );
420 if ( $value !== false ) {
421 $result[$key] = $value;
422 // Force dependent keys to be seen as stale for a while after purging
423 // to reduce race conditions involving stale data getting cached
424 $purgeValues = $purgeValuesForAll;
425 if ( isset( $purgeValuesByKey[$key] ) ) {
426 $purgeValues = array_merge( $purgeValues, $purgeValuesByKey[$key] );
428 foreach ( $purgeValues as $purge ) {
429 $safeTimestamp = $purge[self
::FLD_TIME
] +
$purge[self
::FLD_HOLDOFF
];
430 if ( $safeTimestamp >= $wrappedValues[$vKey][self
::FLD_TIME
] ) {
431 // How long ago this value was invalidated by *this* check key
432 $ago = min( $purge[self
::FLD_TIME
] - $now, self
::TINY_NEGATIVE
);
433 // How long ago this value was invalidated by *any* known check key
434 $curTTL = min( $curTTL, $ago );
438 $curTTLs[$key] = $curTTL;
439 $asOfs[$key] = ( $value !== false ) ?
$wrappedValues[$vKey][self
::FLD_TIME
] : null;
447 * @param array $timeKeys List of prefixed time check keys
448 * @param array $wrappedValues
450 * @return array List of purge value arrays
452 private function processCheckKeys( array $timeKeys, array $wrappedValues, $now ) {
454 foreach ( $timeKeys as $timeKey ) {
455 $purge = isset( $wrappedValues[$timeKey] )
456 ?
$this->parsePurgeValue( $wrappedValues[$timeKey] )
458 if ( $purge === false ) {
459 // Key is not set or malformed; regenerate
460 $newVal = $this->makePurgeValue( $now, self
::HOLDOFF_TTL
);
461 $this->cache
->add( $timeKey, $newVal, self
::CHECK_KEY_TTL
);
462 $purge = $this->parsePurgeValue( $newVal );
464 $purgeValues[] = $purge;
470 * Set the value of a key in cache
472 * Simply calling this method when source data changes is not valid because
473 * the changes do not replicate to the other WAN sites. In that case, delete()
474 * should be used instead. This method is intended for use on cache misses.
476 * If the data was read from a snapshot-isolated transactions (e.g. the default
477 * REPEATABLE-READ in innoDB), use 'since' to avoid the following race condition:
479 * - b) T2 updates a row, calls delete(), and commits
480 * - c) The HOLDOFF_TTL passes, expiring the delete() tombstone
481 * - d) T1 reads the row and calls set() due to a cache miss
482 * - e) Stale value is stuck in cache
484 * Setting 'lag' and 'since' help avoids keys getting stuck in stale states.
486 * Be aware that this does not update the process cache for getWithSetCallback()
487 * callers. Keys accessed via that method are not generally meant to also be set
488 * using this primitive method.
490 * Do not use this method on versioned keys accessed via getWithSetCallback().
494 * $dbr = wfGetDB( DB_REPLICA );
495 * $setOpts = Database::getCacheSetOptions( $dbr );
496 * // Fetch the row from the DB
497 * $row = $dbr->selectRow( ... );
498 * $key = $cache->makeKey( 'building', $buildingId );
499 * $cache->set( $key, $row, $cache::TTL_DAY, $setOpts );
502 * @param string $key Cache key
503 * @param mixed $value
504 * @param int $ttl Seconds to live. Special values are:
505 * - WANObjectCache::TTL_INDEFINITE: Cache forever
506 * @param array $opts Options map:
507 * - lag : Seconds of replica DB lag. Typically, this is either the replica DB lag
508 * before the data was read or, if applicable, the replica DB lag before
509 * the snapshot-isolated transaction the data was read from started.
510 * Use false to indicate that replication is not running.
512 * - since : UNIX timestamp of the data in $value. Typically, this is either
513 * the current time the data was read or (if applicable) the time when
514 * the snapshot-isolated transaction the data was read from started.
516 * - pending : Whether this data is possibly from an uncommitted write transaction.
517 * Generally, other threads should not see values from the future and
518 * they certainly should not see ones that ended up getting rolled back.
520 * - lockTSE : if excessive replication/snapshot lag is detected, then store the value
521 * with this TTL and flag it as stale. This is only useful if the reads for this key
522 * use getWithSetCallback() with "lockTSE" set. Note that if "staleTTL" is set
523 * then it will still add on to this TTL in the excessive lag scenario.
524 * Default: WANObjectCache::TSE_NONE
525 * - staleTTL : Seconds to keep the key around if it is stale. The get()/getMulti()
526 * methods return such stale values with a $curTTL of 0, and getWithSetCallback()
527 * will call the regeneration callback in such cases, passing in the old value
528 * and its as-of time to the callback. This is useful if adaptiveTTL() is used
529 * on the old value's as-of time when it is verified as still being correct.
530 * Default: WANObjectCache::STALE_TTL_NONE.
531 * @note Options added in 1.28: staleTTL
532 * @return bool Success
534 final public function set( $key, $value, $ttl = 0, array $opts = [] ) {
535 $now = $this->getCurrentTime();
536 $lockTSE = $opts['lockTSE'] ?? self
::TSE_NONE
;
537 $staleTTL = $opts['staleTTL'] ?? self
::STALE_TTL_NONE
;
538 $age = isset( $opts['since'] ) ?
max( 0, $now - $opts['since'] ) : 0;
539 $lag = $opts['lag'] ??
0;
541 // Do not cache potentially uncommitted data as it might get rolled back
542 if ( !empty( $opts['pending'] ) ) {
543 $this->logger
->info( 'Rejected set() for {cachekey} due to pending writes.',
544 [ 'cachekey' => $key ] );
546 return true; // no-op the write for being unsafe
549 $wrapExtra = []; // additional wrapped value fields
550 // Check if there's a risk of writing stale data after the purge tombstone expired
551 if ( $lag === false ||
( $lag +
$age ) > self
::MAX_READ_LAG
) {
552 // Case A: read lag with "lockTSE"; save but record value as stale
553 if ( $lockTSE >= 0 ) {
554 $ttl = max( 1, (int)$lockTSE ); // set() expects seconds
555 $wrapExtra[self
::FLD_FLAGS
] = self
::FLG_STALE
; // mark as stale
556 // Case B: any long-running transaction; ignore this set()
557 } elseif ( $age > self
::MAX_READ_LAG
) {
558 $this->logger
->info( 'Rejected set() for {cachekey} due to snapshot lag.',
559 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ] );
561 return true; // no-op the write for being unsafe
562 // Case C: high replication lag; lower TTL instead of ignoring all set()s
563 } elseif ( $lag === false ||
$lag > self
::MAX_READ_LAG
) {
564 $ttl = $ttl ?
min( $ttl, self
::TTL_LAGGED
) : self
::TTL_LAGGED
;
565 $this->logger
->warning( 'Lowered set() TTL for {cachekey} due to replication lag.',
566 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ] );
567 // Case D: medium length request with medium replication lag; ignore this set()
569 $this->logger
->info( 'Rejected set() for {cachekey} due to high read lag.',
570 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ] );
572 return true; // no-op the write for being unsafe
576 // Wrap that value with time/TTL/version metadata
577 $wrapped = $this->wrap( $value, $ttl, $now ) +
$wrapExtra;
579 $func = function ( $cache, $key, $cWrapped ) use ( $wrapped ) {
580 return ( is_string( $cWrapped ) )
581 ?
false // key is tombstoned; do nothing
585 return $this->cache
->merge( self
::VALUE_KEY_PREFIX
. $key, $func, $ttl +
$staleTTL, 1 );
589 * Purge a key from all datacenters
591 * This should only be called when the underlying data (being cached)
592 * changes in a significant way. This deletes the key and starts a hold-off
593 * period where the key cannot be written to for a few seconds (HOLDOFF_TTL).
594 * This is done to avoid the following race condition:
595 * - a) Some DB data changes and delete() is called on a corresponding key
596 * - b) A request refills the key with a stale value from a lagged DB
597 * - c) The stale value is stuck there until the key is expired/evicted
599 * This is implemented by storing a special "tombstone" value at the cache
600 * key that this class recognizes; get() calls will return false for the key
601 * and any set() calls will refuse to replace tombstone values at the key.
602 * For this to always avoid stale value writes, the following must hold:
603 * - a) Replication lag is bounded to being less than HOLDOFF_TTL; or
604 * - b) If lag is higher, the DB will have gone into read-only mode already
606 * Note that set() can also be lag-aware and lower the TTL if it's high.
608 * Be aware that this does not clear the process cache. Even if it did, callbacks
609 * used by getWithSetCallback() might still return stale data in the case of either
610 * uncommitted or not-yet-replicated changes (callback generally use replica DBs).
612 * When using potentially long-running ACID transactions, a good pattern is
613 * to use a pre-commit hook to issue the delete. This means that immediately
614 * after commit, callers will see the tombstone in cache upon purge relay.
615 * It also avoids the following race condition:
616 * - a) T1 begins, changes a row, and calls delete()
617 * - b) The HOLDOFF_TTL passes, expiring the delete() tombstone
618 * - c) T2 starts, reads the row and calls set() due to a cache miss
619 * - d) T1 finally commits
620 * - e) Stale value is stuck in cache
624 * $dbw->startAtomic( __METHOD__ ); // start of request
625 * ... <execute some stuff> ...
626 * // Update the row in the DB
627 * $dbw->update( ... );
628 * $key = $cache->makeKey( 'homes', $homeId );
629 * // Purge the corresponding cache entry just before committing
630 * $dbw->onTransactionPreCommitOrIdle( function() use ( $cache, $key ) {
631 * $cache->delete( $key );
633 * ... <execute some stuff> ...
634 * $dbw->endAtomic( __METHOD__ ); // end of request
637 * The $ttl parameter can be used when purging values that have not actually changed
638 * recently. For example, a cleanup script to purge cache entries does not really need
639 * a hold-off period, so it can use HOLDOFF_NONE. Likewise for user-requested purge.
640 * Note that $ttl limits the effective range of 'lockTSE' for getWithSetCallback().
642 * If called twice on the same key, then the last hold-off TTL takes precedence. For
643 * idempotence, the $ttl should not vary for different delete() calls on the same key.
645 * @param string $key Cache key
646 * @param int $ttl Tombstone TTL; Default: WANObjectCache::HOLDOFF_TTL
647 * @return bool True if the item was purged or not found, false on failure
649 final public function delete( $key, $ttl = self
::HOLDOFF_TTL
) {
650 $key = self
::VALUE_KEY_PREFIX
. $key;
653 // Publish the purge to all datacenters
654 $ok = $this->relayDelete( $key );
656 // Publish the purge to all datacenters
657 $ok = $this->relayPurge( $key, $ttl, self
::HOLDOFF_NONE
);
664 * Fetch the value of a timestamp "check" key
666 * The key will be *initialized* to the current time if not set,
667 * so only call this method if this behavior is actually desired
669 * The timestamp can be used to check whether a cached value is valid.
670 * Callers should not assume that this returns the same timestamp in
671 * all datacenters due to relay delays.
673 * The level of staleness can roughly be estimated from this key, but
674 * if the key was evicted from cache, such calculations may show the
675 * time since expiry as ~0 seconds.
677 * Note that "check" keys won't collide with other regular keys.
680 * @return float UNIX timestamp
682 final public function getCheckKeyTime( $key ) {
683 return $this->getMultiCheckKeyTime( [ $key ] )[$key];
687 * Fetch the values of each timestamp "check" key
689 * This works like getCheckKeyTime() except it takes a list of keys
690 * and returns a map of timestamps instead of just that of one key
692 * This might be useful if both:
693 * - a) a class of entities each depend on hundreds of other entities
694 * - b) these other entities are depended upon by millions of entities
696 * The later entities can each use a "check" key to invalidate their dependee entities.
697 * However, it is expensive for the former entities to verify against all of the relevant
698 * "check" keys during each getWithSetCallback() call. A less expensive approach is to do
699 * these verifications only after a "time-till-verify" (TTV) has passed. This is a middle
700 * ground between using blind TTLs and using constant verification. The adaptiveTTL() method
701 * can be used to dynamically adjust the TTV. Also, the initial TTV can make use of the
702 * last-modified times of the dependant entities (either from the DB or the "check" keys).
706 * $value = $cache->getWithSetCallback(
707 * $cache->makeGlobalKey( 'wikibase-item', $id ),
708 * self::INITIAL_TTV, // initial time-till-verify
709 * function ( $oldValue, &$ttv, &$setOpts, $oldAsOf ) use ( $checkKeys, $cache ) {
710 * $now = microtime( true );
711 * // Use $oldValue if it passes max ultimate age and "check" key comparisons
713 * $oldAsOf > max( $cache->getMultiCheckKeyTime( $checkKeys ) ) &&
714 * ( $now - $oldValue['ctime'] ) <= self::MAX_CACHE_AGE
716 * // Increase time-till-verify by 50% of last time to reduce overhead
717 * $ttv = $cache->adaptiveTTL( $oldAsOf, self::MAX_TTV, self::MIN_TTV, 1.5 );
718 * // Unlike $oldAsOf, "ctime" is the ultimate age of the cached data
722 * $mtimes = []; // dependency last-modified times; passed by reference
723 * $value = [ 'data' => $this->fetchEntityData( $mtimes ), 'ctime' => $now ];
724 * // Guess time-till-change among the dependencies, e.g. 1/(total change rate)
725 * $ttc = 1 / array_sum( array_map(
726 * function ( $mtime ) use ( $now ) {
727 * return 1 / ( $mtime ? ( $now - $mtime ) : 900 );
731 * // The time-to-verify should not be overly pessimistic nor optimistic
732 * $ttv = min( max( $ttc, self::MIN_TTV ), self::MAX_TTV );
736 * [ 'staleTTL' => $cache::TTL_DAY ] // keep around to verify and re-save
740 * @see WANObjectCache::getCheckKeyTime()
741 * @see WANObjectCache::getWithSetCallback()
744 * @return float[] Map of (key => UNIX timestamp)
747 final public function getMultiCheckKeyTime( array $keys ) {
749 foreach ( $keys as $key ) {
750 $rawKeys[$key] = self
::TIME_KEY_PREFIX
. $key;
753 $rawValues = $this->cache
->getMulti( $rawKeys );
754 $rawValues +
= array_fill_keys( $rawKeys, false );
757 foreach ( $rawKeys as $key => $rawKey ) {
758 $purge = $this->parsePurgeValue( $rawValues[$rawKey] );
759 if ( $purge !== false ) {
760 $time = $purge[self
::FLD_TIME
];
762 // Casting assures identical floats for the next getCheckKeyTime() calls
763 $now = (string)$this->getCurrentTime();
766 $this->makePurgeValue( $now, self
::HOLDOFF_TTL
),
772 $times[$key] = $time;
779 * Purge a "check" key from all datacenters, invalidating keys that use it
781 * This should only be called when the underlying data (being cached)
782 * changes in a significant way, and it is impractical to call delete()
783 * on all keys that should be changed. When get() is called on those
784 * keys, the relevant "check" keys must be supplied for this to work.
786 * The "check" key essentially represents a last-modified time of an entity.
787 * When the key is touched, the timestamp will be updated to the current time.
788 * Keys using the "check" key via get(), getMulti(), or getWithSetCallback() will
789 * be invalidated. This approach is useful if many keys depend on a single entity.
791 * The timestamp of the "check" key is treated as being HOLDOFF_TTL seconds in the
792 * future by get*() methods in order to avoid race conditions where keys are updated
793 * with stale values (e.g. from a lagged replica DB). A high TTL is set on the "check"
794 * key, making it possible to know the timestamp of the last change to the corresponding
795 * entities in most cases. This might use more cache space than resetCheckKey().
797 * When a few important keys get a large number of hits, a high cache time is usually
798 * desired as well as "lockTSE" logic. The resetCheckKey() method is less appropriate
799 * in such cases since the "time since expiry" cannot be inferred, causing any get()
800 * after the reset to treat the key as being "hot", resulting in more stale value usage.
802 * Note that "check" keys won't collide with other regular keys.
804 * @see WANObjectCache::get()
805 * @see WANObjectCache::getWithSetCallback()
806 * @see WANObjectCache::resetCheckKey()
808 * @param string $key Cache key
809 * @param int $holdoff HOLDOFF_TTL or HOLDOFF_NONE constant
810 * @return bool True if the item was purged or not found, false on failure
812 final public function touchCheckKey( $key, $holdoff = self
::HOLDOFF_TTL
) {
813 // Publish the purge to all datacenters
814 return $this->relayPurge( self
::TIME_KEY_PREFIX
. $key, self
::CHECK_KEY_TTL
, $holdoff );
818 * Delete a "check" key from all datacenters, invalidating keys that use it
820 * This is similar to touchCheckKey() in that keys using it via get(), getMulti(),
821 * or getWithSetCallback() will be invalidated. The differences are:
822 * - a) The "check" key will be deleted from all caches and lazily
823 * re-initialized when accessed (rather than set everywhere)
824 * - b) Thus, dependent keys will be known to be stale, but not
825 * for how long (they are treated as "just" purged), which
826 * effects any lockTSE logic in getWithSetCallback()
827 * - c) Since "check" keys are initialized only on the server the key hashes
828 * to, any temporary ejection of that server will cause the value to be
829 * seen as purged as a new server will initialize the "check" key.
831 * The advantage here is that the "check" keys, which have high TTLs, will only
832 * be created when a get*() method actually uses that key. This is better when
833 * a large number of "check" keys are invalided in a short period of time.
835 * Note that "check" keys won't collide with other regular keys.
837 * @see WANObjectCache::get()
838 * @see WANObjectCache::getWithSetCallback()
839 * @see WANObjectCache::touchCheckKey()
841 * @param string $key Cache key
842 * @return bool True if the item was purged or not found, false on failure
844 final public function resetCheckKey( $key ) {
845 // Publish the purge to all datacenters
846 return $this->relayDelete( self
::TIME_KEY_PREFIX
. $key );
850 * Method to fetch/regenerate cache keys
852 * On cache miss, the key will be set to the callback result via set()
853 * (unless the callback returns false) and that result will be returned.
854 * The arguments supplied to the callback are:
855 * - $oldValue : current cache value or false if not present
856 * - &$ttl : a reference to the TTL which can be altered
857 * - &$setOpts : a reference to options for set() which can be altered
858 * - $oldAsOf : generation UNIX timestamp of $oldValue or null if not present (since 1.28)
860 * It is strongly recommended to set the 'lag' and 'since' fields to avoid race conditions
861 * that can cause stale values to get stuck at keys. Usually, callbacks ignore the current
862 * value, but it can be used to maintain "most recent X" values that come from time or
863 * sequence based source data, provided that the "as of" id/time is tracked. Note that
864 * preemptive regeneration and $checkKeys can result in a non-false current value.
866 * Usage of $checkKeys is similar to get() and getMulti(). However, rather than the caller
867 * having to inspect a "current time left" variable (e.g. $curTTL, $curTTLs), a cache
868 * regeneration will automatically be triggered using the callback.
870 * The $ttl argument and "hotTTR" option (in $opts) use time-dependant randomization
871 * to avoid stampedes. Keys that are slow to regenerate and either heavily used
872 * or subject to explicit (unpredictable) purges, may need additional mechanisms.
873 * The simplest way to avoid stampedes for such keys is to use 'lockTSE' (in $opts).
874 * If explicit purges are needed, also:
875 * - a) Pass $key into $checkKeys
876 * - b) Use touchCheckKey( $key ) instead of delete( $key )
878 * Example usage (typical key):
880 * $catInfo = $cache->getWithSetCallback(
881 * // Key to store the cached value under
882 * $cache->makeKey( 'cat-attributes', $catId ),
883 * // Time-to-live (in seconds)
884 * $cache::TTL_MINUTE,
885 * // Function that derives the new key value
886 * function ( $oldValue, &$ttl, array &$setOpts ) {
887 * $dbr = wfGetDB( DB_REPLICA );
888 * // Account for any snapshot/replica DB lag
889 * $setOpts += Database::getCacheSetOptions( $dbr );
891 * return $dbr->selectRow( ... );
896 * Example usage (key that is expensive and hot):
898 * $catConfig = $cache->getWithSetCallback(
899 * // Key to store the cached value under
900 * $cache->makeKey( 'site-cat-config' ),
901 * // Time-to-live (in seconds)
903 * // Function that derives the new key value
904 * function ( $oldValue, &$ttl, array &$setOpts ) {
905 * $dbr = wfGetDB( DB_REPLICA );
906 * // Account for any snapshot/replica DB lag
907 * $setOpts += Database::getCacheSetOptions( $dbr );
909 * return CatConfig::newFromRow( $dbr->selectRow( ... ) );
912 * // Calling touchCheckKey() on this key invalidates the cache
913 * 'checkKeys' => [ $cache->makeKey( 'site-cat-config' ) ],
914 * // Try to only let one datacenter thread manage cache updates at a time
916 * // Avoid querying cache servers multiple times in a web request
917 * 'pcTTL' => $cache::TTL_PROC_LONG
922 * Example usage (key with dynamic dependencies):
924 * $catState = $cache->getWithSetCallback(
925 * // Key to store the cached value under
926 * $cache->makeKey( 'cat-state', $cat->getId() ),
927 * // Time-to-live (seconds)
929 * // Function that derives the new key value
930 * function ( $oldValue, &$ttl, array &$setOpts ) {
931 * // Determine new value from the DB
932 * $dbr = wfGetDB( DB_REPLICA );
933 * // Account for any snapshot/replica DB lag
934 * $setOpts += Database::getCacheSetOptions( $dbr );
936 * return CatState::newFromResults( $dbr->select( ... ) );
939 * // The "check" keys that represent things the value depends on;
940 * // Calling touchCheckKey() on any of them invalidates the cache
942 * $cache->makeKey( 'sustenance-bowls', $cat->getRoomId() ),
943 * $cache->makeKey( 'people-present', $cat->getHouseId() ),
944 * $cache->makeKey( 'cat-laws', $cat->getCityId() ),
950 * Example usage (key that is expensive with too many DB dependencies for "check keys"):
952 * $catToys = $cache->getWithSetCallback(
953 * // Key to store the cached value under
954 * $cache->makeKey( 'cat-toys', $catId ),
955 * // Time-to-live (seconds)
957 * // Function that derives the new key value
958 * function ( $oldValue, &$ttl, array &$setOpts ) {
959 * // Determine new value from the DB
960 * $dbr = wfGetDB( DB_REPLICA );
961 * // Account for any snapshot/replica DB lag
962 * $setOpts += Database::getCacheSetOptions( $dbr );
964 * return CatToys::newFromResults( $dbr->select( ... ) );
967 * // Get the highest timestamp of any of the cat's toys
968 * 'touchedCallback' => function ( $value ) use ( $catId ) {
969 * $dbr = wfGetDB( DB_REPLICA );
970 * $ts = $dbr->selectField( 'cat_toys', 'MAX(ct_touched)', ... );
972 * return wfTimestampOrNull( TS_UNIX, $ts );
974 * // Avoid DB queries for repeated access
975 * 'pcTTL' => $cache::TTL_PROC_SHORT
980 * Example usage (hot key holding most recent 100 events):
982 * $lastCatActions = $cache->getWithSetCallback(
983 * // Key to store the cached value under
984 * $cache->makeKey( 'cat-last-actions', 100 ),
985 * // Time-to-live (in seconds)
987 * // Function that derives the new key value
988 * function ( $oldValue, &$ttl, array &$setOpts ) {
989 * $dbr = wfGetDB( DB_REPLICA );
990 * // Account for any snapshot/replica DB lag
991 * $setOpts += Database::getCacheSetOptions( $dbr );
993 * // Start off with the last cached list
994 * $list = $oldValue ?: [];
995 * // Fetch the last 100 relevant rows in descending order;
996 * // only fetch rows newer than $list[0] to reduce scanning
997 * $rows = iterator_to_array( $dbr->select( ... ) );
998 * // Merge them and get the new "last 100" rows
999 * return array_slice( array_merge( $new, $list ), 0, 100 );
1002 * // Try to only let one datacenter thread manage cache updates at a time
1004 * // Use a magic value when no cache value is ready rather than stampeding
1005 * 'busyValue' => 'computing'
1010 * Example usage (key holding an LRU subkey:value map; this can avoid flooding cache with
1011 * keys for an unlimited set of (constraint,situation) pairs, thereby avoiding elevated
1012 * cache evictions and wasted memory):
1014 * $catSituationTolerabilityCache = $this->cache->getWithSetCallback(
1015 * // Group by constraint ID/hash, cat family ID/hash, or something else useful
1016 * $this->cache->makeKey( 'cat-situation-tolerability-checks', $groupKey ),
1017 * WANObjectCache::TTL_DAY, // rarely used groups should fade away
1018 * // The $scenarioKey format is $constraintId:<ID/hash of $situation>
1019 * function ( $cacheMap ) use ( $scenarioKey, $constraintId, $situation ) {
1020 * $lruCache = MapCacheLRU::newFromArray( $cacheMap ?: [], self::CACHE_SIZE );
1021 * $result = $lruCache->get( $scenarioKey ); // triggers LRU bump if present
1022 * if ( $result === null || $this->isScenarioResultExpired( $result ) ) {
1023 * $result = $this->checkScenarioTolerability( $constraintId, $situation );
1024 * $lruCache->set( $scenarioKey, $result, 3 / 8 );
1026 * // Save the new LRU cache map and reset the map's TTL
1027 * return $lruCache->toArray();
1030 * // Once map is > 1 sec old, consider refreshing
1032 * // Update within 5 seconds after "ageNew" given a 1hz cache check rate
1034 * // Avoid querying cache servers multiple times in a request; this also means
1035 * // that a request can only alter the value of any given constraint key once
1036 * 'pcTTL' => WANObjectCache::TTL_PROC_LONG
1039 * $tolerability = isset( $catSituationTolerabilityCache[$scenarioKey] )
1040 * ? $catSituationTolerabilityCache[$scenarioKey]
1041 * : $this->checkScenarioTolerability( $constraintId, $situation );
1044 * @see WANObjectCache::get()
1045 * @see WANObjectCache::set()
1047 * @param string $key Cache key made from makeKey() or makeGlobalKey()
1048 * @param int $ttl Seconds to live for key updates. Special values are:
1049 * - WANObjectCache::TTL_INDEFINITE: Cache forever (subject to LRU-style evictions)
1050 * - WANObjectCache::TTL_UNCACHEABLE: Do not cache (if the key exists, it is not deleted)
1051 * @param callable $callback Value generation function
1052 * @param array $opts Options map:
1053 * - checkKeys: List of "check" keys. The key at $key will be seen as stale when either
1054 * touchCheckKey() or resetCheckKey() is called on any of the keys in this list. This
1055 * is useful if thousands or millions of keys depend on the same entity. The entity can
1056 * simply have its "check" key updated whenever the entity is modified.
1058 * - graceTTL: If the key is invalidated (by "checkKeys"/"touchedCallback") less than this
1059 * many seconds ago, consider reusing the stale value. The odds of a refresh becomes
1060 * more likely over time, becoming certain once the grace period is reached. This can
1061 * reduce traffic spikes when millions of keys are compared to the same "check" key and
1062 * touchCheckKey() or resetCheckKey() is called on that "check" key. This option is not
1063 * useful for avoiding traffic spikes in the case of the key simply expiring on account
1064 * of its TTL (use "lowTTL" instead).
1065 * Default: WANObjectCache::GRACE_TTL_NONE.
1066 * - lockTSE: If the key is tombstoned or invalidated (by "checkKeys"/"touchedCallback")
1067 * less than this many seconds ago, try to have a single thread handle cache regeneration
1068 * at any given time. Other threads will use stale values if possible. If, on miss,
1069 * the time since expiration is low, the assumption is that the key is hot and that a
1070 * stampede is worth avoiding. Note that if the key falls out of cache then concurrent
1071 * threads will all run the callback on cache miss until the value is saved in cache.
1072 * The only stampede protection in that case is from duplicate cache sets when the
1073 * callback takes longer than WANObjectCache::SET_DELAY_HIGH_SEC seconds; consider
1074 * using "busyValue" if such stampedes are a problem. Note that the higher "lockTSE" is
1075 * set, the higher the worst-case staleness of returned values can be. Also note that
1076 * this option does not by itself handle the case of the key simply expiring on account
1077 * of its TTL, so make sure that "lowTTL" is not disabled when using this option. Avoid
1078 * combining this option with delete() as it can always cause a stampede due to their
1079 * being no stale value available until after a thread completes the callback.
1080 * Use WANObjectCache::TSE_NONE to disable this logic.
1081 * Default: WANObjectCache::TSE_NONE.
1082 * - busyValue: If no value exists and another thread is currently regenerating it, use this
1083 * as a fallback value (or a callback to generate such a value). This assures that cache
1084 * stampedes cannot happen if the value falls out of cache. This can be used as insurance
1085 * against cache regeneration becoming very slow for some reason (greater than the TTL).
1087 * - pcTTL: Process cache the value in this PHP instance for this many seconds. This avoids
1088 * network I/O when a key is read several times. This will not cache when the callback
1089 * returns false, however. Note that any purges will not be seen while process cached;
1090 * since the callback should use replica DBs and they may be lagged or have snapshot
1091 * isolation anyway, this should not typically matter.
1092 * Default: WANObjectCache::TTL_UNCACHEABLE.
1093 * - pcGroup: Process cache group to use instead of the primary one. If set, this must be
1094 * of the format ALPHANUMERIC_NAME:MAX_KEY_SIZE, e.g. "mydata:10". Use this for storing
1095 * large values, small yet numerous values, or some values with a high cost of eviction.
1096 * It is generally preferable to use a class constant when setting this value.
1097 * This has no effect unless pcTTL is used.
1098 * Default: WANObjectCache::PC_PRIMARY.
1099 * - version: Integer version number. This allows for callers to make breaking changes to
1100 * how values are stored while maintaining compatability and correct cache purges. New
1101 * versions are stored alongside older versions concurrently. Avoid storing class objects
1102 * however, as this reduces compatibility (due to serialization).
1104 * - minAsOf: Reject values if they were generated before this UNIX timestamp.
1105 * This is useful if the source of a key is suspected of having possibly changed
1106 * recently, and the caller wants any such changes to be reflected.
1107 * Default: WANObjectCache::MIN_TIMESTAMP_NONE.
1108 * - hotTTR: Expected time-till-refresh (TTR) in seconds for keys that average ~1 hit per
1109 * second (e.g. 1Hz). Keys with a hit rate higher than 1Hz will refresh sooner than this
1110 * TTR and vise versa. Such refreshes won't happen until keys are "ageNew" seconds old.
1111 * This uses randomization to avoid triggering cache stampedes. The TTR is useful at
1112 * reducing the impact of missed cache purges, since the effect of a heavily referenced
1113 * key being stale is worse than that of a rarely referenced key. Unlike simply lowering
1114 * $ttl, seldomly used keys are largely unaffected by this option, which makes it
1115 * possible to have a high hit rate for the "long-tail" of less-used keys.
1116 * Default: WANObjectCache::HOT_TTR.
1117 * - lowTTL: Consider pre-emptive updates when the current TTL (seconds) of the key is less
1118 * than this. It becomes more likely over time, becoming certain once the key is expired.
1119 * This helps avoid cache stampedes that might be triggered due to the key expiring.
1120 * Default: WANObjectCache::LOW_TTL.
1121 * - ageNew: Consider popularity refreshes only once a key reaches this age in seconds.
1122 * Default: WANObjectCache::AGE_NEW.
1123 * - staleTTL: Seconds to keep the key around if it is stale. This means that on cache
1124 * miss the callback may get $oldValue/$oldAsOf values for keys that have already been
1125 * expired for this specified time. This is useful if adaptiveTTL() is used on the old
1126 * value's as-of time when it is verified as still being correct.
1127 * Default: WANObjectCache::STALE_TTL_NONE
1128 * - touchedCallback: A callback that takes the current value and returns a UNIX timestamp
1129 * indicating the last time a dynamic dependency changed. Null can be returned if there
1130 * are no relevant dependency changes to check. This can be used to check against things
1131 * like last-modified times of files or DB timestamp fields. This should generally not be
1132 * used for small and easily queried values in a DB if the callback itself ends up doing
1133 * a similarly expensive DB query to check a timestamp. Usages of this option makes the
1134 * most sense for values that are moderately to highly expensive to regenerate and easy
1135 * to query for dependency timestamps. The use of "pcTTL" reduces timestamp queries.
1137 * @return mixed Value found or written to the key
1138 * @note Options added in 1.28: version, busyValue, hotTTR, ageNew, pcGroup, minAsOf
1139 * @note Options added in 1.31: staleTTL, graceTTL
1140 * @note Options added in 1.33: touchedCallback
1141 * @note Callable type hints are not used to avoid class-autoloading
1143 final public function getWithSetCallback( $key, $ttl, $callback, array $opts = [] ) {
1144 $pcTTL = $opts['pcTTL'] ?? self
::TTL_UNCACHEABLE
;
1146 // Try the process cache if enabled and the cache callback is not within a cache callback.
1147 // Process cache use in nested callbacks is not lag-safe with regard to HOLDOFF_TTL since
1148 // the in-memory value is further lagged than the shared one since it uses a blind TTL.
1149 if ( $pcTTL >= 0 && $this->callbackDepth
== 0 ) {
1150 $group = $opts['pcGroup'] ?? self
::PC_PRIMARY
;
1151 $procCache = $this->getProcessCache( $group );
1152 $value = $procCache->has( $key, $pcTTL ) ?
$procCache->get( $key ) : false;
1158 if ( $value === false ) {
1159 // Fetch the value over the network
1160 if ( isset( $opts['version'] ) ) {
1161 $version = $opts['version'];
1163 $cur = $this->doGetWithSetCallback(
1166 function ( $oldValue, &$ttl, &$setOpts, $oldAsOf )
1167 use ( $callback, $version ) {
1168 if ( is_array( $oldValue )
1169 && array_key_exists( self
::VFLD_DATA
, $oldValue )
1170 && array_key_exists( self
::VFLD_VERSION
, $oldValue )
1171 && $oldValue[self
::VFLD_VERSION
] === $version
1173 $oldData = $oldValue[self
::VFLD_DATA
];
1175 // VFLD_DATA is not set if an old, unversioned, key is present
1181 self
::VFLD_DATA
=> $callback( $oldData, $ttl, $setOpts, $oldAsOf ),
1182 self
::VFLD_VERSION
=> $version
1188 if ( $cur[self
::VFLD_VERSION
] === $version ) {
1189 // Value created or existed before with version; use it
1190 $value = $cur[self
::VFLD_DATA
];
1192 // Value existed before with a different version; use variant key.
1193 // Reflect purges to $key by requiring that this key value be newer.
1194 $value = $this->doGetWithSetCallback(
1195 $this->makeGlobalKey( 'WANCache-key-variant', md5( $key ), $version ),
1198 // Regenerate value if not newer than $key
1199 [ 'version' => null, 'minAsOf' => $asOf ] +
$opts
1203 $value = $this->doGetWithSetCallback( $key, $ttl, $callback, $opts );
1206 // Update the process cache if enabled
1207 if ( $procCache && $value !== false ) {
1208 $procCache->set( $key, $value );
1216 * Do the actual I/O for getWithSetCallback() when needed
1218 * @see WANObjectCache::getWithSetCallback()
1220 * @param string $key
1222 * @param callable $callback
1223 * @param array $opts Options map for getWithSetCallback()
1224 * @param float|null &$asOf Cache generation timestamp of returned value [returned]
1226 * @note Callable type hints are not used to avoid class-autoloading
1228 protected function doGetWithSetCallback( $key, $ttl, $callback, array $opts, &$asOf = null ) {
1229 $lowTTL = $opts['lowTTL'] ??
min( self
::LOW_TTL
, $ttl );
1230 $lockTSE = $opts['lockTSE'] ?? self
::TSE_NONE
;
1231 $staleTTL = $opts['staleTTL'] ?? self
::STALE_TTL_NONE
;
1232 $graceTTL = $opts['graceTTL'] ?? self
::GRACE_TTL_NONE
;
1233 $checkKeys = $opts['checkKeys'] ??
[];
1234 $busyValue = $opts['busyValue'] ??
null;
1235 $popWindow = $opts['hotTTR'] ?? self
::HOT_TTR
;
1236 $ageNew = $opts['ageNew'] ?? self
::AGE_NEW
;
1237 $minTime = $opts['minAsOf'] ?? self
::MIN_TIMESTAMP_NONE
;
1238 $versioned = isset( $opts['version'] );
1239 $touchedCallback = $opts['touchedCallback'] ??
null;
1241 // Get a collection name to describe this class of key
1242 $kClass = $this->determineKeyClass( $key );
1244 // Get the current key value
1246 $cValue = $this->get( $key, $curTTL, $checkKeys, $asOf ); // current value
1247 $value = $cValue; // return value
1249 // Apply additional dynamic expiration logic if supplied
1250 $curTTL = $this->applyTouchedCallback( $value, $asOf, $curTTL, $touchedCallback );
1252 $preCallbackTime = $this->getCurrentTime();
1253 // Determine if a cached value regeneration is needed or desired
1255 $this->isValid( $value, $versioned, $asOf, $minTime ) &&
1256 $this->isAliveOrInGracePeriod( $curTTL, $graceTTL )
1258 $preemptiveRefresh = (
1259 $this->worthRefreshExpiring( $curTTL, $lowTTL ) ||
1260 $this->worthRefreshPopular( $asOf, $ageNew, $popWindow, $preCallbackTime )
1263 if ( !$preemptiveRefresh ) {
1264 $this->stats
->increment( "wanobjectcache.$kClass.hit.good" );
1267 } elseif ( $this->asyncHandler
) {
1268 // Update the cache value later, such during post-send of an HTTP request
1269 $func = $this->asyncHandler
;
1270 $func( function () use ( $key, $ttl, $callback, $opts, $asOf ) {
1271 $opts['minAsOf'] = INF
; // force a refresh
1272 $this->doGetWithSetCallback( $key, $ttl, $callback, $opts, $asOf );
1274 $this->stats
->increment( "wanobjectcache.$kClass.hit.refresh" );
1280 // Only a tombstoned key yields no value yet has a (negative) "current time left"
1281 $isKeyTombstoned = ( $curTTL !== null && $value === false );
1282 // Decide if only one thread should handle regeneration at a time
1284 // Note that since tombstones no-op set(), $lockTSE and $curTTL cannot be used to
1285 // deduce the key hotness because $curTTL will always keep increasing until the
1286 // tombstone expires or is overwritten by a new tombstone. Also, even if $lockTSE
1287 // is not set, constant regeneration of a key for the tombstone lifetime might be
1288 // very expensive. Assume tombstoned keys are possibly hot in order to reduce
1289 // the risk of high regeneration load after the delete() method is called.
1291 // Assume a key is hot if requested soon ($lockTSE seconds) after invalidation.
1292 // This avoids stampedes when timestamps from $checkKeys/$touchedCallback bump.
1293 ( $curTTL !== null && $curTTL <= 0 && abs( $curTTL ) <= $lockTSE ) ||
1294 // Assume a key is hot if there is no value and a busy fallback is given.
1295 // This avoids stampedes on eviction or preemptive regeneration taking too long.
1296 ( $busyValue !== null && $value === false );
1300 // Acquire a datacenter-local non-blocking lock
1301 if ( $this->cache
->add( self
::MUTEX_KEY_PREFIX
. $key, 1, self
::LOCK_TTL
) ) {
1302 // Lock acquired; this thread will recompute the value and update cache
1304 } elseif ( $this->isValid( $value, $versioned, $asOf, $minTime ) ) {
1305 // Lock not acquired and a stale value exists; use the stale value
1306 $this->stats
->increment( "wanobjectcache.$kClass.hit.stale" );
1310 // Lock not acquired and no stale value exists
1311 if ( $isKeyTombstoned ) {
1312 // Use the INTERIM value from the last thread that regenerated it if possible
1313 $value = $this->getInterimValue( $key, $versioned, $minTime, $asOf );
1314 if ( $value !== false ) {
1315 $this->stats
->increment( "wanobjectcache.$kClass.hit.volatile" );
1321 if ( $busyValue !== null ) {
1322 // Use the busy fallback value if nothing else
1323 $miss = is_infinite( $minTime ) ?
'renew' : 'miss';
1324 $this->stats
->increment( "wanobjectcache.$kClass.$miss.busy" );
1326 return is_callable( $busyValue ) ?
$busyValue() : $busyValue;
1331 if ( !is_callable( $callback ) ) {
1332 throw new InvalidArgumentException( "Invalid cache miss callback provided." );
1335 // Generate the new value from the callback...
1337 ++
$this->callbackDepth
;
1339 $value = call_user_func_array( $callback, [ $cValue, &$ttl, &$setOpts, $asOf ] );
1341 --$this->callbackDepth
;
1343 $valueIsCacheable = ( $value !== false && $ttl >= 0 );
1345 if ( $valueIsCacheable ) {
1346 $ago = max( $this->getCurrentTime() - $preCallbackTime, 0.0 );
1347 if ( $isKeyTombstoned ) {
1348 if ( $this->checkAndSetCooloff( $key, $kClass, $ago, $lockTSE, $hasLock ) ) {
1349 // When delete() is called, writes are write-holed by the tombstone,
1350 // so use a special INTERIM key to pass the new value among threads.
1351 $tempTTL = max( self
::INTERIM_KEY_TTL
, (int)$lockTSE ); // set() expects seconds
1352 $newAsOf = $this->getCurrentTime();
1353 $wrapped = $this->wrap( $value, $tempTTL, $newAsOf );
1354 // Avoid using set() to avoid pointless mcrouter broadcasting
1355 $this->setInterimValue( $key, $wrapped, $tempTTL );
1357 } elseif ( !$useMutex ||
$hasLock ) {
1358 if ( $this->checkAndSetCooloff( $key, $kClass, $ago, $lockTSE, $hasLock ) ) {
1359 // Save the value unless a lock-winning thread is already expected to do that
1360 $setOpts['lockTSE'] = $lockTSE;
1361 $setOpts['staleTTL'] = $staleTTL;
1362 // Use best known "since" timestamp if not provided
1363 $setOpts +
= [ 'since' => $preCallbackTime ];
1364 // Update the cache; this will fail if the key is tombstoned
1365 $this->set( $key, $value, $ttl, $setOpts );
1371 // Avoid using delete() to avoid pointless mcrouter broadcasting
1372 $this->cache
->changeTTL( self
::MUTEX_KEY_PREFIX
. $key, (int)$preCallbackTime - 60 );
1375 $miss = is_infinite( $minTime ) ?
'renew' : 'miss';
1376 $this->stats
->increment( "wanobjectcache.$kClass.$miss.compute" );
1382 * @param string $key
1383 * @param string $kClass
1384 * @param float $elapsed Seconds spent regenerating the value
1385 * @param float $lockTSE
1386 * @param $hasLock bool
1387 * @return bool Whether it is OK to proceed with a key set operation
1389 private function checkAndSetCooloff( $key, $kClass, $elapsed, $lockTSE, $hasLock ) {
1390 // If $lockTSE is set, the lock was bypassed because there was no stale/interim value,
1391 // and $elapsed indicates that regeration is slow, then there is a risk of set()
1392 // stampedes with large blobs. With a typical scale-out infrastructure, CPU and query
1393 // load from $callback invocations is distributed among appservers and replica DBs,
1394 // but cache operations for a given key route to a single cache server (e.g. striped
1395 // consistent hashing).
1396 if ( $lockTSE < 0 ||
$hasLock ) {
1397 return true; // either not a priori hot or thread has the lock
1398 } elseif ( $elapsed <= self
::SET_DELAY_HIGH_SEC
) {
1399 return true; // not enough time for threads to pile up
1402 $this->cache
->clearLastError();
1404 !$this->cache
->add( self
::COOLOFF_KEY_PREFIX
. $key, 1, self
::COOLOFF_TTL
) &&
1405 // Don't treat failures due to I/O errors as the key being in cooloff
1406 $this->cache
->getLastError() === BagOStuff
::ERR_NONE
1408 $this->stats
->increment( "wanobjectcache.$kClass.cooloff_bounce" );
1417 * @param mixed $value
1418 * @param float $asOf
1419 * @param float $curTTL
1420 * @param callable|null $callback
1423 protected function applyTouchedCallback( $value, $asOf, $curTTL, $callback ) {
1424 if ( $callback === null ) {
1428 if ( !is_callable( $callback ) ) {
1429 throw new InvalidArgumentException( "Invalid expiration callback provided." );
1432 if ( $value !== false ) {
1433 $touched = $callback( $value );
1434 if ( $touched !== null && $touched >= $asOf ) {
1435 $curTTL = min( $curTTL, self
::TINY_NEGATIVE
, $asOf - $touched );
1443 * @param string $key
1444 * @param bool $versioned
1445 * @param float $minTime
1446 * @param mixed &$asOf
1449 protected function getInterimValue( $key, $versioned, $minTime, &$asOf ) {
1450 if ( !$this->useInterimHoldOffCaching
) {
1451 return false; // disabled
1454 $wrapped = $this->cache
->get( self
::INTERIM_KEY_PREFIX
. $key );
1455 list( $value ) = $this->unwrap( $wrapped, $this->getCurrentTime() );
1456 if ( $this->isValid( $value, $versioned, $asOf, $minTime ) ) {
1457 $asOf = $wrapped[self
::FLD_TIME
];
1466 * @param string $key
1467 * @param array $wrapped
1468 * @param int $tempTTL
1470 protected function setInterimValue( $key, $wrapped, $tempTTL ) {
1471 $this->cache
->merge(
1472 self
::INTERIM_KEY_PREFIX
. $key,
1473 function () use ( $wrapped ) {
1482 * Method to fetch multiple cache keys at once with regeneration
1484 * This works the same as getWithSetCallback() except:
1485 * - a) The $keys argument expects the result of WANObjectCache::makeMultiKeys()
1486 * - b) The $callback argument expects a callback taking the following arguments:
1487 * - $id: ID of an entity to query
1488 * - $oldValue : the prior cache value or false if none was present
1489 * - &$ttl : a reference to the new value TTL in seconds
1490 * - &$setOpts : a reference to options for set() which can be altered
1491 * - $oldAsOf : generation UNIX timestamp of $oldValue or null if not present
1492 * Aside from the additional $id argument, the other arguments function the same
1493 * way they do in getWithSetCallback().
1494 * - c) The return value is a map of (cache key => value) in the order of $keyedIds
1496 * @see WANObjectCache::getWithSetCallback()
1497 * @see WANObjectCache::getMultiWithUnionSetCallback()
1501 * $rows = $cache->getMultiWithSetCallback(
1502 * // Map of cache keys to entity IDs
1503 * $cache->makeMultiKeys(
1504 * $this->fileVersionIds(),
1505 * function ( $id, WANObjectCache $cache ) {
1506 * return $cache->makeKey( 'file-version', $id );
1509 * // Time-to-live (in seconds)
1511 * // Function that derives the new key value
1512 * function ( $id, $oldValue, &$ttl, array &$setOpts ) {
1513 * $dbr = wfGetDB( DB_REPLICA );
1514 * // Account for any snapshot/replica DB lag
1515 * $setOpts += Database::getCacheSetOptions( $dbr );
1517 * // Load the row for this file
1518 * $queryInfo = File::getQueryInfo();
1519 * $row = $dbr->selectRow(
1520 * $queryInfo['tables'],
1521 * $queryInfo['fields'],
1525 * $queryInfo['joins']
1528 * return $row ? (array)$row : false;
1531 * // Process cache for 30 seconds
1533 * // Use a dedicated 500 item cache (initialized on-the-fly)
1534 * 'pcGroup' => 'file-versions:500'
1537 * $files = array_map( [ __CLASS__, 'newFromRow' ], $rows );
1540 * @param ArrayIterator $keyedIds Result of WANObjectCache::makeMultiKeys()
1541 * @param int $ttl Seconds to live for key updates
1542 * @param callable $callback Callback the yields entity regeneration callbacks
1543 * @param array $opts Options map
1544 * @return array Map of (cache key => value) in the same order as $keyedIds
1547 final public function getMultiWithSetCallback(
1548 ArrayIterator
$keyedIds, $ttl, callable
$callback, array $opts = []
1550 $valueKeys = array_keys( $keyedIds->getArrayCopy() );
1551 $checkKeys = $opts['checkKeys'] ??
[];
1552 $pcTTL = $opts['pcTTL'] ?? self
::TTL_UNCACHEABLE
;
1554 // Load required keys into process cache in one go
1555 $this->warmupCache
= $this->getRawKeysForWarmup(
1556 $this->getNonProcessCachedKeys( $valueKeys, $opts, $pcTTL ),
1559 $this->warmupKeyMisses
= 0;
1561 // Wrap $callback to match the getWithSetCallback() format while passing $id to $callback
1562 $id = null; // current entity ID
1563 $func = function ( $oldValue, &$ttl, &$setOpts, $oldAsOf ) use ( $callback, &$id ) {
1564 return $callback( $id, $oldValue, $ttl, $setOpts, $oldAsOf );
1568 foreach ( $keyedIds as $key => $id ) { // preserve order
1569 $values[$key] = $this->getWithSetCallback( $key, $ttl, $func, $opts );
1572 $this->warmupCache
= [];
1578 * Method to fetch/regenerate multiple cache keys at once
1580 * This works the same as getWithSetCallback() except:
1581 * - a) The $keys argument expects the result of WANObjectCache::makeMultiKeys()
1582 * - b) The $callback argument expects a callback returning a map of (ID => new value)
1583 * for all entity IDs in $ids and it takes the following arguments:
1584 * - $ids: a list of entity IDs that require cache regeneration
1585 * - &$ttls: a reference to the (entity ID => new TTL) map
1586 * - &$setOpts: a reference to options for set() which can be altered
1587 * - c) The return value is a map of (cache key => value) in the order of $keyedIds
1588 * - d) The "lockTSE" and "busyValue" options are ignored
1590 * @see WANObjectCache::getWithSetCallback()
1591 * @see WANObjectCache::getMultiWithSetCallback()
1595 * $rows = $cache->getMultiWithUnionSetCallback(
1596 * // Map of cache keys to entity IDs
1597 * $cache->makeMultiKeys(
1598 * $this->fileVersionIds(),
1599 * function ( $id, WANObjectCache $cache ) {
1600 * return $cache->makeKey( 'file-version', $id );
1603 * // Time-to-live (in seconds)
1605 * // Function that derives the new key value
1606 * function ( array $ids, array &$ttls, array &$setOpts ) {
1607 * $dbr = wfGetDB( DB_REPLICA );
1608 * // Account for any snapshot/replica DB lag
1609 * $setOpts += Database::getCacheSetOptions( $dbr );
1611 * // Load the rows for these files
1613 * $queryInfo = File::getQueryInfo();
1614 * $res = $dbr->select(
1615 * $queryInfo['tables'],
1616 * $queryInfo['fields'],
1620 * $queryInfo['joins']
1622 * foreach ( $res as $row ) {
1623 * $rows[$row->id] = $row;
1624 * $mtime = wfTimestamp( TS_UNIX, $row->timestamp );
1625 * $ttls[$row->id] = $this->adaptiveTTL( $mtime, $ttls[$row->id] );
1632 * $files = array_map( [ __CLASS__, 'newFromRow' ], $rows );
1635 * @param ArrayIterator $keyedIds Result of WANObjectCache::makeMultiKeys()
1636 * @param int $ttl Seconds to live for key updates
1637 * @param callable $callback Callback the yields entity regeneration callbacks
1638 * @param array $opts Options map
1639 * @return array Map of (cache key => value) in the same order as $keyedIds
1642 final public function getMultiWithUnionSetCallback(
1643 ArrayIterator
$keyedIds, $ttl, callable
$callback, array $opts = []
1645 $idsByValueKey = $keyedIds->getArrayCopy();
1646 $valueKeys = array_keys( $idsByValueKey );
1647 $checkKeys = $opts['checkKeys'] ??
[];
1648 $pcTTL = $opts['pcTTL'] ?? self
::TTL_UNCACHEABLE
;
1649 unset( $opts['lockTSE'] ); // incompatible
1650 unset( $opts['busyValue'] ); // incompatible
1652 // Load required keys into process cache in one go
1653 $keysGet = $this->getNonProcessCachedKeys( $valueKeys, $opts, $pcTTL );
1654 $this->warmupCache
= $this->getRawKeysForWarmup( $keysGet, $checkKeys );
1655 $this->warmupKeyMisses
= 0;
1657 // IDs of entities known to be in need of regeneration
1660 // Find out which keys are missing/deleted/stale
1663 $curByKey = $this->getMulti( $keysGet, $curTTLs, $checkKeys, $asOfs );
1664 foreach ( $keysGet as $key ) {
1665 if ( !array_key_exists( $key, $curByKey ) ||
$curTTLs[$key] < 0 ) {
1666 $idsRegen[] = $idsByValueKey[$key];
1670 // Run the callback to populate the regeneration value map for all required IDs
1672 $newTTLsById = array_fill_keys( $idsRegen, $ttl );
1673 $newValsById = $idsRegen ?
$callback( $idsRegen, $newTTLsById, $newSetOpts ) : [];
1675 // Wrap $callback to match the getWithSetCallback() format while passing $id to $callback
1676 $id = null; // current entity ID
1677 $func = function ( $oldValue, &$ttl, &$setOpts, $oldAsOf )
1678 use ( $callback, &$id, $newValsById, $newTTLsById, $newSetOpts )
1680 if ( array_key_exists( $id, $newValsById ) ) {
1681 // Value was already regerated as expected, so use the value in $newValsById
1682 $newValue = $newValsById[$id];
1683 $ttl = $newTTLsById[$id];
1684 $setOpts = $newSetOpts;
1686 // Pre-emptive/popularity refresh and version mismatch cases are not detected
1687 // above and thus $newValsById has no entry. Run $callback on this single entity.
1688 $ttls = [ $id => $ttl ];
1689 $newValue = $callback( [ $id ], $ttls, $setOpts )[$id];
1696 // Run the cache-aside logic using warmupCache instead of persistent cache queries
1698 foreach ( $idsByValueKey as $key => $id ) { // preserve order
1699 $values[$key] = $this->getWithSetCallback( $key, $ttl, $func, $opts );
1702 $this->warmupCache
= [];
1708 * Set a key to soon expire in the local cluster if it pre-dates $purgeTimestamp
1710 * This sets stale keys' time-to-live at HOLDOFF_TTL seconds, which both avoids
1711 * broadcasting in mcrouter setups and also avoids races with new tombstones.
1713 * @param string $key Cache key
1714 * @param int $purgeTimestamp UNIX timestamp of purge
1715 * @param bool &$isStale Whether the key is stale
1716 * @return bool Success
1719 final public function reap( $key, $purgeTimestamp, &$isStale = false ) {
1720 $minAsOf = $purgeTimestamp + self
::HOLDOFF_TTL
;
1721 $wrapped = $this->cache
->get( self
::VALUE_KEY_PREFIX
. $key );
1722 if ( is_array( $wrapped ) && $wrapped[self
::FLD_TIME
] < $minAsOf ) {
1724 $this->logger
->warning( "Reaping stale value key '$key'." );
1725 $ttlReap = self
::HOLDOFF_TTL
; // avoids races with tombstone creation
1726 $ok = $this->cache
->changeTTL( self
::VALUE_KEY_PREFIX
. $key, $ttlReap );
1728 $this->logger
->error( "Could not complete reap of key '$key'." );
1740 * Set a "check" key to soon expire in the local cluster if it pre-dates $purgeTimestamp
1742 * @param string $key Cache key
1743 * @param int $purgeTimestamp UNIX timestamp of purge
1744 * @param bool &$isStale Whether the key is stale
1745 * @return bool Success
1748 final public function reapCheckKey( $key, $purgeTimestamp, &$isStale = false ) {
1749 $purge = $this->parsePurgeValue( $this->cache
->get( self
::TIME_KEY_PREFIX
. $key ) );
1750 if ( $purge && $purge[self
::FLD_TIME
] < $purgeTimestamp ) {
1752 $this->logger
->warning( "Reaping stale check key '$key'." );
1753 $ok = $this->cache
->changeTTL( self
::TIME_KEY_PREFIX
. $key, self
::TTL_SECOND
);
1755 $this->logger
->error( "Could not complete reap of check key '$key'." );
1767 * @see BagOStuff::makeKey()
1768 * @param string $class Key class
1769 * @param string|null $component [optional] Key component (starting with a key collection name)
1770 * @return string Colon-delimited list of $keyspace followed by escaped components of $args
1773 public function makeKey( $class, $component = null ) {
1774 return $this->cache
->makeKey( ...func_get_args() );
1778 * @see BagOStuff::makeGlobalKey()
1779 * @param string $class Key class
1780 * @param string|null $component [optional] Key component (starting with a key collection name)
1781 * @return string Colon-delimited list of $keyspace followed by escaped components of $args
1784 public function makeGlobalKey( $class, $component = null ) {
1785 return $this->cache
->makeGlobalKey( ...func_get_args() );
1789 * @param array $entities List of entity IDs
1790 * @param callable $keyFunc Callback yielding a key from (entity ID, this WANObjectCache)
1791 * @return ArrayIterator Iterator yielding (cache key => entity ID) in $entities order
1794 final public function makeMultiKeys( array $entities, callable
$keyFunc ) {
1796 foreach ( $entities as $entity ) {
1797 $map[$keyFunc( $entity, $this )] = $entity;
1800 return new ArrayIterator( $map );
1804 * Get the "last error" registered; clearLastError() should be called manually
1805 * @return int ERR_* class constant for the "last error" registry
1807 final public function getLastError() {
1808 if ( $this->lastRelayError
) {
1809 // If the cache and the relayer failed, focus on the latter.
1810 // An update not making it to the relayer means it won't show up
1811 // in other DCs (nor will consistent re-hashing see up-to-date values).
1812 // On the other hand, if just the cache update failed, then it should
1813 // eventually be applied by the relayer.
1814 return $this->lastRelayError
;
1817 $code = $this->cache
->getLastError();
1819 case BagOStuff
::ERR_NONE
:
1820 return self
::ERR_NONE
;
1821 case BagOStuff
::ERR_NO_RESPONSE
:
1822 return self
::ERR_NO_RESPONSE
;
1823 case BagOStuff
::ERR_UNREACHABLE
:
1824 return self
::ERR_UNREACHABLE
;
1826 return self
::ERR_UNEXPECTED
;
1831 * Clear the "last error" registry
1833 final public function clearLastError() {
1834 $this->cache
->clearLastError();
1835 $this->lastRelayError
= self
::ERR_NONE
;
1839 * Clear the in-process caches; useful for testing
1843 public function clearProcessCache() {
1844 $this->processCaches
= [];
1848 * Enable or disable the use of brief caching for tombstoned keys
1850 * When a key is purged via delete(), there normally is a period where caching
1851 * is hold-off limited to an extremely short time. This method will disable that
1852 * caching, forcing the callback to run for any of:
1853 * - WANObjectCache::getWithSetCallback()
1854 * - WANObjectCache::getMultiWithSetCallback()
1855 * - WANObjectCache::getMultiWithUnionSetCallback()
1857 * This is useful when both:
1858 * - a) the database used by the callback is known to be up-to-date enough
1859 * for some particular purpose (e.g. replica DB has applied transaction X)
1860 * - b) the caller needs to exploit that fact, and therefore needs to avoid the
1861 * use of inherently volatile and possibly stale interim keys
1863 * @see WANObjectCache::delete()
1864 * @param bool $enabled Whether to enable interim caching
1867 final public function useInterimHoldOffCaching( $enabled ) {
1868 $this->useInterimHoldOffCaching
= $enabled;
1872 * @param int $flag ATTR_* class constant
1873 * @return int QOS_* class constant
1876 public function getQoS( $flag ) {
1877 return $this->cache
->getQoS( $flag );
1881 * Get a TTL that is higher for objects that have not changed recently
1883 * This is useful for keys that get explicit purges and DB or purge relay
1884 * lag is a potential concern (especially how it interacts with CDN cache)
1888 * // Last-modified time of page
1889 * $mtime = wfTimestamp( TS_UNIX, $page->getTimestamp() );
1890 * // Get adjusted TTL. If $mtime is 3600 seconds ago and $minTTL/$factor left at
1891 * // defaults, then $ttl is 3600 * .2 = 720. If $minTTL was greater than 720, then
1892 * // $ttl would be $minTTL. If $maxTTL was smaller than 720, $ttl would be $maxTTL.
1893 * $ttl = $cache->adaptiveTTL( $mtime, $cache::TTL_DAY );
1896 * Another use case is when there are no applicable "last modified" fields in the DB,
1897 * and there are too many dependencies for explicit purges to be viable, and the rate of
1898 * change to relevant content is unstable, and it is highly valued to have the cached value
1899 * be as up-to-date as possible.
1903 * $query = "<some complex query>";
1904 * $idListFromComplexQuery = $cache->getWithSetCallback(
1905 * $cache->makeKey( 'complex-graph-query', $hashOfQuery ),
1906 * GraphQueryClass::STARTING_TTL,
1907 * function ( $oldValue, &$ttl, array &$setOpts, $oldAsOf ) use ( $query, $cache ) {
1908 * $gdb = $this->getReplicaGraphDbConnection();
1909 * // Account for any snapshot/replica DB lag
1910 * $setOpts += GraphDatabase::getCacheSetOptions( $gdb );
1912 * $newList = iterator_to_array( $gdb->query( $query ) );
1913 * sort( $newList, SORT_NUMERIC ); // normalize
1915 * $minTTL = GraphQueryClass::MIN_TTL;
1916 * $maxTTL = GraphQueryClass::MAX_TTL;
1917 * if ( $oldValue !== false ) {
1918 * // Note that $oldAsOf is the last time this callback ran
1919 * $ttl = ( $newList === $oldValue )
1920 * // No change: cache for 150% of the age of $oldValue
1921 * ? $cache->adaptiveTTL( $oldAsOf, $maxTTL, $minTTL, 1.5 )
1922 * // Changed: cache for 50% of the age of $oldValue
1923 * : $cache->adaptiveTTL( $oldAsOf, $maxTTL, $minTTL, .5 );
1929 * // Keep stale values around for doing comparisons for TTL calculations.
1930 * // High values improve long-tail keys hit-rates, though might waste space.
1931 * 'staleTTL' => GraphQueryClass::GRACE_TTL
1936 * @param int|float $mtime UNIX timestamp
1937 * @param int $maxTTL Maximum TTL (seconds)
1938 * @param int $minTTL Minimum TTL (seconds); Default: 30
1939 * @param float $factor Value in the range (0,1); Default: .2
1940 * @return int Adaptive TTL
1943 public function adaptiveTTL( $mtime, $maxTTL, $minTTL = 30, $factor = 0.2 ) {
1944 if ( is_float( $mtime ) ||
ctype_digit( $mtime ) ) {
1945 $mtime = (int)$mtime; // handle fractional seconds and string integers
1948 if ( !is_int( $mtime ) ||
$mtime <= 0 ) {
1949 return $minTTL; // no last-modified time provided
1952 $age = $this->getCurrentTime() - $mtime;
1954 return (int)min( $maxTTL, max( $minTTL, $factor * $age ) );
1958 * @return int Number of warmup key cache misses last round
1961 final public function getWarmupKeyMisses() {
1962 return $this->warmupKeyMisses
;
1966 * Do the actual async bus purge of a key
1968 * This must set the key to "PURGED:<UNIX timestamp>:<holdoff>"
1970 * @param string $key Cache key
1971 * @param int $ttl How long to keep the tombstone [seconds]
1972 * @param int $holdoff HOLDOFF_* constant controlling how long to ignore sets for this key
1973 * @return bool Success
1975 protected function relayPurge( $key, $ttl, $holdoff ) {
1976 if ( $this->mcrouterAware
) {
1977 // See https://github.com/facebook/mcrouter/wiki/Multi-cluster-broadcast-setup
1978 // Wildcards select all matching routes, e.g. the WAN cluster on all DCs
1979 $ok = $this->cache
->set(
1980 "/*/{$this->cluster}/{$key}",
1981 $this->makePurgeValue( $this->getCurrentTime(), self
::HOLDOFF_NONE
),
1984 } elseif ( $this->purgeRelayer
instanceof EventRelayerNull
) {
1985 // This handles the mcrouter and the single-DC case
1986 $ok = $this->cache
->set(
1988 $this->makePurgeValue( $this->getCurrentTime(), self
::HOLDOFF_NONE
),
1992 $event = $this->cache
->modifySimpleRelayEvent( [
1995 'val' => 'PURGED:$UNIXTIME$:' . (int)$holdoff,
1996 'ttl' => max( $ttl, self
::TTL_SECOND
),
1997 'sbt' => true, // substitute $UNIXTIME$ with actual microtime
2000 $ok = $this->purgeRelayer
->notify( $this->purgeChannel
, $event );
2002 $this->lastRelayError
= self
::ERR_RELAY
;
2010 * Do the actual async bus delete of a key
2012 * @param string $key Cache key
2013 * @return bool Success
2015 protected function relayDelete( $key ) {
2016 if ( $this->mcrouterAware
) {
2017 // See https://github.com/facebook/mcrouter/wiki/Multi-cluster-broadcast-setup
2018 // Wildcards select all matching routes, e.g. the WAN cluster on all DCs
2019 $ok = $this->cache
->delete( "/*/{$this->cluster}/{$key}" );
2020 } elseif ( $this->purgeRelayer
instanceof EventRelayerNull
) {
2021 // Some other proxy handles broadcasting or there is only one datacenter
2022 $ok = $this->cache
->delete( $key );
2024 $event = $this->cache
->modifySimpleRelayEvent( [
2029 $ok = $this->purgeRelayer
->notify( $this->purgeChannel
, $event );
2031 $this->lastRelayError
= self
::ERR_RELAY
;
2039 * Check if a key is fresh or in the grace window and thus due for randomized reuse
2041 * If $curTTL > 0 (e.g. not expired) this returns true. Otherwise, the chance of returning
2042 * true decrease steadily from 100% to 0% as the |$curTTL| moves from 0 to $graceTTL seconds.
2043 * This handles widely varying levels of cache access traffic.
2045 * If $curTTL <= -$graceTTL (e.g. already expired), then this returns false.
2047 * @param float $curTTL Approximate TTL left on the key if present
2048 * @param int $graceTTL Consider using stale values if $curTTL is greater than this
2051 protected function isAliveOrInGracePeriod( $curTTL, $graceTTL ) {
2052 if ( $curTTL > 0 ) {
2054 } elseif ( $graceTTL <= 0 ) {
2058 $ageStale = abs( $curTTL ); // seconds of staleness
2059 $curGTTL = ( $graceTTL - $ageStale ); // current grace-time-to-live
2060 if ( $curGTTL <= 0 ) {
2061 return false; // already out of grace period
2064 // Chance of using a stale value is the complement of the chance of refreshing it
2065 return !$this->worthRefreshExpiring( $curGTTL, $graceTTL );
2069 * Check if a key is nearing expiration and thus due for randomized regeneration
2071 * This returns false if $curTTL >= $lowTTL. Otherwise, the chance of returning true
2072 * increases steadily from 0% to 100% as the $curTTL moves from $lowTTL to 0 seconds.
2073 * This handles widely varying levels of cache access traffic.
2075 * If $curTTL <= 0 (e.g. already expired), then this returns false.
2077 * @param float $curTTL Approximate TTL left on the key if present
2078 * @param float $lowTTL Consider a refresh when $curTTL is less than this
2081 protected function worthRefreshExpiring( $curTTL, $lowTTL ) {
2082 if ( $lowTTL <= 0 ) {
2084 } elseif ( $curTTL >= $lowTTL ) {
2086 } elseif ( $curTTL <= 0 ) {
2090 $chance = ( 1 - $curTTL / $lowTTL );
2092 return mt_rand( 1, 1e9
) <= 1e9
* $chance;
2096 * Check if a key is due for randomized regeneration due to its popularity
2098 * This is used so that popular keys can preemptively refresh themselves for higher
2099 * consistency (especially in the case of purge loss/delay). Unpopular keys can remain
2100 * in cache with their high nominal TTL. This means popular keys keep good consistency,
2101 * whether the data changes frequently or not, and long-tail keys get to stay in cache
2102 * and get hits too. Similar to worthRefreshExpiring(), randomization is used.
2104 * @param float $asOf UNIX timestamp of the value
2105 * @param int $ageNew Age of key when this might recommend refreshing (seconds)
2106 * @param int $timeTillRefresh Age of key when it should be refreshed if popular (seconds)
2107 * @param float $now The current UNIX timestamp
2110 protected function worthRefreshPopular( $asOf, $ageNew, $timeTillRefresh, $now ) {
2111 if ( $ageNew < 0 ||
$timeTillRefresh <= 0 ) {
2115 $age = $now - $asOf;
2116 $timeOld = $age - $ageNew;
2117 if ( $timeOld <= 0 ) {
2121 // Lifecycle is: new, ramp-up refresh chance, full refresh chance.
2122 // Note that the "expected # of refreshes" for the ramp-up time range is half of what it
2123 // would be if P(refresh) was at its full value during that time range.
2124 $refreshWindowSec = max( $timeTillRefresh - $ageNew - self
::RAMPUP_TTL
/ 2, 1 );
2125 // P(refresh) * (# hits in $refreshWindowSec) = (expected # of refreshes)
2126 // P(refresh) * ($refreshWindowSec * $popularHitsPerSec) = 1
2127 // P(refresh) = 1/($refreshWindowSec * $popularHitsPerSec)
2128 $chance = 1 / ( self
::HIT_RATE_HIGH
* $refreshWindowSec );
2130 // Ramp up $chance from 0 to its nominal value over RAMPUP_TTL seconds to avoid stampedes
2131 $chance *= ( $timeOld <= self
::RAMPUP_TTL
) ?
$timeOld / self
::RAMPUP_TTL
: 1;
2133 return mt_rand( 1, 1e9
) <= 1e9
* $chance;
2137 * Check if $value is not false, versioned (if needed), and not older than $minTime (if set)
2139 * @param array|bool $value
2140 * @param bool $versioned
2141 * @param float $asOf The time $value was generated
2142 * @param float $minTime The last time the main value was generated (0.0 if unknown)
2145 protected function isValid( $value, $versioned, $asOf, $minTime ) {
2146 if ( $value === false ) {
2148 } elseif ( $versioned && !isset( $value[self
::VFLD_VERSION
] ) ) {
2150 } elseif ( $minTime > 0 && $asOf < $minTime ) {
2158 * Do not use this method outside WANObjectCache
2160 * @param mixed $value
2161 * @param int $ttl [0=forever]
2162 * @param float $now Unix Current timestamp just before calling set()
2165 protected function wrap( $value, $ttl, $now ) {
2167 self
::FLD_VERSION
=> self
::VERSION
,
2168 self
::FLD_VALUE
=> $value,
2169 self
::FLD_TTL
=> $ttl,
2170 self
::FLD_TIME
=> $now
2175 * Do not use this method outside WANObjectCache
2177 * @param array|string|bool $wrapped
2178 * @param float $now Unix Current timestamp (preferrably pre-query)
2179 * @return array (mixed; false if absent/tombstoned/malformed, current time left)
2181 protected function unwrap( $wrapped, $now ) {
2182 // Check if the value is a tombstone
2183 $purge = $this->parsePurgeValue( $wrapped );
2184 if ( $purge !== false ) {
2185 // Purged values should always have a negative current $ttl
2186 $curTTL = min( $purge[self
::FLD_TIME
] - $now, self
::TINY_NEGATIVE
);
2187 return [ false, $curTTL ];
2190 if ( !is_array( $wrapped ) // not found
2191 ||
!isset( $wrapped[self
::FLD_VERSION
] ) // wrong format
2192 ||
$wrapped[self
::FLD_VERSION
] !== self
::VERSION
// wrong version
2194 return [ false, null ];
2197 $flags = $wrapped[self
::FLD_FLAGS
] ??
0;
2198 if ( ( $flags & self
::FLG_STALE
) == self
::FLG_STALE
) {
2199 // Treat as expired, with the cache time as the expiration
2200 $age = $now - $wrapped[self
::FLD_TIME
];
2201 $curTTL = min( -$age, self
::TINY_NEGATIVE
);
2202 } elseif ( $wrapped[self
::FLD_TTL
] > 0 ) {
2203 // Get the approximate time left on the key
2204 $age = $now - $wrapped[self
::FLD_TIME
];
2205 $curTTL = max( $wrapped[self
::FLD_TTL
] - $age, 0.0 );
2207 // Key had no TTL, so the time left is unbounded
2211 if ( $wrapped[self
::FLD_TIME
] < $this->epoch
) {
2212 // Values this old are ignored
2213 return [ false, null ];
2216 return [ $wrapped[self
::FLD_VALUE
], $curTTL ];
2220 * @param array $keys
2221 * @param string $prefix
2224 protected static function prefixCacheKeys( array $keys, $prefix ) {
2226 foreach ( $keys as $key ) {
2227 $res[] = $prefix . $key;
2234 * @param string $key String of the format <scope>:<class>[:<class or variable>]...
2237 protected function determineKeyClass( $key ) {
2238 $parts = explode( ':', $key );
2240 return $parts[1] ??
$parts[0]; // sanity
2244 * @param string|array|bool $value Possible string of the form "PURGED:<timestamp>:<holdoff>"
2245 * @return array|bool Array containing a UNIX timestamp (float) and holdoff period (integer),
2246 * or false if value isn't a valid purge value
2248 protected function parsePurgeValue( $value ) {
2249 if ( !is_string( $value ) ) {
2253 $segments = explode( ':', $value, 3 );
2254 if ( !isset( $segments[0] ) ||
!isset( $segments[1] )
2255 ||
"{$segments[0]}:" !== self
::PURGE_VAL_PREFIX
2260 if ( !isset( $segments[2] ) ) {
2261 // Back-compat with old purge values without holdoff
2262 $segments[2] = self
::HOLDOFF_TTL
;
2265 if ( $segments[1] < $this->epoch
) {
2266 // Values this old are ignored
2271 self
::FLD_TIME
=> (float)$segments[1],
2272 self
::FLD_HOLDOFF
=> (int)$segments[2],
2277 * @param float $timestamp
2278 * @param int $holdoff In seconds
2279 * @return string Wrapped purge value
2281 protected function makePurgeValue( $timestamp, $holdoff ) {
2282 return self
::PURGE_VAL_PREFIX
. (float)$timestamp . ':' . (int)$holdoff;
2286 * @param string $group
2287 * @return MapCacheLRU
2289 protected function getProcessCache( $group ) {
2290 if ( !isset( $this->processCaches
[$group] ) ) {
2291 list( , $n ) = explode( ':', $group );
2292 $this->processCaches
[$group] = new MapCacheLRU( (int)$n );
2295 return $this->processCaches
[$group];
2299 * @param array $keys
2300 * @param array $opts
2302 * @return array List of keys
2304 private function getNonProcessCachedKeys( array $keys, array $opts, $pcTTL ) {
2306 if ( isset( $opts['pcTTL'] ) && $opts['pcTTL'] > 0 && $this->callbackDepth
== 0 ) {
2307 $pcGroup = $opts['pcGroup'] ?? self
::PC_PRIMARY
;
2308 $procCache = $this->getProcessCache( $pcGroup );
2309 foreach ( $keys as $key ) {
2310 if ( $procCache->has( $key, $pcTTL ) ) {
2311 $keysFound[] = $key;
2316 return array_diff( $keys, $keysFound );
2320 * @param array $keys
2321 * @param array $checkKeys
2322 * @return array Map of (cache key => mixed)
2324 private function getRawKeysForWarmup( array $keys, array $checkKeys ) {
2330 // Get all the value keys to fetch...
2331 foreach ( $keys as $key ) {
2332 $keysWarmUp[] = self
::VALUE_KEY_PREFIX
. $key;
2334 // Get all the check keys to fetch...
2335 foreach ( $checkKeys as $i => $checkKeyOrKeys ) {
2336 if ( is_int( $i ) ) {
2337 // Single check key that applies to all value keys
2338 $keysWarmUp[] = self
::TIME_KEY_PREFIX
. $checkKeyOrKeys;
2340 // List of check keys that apply to value key $i
2341 $keysWarmUp = array_merge(
2343 self
::prefixCacheKeys( $checkKeyOrKeys, self
::TIME_KEY_PREFIX
)
2348 $warmupCache = $this->cache
->getMulti( $keysWarmUp );
2349 $warmupCache +
= array_fill_keys( $keysWarmUp, false );
2351 return $warmupCache;
2355 * @return float UNIX timestamp
2356 * @codeCoverageIgnore
2358 protected function getCurrentTime() {
2359 if ( $this->wallClockOverride
) {
2360 return $this->wallClockOverride
;
2363 $clockTime = (float)time(); // call this first
2364 // microtime() uses an initial gettimeofday() call added to usage clocks.
2365 // This can severely drift from time() and the microtime() value of other threads
2366 // due to undercounting of the amount of time elapsed. Instead of seeing the current
2367 // time as being in the past, use the value of time(). This avoids setting cache values
2368 // that will immediately be seen as expired and possibly cause stampedes.
2369 return max( microtime( true ), $clockTime );
2373 * @param float|null &$time Mock UNIX timestamp for testing
2374 * @codeCoverageIgnore
2376 public function setMockTime( &$time ) {
2377 $this->wallClockOverride
=& $time;
2378 $this->cache
->setMockTime( $time );