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 If outdated, the value is recomputed.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 /** Default remaining TTL at which to consider pre-emptive regeneration */
174 /** Never consider performing "popularity" refreshes until a key reaches this age */
176 /** The time length of the "popularity" refresh window for hot keys */
178 /** Hits/second for a refresh to be expected within the "popularity" window */
179 const HIT_RATE_HIGH
= 1;
180 /** Seconds to ramp up to the "popularity" refresh chance after a key is no longer new */
181 const RAMPUP_TTL
= 30;
183 /** Idiom for getWithSetCallback() callbacks to avoid calling set() */
184 const TTL_UNCACHEABLE
= -1;
185 /** Idiom for getWithSetCallback() callbacks to 'lockTSE' logic */
187 /** Max TTL to store keys when a data sourced is lagged */
188 const TTL_LAGGED
= 30;
189 /** Idiom for delete() for "no hold-off" */
190 const HOLDOFF_NONE
= 0;
191 /** Idiom for set()/getWithSetCallback() for "do not augment the storage medium TTL" */
192 const STALE_TTL_NONE
= 0;
193 /** Idiom for set()/getWithSetCallback() for "no post-expired grace period" */
194 const GRACE_TTL_NONE
= 0;
196 /** Idiom for getWithSetCallback() for "no minimum required as-of timestamp" */
197 const MIN_TIMESTAMP_NONE
= 0.0;
199 /** Tiny negative float to use when CTL comes up >= 0 due to clock skew */
200 const TINY_NEGATIVE
= -0.000001;
202 /** Cache format version number */
205 const FLD_VERSION
= 0; // key to cache version number
206 const FLD_VALUE
= 1; // key to the cached value
207 const FLD_TTL
= 2; // key to the original TTL
208 const FLD_TIME
= 3; // key to the cache time
209 const FLD_FLAGS
= 4; // key to the flags bitfield
210 const FLD_HOLDOFF
= 5; // key to any hold-off TTL
212 /** @var int Treat this value as expired-on-arrival */
215 const ERR_NONE
= 0; // no error
216 const ERR_NO_RESPONSE
= 1; // no response
217 const ERR_UNREACHABLE
= 2; // can't connect
218 const ERR_UNEXPECTED
= 3; // response gave some error
219 const ERR_RELAY
= 4; // relay broadcast failed
221 const VALUE_KEY_PREFIX
= 'WANCache:v:';
222 const INTERIM_KEY_PREFIX
= 'WANCache:i:';
223 const TIME_KEY_PREFIX
= 'WANCache:t:';
224 const MUTEX_KEY_PREFIX
= 'WANCache:m:';
226 const PURGE_VAL_PREFIX
= 'PURGED:';
228 const VFLD_DATA
= 'WOC:d'; // key to the value of versioned data
229 const VFLD_VERSION
= 'WOC:v'; // key to the version of the value present
231 const PC_PRIMARY
= 'primary:1000'; // process cache name and max key count
233 const DEFAULT_PURGE_CHANNEL
= 'wancache-purge';
236 * @param array $params
237 * - cache : BagOStuff object for a persistent cache
238 * - channels : Map of (action => channel string). Actions include "purge".
239 * - relayers : Map of (action => EventRelayer object). Actions include "purge".
240 * - logger : LoggerInterface object
241 * - stats : LoggerInterface object
242 * - asyncHandler : A function that takes a callback and runs it later. If supplied,
243 * whenever a preemptive refresh would be triggered in getWithSetCallback(), the
244 * current cache value is still used instead. However, the async-handler function
245 * receives a WAN cache callback that, when run, will execute the value generation
246 * callback supplied by the getWithSetCallback() caller. The result will be saved
247 * as normal. The handler is expected to call the WAN cache callback at an opportune
248 * time (e.g. HTTP post-send), though generally within a few 100ms. [optional]
249 * - region: the current physical region. This is required when using mcrouter as the
250 * backing store proxy. [optional]
251 * - cluster: name of the cache cluster used by this WAN cache. The name must be the
252 * same in all datacenters; the ("region","cluster") tuple is what distinguishes
253 * the counterpart cache clusters among all the datacenter. The contents of
254 * https://github.com/facebook/mcrouter/wiki/Config-Files give background on this.
255 * This is required when using mcrouter as the backing store proxy. [optional]
256 * - mcrouterAware: set as true if mcrouter is the backing store proxy and mcrouter
257 * is configured to interpret /<region>/<cluster>/ key prefixes as routes. This
258 * requires that "region" and "cluster" are both set above. [optional]
259 * - epoch: lowest UNIX timestamp a value/tombstone must have to be valid. [optional]
261 public function __construct( array $params ) {
262 $this->cache
= $params['cache'];
263 $this->purgeChannel
= $params['channels']['purge'] ?? self
::DEFAULT_PURGE_CHANNEL
;
264 $this->purgeRelayer
= $params['relayers']['purge'] ??
new EventRelayerNull( [] );
265 $this->region
= $params['region'] ??
'main';
266 $this->cluster
= $params['cluster'] ??
'wan-main';
267 $this->mcrouterAware
= !empty( $params['mcrouterAware'] );
268 $this->epoch
= $params['epoch'] ??
1.0;
270 $this->setLogger( $params['logger'] ??
new NullLogger() );
271 $this->stats
= $params['stats'] ??
new NullStatsdDataFactory();
272 $this->asyncHandler
= $params['asyncHandler'] ??
null;
276 * @param LoggerInterface $logger
278 public function setLogger( LoggerInterface
$logger ) {
279 $this->logger
= $logger;
283 * Get an instance that wraps EmptyBagOStuff
285 * @return WANObjectCache
287 public static function newEmpty() {
289 'cache' => new EmptyBagOStuff()
294 * Fetch the value of a key from cache
296 * If supplied, $curTTL is set to the remaining TTL (current time left):
297 * - a) INF; if $key exists, has no TTL, and is not invalidated by $checkKeys
298 * - b) float (>=0); if $key exists, has a TTL, and is not invalidated by $checkKeys
299 * - c) float (<0); if $key is tombstoned, stale, or existing but invalidated by $checkKeys
300 * - d) null; if $key does not exist and is not tombstoned
302 * If a key is tombstoned, $curTTL will reflect the time since delete().
304 * The timestamp of $key will be checked against the last-purge timestamp
305 * of each of $checkKeys. Those $checkKeys not in cache will have the last-purge
306 * initialized to the current timestamp. If any of $checkKeys have a timestamp
307 * greater than that of $key, then $curTTL will reflect how long ago $key
308 * became invalid. Callers can use $curTTL to know when the value is stale.
309 * The $checkKeys parameter allow mass invalidations by updating a single key:
310 * - a) Each "check" key represents "last purged" of some source data
311 * - b) Callers pass in relevant "check" keys as $checkKeys in get()
312 * - c) When the source data that "check" keys represent changes,
313 * the touchCheckKey() method is called on them
315 * Source data entities might exists in a DB that uses snapshot isolation
316 * (e.g. the default REPEATABLE-READ in innoDB). Even for mutable data, that
317 * isolation can largely be maintained by doing the following:
318 * - a) Calling delete() on entity change *and* creation, before DB commit
319 * - b) Keeping transaction duration shorter than the delete() hold-off TTL
320 * - c) Disabling interim key caching via useInterimHoldOffCaching() before get() calls
322 * However, pre-snapshot values might still be seen if an update was made
323 * in a remote datacenter but the purge from delete() didn't relay yet.
325 * Consider using getWithSetCallback() instead of get() and set() cycles.
326 * That method has cache slam avoiding features for hot/expensive keys.
328 * @param string $key Cache key made from makeKey() or makeGlobalKey()
329 * @param mixed|null &$curTTL Approximate TTL left on the key if present/tombstoned [returned]
330 * @param array $checkKeys List of "check" keys
331 * @param float|null &$asOf UNIX timestamp of cached value; null on failure [returned]
332 * @return mixed Value of cache key or false on failure
334 final public function get( $key, &$curTTL = null, array $checkKeys = [], &$asOf = null ) {
337 $values = $this->getMulti( [ $key ], $curTTLs, $checkKeys, $asOfs );
338 $curTTL = $curTTLs[$key] ??
null;
339 $asOf = $asOfs[$key] ??
null;
341 return $values[$key] ??
false;
345 * Fetch the value of several keys from cache
347 * @see WANObjectCache::get()
349 * @param array $keys List of cache keys made from makeKey() or makeGlobalKey()
350 * @param array &$curTTLs Map of (key => approximate TTL left) for existing keys [returned]
351 * @param array $checkKeys List of check keys to apply to all $keys. May also apply "check"
352 * keys to specific cache keys only by using cache keys as keys in the $checkKeys array.
353 * @param float[] &$asOfs Map of (key => UNIX timestamp of cached value; null on failure)
354 * @return array Map of (key => value) for keys that exist and are not tombstoned
356 final public function getMulti(
357 array $keys, &$curTTLs = [], array $checkKeys = [], array &$asOfs = []
363 $vPrefixLen = strlen( self
::VALUE_KEY_PREFIX
);
364 $valueKeys = self
::prefixCacheKeys( $keys, self
::VALUE_KEY_PREFIX
);
366 $checkKeysForAll = [];
367 $checkKeysByKey = [];
369 foreach ( $checkKeys as $i => $checkKeyGroup ) {
370 $prefixed = self
::prefixCacheKeys( (array)$checkKeyGroup, self
::TIME_KEY_PREFIX
);
371 $checkKeysFlat = array_merge( $checkKeysFlat, $prefixed );
372 // Is this check keys for a specific cache key, or for all keys being fetched?
373 if ( is_int( $i ) ) {
374 $checkKeysForAll = array_merge( $checkKeysForAll, $prefixed );
376 $checkKeysByKey[$i] = isset( $checkKeysByKey[$i] )
377 ?
array_merge( $checkKeysByKey[$i], $prefixed )
382 // Fetch all of the raw values
383 $keysGet = array_merge( $valueKeys, $checkKeysFlat );
384 if ( $this->warmupCache
) {
385 $wrappedValues = array_intersect_key( $this->warmupCache
, array_flip( $keysGet ) );
386 $keysGet = array_diff( $keysGet, array_keys( $wrappedValues ) ); // keys left to fetch
387 $this->warmupKeyMisses +
= count( $keysGet );
392 $wrappedValues +
= $this->cache
->getMulti( $keysGet );
394 // Time used to compare/init "check" keys (derived after getMulti() to be pessimistic)
395 $now = $this->getCurrentTime();
397 // Collect timestamps from all "check" keys
398 $purgeValuesForAll = $this->processCheckKeys( $checkKeysForAll, $wrappedValues, $now );
399 $purgeValuesByKey = [];
400 foreach ( $checkKeysByKey as $cacheKey => $checks ) {
401 $purgeValuesByKey[$cacheKey] =
402 $this->processCheckKeys( $checks, $wrappedValues, $now );
405 // Get the main cache value for each key and validate them
406 foreach ( $valueKeys as $vKey ) {
407 if ( !isset( $wrappedValues[$vKey] ) ) {
408 continue; // not found
411 $key = substr( $vKey, $vPrefixLen ); // unprefix
413 list( $value, $curTTL ) = $this->unwrap( $wrappedValues[$vKey], $now );
414 if ( $value !== false ) {
415 $result[$key] = $value;
416 // Force dependent keys to be seen as stale for a while after purging
417 // to reduce race conditions involving stale data getting cached
418 $purgeValues = $purgeValuesForAll;
419 if ( isset( $purgeValuesByKey[$key] ) ) {
420 $purgeValues = array_merge( $purgeValues, $purgeValuesByKey[$key] );
422 foreach ( $purgeValues as $purge ) {
423 $safeTimestamp = $purge[self
::FLD_TIME
] +
$purge[self
::FLD_HOLDOFF
];
424 if ( $safeTimestamp >= $wrappedValues[$vKey][self
::FLD_TIME
] ) {
425 // How long ago this value was invalidated by *this* check key
426 $ago = min( $purge[self
::FLD_TIME
] - $now, self
::TINY_NEGATIVE
);
427 // How long ago this value was invalidated by *any* known check key
428 $curTTL = min( $curTTL, $ago );
432 $curTTLs[$key] = $curTTL;
433 $asOfs[$key] = ( $value !== false ) ?
$wrappedValues[$vKey][self
::FLD_TIME
] : null;
441 * @param array $timeKeys List of prefixed time check keys
442 * @param array $wrappedValues
444 * @return array List of purge value arrays
446 private function processCheckKeys( array $timeKeys, array $wrappedValues, $now ) {
448 foreach ( $timeKeys as $timeKey ) {
449 $purge = isset( $wrappedValues[$timeKey] )
450 ?
$this->parsePurgeValue( $wrappedValues[$timeKey] )
452 if ( $purge === false ) {
453 // Key is not set or malformed; regenerate
454 $newVal = $this->makePurgeValue( $now, self
::HOLDOFF_TTL
);
455 $this->cache
->add( $timeKey, $newVal, self
::CHECK_KEY_TTL
);
456 $purge = $this->parsePurgeValue( $newVal );
458 $purgeValues[] = $purge;
464 * Set the value of a key in cache
466 * Simply calling this method when source data changes is not valid because
467 * the changes do not replicate to the other WAN sites. In that case, delete()
468 * should be used instead. This method is intended for use on cache misses.
470 * If the data was read from a snapshot-isolated transactions (e.g. the default
471 * REPEATABLE-READ in innoDB), use 'since' to avoid the following race condition:
473 * - b) T2 updates a row, calls delete(), and commits
474 * - c) The HOLDOFF_TTL passes, expiring the delete() tombstone
475 * - d) T1 reads the row and calls set() due to a cache miss
476 * - e) Stale value is stuck in cache
478 * Setting 'lag' and 'since' help avoids keys getting stuck in stale states.
480 * Be aware that this does not update the process cache for getWithSetCallback()
481 * callers. Keys accessed via that method are not generally meant to also be set
482 * using this primitive method.
484 * Do not use this method on versioned keys accessed via getWithSetCallback().
488 * $dbr = wfGetDB( DB_REPLICA );
489 * $setOpts = Database::getCacheSetOptions( $dbr );
490 * // Fetch the row from the DB
491 * $row = $dbr->selectRow( ... );
492 * $key = $cache->makeKey( 'building', $buildingId );
493 * $cache->set( $key, $row, $cache::TTL_DAY, $setOpts );
496 * @param string $key Cache key
497 * @param mixed $value
498 * @param int $ttl Seconds to live. Special values are:
499 * - WANObjectCache::TTL_INDEFINITE: Cache forever
500 * @param array $opts Options map:
501 * - lag : Seconds of replica DB lag. Typically, this is either the replica DB lag
502 * before the data was read or, if applicable, the replica DB lag before
503 * the snapshot-isolated transaction the data was read from started.
504 * Use false to indicate that replication is not running.
506 * - since : UNIX timestamp of the data in $value. Typically, this is either
507 * the current time the data was read or (if applicable) the time when
508 * the snapshot-isolated transaction the data was read from started.
510 * - pending : Whether this data is possibly from an uncommitted write transaction.
511 * Generally, other threads should not see values from the future and
512 * they certainly should not see ones that ended up getting rolled back.
514 * - lockTSE : if excessive replication/snapshot lag is detected, then store the value
515 * with this TTL and flag it as stale. This is only useful if the reads for this key
516 * use getWithSetCallback() with "lockTSE" set. Note that if "staleTTL" is set
517 * then it will still add on to this TTL in the excessive lag scenario.
518 * Default: WANObjectCache::TSE_NONE
519 * - staleTTL : Seconds to keep the key around if it is stale. The get()/getMulti()
520 * methods return such stale values with a $curTTL of 0, and getWithSetCallback()
521 * will call the regeneration callback in such cases, passing in the old value
522 * and its as-of time to the callback. This is useful if adaptiveTTL() is used
523 * on the old value's as-of time when it is verified as still being correct.
524 * Default: WANObjectCache::STALE_TTL_NONE.
525 * @note Options added in 1.28: staleTTL
526 * @return bool Success
528 final public function set( $key, $value, $ttl = 0, array $opts = [] ) {
529 $now = $this->getCurrentTime();
530 $lockTSE = $opts['lockTSE'] ?? self
::TSE_NONE
;
531 $staleTTL = $opts['staleTTL'] ?? self
::STALE_TTL_NONE
;
532 $age = isset( $opts['since'] ) ?
max( 0, $now - $opts['since'] ) : 0;
533 $lag = $opts['lag'] ??
0;
535 // Do not cache potentially uncommitted data as it might get rolled back
536 if ( !empty( $opts['pending'] ) ) {
537 $this->logger
->info( 'Rejected set() for {cachekey} due to pending writes.',
538 [ 'cachekey' => $key ] );
540 return true; // no-op the write for being unsafe
543 $wrapExtra = []; // additional wrapped value fields
544 // Check if there's a risk of writing stale data after the purge tombstone expired
545 if ( $lag === false ||
( $lag +
$age ) > self
::MAX_READ_LAG
) {
546 // Case A: read lag with "lockTSE"; save but record value as stale
547 if ( $lockTSE >= 0 ) {
548 $ttl = max( 1, (int)$lockTSE ); // set() expects seconds
549 $wrapExtra[self
::FLD_FLAGS
] = self
::FLG_STALE
; // mark as stale
550 // Case B: any long-running transaction; ignore this set()
551 } elseif ( $age > self
::MAX_READ_LAG
) {
552 $this->logger
->info( 'Rejected set() for {cachekey} due to snapshot lag.',
553 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ] );
555 return true; // no-op the write for being unsafe
556 // Case C: high replication lag; lower TTL instead of ignoring all set()s
557 } elseif ( $lag === false ||
$lag > self
::MAX_READ_LAG
) {
558 $ttl = $ttl ?
min( $ttl, self
::TTL_LAGGED
) : self
::TTL_LAGGED
;
559 $this->logger
->warning( 'Lowered set() TTL for {cachekey} due to replication lag.',
560 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ] );
561 // Case D: medium length request with medium replication lag; ignore this set()
563 $this->logger
->info( 'Rejected set() for {cachekey} due to high read lag.',
564 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ] );
566 return true; // no-op the write for being unsafe
570 // Wrap that value with time/TTL/version metadata
571 $wrapped = $this->wrap( $value, $ttl, $now ) +
$wrapExtra;
573 $func = function ( $cache, $key, $cWrapped ) use ( $wrapped ) {
574 return ( is_string( $cWrapped ) )
575 ?
false // key is tombstoned; do nothing
579 return $this->cache
->merge( self
::VALUE_KEY_PREFIX
. $key, $func, $ttl +
$staleTTL, 1 );
583 * Purge a key from all datacenters
585 * This should only be called when the underlying data (being cached)
586 * changes in a significant way. This deletes the key and starts a hold-off
587 * period where the key cannot be written to for a few seconds (HOLDOFF_TTL).
588 * This is done to avoid the following race condition:
589 * - a) Some DB data changes and delete() is called on a corresponding key
590 * - b) A request refills the key with a stale value from a lagged DB
591 * - c) The stale value is stuck there until the key is expired/evicted
593 * This is implemented by storing a special "tombstone" value at the cache
594 * key that this class recognizes; get() calls will return false for the key
595 * and any set() calls will refuse to replace tombstone values at the key.
596 * For this to always avoid stale value writes, the following must hold:
597 * - a) Replication lag is bounded to being less than HOLDOFF_TTL; or
598 * - b) If lag is higher, the DB will have gone into read-only mode already
600 * Note that set() can also be lag-aware and lower the TTL if it's high.
602 * Be aware that this does not clear the process cache. Even if it did, callbacks
603 * used by getWithSetCallback() might still return stale data in the case of either
604 * uncommitted or not-yet-replicated changes (callback generally use replica DBs).
606 * When using potentially long-running ACID transactions, a good pattern is
607 * to use a pre-commit hook to issue the delete. This means that immediately
608 * after commit, callers will see the tombstone in cache upon purge relay.
609 * It also avoids the following race condition:
610 * - a) T1 begins, changes a row, and calls delete()
611 * - b) The HOLDOFF_TTL passes, expiring the delete() tombstone
612 * - c) T2 starts, reads the row and calls set() due to a cache miss
613 * - d) T1 finally commits
614 * - e) Stale value is stuck in cache
618 * $dbw->startAtomic( __METHOD__ ); // start of request
619 * ... <execute some stuff> ...
620 * // Update the row in the DB
621 * $dbw->update( ... );
622 * $key = $cache->makeKey( 'homes', $homeId );
623 * // Purge the corresponding cache entry just before committing
624 * $dbw->onTransactionPreCommitOrIdle( function() use ( $cache, $key ) {
625 * $cache->delete( $key );
627 * ... <execute some stuff> ...
628 * $dbw->endAtomic( __METHOD__ ); // end of request
631 * The $ttl parameter can be used when purging values that have not actually changed
632 * recently. For example, a cleanup script to purge cache entries does not really need
633 * a hold-off period, so it can use HOLDOFF_NONE. Likewise for user-requested purge.
634 * Note that $ttl limits the effective range of 'lockTSE' for getWithSetCallback().
636 * If called twice on the same key, then the last hold-off TTL takes precedence. For
637 * idempotence, the $ttl should not vary for different delete() calls on the same key.
639 * @param string $key Cache key
640 * @param int $ttl Tombstone TTL; Default: WANObjectCache::HOLDOFF_TTL
641 * @return bool True if the item was purged or not found, false on failure
643 final public function delete( $key, $ttl = self
::HOLDOFF_TTL
) {
644 $key = self
::VALUE_KEY_PREFIX
. $key;
647 // Publish the purge to all datacenters
648 $ok = $this->relayDelete( $key );
650 // Publish the purge to all datacenters
651 $ok = $this->relayPurge( $key, $ttl, self
::HOLDOFF_NONE
);
658 * Fetch the value of a timestamp "check" key
660 * The key will be *initialized* to the current time if not set,
661 * so only call this method if this behavior is actually desired
663 * The timestamp can be used to check whether a cached value is valid.
664 * Callers should not assume that this returns the same timestamp in
665 * all datacenters due to relay delays.
667 * The level of staleness can roughly be estimated from this key, but
668 * if the key was evicted from cache, such calculations may show the
669 * time since expiry as ~0 seconds.
671 * Note that "check" keys won't collide with other regular keys.
674 * @return float UNIX timestamp
676 final public function getCheckKeyTime( $key ) {
677 return $this->getMultiCheckKeyTime( [ $key ] )[$key];
681 * Fetch the values of each timestamp "check" key
683 * This works like getCheckKeyTime() except it takes a list of keys
684 * and returns a map of timestamps instead of just that of one key
686 * This might be useful if both:
687 * - a) a class of entities each depend on hundreds of other entities
688 * - b) these other entities are depended upon by millions of entities
690 * The later entities can each use a "check" key to invalidate their dependee entities.
691 * However, it is expensive for the former entities to verify against all of the relevant
692 * "check" keys during each getWithSetCallback() call. A less expensive approach is to do
693 * these verifications only after a "time-till-verify" (TTV) has passed. This is a middle
694 * ground between using blind TTLs and using constant verification. The adaptiveTTL() method
695 * can be used to dynamically adjust the TTV. Also, the initial TTV can make use of the
696 * last-modified times of the dependant entities (either from the DB or the "check" keys).
700 * $value = $cache->getWithSetCallback(
701 * $cache->makeGlobalKey( 'wikibase-item', $id ),
702 * self::INITIAL_TTV, // initial time-till-verify
703 * function ( $oldValue, &$ttv, &$setOpts, $oldAsOf ) use ( $checkKeys, $cache ) {
704 * $now = microtime( true );
705 * // Use $oldValue if it passes max ultimate age and "check" key comparisons
707 * $oldAsOf > max( $cache->getMultiCheckKeyTime( $checkKeys ) ) &&
708 * ( $now - $oldValue['ctime'] ) <= self::MAX_CACHE_AGE
710 * // Increase time-till-verify by 50% of last time to reduce overhead
711 * $ttv = $cache->adaptiveTTL( $oldAsOf, self::MAX_TTV, self::MIN_TTV, 1.5 );
712 * // Unlike $oldAsOf, "ctime" is the ultimate age of the cached data
716 * $mtimes = []; // dependency last-modified times; passed by reference
717 * $value = [ 'data' => $this->fetchEntityData( $mtimes ), 'ctime' => $now ];
718 * // Guess time-till-change among the dependencies, e.g. 1/(total change rate)
719 * $ttc = 1 / array_sum( array_map(
720 * function ( $mtime ) use ( $now ) {
721 * return 1 / ( $mtime ? ( $now - $mtime ) : 900 );
725 * // The time-to-verify should not be overly pessimistic nor optimistic
726 * $ttv = min( max( $ttc, self::MIN_TTV ), self::MAX_TTV );
730 * [ 'staleTTL' => $cache::TTL_DAY ] // keep around to verify and re-save
734 * @see WANObjectCache::getCheckKeyTime()
735 * @see WANObjectCache::getWithSetCallback()
738 * @return float[] Map of (key => UNIX timestamp)
741 final public function getMultiCheckKeyTime( array $keys ) {
743 foreach ( $keys as $key ) {
744 $rawKeys[$key] = self
::TIME_KEY_PREFIX
. $key;
747 $rawValues = $this->cache
->getMulti( $rawKeys );
748 $rawValues +
= array_fill_keys( $rawKeys, false );
751 foreach ( $rawKeys as $key => $rawKey ) {
752 $purge = $this->parsePurgeValue( $rawValues[$rawKey] );
753 if ( $purge !== false ) {
754 $time = $purge[self
::FLD_TIME
];
756 // Casting assures identical floats for the next getCheckKeyTime() calls
757 $now = (string)$this->getCurrentTime();
760 $this->makePurgeValue( $now, self
::HOLDOFF_TTL
),
766 $times[$key] = $time;
773 * Purge a "check" key from all datacenters, invalidating keys that use it
775 * This should only be called when the underlying data (being cached)
776 * changes in a significant way, and it is impractical to call delete()
777 * on all keys that should be changed. When get() is called on those
778 * keys, the relevant "check" keys must be supplied for this to work.
780 * The "check" key essentially represents a last-modified time of an entity.
781 * When the key is touched, the timestamp will be updated to the current time.
782 * Keys using the "check" key via get(), getMulti(), or getWithSetCallback() will
783 * be invalidated. This approach is useful if many keys depend on a single entity.
785 * The timestamp of the "check" key is treated as being HOLDOFF_TTL seconds in the
786 * future by get*() methods in order to avoid race conditions where keys are updated
787 * with stale values (e.g. from a lagged replica DB). A high TTL is set on the "check"
788 * key, making it possible to know the timestamp of the last change to the corresponding
789 * entities in most cases. This might use more cache space than resetCheckKey().
791 * When a few important keys get a large number of hits, a high cache time is usually
792 * desired as well as "lockTSE" logic. The resetCheckKey() method is less appropriate
793 * in such cases since the "time since expiry" cannot be inferred, causing any get()
794 * after the reset to treat the key as being "hot", resulting in more stale value usage.
796 * Note that "check" keys won't collide with other regular keys.
798 * @see WANObjectCache::get()
799 * @see WANObjectCache::getWithSetCallback()
800 * @see WANObjectCache::resetCheckKey()
802 * @param string $key Cache key
803 * @param int $holdoff HOLDOFF_TTL or HOLDOFF_NONE constant
804 * @return bool True if the item was purged or not found, false on failure
806 final public function touchCheckKey( $key, $holdoff = self
::HOLDOFF_TTL
) {
807 // Publish the purge to all datacenters
808 return $this->relayPurge( self
::TIME_KEY_PREFIX
. $key, self
::CHECK_KEY_TTL
, $holdoff );
812 * Delete a "check" key from all datacenters, invalidating keys that use it
814 * This is similar to touchCheckKey() in that keys using it via get(), getMulti(),
815 * or getWithSetCallback() will be invalidated. The differences are:
816 * - a) The "check" key will be deleted from all caches and lazily
817 * re-initialized when accessed (rather than set everywhere)
818 * - b) Thus, dependent keys will be known to be stale, but not
819 * for how long (they are treated as "just" purged), which
820 * effects any lockTSE logic in getWithSetCallback()
821 * - c) Since "check" keys are initialized only on the server the key hashes
822 * to, any temporary ejection of that server will cause the value to be
823 * seen as purged as a new server will initialize the "check" key.
825 * The advantage here is that the "check" keys, which have high TTLs, will only
826 * be created when a get*() method actually uses that key. This is better when
827 * a large number of "check" keys are invalided in a short period of time.
829 * Note that "check" keys won't collide with other regular keys.
831 * @see WANObjectCache::get()
832 * @see WANObjectCache::getWithSetCallback()
833 * @see WANObjectCache::touchCheckKey()
835 * @param string $key Cache key
836 * @return bool True if the item was purged or not found, false on failure
838 final public function resetCheckKey( $key ) {
839 // Publish the purge to all datacenters
840 return $this->relayDelete( self
::TIME_KEY_PREFIX
. $key );
844 * Method to fetch/regenerate cache keys
846 * On cache miss, the key will be set to the callback result via set()
847 * (unless the callback returns false) and that result will be returned.
848 * The arguments supplied to the callback are:
849 * - $oldValue : current cache value or false if not present
850 * - &$ttl : a reference to the TTL which can be altered
851 * - &$setOpts : a reference to options for set() which can be altered
852 * - $oldAsOf : generation UNIX timestamp of $oldValue or null if not present (since 1.28)
854 * It is strongly recommended to set the 'lag' and 'since' fields to avoid race conditions
855 * that can cause stale values to get stuck at keys. Usually, callbacks ignore the current
856 * value, but it can be used to maintain "most recent X" values that come from time or
857 * sequence based source data, provided that the "as of" id/time is tracked. Note that
858 * preemptive regeneration and $checkKeys can result in a non-false current value.
860 * Usage of $checkKeys is similar to get() and getMulti(). However, rather than the caller
861 * having to inspect a "current time left" variable (e.g. $curTTL, $curTTLs), a cache
862 * regeneration will automatically be triggered using the callback.
864 * The $ttl argument and "hotTTR" option (in $opts) use time-dependant randomization
865 * to avoid stampedes. Keys that are slow to regenerate and either heavily used
866 * or subject to explicit (unpredictable) purges, may need additional mechanisms.
867 * The simplest way to avoid stampedes for such keys is to use 'lockTSE' (in $opts).
868 * If explicit purges are needed, also:
869 * - a) Pass $key into $checkKeys
870 * - b) Use touchCheckKey( $key ) instead of delete( $key )
872 * Example usage (typical key):
874 * $catInfo = $cache->getWithSetCallback(
875 * // Key to store the cached value under
876 * $cache->makeKey( 'cat-attributes', $catId ),
877 * // Time-to-live (in seconds)
878 * $cache::TTL_MINUTE,
879 * // Function that derives the new key value
880 * function ( $oldValue, &$ttl, array &$setOpts ) {
881 * $dbr = wfGetDB( DB_REPLICA );
882 * // Account for any snapshot/replica DB lag
883 * $setOpts += Database::getCacheSetOptions( $dbr );
885 * return $dbr->selectRow( ... );
890 * Example usage (key that is expensive and hot):
892 * $catConfig = $cache->getWithSetCallback(
893 * // Key to store the cached value under
894 * $cache->makeKey( 'site-cat-config' ),
895 * // Time-to-live (in seconds)
897 * // Function that derives the new key value
898 * function ( $oldValue, &$ttl, array &$setOpts ) {
899 * $dbr = wfGetDB( DB_REPLICA );
900 * // Account for any snapshot/replica DB lag
901 * $setOpts += Database::getCacheSetOptions( $dbr );
903 * return CatConfig::newFromRow( $dbr->selectRow( ... ) );
906 * // Calling touchCheckKey() on this key invalidates the cache
907 * 'checkKeys' => [ $cache->makeKey( 'site-cat-config' ) ],
908 * // Try to only let one datacenter thread manage cache updates at a time
910 * // Avoid querying cache servers multiple times in a web request
911 * 'pcTTL' => $cache::TTL_PROC_LONG
916 * Example usage (key with dynamic dependencies):
918 * $catState = $cache->getWithSetCallback(
919 * // Key to store the cached value under
920 * $cache->makeKey( 'cat-state', $cat->getId() ),
921 * // Time-to-live (seconds)
923 * // Function that derives the new key value
924 * function ( $oldValue, &$ttl, array &$setOpts ) {
925 * // Determine new value from the DB
926 * $dbr = wfGetDB( DB_REPLICA );
927 * // Account for any snapshot/replica DB lag
928 * $setOpts += Database::getCacheSetOptions( $dbr );
930 * return CatState::newFromResults( $dbr->select( ... ) );
933 * // The "check" keys that represent things the value depends on;
934 * // Calling touchCheckKey() on any of them invalidates the cache
936 * $cache->makeKey( 'sustenance-bowls', $cat->getRoomId() ),
937 * $cache->makeKey( 'people-present', $cat->getHouseId() ),
938 * $cache->makeKey( 'cat-laws', $cat->getCityId() ),
944 * Example usage (key that is expensive with too many DB dependencies for "check keys"):
946 * $catToys = $cache->getWithSetCallback(
947 * // Key to store the cached value under
948 * $cache->makeKey( 'cat-toys', $catId ),
949 * // Time-to-live (seconds)
951 * // Function that derives the new key value
952 * function ( $oldValue, &$ttl, array &$setOpts ) {
953 * // Determine new value from the DB
954 * $dbr = wfGetDB( DB_REPLICA );
955 * // Account for any snapshot/replica DB lag
956 * $setOpts += Database::getCacheSetOptions( $dbr );
958 * return CatToys::newFromResults( $dbr->select( ... ) );
961 * // Get the highest timestamp of any of the cat's toys
962 * 'touchedCallback' => function ( $value ) use ( $catId ) {
963 * $dbr = wfGetDB( DB_REPLICA );
964 * $ts = $dbr->selectField( 'cat_toys', 'MAX(ct_touched)', ... );
966 * return wfTimestampOrNull( TS_UNIX, $ts );
968 * // Avoid DB queries for repeated access
969 * 'pcTTL' => $cache::TTL_PROC_SHORT
974 * Example usage (hot key holding most recent 100 events):
976 * $lastCatActions = $cache->getWithSetCallback(
977 * // Key to store the cached value under
978 * $cache->makeKey( 'cat-last-actions', 100 ),
979 * // Time-to-live (in seconds)
981 * // Function that derives the new key value
982 * function ( $oldValue, &$ttl, array &$setOpts ) {
983 * $dbr = wfGetDB( DB_REPLICA );
984 * // Account for any snapshot/replica DB lag
985 * $setOpts += Database::getCacheSetOptions( $dbr );
987 * // Start off with the last cached list
988 * $list = $oldValue ?: [];
989 * // Fetch the last 100 relevant rows in descending order;
990 * // only fetch rows newer than $list[0] to reduce scanning
991 * $rows = iterator_to_array( $dbr->select( ... ) );
992 * // Merge them and get the new "last 100" rows
993 * return array_slice( array_merge( $new, $list ), 0, 100 );
996 * // Try to only let one datacenter thread manage cache updates at a time
998 * // Use a magic value when no cache value is ready rather than stampeding
999 * 'busyValue' => 'computing'
1004 * Example usage (key holding an LRU subkey:value map; this can avoid flooding cache with
1005 * keys for an unlimited set of (constraint,situation) pairs, thereby avoiding elevated
1006 * cache evictions and wasted memory):
1008 * $catSituationTolerabilityCache = $this->cache->getWithSetCallback(
1009 * // Group by constraint ID/hash, cat family ID/hash, or something else useful
1010 * $this->cache->makeKey( 'cat-situation-tolerability-checks', $groupKey ),
1011 * WANObjectCache::TTL_DAY, // rarely used groups should fade away
1012 * // The $scenarioKey format is $constraintId:<ID/hash of $situation>
1013 * function ( $cacheMap ) use ( $scenarioKey, $constraintId, $situation ) {
1014 * $lruCache = MapCacheLRU::newFromArray( $cacheMap ?: [], self::CACHE_SIZE );
1015 * $result = $lruCache->get( $scenarioKey ); // triggers LRU bump if present
1016 * if ( $result === null || $this->isScenarioResultExpired( $result ) ) {
1017 * $result = $this->checkScenarioTolerability( $constraintId, $situation );
1018 * $lruCache->set( $scenarioKey, $result, 3 / 8 );
1020 * // Save the new LRU cache map and reset the map's TTL
1021 * return $lruCache->toArray();
1024 * // Once map is > 1 sec old, consider refreshing
1026 * // Update within 5 seconds after "ageNew" given a 1hz cache check rate
1028 * // Avoid querying cache servers multiple times in a request; this also means
1029 * // that a request can only alter the value of any given constraint key once
1030 * 'pcTTL' => WANObjectCache::TTL_PROC_LONG
1033 * $tolerability = isset( $catSituationTolerabilityCache[$scenarioKey] )
1034 * ? $catSituationTolerabilityCache[$scenarioKey]
1035 * : $this->checkScenarioTolerability( $constraintId, $situation );
1038 * @see WANObjectCache::get()
1039 * @see WANObjectCache::set()
1041 * @param string $key Cache key made from makeKey() or makeGlobalKey()
1042 * @param int $ttl Seconds to live for key updates. Special values are:
1043 * - WANObjectCache::TTL_INDEFINITE: Cache forever (subject to LRU-style evictions)
1044 * - WANObjectCache::TTL_UNCACHEABLE: Do not cache (if the key exists, it is not deleted)
1045 * @param callable $callback Value generation function
1046 * @param array $opts Options map:
1047 * - checkKeys: List of "check" keys. The key at $key will be seen as stale when either
1048 * touchCheckKey() or resetCheckKey() is called on any of the keys in this list. This
1049 * is useful if thousands or millions of keys depend on the same entity. The entity can
1050 * simply have its "check" key updated whenever the entity is modified.
1052 * - graceTTL: If the key is invalidated (by "checkKeys") less than this many seconds ago,
1053 * consider reusing the stale value. The odds of a refresh becomes more likely over time,
1054 * becoming certain once the grace period is reached. This can reduce traffic spikes
1055 * when millions of keys are compared to the same "check" key and touchCheckKey() or
1056 * resetCheckKey() is called on that "check" key. This option is not useful for the
1057 * case of the key simply expiring on account of its TTL (use "lowTTL" instead).
1058 * Default: WANObjectCache::GRACE_TTL_NONE.
1059 * - lockTSE: If the key is tombstoned or invalidated (by "checkKeys") less than this many
1060 * seconds ago, try to have a single thread handle cache regeneration at any given time.
1061 * Other threads will try to use stale values if possible. If, on miss, the time since
1062 * expiration is low, the assumption is that the key is hot and that a stampede is worth
1063 * avoiding. Setting this above WANObjectCache::HOLDOFF_TTL makes no difference. The
1064 * higher this is set, the higher the worst-case staleness can be. This option does not
1065 * by itself handle the case of the key simply expiring on account of its TTL, so make
1066 * sure that "lowTTL" is not disabled when using this option.
1067 * Use WANObjectCache::TSE_NONE to disable this logic.
1068 * Default: WANObjectCache::TSE_NONE.
1069 * - busyValue: If no value exists and another thread is currently regenerating it, use this
1070 * as a fallback value (or a callback to generate such a value). This assures that cache
1071 * stampedes cannot happen if the value falls out of cache. This can be used as insurance
1072 * against cache regeneration becoming very slow for some reason (greater than the TTL).
1074 * - pcTTL: Process cache the value in this PHP instance for this many seconds. This avoids
1075 * network I/O when a key is read several times. This will not cache when the callback
1076 * returns false, however. Note that any purges will not be seen while process cached;
1077 * since the callback should use replica DBs and they may be lagged or have snapshot
1078 * isolation anyway, this should not typically matter.
1079 * Default: WANObjectCache::TTL_UNCACHEABLE.
1080 * - pcGroup: Process cache group to use instead of the primary one. If set, this must be
1081 * of the format ALPHANUMERIC_NAME:MAX_KEY_SIZE, e.g. "mydata:10". Use this for storing
1082 * large values, small yet numerous values, or some values with a high cost of eviction.
1083 * It is generally preferable to use a class constant when setting this value.
1084 * This has no effect unless pcTTL is used.
1085 * Default: WANObjectCache::PC_PRIMARY.
1086 * - version: Integer version number. This allows for callers to make breaking changes to
1087 * how values are stored while maintaining compatability and correct cache purges. New
1088 * versions are stored alongside older versions concurrently. Avoid storing class objects
1089 * however, as this reduces compatibility (due to serialization).
1091 * - minAsOf: Reject values if they were generated before this UNIX timestamp.
1092 * This is useful if the source of a key is suspected of having possibly changed
1093 * recently, and the caller wants any such changes to be reflected.
1094 * Default: WANObjectCache::MIN_TIMESTAMP_NONE.
1095 * - hotTTR: Expected time-till-refresh (TTR) in seconds for keys that average ~1 hit per
1096 * second (e.g. 1Hz). Keys with a hit rate higher than 1Hz will refresh sooner than this
1097 * TTR and vise versa. Such refreshes won't happen until keys are "ageNew" seconds old.
1098 * This uses randomization to avoid triggering cache stampedes. The TTR is useful at
1099 * reducing the impact of missed cache purges, since the effect of a heavily referenced
1100 * key being stale is worse than that of a rarely referenced key. Unlike simply lowering
1101 * $ttl, seldomly used keys are largely unaffected by this option, which makes it
1102 * possible to have a high hit rate for the "long-tail" of less-used keys.
1103 * Default: WANObjectCache::HOT_TTR.
1104 * - lowTTL: Consider pre-emptive updates when the current TTL (seconds) of the key is less
1105 * than this. It becomes more likely over time, becoming certain once the key is expired.
1106 * This helps avoid cache stampedes that might be triggered due to the key expiring.
1107 * Default: WANObjectCache::LOW_TTL.
1108 * - ageNew: Consider popularity refreshes only once a key reaches this age in seconds.
1109 * Default: WANObjectCache::AGE_NEW.
1110 * - staleTTL: Seconds to keep the key around if it is stale. This means that on cache
1111 * miss the callback may get $oldValue/$oldAsOf values for keys that have already been
1112 * expired for this specified time. This is useful if adaptiveTTL() is used on the old
1113 * value's as-of time when it is verified as still being correct.
1114 * Default: WANObjectCache::STALE_TTL_NONE
1115 * - touchedCallback: A callback that takes the current value and returns a UNIX timestamp
1116 * indicating the last time a dynamic dependency changed. Null can be returned if there
1117 * are no relevant dependency changes to check. This can be used to check against things
1118 * like last-modified times of files or DB timestamp fields. This should generally not be
1119 * used for small and easily queried values in a DB if the callback itself ends up doing
1120 * a similarly expensive DB query to check a timestamp. Usages of this option makes the
1121 * most sense for values that are moderately to highly expensive to regenerate and easy
1122 * to query for dependency timestamps. The use of "pcTTL" reduces timestamp queries.
1124 * @return mixed Value found or written to the key
1125 * @note Options added in 1.28: version, busyValue, hotTTR, ageNew, pcGroup, minAsOf
1126 * @note Options added in 1.31: staleTTL, graceTTL
1127 * @note Options added in 1.33: touchedCallback
1128 * @note Callable type hints are not used to avoid class-autoloading
1130 final public function getWithSetCallback( $key, $ttl, $callback, array $opts = [] ) {
1131 $pcTTL = $opts['pcTTL'] ?? self
::TTL_UNCACHEABLE
;
1133 // Try the process cache if enabled and the cache callback is not within a cache callback.
1134 // Process cache use in nested callbacks is not lag-safe with regard to HOLDOFF_TTL since
1135 // the in-memory value is further lagged than the shared one since it uses a blind TTL.
1136 if ( $pcTTL >= 0 && $this->callbackDepth
== 0 ) {
1137 $group = $opts['pcGroup'] ?? self
::PC_PRIMARY
;
1138 $procCache = $this->getProcessCache( $group );
1139 $value = $procCache->has( $key, $pcTTL ) ?
$procCache->get( $key ) : false;
1145 if ( $value === false ) {
1146 // Fetch the value over the network
1147 if ( isset( $opts['version'] ) ) {
1148 $version = $opts['version'];
1150 $cur = $this->doGetWithSetCallback(
1153 function ( $oldValue, &$ttl, &$setOpts, $oldAsOf )
1154 use ( $callback, $version ) {
1155 if ( is_array( $oldValue )
1156 && array_key_exists( self
::VFLD_DATA
, $oldValue )
1157 && array_key_exists( self
::VFLD_VERSION
, $oldValue )
1158 && $oldValue[self
::VFLD_VERSION
] === $version
1160 $oldData = $oldValue[self
::VFLD_DATA
];
1162 // VFLD_DATA is not set if an old, unversioned, key is present
1168 self
::VFLD_DATA
=> $callback( $oldData, $ttl, $setOpts, $oldAsOf ),
1169 self
::VFLD_VERSION
=> $version
1175 if ( $cur[self
::VFLD_VERSION
] === $version ) {
1176 // Value created or existed before with version; use it
1177 $value = $cur[self
::VFLD_DATA
];
1179 // Value existed before with a different version; use variant key.
1180 // Reflect purges to $key by requiring that this key value be newer.
1181 $value = $this->doGetWithSetCallback(
1182 $this->makeGlobalKey( 'WANCache-key-variant', md5( $key ), $version ),
1185 // Regenerate value if not newer than $key
1186 [ 'version' => null, 'minAsOf' => $asOf ] +
$opts
1190 $value = $this->doGetWithSetCallback( $key, $ttl, $callback, $opts );
1193 // Update the process cache if enabled
1194 if ( $procCache && $value !== false ) {
1195 $procCache->set( $key, $value );
1203 * Do the actual I/O for getWithSetCallback() when needed
1205 * @see WANObjectCache::getWithSetCallback()
1207 * @param string $key
1209 * @param callable $callback
1210 * @param array $opts Options map for getWithSetCallback()
1211 * @param float|null &$asOf Cache generation timestamp of returned value [returned]
1213 * @note Callable type hints are not used to avoid class-autoloading
1215 protected function doGetWithSetCallback( $key, $ttl, $callback, array $opts, &$asOf = null ) {
1216 $lowTTL = $opts['lowTTL'] ??
min( self
::LOW_TTL
, $ttl );
1217 $lockTSE = $opts['lockTSE'] ?? self
::TSE_NONE
;
1218 $staleTTL = $opts['staleTTL'] ?? self
::STALE_TTL_NONE
;
1219 $graceTTL = $opts['graceTTL'] ?? self
::GRACE_TTL_NONE
;
1220 $checkKeys = $opts['checkKeys'] ??
[];
1221 $busyValue = $opts['busyValue'] ??
null;
1222 $popWindow = $opts['hotTTR'] ?? self
::HOT_TTR
;
1223 $ageNew = $opts['ageNew'] ?? self
::AGE_NEW
;
1224 $minTime = $opts['minAsOf'] ?? self
::MIN_TIMESTAMP_NONE
;
1225 $versioned = isset( $opts['version'] );
1226 $touchedCallback = $opts['touchedCallback'] ??
null;
1228 // Get a collection name to describe this class of key
1229 $kClass = $this->determineKeyClass( $key );
1231 // Get the current key value
1233 $cValue = $this->get( $key, $curTTL, $checkKeys, $asOf ); // current value
1234 $value = $cValue; // return value
1236 // Apply additional dynamic expiration logic if supplied
1237 $curTTL = $this->applyTouchedCallback( $value, $asOf, $curTTL, $touchedCallback );
1239 $preCallbackTime = $this->getCurrentTime();
1240 // Determine if a cached value regeneration is needed or desired
1241 if ( $value !== false
1242 && $this->isAliveOrInGracePeriod( $curTTL, $graceTTL )
1243 && $this->isValid( $value, $versioned, $asOf, $minTime )
1245 $preemptiveRefresh = (
1246 $this->worthRefreshExpiring( $curTTL, $lowTTL ) ||
1247 $this->worthRefreshPopular( $asOf, $ageNew, $popWindow, $preCallbackTime )
1250 if ( !$preemptiveRefresh ) {
1251 $this->stats
->increment( "wanobjectcache.$kClass.hit.good" );
1254 } elseif ( $this->asyncHandler
) {
1255 // Update the cache value later, such during post-send of an HTTP request
1256 $func = $this->asyncHandler
;
1257 $func( function () use ( $key, $ttl, $callback, $opts, $asOf ) {
1258 $opts['minAsOf'] = INF
; // force a refresh
1259 $this->doGetWithSetCallback( $key, $ttl, $callback, $opts, $asOf );
1261 $this->stats
->increment( "wanobjectcache.$kClass.hit.refresh" );
1267 // A deleted key with a negative TTL left must be tombstoned
1268 $isTombstone = ( $curTTL !== null && $value === false );
1269 if ( $isTombstone && $lockTSE <= 0 ) {
1270 // Use the INTERIM value for tombstoned keys to reduce regeneration load
1271 $lockTSE = self
::INTERIM_KEY_TTL
;
1273 // Assume a key is hot if requested soon after invalidation
1274 $isHot = ( $curTTL !== null && $curTTL <= 0 && abs( $curTTL ) <= $lockTSE );
1275 // Use the mutex if there is no value and a busy fallback is given
1276 $checkBusy = ( $busyValue !== null && $value === false );
1277 // Decide whether a single thread should handle regenerations.
1278 // This avoids stampedes when $checkKeys are bumped and when preemptive
1279 // renegerations take too long. It also reduces regenerations while $key
1280 // is tombstoned. This balances cache freshness with avoiding DB load.
1281 $useMutex = ( $isHot ||
( $isTombstone && $lockTSE > 0 ) ||
$checkBusy );
1283 $lockAcquired = false;
1285 // Acquire a datacenter-local non-blocking lock
1286 if ( $this->cache
->add( self
::MUTEX_KEY_PREFIX
. $key, 1, self
::LOCK_TTL
) ) {
1287 // Lock acquired; this thread should update the key
1288 $lockAcquired = true;
1289 } elseif ( $value !== false && $this->isValid( $value, $versioned, $asOf, $minTime ) ) {
1290 $this->stats
->increment( "wanobjectcache.$kClass.hit.stale" );
1291 // If it cannot be acquired; then the stale value can be used
1294 // Use the INTERIM value for tombstoned keys to reduce regeneration load.
1295 // For hot keys, either another thread has the lock or the lock failed;
1296 // use the INTERIM value from the last thread that regenerated it.
1297 $value = $this->getInterimValue( $key, $versioned, $minTime, $asOf );
1298 if ( $value !== false ) {
1299 $this->stats
->increment( "wanobjectcache.$kClass.hit.volatile" );
1303 // Use the busy fallback value if nothing else
1304 if ( $busyValue !== null ) {
1305 $miss = is_infinite( $minTime ) ?
'renew' : 'miss';
1306 $this->stats
->increment( "wanobjectcache.$kClass.$miss.busy" );
1308 return is_callable( $busyValue ) ?
$busyValue() : $busyValue;
1313 if ( !is_callable( $callback ) ) {
1314 throw new InvalidArgumentException( "Invalid cache miss callback provided." );
1317 // Generate the new value from the callback...
1319 ++
$this->callbackDepth
;
1321 $value = call_user_func_array( $callback, [ $cValue, &$ttl, &$setOpts, $asOf ] );
1323 --$this->callbackDepth
;
1325 $valueIsCacheable = ( $value !== false && $ttl >= 0 );
1327 // When delete() is called, writes are write-holed by the tombstone,
1328 // so use a special INTERIM key to pass the new value around threads.
1329 if ( ( $isTombstone && $lockTSE > 0 ) && $valueIsCacheable ) {
1330 $tempTTL = max( 1, (int)$lockTSE ); // set() expects seconds
1331 $newAsOf = $this->getCurrentTime();
1332 $wrapped = $this->wrap( $value, $tempTTL, $newAsOf );
1333 // Avoid using set() to avoid pointless mcrouter broadcasting
1334 $this->setInterimValue( $key, $wrapped, $tempTTL );
1337 if ( $valueIsCacheable ) {
1338 $setOpts['lockTSE'] = $lockTSE;
1339 $setOpts['staleTTL'] = $staleTTL;
1340 // Use best known "since" timestamp if not provided
1341 $setOpts +
= [ 'since' => $preCallbackTime ];
1342 // Update the cache; this will fail if the key is tombstoned
1343 $this->set( $key, $value, $ttl, $setOpts );
1346 if ( $lockAcquired ) {
1347 // Avoid using delete() to avoid pointless mcrouter broadcasting
1348 $this->cache
->changeTTL( self
::MUTEX_KEY_PREFIX
. $key, (int)$preCallbackTime - 60 );
1351 $miss = is_infinite( $minTime ) ?
'renew' : 'miss';
1352 $this->stats
->increment( "wanobjectcache.$kClass.$miss.compute" );
1358 * @param mixed $value
1359 * @param float $asOf
1360 * @param float $curTTL
1361 * @param callable|null $callback
1364 protected function applyTouchedCallback( $value, $asOf, $curTTL, $callback ) {
1365 if ( $callback === null ) {
1369 if ( !is_callable( $callback ) ) {
1370 throw new InvalidArgumentException( "Invalid expiration callback provided." );
1373 if ( $value !== false ) {
1374 $touched = $callback( $value );
1375 if ( $touched !== null && $touched >= $asOf ) {
1376 $curTTL = min( $curTTL, self
::TINY_NEGATIVE
, $asOf - $touched );
1384 * @param string $key
1385 * @param bool $versioned
1386 * @param float $minTime
1387 * @param mixed &$asOf
1390 protected function getInterimValue( $key, $versioned, $minTime, &$asOf ) {
1391 if ( !$this->useInterimHoldOffCaching
) {
1392 return false; // disabled
1395 $wrapped = $this->cache
->get( self
::INTERIM_KEY_PREFIX
. $key );
1396 list( $value ) = $this->unwrap( $wrapped, $this->getCurrentTime() );
1397 if ( $value !== false && $this->isValid( $value, $versioned, $asOf, $minTime ) ) {
1398 $asOf = $wrapped[self
::FLD_TIME
];
1407 * @param string $key
1408 * @param array $wrapped
1409 * @param int $tempTTL
1411 protected function setInterimValue( $key, $wrapped, $tempTTL ) {
1412 $this->cache
->merge(
1413 self
::INTERIM_KEY_PREFIX
. $key,
1414 function () use ( $wrapped ) {
1423 * Method to fetch multiple cache keys at once with regeneration
1425 * This works the same as getWithSetCallback() except:
1426 * - a) The $keys argument expects the result of WANObjectCache::makeMultiKeys()
1427 * - b) The $callback argument expects a callback taking the following arguments:
1428 * - $id: ID of an entity to query
1429 * - $oldValue : the prior cache value or false if none was present
1430 * - &$ttl : a reference to the new value TTL in seconds
1431 * - &$setOpts : a reference to options for set() which can be altered
1432 * - $oldAsOf : generation UNIX timestamp of $oldValue or null if not present
1433 * Aside from the additional $id argument, the other arguments function the same
1434 * way they do in getWithSetCallback().
1435 * - c) The return value is a map of (cache key => value) in the order of $keyedIds
1437 * @see WANObjectCache::getWithSetCallback()
1438 * @see WANObjectCache::getMultiWithUnionSetCallback()
1442 * $rows = $cache->getMultiWithSetCallback(
1443 * // Map of cache keys to entity IDs
1444 * $cache->makeMultiKeys(
1445 * $this->fileVersionIds(),
1446 * function ( $id, WANObjectCache $cache ) {
1447 * return $cache->makeKey( 'file-version', $id );
1450 * // Time-to-live (in seconds)
1452 * // Function that derives the new key value
1453 * function ( $id, $oldValue, &$ttl, array &$setOpts ) {
1454 * $dbr = wfGetDB( DB_REPLICA );
1455 * // Account for any snapshot/replica DB lag
1456 * $setOpts += Database::getCacheSetOptions( $dbr );
1458 * // Load the row for this file
1459 * $queryInfo = File::getQueryInfo();
1460 * $row = $dbr->selectRow(
1461 * $queryInfo['tables'],
1462 * $queryInfo['fields'],
1466 * $queryInfo['joins']
1469 * return $row ? (array)$row : false;
1472 * // Process cache for 30 seconds
1474 * // Use a dedicated 500 item cache (initialized on-the-fly)
1475 * 'pcGroup' => 'file-versions:500'
1478 * $files = array_map( [ __CLASS__, 'newFromRow' ], $rows );
1481 * @param ArrayIterator $keyedIds Result of WANObjectCache::makeMultiKeys()
1482 * @param int $ttl Seconds to live for key updates
1483 * @param callable $callback Callback the yields entity regeneration callbacks
1484 * @param array $opts Options map
1485 * @return array Map of (cache key => value) in the same order as $keyedIds
1488 final public function getMultiWithSetCallback(
1489 ArrayIterator
$keyedIds, $ttl, callable
$callback, array $opts = []
1491 $valueKeys = array_keys( $keyedIds->getArrayCopy() );
1492 $checkKeys = $opts['checkKeys'] ??
[];
1493 $pcTTL = $opts['pcTTL'] ?? self
::TTL_UNCACHEABLE
;
1495 // Load required keys into process cache in one go
1496 $this->warmupCache
= $this->getRawKeysForWarmup(
1497 $this->getNonProcessCachedKeys( $valueKeys, $opts, $pcTTL ),
1500 $this->warmupKeyMisses
= 0;
1502 // Wrap $callback to match the getWithSetCallback() format while passing $id to $callback
1503 $id = null; // current entity ID
1504 $func = function ( $oldValue, &$ttl, &$setOpts, $oldAsOf ) use ( $callback, &$id ) {
1505 return $callback( $id, $oldValue, $ttl, $setOpts, $oldAsOf );
1509 foreach ( $keyedIds as $key => $id ) { // preserve order
1510 $values[$key] = $this->getWithSetCallback( $key, $ttl, $func, $opts );
1513 $this->warmupCache
= [];
1519 * Method to fetch/regenerate multiple cache keys at once
1521 * This works the same as getWithSetCallback() except:
1522 * - a) The $keys argument expects the result of WANObjectCache::makeMultiKeys()
1523 * - b) The $callback argument expects a callback returning a map of (ID => new value)
1524 * for all entity IDs in $ids and it takes the following arguments:
1525 * - $ids: a list of entity IDs that require cache regeneration
1526 * - &$ttls: a reference to the (entity ID => new TTL) map
1527 * - &$setOpts: a reference to options for set() which can be altered
1528 * - c) The return value is a map of (cache key => value) in the order of $keyedIds
1529 * - d) The "lockTSE" and "busyValue" options are ignored
1531 * @see WANObjectCache::getWithSetCallback()
1532 * @see WANObjectCache::getMultiWithSetCallback()
1536 * $rows = $cache->getMultiWithUnionSetCallback(
1537 * // Map of cache keys to entity IDs
1538 * $cache->makeMultiKeys(
1539 * $this->fileVersionIds(),
1540 * function ( $id, WANObjectCache $cache ) {
1541 * return $cache->makeKey( 'file-version', $id );
1544 * // Time-to-live (in seconds)
1546 * // Function that derives the new key value
1547 * function ( array $ids, array &$ttls, array &$setOpts ) {
1548 * $dbr = wfGetDB( DB_REPLICA );
1549 * // Account for any snapshot/replica DB lag
1550 * $setOpts += Database::getCacheSetOptions( $dbr );
1552 * // Load the rows for these files
1554 * $queryInfo = File::getQueryInfo();
1555 * $res = $dbr->select(
1556 * $queryInfo['tables'],
1557 * $queryInfo['fields'],
1561 * $queryInfo['joins']
1563 * foreach ( $res as $row ) {
1564 * $rows[$row->id] = $row;
1565 * $mtime = wfTimestamp( TS_UNIX, $row->timestamp );
1566 * $ttls[$row->id] = $this->adaptiveTTL( $mtime, $ttls[$row->id] );
1573 * $files = array_map( [ __CLASS__, 'newFromRow' ], $rows );
1576 * @param ArrayIterator $keyedIds Result of WANObjectCache::makeMultiKeys()
1577 * @param int $ttl Seconds to live for key updates
1578 * @param callable $callback Callback the yields entity regeneration callbacks
1579 * @param array $opts Options map
1580 * @return array Map of (cache key => value) in the same order as $keyedIds
1583 final public function getMultiWithUnionSetCallback(
1584 ArrayIterator
$keyedIds, $ttl, callable
$callback, array $opts = []
1586 $idsByValueKey = $keyedIds->getArrayCopy();
1587 $valueKeys = array_keys( $idsByValueKey );
1588 $checkKeys = $opts['checkKeys'] ??
[];
1589 $pcTTL = $opts['pcTTL'] ?? self
::TTL_UNCACHEABLE
;
1590 unset( $opts['lockTSE'] ); // incompatible
1591 unset( $opts['busyValue'] ); // incompatible
1593 // Load required keys into process cache in one go
1594 $keysGet = $this->getNonProcessCachedKeys( $valueKeys, $opts, $pcTTL );
1595 $this->warmupCache
= $this->getRawKeysForWarmup( $keysGet, $checkKeys );
1596 $this->warmupKeyMisses
= 0;
1598 // IDs of entities known to be in need of regeneration
1601 // Find out which keys are missing/deleted/stale
1604 $curByKey = $this->getMulti( $keysGet, $curTTLs, $checkKeys, $asOfs );
1605 foreach ( $keysGet as $key ) {
1606 if ( !array_key_exists( $key, $curByKey ) ||
$curTTLs[$key] < 0 ) {
1607 $idsRegen[] = $idsByValueKey[$key];
1611 // Run the callback to populate the regeneration value map for all required IDs
1613 $newTTLsById = array_fill_keys( $idsRegen, $ttl );
1614 $newValsById = $idsRegen ?
$callback( $idsRegen, $newTTLsById, $newSetOpts ) : [];
1616 // Wrap $callback to match the getWithSetCallback() format while passing $id to $callback
1617 $id = null; // current entity ID
1618 $func = function ( $oldValue, &$ttl, &$setOpts, $oldAsOf )
1619 use ( $callback, &$id, $newValsById, $newTTLsById, $newSetOpts )
1621 if ( array_key_exists( $id, $newValsById ) ) {
1622 // Value was already regerated as expected, so use the value in $newValsById
1623 $newValue = $newValsById[$id];
1624 $ttl = $newTTLsById[$id];
1625 $setOpts = $newSetOpts;
1627 // Pre-emptive/popularity refresh and version mismatch cases are not detected
1628 // above and thus $newValsById has no entry. Run $callback on this single entity.
1629 $ttls = [ $id => $ttl ];
1630 $newValue = $callback( [ $id ], $ttls, $setOpts )[$id];
1637 // Run the cache-aside logic using warmupCache instead of persistent cache queries
1639 foreach ( $idsByValueKey as $key => $id ) { // preserve order
1640 $values[$key] = $this->getWithSetCallback( $key, $ttl, $func, $opts );
1643 $this->warmupCache
= [];
1649 * Set a key to soon expire in the local cluster if it pre-dates $purgeTimestamp
1651 * This sets stale keys' time-to-live at HOLDOFF_TTL seconds, which both avoids
1652 * broadcasting in mcrouter setups and also avoids races with new tombstones.
1654 * @param string $key Cache key
1655 * @param int $purgeTimestamp UNIX timestamp of purge
1656 * @param bool &$isStale Whether the key is stale
1657 * @return bool Success
1660 final public function reap( $key, $purgeTimestamp, &$isStale = false ) {
1661 $minAsOf = $purgeTimestamp + self
::HOLDOFF_TTL
;
1662 $wrapped = $this->cache
->get( self
::VALUE_KEY_PREFIX
. $key );
1663 if ( is_array( $wrapped ) && $wrapped[self
::FLD_TIME
] < $minAsOf ) {
1665 $this->logger
->warning( "Reaping stale value key '$key'." );
1666 $ttlReap = self
::HOLDOFF_TTL
; // avoids races with tombstone creation
1667 $ok = $this->cache
->changeTTL( self
::VALUE_KEY_PREFIX
. $key, $ttlReap );
1669 $this->logger
->error( "Could not complete reap of key '$key'." );
1681 * Set a "check" key to soon expire in the local cluster if it pre-dates $purgeTimestamp
1683 * @param string $key Cache key
1684 * @param int $purgeTimestamp UNIX timestamp of purge
1685 * @param bool &$isStale Whether the key is stale
1686 * @return bool Success
1689 final public function reapCheckKey( $key, $purgeTimestamp, &$isStale = false ) {
1690 $purge = $this->parsePurgeValue( $this->cache
->get( self
::TIME_KEY_PREFIX
. $key ) );
1691 if ( $purge && $purge[self
::FLD_TIME
] < $purgeTimestamp ) {
1693 $this->logger
->warning( "Reaping stale check key '$key'." );
1694 $ok = $this->cache
->changeTTL( self
::TIME_KEY_PREFIX
. $key, self
::TTL_SECOND
);
1696 $this->logger
->error( "Could not complete reap of check key '$key'." );
1708 * @see BagOStuff::makeKey()
1709 * @param string $class Key class
1710 * @param string|null $component [optional] Key component (starting with a key collection name)
1711 * @return string Colon-delimited list of $keyspace followed by escaped components of $args
1714 public function makeKey( $class, $component = null ) {
1715 return $this->cache
->makeKey( ...func_get_args() );
1719 * @see BagOStuff::makeGlobalKey()
1720 * @param string $class Key class
1721 * @param string|null $component [optional] Key component (starting with a key collection name)
1722 * @return string Colon-delimited list of $keyspace followed by escaped components of $args
1725 public function makeGlobalKey( $class, $component = null ) {
1726 return $this->cache
->makeGlobalKey( ...func_get_args() );
1730 * @param array $entities List of entity IDs
1731 * @param callable $keyFunc Callback yielding a key from (entity ID, this WANObjectCache)
1732 * @return ArrayIterator Iterator yielding (cache key => entity ID) in $entities order
1735 final public function makeMultiKeys( array $entities, callable
$keyFunc ) {
1737 foreach ( $entities as $entity ) {
1738 $map[$keyFunc( $entity, $this )] = $entity;
1741 return new ArrayIterator( $map );
1745 * Get the "last error" registered; clearLastError() should be called manually
1746 * @return int ERR_* class constant for the "last error" registry
1748 final public function getLastError() {
1749 if ( $this->lastRelayError
) {
1750 // If the cache and the relayer failed, focus on the latter.
1751 // An update not making it to the relayer means it won't show up
1752 // in other DCs (nor will consistent re-hashing see up-to-date values).
1753 // On the other hand, if just the cache update failed, then it should
1754 // eventually be applied by the relayer.
1755 return $this->lastRelayError
;
1758 $code = $this->cache
->getLastError();
1760 case BagOStuff
::ERR_NONE
:
1761 return self
::ERR_NONE
;
1762 case BagOStuff
::ERR_NO_RESPONSE
:
1763 return self
::ERR_NO_RESPONSE
;
1764 case BagOStuff
::ERR_UNREACHABLE
:
1765 return self
::ERR_UNREACHABLE
;
1767 return self
::ERR_UNEXPECTED
;
1772 * Clear the "last error" registry
1774 final public function clearLastError() {
1775 $this->cache
->clearLastError();
1776 $this->lastRelayError
= self
::ERR_NONE
;
1780 * Clear the in-process caches; useful for testing
1784 public function clearProcessCache() {
1785 $this->processCaches
= [];
1789 * Enable or disable the use of brief caching for tombstoned keys
1791 * When a key is purged via delete(), there normally is a period where caching
1792 * is hold-off limited to an extremely short time. This method will disable that
1793 * caching, forcing the callback to run for any of:
1794 * - WANObjectCache::getWithSetCallback()
1795 * - WANObjectCache::getMultiWithSetCallback()
1796 * - WANObjectCache::getMultiWithUnionSetCallback()
1798 * This is useful when both:
1799 * - a) the database used by the callback is known to be up-to-date enough
1800 * for some particular purpose (e.g. replica DB has applied transaction X)
1801 * - b) the caller needs to exploit that fact, and therefore needs to avoid the
1802 * use of inherently volatile and possibly stale interim keys
1804 * @see WANObjectCache::delete()
1805 * @param bool $enabled Whether to enable interim caching
1808 final public function useInterimHoldOffCaching( $enabled ) {
1809 $this->useInterimHoldOffCaching
= $enabled;
1813 * @param int $flag ATTR_* class constant
1814 * @return int QOS_* class constant
1817 public function getQoS( $flag ) {
1818 return $this->cache
->getQoS( $flag );
1822 * Get a TTL that is higher for objects that have not changed recently
1824 * This is useful for keys that get explicit purges and DB or purge relay
1825 * lag is a potential concern (especially how it interacts with CDN cache)
1829 * // Last-modified time of page
1830 * $mtime = wfTimestamp( TS_UNIX, $page->getTimestamp() );
1831 * // Get adjusted TTL. If $mtime is 3600 seconds ago and $minTTL/$factor left at
1832 * // defaults, then $ttl is 3600 * .2 = 720. If $minTTL was greater than 720, then
1833 * // $ttl would be $minTTL. If $maxTTL was smaller than 720, $ttl would be $maxTTL.
1834 * $ttl = $cache->adaptiveTTL( $mtime, $cache::TTL_DAY );
1837 * Another use case is when there are no applicable "last modified" fields in the DB,
1838 * and there are too many dependencies for explicit purges to be viable, and the rate of
1839 * change to relevant content is unstable, and it is highly valued to have the cached value
1840 * be as up-to-date as possible.
1844 * $query = "<some complex query>";
1845 * $idListFromComplexQuery = $cache->getWithSetCallback(
1846 * $cache->makeKey( 'complex-graph-query', $hashOfQuery ),
1847 * GraphQueryClass::STARTING_TTL,
1848 * function ( $oldValue, &$ttl, array &$setOpts, $oldAsOf ) use ( $query, $cache ) {
1849 * $gdb = $this->getReplicaGraphDbConnection();
1850 * // Account for any snapshot/replica DB lag
1851 * $setOpts += GraphDatabase::getCacheSetOptions( $gdb );
1853 * $newList = iterator_to_array( $gdb->query( $query ) );
1854 * sort( $newList, SORT_NUMERIC ); // normalize
1856 * $minTTL = GraphQueryClass::MIN_TTL;
1857 * $maxTTL = GraphQueryClass::MAX_TTL;
1858 * if ( $oldValue !== false ) {
1859 * // Note that $oldAsOf is the last time this callback ran
1860 * $ttl = ( $newList === $oldValue )
1861 * // No change: cache for 150% of the age of $oldValue
1862 * ? $cache->adaptiveTTL( $oldAsOf, $maxTTL, $minTTL, 1.5 )
1863 * // Changed: cache for 50% of the age of $oldValue
1864 * : $cache->adaptiveTTL( $oldAsOf, $maxTTL, $minTTL, .5 );
1870 * // Keep stale values around for doing comparisons for TTL calculations.
1871 * // High values improve long-tail keys hit-rates, though might waste space.
1872 * 'staleTTL' => GraphQueryClass::GRACE_TTL
1877 * @param int|float $mtime UNIX timestamp
1878 * @param int $maxTTL Maximum TTL (seconds)
1879 * @param int $minTTL Minimum TTL (seconds); Default: 30
1880 * @param float $factor Value in the range (0,1); Default: .2
1881 * @return int Adaptive TTL
1884 public function adaptiveTTL( $mtime, $maxTTL, $minTTL = 30, $factor = 0.2 ) {
1885 if ( is_float( $mtime ) ||
ctype_digit( $mtime ) ) {
1886 $mtime = (int)$mtime; // handle fractional seconds and string integers
1889 if ( !is_int( $mtime ) ||
$mtime <= 0 ) {
1890 return $minTTL; // no last-modified time provided
1893 $age = $this->getCurrentTime() - $mtime;
1895 return (int)min( $maxTTL, max( $minTTL, $factor * $age ) );
1899 * @return int Number of warmup key cache misses last round
1902 final public function getWarmupKeyMisses() {
1903 return $this->warmupKeyMisses
;
1907 * Do the actual async bus purge of a key
1909 * This must set the key to "PURGED:<UNIX timestamp>:<holdoff>"
1911 * @param string $key Cache key
1912 * @param int $ttl How long to keep the tombstone [seconds]
1913 * @param int $holdoff HOLDOFF_* constant controlling how long to ignore sets for this key
1914 * @return bool Success
1916 protected function relayPurge( $key, $ttl, $holdoff ) {
1917 if ( $this->mcrouterAware
) {
1918 // See https://github.com/facebook/mcrouter/wiki/Multi-cluster-broadcast-setup
1919 // Wildcards select all matching routes, e.g. the WAN cluster on all DCs
1920 $ok = $this->cache
->set(
1921 "/*/{$this->cluster}/{$key}",
1922 $this->makePurgeValue( $this->getCurrentTime(), self
::HOLDOFF_NONE
),
1925 } elseif ( $this->purgeRelayer
instanceof EventRelayerNull
) {
1926 // This handles the mcrouter and the single-DC case
1927 $ok = $this->cache
->set(
1929 $this->makePurgeValue( $this->getCurrentTime(), self
::HOLDOFF_NONE
),
1933 $event = $this->cache
->modifySimpleRelayEvent( [
1936 'val' => 'PURGED:$UNIXTIME$:' . (int)$holdoff,
1937 'ttl' => max( $ttl, self
::TTL_SECOND
),
1938 'sbt' => true, // substitute $UNIXTIME$ with actual microtime
1941 $ok = $this->purgeRelayer
->notify( $this->purgeChannel
, $event );
1943 $this->lastRelayError
= self
::ERR_RELAY
;
1951 * Do the actual async bus delete of a key
1953 * @param string $key Cache key
1954 * @return bool Success
1956 protected function relayDelete( $key ) {
1957 if ( $this->mcrouterAware
) {
1958 // See https://github.com/facebook/mcrouter/wiki/Multi-cluster-broadcast-setup
1959 // Wildcards select all matching routes, e.g. the WAN cluster on all DCs
1960 $ok = $this->cache
->delete( "/*/{$this->cluster}/{$key}" );
1961 } elseif ( $this->purgeRelayer
instanceof EventRelayerNull
) {
1962 // Some other proxy handles broadcasting or there is only one datacenter
1963 $ok = $this->cache
->delete( $key );
1965 $event = $this->cache
->modifySimpleRelayEvent( [
1970 $ok = $this->purgeRelayer
->notify( $this->purgeChannel
, $event );
1972 $this->lastRelayError
= self
::ERR_RELAY
;
1980 * Check if a key is fresh or in the grace window and thus due for randomized reuse
1982 * If $curTTL > 0 (e.g. not expired) this returns true. Otherwise, the chance of returning
1983 * true decrease steadily from 100% to 0% as the |$curTTL| moves from 0 to $graceTTL seconds.
1984 * This handles widely varying levels of cache access traffic.
1986 * If $curTTL <= -$graceTTL (e.g. already expired), then this returns false.
1988 * @param float $curTTL Approximate TTL left on the key if present
1989 * @param int $graceTTL Consider using stale values if $curTTL is greater than this
1992 protected function isAliveOrInGracePeriod( $curTTL, $graceTTL ) {
1993 if ( $curTTL > 0 ) {
1995 } elseif ( $graceTTL <= 0 ) {
1999 $ageStale = abs( $curTTL ); // seconds of staleness
2000 $curGTTL = ( $graceTTL - $ageStale ); // current grace-time-to-live
2001 if ( $curGTTL <= 0 ) {
2002 return false; // already out of grace period
2005 // Chance of using a stale value is the complement of the chance of refreshing it
2006 return !$this->worthRefreshExpiring( $curGTTL, $graceTTL );
2010 * Check if a key is nearing expiration and thus due for randomized regeneration
2012 * This returns false if $curTTL >= $lowTTL. Otherwise, the chance of returning true
2013 * increases steadily from 0% to 100% as the $curTTL moves from $lowTTL to 0 seconds.
2014 * This handles widely varying levels of cache access traffic.
2016 * If $curTTL <= 0 (e.g. already expired), then this returns false.
2018 * @param float $curTTL Approximate TTL left on the key if present
2019 * @param float $lowTTL Consider a refresh when $curTTL is less than this
2022 protected function worthRefreshExpiring( $curTTL, $lowTTL ) {
2023 if ( $lowTTL <= 0 ) {
2025 } elseif ( $curTTL >= $lowTTL ) {
2027 } elseif ( $curTTL <= 0 ) {
2031 $chance = ( 1 - $curTTL / $lowTTL );
2033 return mt_rand( 1, 1e9
) <= 1e9
* $chance;
2037 * Check if a key is due for randomized regeneration due to its popularity
2039 * This is used so that popular keys can preemptively refresh themselves for higher
2040 * consistency (especially in the case of purge loss/delay). Unpopular keys can remain
2041 * in cache with their high nominal TTL. This means popular keys keep good consistency,
2042 * whether the data changes frequently or not, and long-tail keys get to stay in cache
2043 * and get hits too. Similar to worthRefreshExpiring(), randomization is used.
2045 * @param float $asOf UNIX timestamp of the value
2046 * @param int $ageNew Age of key when this might recommend refreshing (seconds)
2047 * @param int $timeTillRefresh Age of key when it should be refreshed if popular (seconds)
2048 * @param float $now The current UNIX timestamp
2051 protected function worthRefreshPopular( $asOf, $ageNew, $timeTillRefresh, $now ) {
2052 if ( $ageNew < 0 ||
$timeTillRefresh <= 0 ) {
2056 $age = $now - $asOf;
2057 $timeOld = $age - $ageNew;
2058 if ( $timeOld <= 0 ) {
2062 // Lifecycle is: new, ramp-up refresh chance, full refresh chance.
2063 // Note that the "expected # of refreshes" for the ramp-up time range is half of what it
2064 // would be if P(refresh) was at its full value during that time range.
2065 $refreshWindowSec = max( $timeTillRefresh - $ageNew - self
::RAMPUP_TTL
/ 2, 1 );
2066 // P(refresh) * (# hits in $refreshWindowSec) = (expected # of refreshes)
2067 // P(refresh) * ($refreshWindowSec * $popularHitsPerSec) = 1
2068 // P(refresh) = 1/($refreshWindowSec * $popularHitsPerSec)
2069 $chance = 1 / ( self
::HIT_RATE_HIGH
* $refreshWindowSec );
2071 // Ramp up $chance from 0 to its nominal value over RAMPUP_TTL seconds to avoid stampedes
2072 $chance *= ( $timeOld <= self
::RAMPUP_TTL
) ?
$timeOld / self
::RAMPUP_TTL
: 1;
2074 return mt_rand( 1, 1e9
) <= 1e9
* $chance;
2078 * Check whether $value is appropriately versioned and not older than $minTime (if set)
2080 * @param array $value
2081 * @param bool $versioned
2082 * @param float $asOf The time $value was generated
2083 * @param float $minTime The last time the main value was generated (0.0 if unknown)
2086 protected function isValid( $value, $versioned, $asOf, $minTime ) {
2087 if ( $versioned && !isset( $value[self
::VFLD_VERSION
] ) ) {
2089 } elseif ( $minTime > 0 && $asOf < $minTime ) {
2097 * Do not use this method outside WANObjectCache
2099 * @param mixed $value
2100 * @param int $ttl [0=forever]
2101 * @param float $now Unix Current timestamp just before calling set()
2104 protected function wrap( $value, $ttl, $now ) {
2106 self
::FLD_VERSION
=> self
::VERSION
,
2107 self
::FLD_VALUE
=> $value,
2108 self
::FLD_TTL
=> $ttl,
2109 self
::FLD_TIME
=> $now
2114 * Do not use this method outside WANObjectCache
2116 * @param array|string|bool $wrapped
2117 * @param float $now Unix Current timestamp (preferrably pre-query)
2118 * @return array (mixed; false if absent/tombstoned/malformed, current time left)
2120 protected function unwrap( $wrapped, $now ) {
2121 // Check if the value is a tombstone
2122 $purge = $this->parsePurgeValue( $wrapped );
2123 if ( $purge !== false ) {
2124 // Purged values should always have a negative current $ttl
2125 $curTTL = min( $purge[self
::FLD_TIME
] - $now, self
::TINY_NEGATIVE
);
2126 return [ false, $curTTL ];
2129 if ( !is_array( $wrapped ) // not found
2130 ||
!isset( $wrapped[self
::FLD_VERSION
] ) // wrong format
2131 ||
$wrapped[self
::FLD_VERSION
] !== self
::VERSION
// wrong version
2133 return [ false, null ];
2136 $flags = $wrapped[self
::FLD_FLAGS
] ??
0;
2137 if ( ( $flags & self
::FLG_STALE
) == self
::FLG_STALE
) {
2138 // Treat as expired, with the cache time as the expiration
2139 $age = $now - $wrapped[self
::FLD_TIME
];
2140 $curTTL = min( -$age, self
::TINY_NEGATIVE
);
2141 } elseif ( $wrapped[self
::FLD_TTL
] > 0 ) {
2142 // Get the approximate time left on the key
2143 $age = $now - $wrapped[self
::FLD_TIME
];
2144 $curTTL = max( $wrapped[self
::FLD_TTL
] - $age, 0.0 );
2146 // Key had no TTL, so the time left is unbounded
2150 if ( $wrapped[self
::FLD_TIME
] < $this->epoch
) {
2151 // Values this old are ignored
2152 return [ false, null ];
2155 return [ $wrapped[self
::FLD_VALUE
], $curTTL ];
2159 * @param array $keys
2160 * @param string $prefix
2163 protected static function prefixCacheKeys( array $keys, $prefix ) {
2165 foreach ( $keys as $key ) {
2166 $res[] = $prefix . $key;
2173 * @param string $key String of the format <scope>:<class>[:<class or variable>]...
2176 protected function determineKeyClass( $key ) {
2177 $parts = explode( ':', $key );
2179 return $parts[1] ??
$parts[0]; // sanity
2183 * @param string|array|bool $value Possible string of the form "PURGED:<timestamp>:<holdoff>"
2184 * @return array|bool Array containing a UNIX timestamp (float) and holdoff period (integer),
2185 * or false if value isn't a valid purge value
2187 protected function parsePurgeValue( $value ) {
2188 if ( !is_string( $value ) ) {
2192 $segments = explode( ':', $value, 3 );
2193 if ( !isset( $segments[0] ) ||
!isset( $segments[1] )
2194 ||
"{$segments[0]}:" !== self
::PURGE_VAL_PREFIX
2199 if ( !isset( $segments[2] ) ) {
2200 // Back-compat with old purge values without holdoff
2201 $segments[2] = self
::HOLDOFF_TTL
;
2204 if ( $segments[1] < $this->epoch
) {
2205 // Values this old are ignored
2210 self
::FLD_TIME
=> (float)$segments[1],
2211 self
::FLD_HOLDOFF
=> (int)$segments[2],
2216 * @param float $timestamp
2217 * @param int $holdoff In seconds
2218 * @return string Wrapped purge value
2220 protected function makePurgeValue( $timestamp, $holdoff ) {
2221 return self
::PURGE_VAL_PREFIX
. (float)$timestamp . ':' . (int)$holdoff;
2225 * @param string $group
2226 * @return MapCacheLRU
2228 protected function getProcessCache( $group ) {
2229 if ( !isset( $this->processCaches
[$group] ) ) {
2230 list( , $n ) = explode( ':', $group );
2231 $this->processCaches
[$group] = new MapCacheLRU( (int)$n );
2234 return $this->processCaches
[$group];
2238 * @param array $keys
2239 * @param array $opts
2241 * @return array List of keys
2243 private function getNonProcessCachedKeys( array $keys, array $opts, $pcTTL ) {
2245 if ( isset( $opts['pcTTL'] ) && $opts['pcTTL'] > 0 && $this->callbackDepth
== 0 ) {
2246 $pcGroup = $opts['pcGroup'] ?? self
::PC_PRIMARY
;
2247 $procCache = $this->getProcessCache( $pcGroup );
2248 foreach ( $keys as $key ) {
2249 if ( $procCache->has( $key, $pcTTL ) ) {
2250 $keysFound[] = $key;
2255 return array_diff( $keys, $keysFound );
2259 * @param array $keys
2260 * @param array $checkKeys
2261 * @return array Map of (cache key => mixed)
2263 private function getRawKeysForWarmup( array $keys, array $checkKeys ) {
2269 // Get all the value keys to fetch...
2270 foreach ( $keys as $key ) {
2271 $keysWarmUp[] = self
::VALUE_KEY_PREFIX
. $key;
2273 // Get all the check keys to fetch...
2274 foreach ( $checkKeys as $i => $checkKeyOrKeys ) {
2275 if ( is_int( $i ) ) {
2276 // Single check key that applies to all value keys
2277 $keysWarmUp[] = self
::TIME_KEY_PREFIX
. $checkKeyOrKeys;
2279 // List of check keys that apply to value key $i
2280 $keysWarmUp = array_merge(
2282 self
::prefixCacheKeys( $checkKeyOrKeys, self
::TIME_KEY_PREFIX
)
2287 $warmupCache = $this->cache
->getMulti( $keysWarmUp );
2288 $warmupCache +
= array_fill_keys( $keysWarmUp, false );
2290 return $warmupCache;
2294 * @return float UNIX timestamp
2295 * @codeCoverageIgnore
2297 protected function getCurrentTime() {
2298 if ( $this->wallClockOverride
) {
2299 return $this->wallClockOverride
;
2302 $clockTime = (float)time(); // call this first
2303 // microtime() uses an initial gettimeofday() call added to usage clocks.
2304 // This can severely drift from time() and the microtime() value of other threads
2305 // due to undercounting of the amount of time elapsed. Instead of seeing the current
2306 // time as being in the past, use the value of time(). This avoids setting cache values
2307 // that will immediately be seen as expired and possibly cause stampedes.
2308 return max( microtime( true ), $clockTime );
2312 * @param float|null &$time Mock UNIX timestamp for testing
2313 * @codeCoverageIgnore
2315 public function setMockTime( &$time ) {
2316 $this->wallClockOverride
=& $time;
2317 $this->cache
->setMockTime( $time );