3 * This program is free software; you can redistribute it and/or modify
4 * it under the terms of the GNU General Public License as published by
5 * the Free Software Foundation; either version 2 of the License, or
6 * (at your option) any later version.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 * GNU General Public License for more details.
13 * You should have received a copy of the GNU General Public License along
14 * with this program; if not, write to the Free Software Foundation, Inc.,
15 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
16 * http://www.gnu.org/copyleft/gpl.html
22 use Liuggio\StatsdClient\Factory\StatsdDataFactoryInterface
;
23 use Psr\Log\LoggerAwareInterface
;
24 use Psr\Log\LoggerInterface
;
25 use Psr\Log\NullLogger
;
28 * Multi-datacenter aware caching interface
30 * ### Using WANObjectCache
32 * All operations go to the local datacenter cache, except for delete(),
33 * touchCheckKey(), and resetCheckKey(), which broadcast to all datacenters.
35 * This class is intended for caching data from primary stores.
36 * If the get() method does not return a value, then the caller
37 * should query the new value and backfill the cache using set().
38 * The preferred way to do this logic is through getWithSetCallback().
39 * When querying the store on cache miss, the closest DB replica
40 * should be used. Try to avoid heavyweight DB master or quorum reads.
42 * To ensure consumers of the cache see new values in a timely manner,
43 * you either need to follow either the validation strategy, or the
46 * The validation strategy refers to the natural avoidance of stale data
47 * by one of the following means:
49 * - A) The cached value is immutable.
50 * If the consumer has access to an identifier that uniquely describes a value,
51 * cached value need not change. Instead, the key can change. This also allows
52 * all servers to access their perceived current version. This is important
53 * in context of multiple deployed versions of your application and/or cross-dc
54 * database replication, to ensure deterministic values without oscillation.
55 * - B) Validity is checked against the source after get().
56 * This is the inverse of A. The unique identifier is embedded inside the value
57 * and validated after on retreival. If outdated, the value is recomputed.
58 * - C) The value is cached with a modest TTL (without validation).
59 * If value recomputation is reasonably performant, and the value is allowed to
60 * be stale, one should consider using TTL only – using the value's age as
61 * method of validation.
63 * The purge strategy refers to the the approach whereby your application knows that
64 * source data has changed and can react by purging the relevant cache keys.
65 * As purges are expensive, this strategy should be avoided if possible.
66 * The simplest purge method is delete().
68 * No matter which strategy you choose, callers must not rely on updates or purges
69 * being immediately visible to other servers. It should be treated similarly as
70 * one would a database replica.
72 * The need for immediate updates should be avoided. If needed, solutions must be
73 * sought outside WANObjectCache.
75 * ### Deploying WANObjectCache
77 * There are two supported ways to set up broadcasted operations:
79 * - A) Set up mcrouter as the underlying cache backend, using a memcached BagOStuff class
80 * for the 'cache' parameter. The 'region' and 'cluster' parameters must be provided
81 * and 'mcrouterAware' must be set to `true`.
82 * Configure mcrouter as follows:
83 * - 1) Use Route Prefixing based on region (datacenter) and cache cluster.
84 * See https://github.com/facebook/mcrouter/wiki/Routing-Prefix and
85 * https://github.com/facebook/mcrouter/wiki/Multi-cluster-broadcast-setup.
86 * - 2) To increase the consistency of delete() and touchCheckKey() during cache
87 * server membership changes, you can use the OperationSelectorRoute to
88 * configure 'set' and 'delete' operations to go to all servers in the cache
89 * cluster, instead of just one server determined by hashing.
90 * See https://github.com/facebook/mcrouter/wiki/List-of-Route-Handles.
91 * - B) Set up dynomite as a cache middleware between the web servers and either memcached
92 * or redis and use it as the underlying cache backend, using a memcached BagOStuff
93 * class for the 'cache' parameter. This will broadcast all key setting operations,
94 * not just purges, which can be useful for cache warming. Writes are eventually
95 * consistent via the Dynamo replication model. See https://github.com/Netflix/dynomite.
97 * Broadcasted operations like delete() and touchCheckKey() are done asynchronously
98 * in all datacenters this way, though the local one should likely be near immediate.
100 * This means that callers in all datacenters may see older values for however many
101 * milliseconds that the purge took to reach that datacenter. As with any cache, this
102 * should not be relied on for cases where reads are used to determine writes to source
103 * (e.g. non-cache) data stores, except when reading immutable data.
105 * All values are wrapped in metadata arrays. Keys use a "WANCache:" prefix
106 * to avoid collisions with keys that are not wrapped as metadata arrays. The
107 * prefixes are as follows:
108 * - a) "WANCache:v" : used for regular value keys
109 * - b) "WANCache:i" : used for temporarily storing values of tombstoned keys
110 * - c) "WANCache:t" : used for storing timestamp "check" keys
111 * - d) "WANCache:m" : used for temporary mutex keys to avoid cache stampedes
116 class WANObjectCache
implements IExpiringStore
, LoggerAwareInterface
{
117 /** @var BagOStuff The local datacenter cache */
119 /** @var MapCacheLRU[] Map of group PHP instance caches */
120 protected $processCaches = [];
121 /** @bar bool Whether to use mcrouter key prefixing for routing */
122 protected $mcrouterAware;
123 /** @var string Physical region for mcrouter use */
125 /** @var string Cache cluster name for mcrouter use */
127 /** @var LoggerInterface */
129 /** @var StatsdDataFactoryInterface */
131 /** @var bool Whether to use "interim" caching while keys are tombstoned */
132 protected $useInterimHoldOffCaching = true;
133 /** @var callable|null Function that takes a WAN cache callback and runs it later */
134 protected $asyncHandler;
135 /** @var float Unix timestamp of the oldest possible valid values */
138 /** @var int Callback stack depth for getWithSetCallback() */
139 private $callbackDepth = 0;
140 /** @var mixed[] Temporary warm-up cache */
141 private $warmupCache = [];
142 /** @var int Key fetched */
143 private $warmupKeyMisses = 0;
145 /** @var float|null */
146 private $wallClockOverride;
148 /** Max time expected to pass between delete() and DB commit finishing */
149 const MAX_COMMIT_DELAY
= 3;
150 /** Max replication+snapshot lag before applying TTL_LAGGED or disallowing set() */
151 const MAX_READ_LAG
= 7;
152 /** Seconds to tombstone keys on delete() */
153 const HOLDOFF_TTL
= 11; // MAX_COMMIT_DELAY + MAX_READ_LAG + 1
155 /** Seconds to keep dependency purge keys around */
156 const CHECK_KEY_TTL
= self
::TTL_YEAR
;
157 /** Seconds to keep interim value keys for tombstoned keys around */
158 const INTERIM_KEY_TTL
= 1;
160 /** Seconds to keep lock keys around */
162 /** Default remaining TTL at which to consider pre-emptive regeneration */
165 /** Never consider performing "popularity" refreshes until a key reaches this age */
167 /** The time length of the "popularity" refresh window for hot keys */
169 /** Hits/second for a refresh to be expected within the "popularity" window */
170 const HIT_RATE_HIGH
= 1;
171 /** Seconds to ramp up to the "popularity" refresh chance after a key is no longer new */
172 const RAMPUP_TTL
= 30;
174 /** Idiom for getWithSetCallback() callbacks to avoid calling set() */
175 const TTL_UNCACHEABLE
= -1;
176 /** Idiom for getWithSetCallback() callbacks to 'lockTSE' logic */
178 /** Max TTL to store keys when a data sourced is lagged */
179 const TTL_LAGGED
= 30;
180 /** Idiom for delete() for "no hold-off" */
181 const HOLDOFF_NONE
= 0;
182 /** Idiom for set()/getWithSetCallback() for "do not augment the storage medium TTL" */
183 const STALE_TTL_NONE
= 0;
184 /** Idiom for set()/getWithSetCallback() for "no post-expired grace period" */
185 const GRACE_TTL_NONE
= 0;
187 /** Idiom for getWithSetCallback() for "no minimum required as-of timestamp" */
188 const MIN_TIMESTAMP_NONE
= 0.0;
190 /** Tiny negative float to use when CTL comes up >= 0 due to clock skew */
191 const TINY_NEGATIVE
= -0.000001;
193 /** Cache format version number */
196 const FLD_VERSION
= 0; // key to cache version number
197 const FLD_VALUE
= 1; // key to the cached value
198 const FLD_TTL
= 2; // key to the original TTL
199 const FLD_TIME
= 3; // key to the cache time
200 const FLD_FLAGS
= 4; // key to the flags bitfield
201 const FLD_HOLDOFF
= 5; // key to any hold-off TTL
203 /** @var int Treat this value as expired-on-arrival */
206 const ERR_NONE
= 0; // no error
207 const ERR_NO_RESPONSE
= 1; // no response
208 const ERR_UNREACHABLE
= 2; // can't connect
209 const ERR_UNEXPECTED
= 3; // response gave some error
210 const ERR_RELAY
= 4; // relay broadcast failed
212 const VALUE_KEY_PREFIX
= 'WANCache:v:';
213 const INTERIM_KEY_PREFIX
= 'WANCache:i:';
214 const TIME_KEY_PREFIX
= 'WANCache:t:';
215 const MUTEX_KEY_PREFIX
= 'WANCache:m:';
217 const PURGE_VAL_PREFIX
= 'PURGED:';
219 const VFLD_DATA
= 'WOC:d'; // key to the value of versioned data
220 const VFLD_VERSION
= 'WOC:v'; // key to the version of the value present
222 const PC_PRIMARY
= 'primary:1000'; // process cache name and max key count
224 const DEFAULT_PURGE_CHANNEL
= 'wancache-purge';
227 * @param array $params
228 * - cache : BagOStuff object for a persistent cache
229 * - logger : LoggerInterface object
230 * - stats : LoggerInterface object
231 * - asyncHandler : A function that takes a callback and runs it later. If supplied,
232 * whenever a preemptive refresh would be triggered in getWithSetCallback(), the
233 * current cache value is still used instead. However, the async-handler function
234 * receives a WAN cache callback that, when run, will execute the value generation
235 * callback supplied by the getWithSetCallback() caller. The result will be saved
236 * as normal. The handler is expected to call the WAN cache callback at an opportune
237 * time (e.g. HTTP post-send), though generally within a few 100ms. [optional]
238 * - region: the current physical region. This is required when using mcrouter as the
239 * backing store proxy. [optional]
240 * - cluster: name of the cache cluster used by this WAN cache. The name must be the
241 * same in all datacenters; the ("region","cluster") tuple is what distinguishes
242 * the counterpart cache clusters among all the datacenter. The contents of
243 * https://github.com/facebook/mcrouter/wiki/Config-Files give background on this.
244 * This is required when using mcrouter as the backing store proxy. [optional]
245 * - mcrouterAware: set as true if mcrouter is the backing store proxy and mcrouter
246 * is configured to interpret /<region>/<cluster>/ key prefixes as routes. This
247 * requires that "region" and "cluster" are both set above. [optional]
248 * - epoch: lowest UNIX timestamp a value/tombstone must have to be valid. [optional]
250 public function __construct( array $params ) {
251 $this->cache
= $params['cache'];
252 $this->region
= $params['region'] ??
'main';
253 $this->cluster
= $params['cluster'] ??
'wan-main';
254 $this->mcrouterAware
= !empty( $params['mcrouterAware'] );
255 $this->epoch
= $params['epoch'] ??
1.0;
257 $this->setLogger( $params['logger'] ??
new NullLogger() );
258 $this->stats
= $params['stats'] ??
new NullStatsdDataFactory();
259 $this->asyncHandler
= $params['asyncHandler'] ??
null;
263 * @param LoggerInterface $logger
265 public function setLogger( LoggerInterface
$logger ) {
266 $this->logger
= $logger;
270 * Get an instance that wraps EmptyBagOStuff
272 * @return WANObjectCache
274 public static function newEmpty() {
276 'cache' => new EmptyBagOStuff()
281 * Fetch the value of a key from cache
283 * If supplied, $curTTL is set to the remaining TTL (current time left):
284 * - a) INF; if $key exists, has no TTL, and is not invalidated by $checkKeys
285 * - b) float (>=0); if $key exists, has a TTL, and is not invalidated by $checkKeys
286 * - c) float (<0); if $key is tombstoned, stale, or existing but invalidated by $checkKeys
287 * - d) null; if $key does not exist and is not tombstoned
289 * If a key is tombstoned, $curTTL will reflect the time since delete().
291 * The timestamp of $key will be checked against the last-purge timestamp
292 * of each of $checkKeys. Those $checkKeys not in cache will have the last-purge
293 * initialized to the current timestamp. If any of $checkKeys have a timestamp
294 * greater than that of $key, then $curTTL will reflect how long ago $key
295 * became invalid. Callers can use $curTTL to know when the value is stale.
296 * The $checkKeys parameter allow mass invalidations by updating a single key:
297 * - a) Each "check" key represents "last purged" of some source data
298 * - b) Callers pass in relevant "check" keys as $checkKeys in get()
299 * - c) When the source data that "check" keys represent changes,
300 * the touchCheckKey() method is called on them
302 * Source data entities might exists in a DB that uses snapshot isolation
303 * (e.g. the default REPEATABLE-READ in innoDB). Even for mutable data, that
304 * isolation can largely be maintained by doing the following:
305 * - a) Calling delete() on entity change *and* creation, before DB commit
306 * - b) Keeping transaction duration shorter than the delete() hold-off TTL
307 * - c) Disabling interim key caching via useInterimHoldOffCaching() before get() calls
309 * However, pre-snapshot values might still be seen if an update was made
310 * in a remote datacenter but the purge from delete() didn't relay yet.
312 * Consider using getWithSetCallback() instead of get() and set() cycles.
313 * That method has cache slam avoiding features for hot/expensive keys.
315 * @param string $key Cache key made from makeKey() or makeGlobalKey()
316 * @param mixed|null &$curTTL Approximate TTL left on the key if present/tombstoned [returned]
317 * @param array $checkKeys List of "check" keys
318 * @param float|null &$asOf UNIX timestamp of cached value; null on failure [returned]
319 * @return mixed Value of cache key or false on failure
321 final public function get( $key, &$curTTL = null, array $checkKeys = [], &$asOf = null ) {
324 $values = $this->getMulti( [ $key ], $curTTLs, $checkKeys, $asOfs );
325 $curTTL = $curTTLs[$key] ??
null;
326 $asOf = $asOfs[$key] ??
null;
328 return $values[$key] ??
false;
332 * Fetch the value of several keys from cache
334 * @see WANObjectCache::get()
336 * @param array $keys List of cache keys made from makeKey() or makeGlobalKey()
337 * @param array &$curTTLs Map of (key => approximate TTL left) for existing keys [returned]
338 * @param array $checkKeys List of check keys to apply to all $keys. May also apply "check"
339 * keys to specific cache keys only by using cache keys as keys in the $checkKeys array.
340 * @param float[] &$asOfs Map of (key => UNIX timestamp of cached value; null on failure)
341 * @return array Map of (key => value) for keys that exist and are not tombstoned
343 final public function getMulti(
344 array $keys, &$curTTLs = [], array $checkKeys = [], array &$asOfs = []
350 $vPrefixLen = strlen( self
::VALUE_KEY_PREFIX
);
351 $valueKeys = self
::prefixCacheKeys( $keys, self
::VALUE_KEY_PREFIX
);
353 $checkKeysForAll = [];
354 $checkKeysByKey = [];
356 foreach ( $checkKeys as $i => $checkKeyGroup ) {
357 $prefixed = self
::prefixCacheKeys( (array)$checkKeyGroup, self
::TIME_KEY_PREFIX
);
358 $checkKeysFlat = array_merge( $checkKeysFlat, $prefixed );
359 // Is this check keys for a specific cache key, or for all keys being fetched?
360 if ( is_int( $i ) ) {
361 $checkKeysForAll = array_merge( $checkKeysForAll, $prefixed );
363 $checkKeysByKey[$i] = isset( $checkKeysByKey[$i] )
364 ?
array_merge( $checkKeysByKey[$i], $prefixed )
369 // Fetch all of the raw values
370 $keysGet = array_merge( $valueKeys, $checkKeysFlat );
371 if ( $this->warmupCache
) {
372 $wrappedValues = array_intersect_key( $this->warmupCache
, array_flip( $keysGet ) );
373 $keysGet = array_diff( $keysGet, array_keys( $wrappedValues ) ); // keys left to fetch
374 $this->warmupKeyMisses +
= count( $keysGet );
379 $wrappedValues +
= $this->cache
->getMulti( $keysGet );
381 // Time used to compare/init "check" keys (derived after getMulti() to be pessimistic)
382 $now = $this->getCurrentTime();
384 // Collect timestamps from all "check" keys
385 $purgeValuesForAll = $this->processCheckKeys( $checkKeysForAll, $wrappedValues, $now );
386 $purgeValuesByKey = [];
387 foreach ( $checkKeysByKey as $cacheKey => $checks ) {
388 $purgeValuesByKey[$cacheKey] =
389 $this->processCheckKeys( $checks, $wrappedValues, $now );
392 // Get the main cache value for each key and validate them
393 foreach ( $valueKeys as $vKey ) {
394 if ( !isset( $wrappedValues[$vKey] ) ) {
395 continue; // not found
398 $key = substr( $vKey, $vPrefixLen ); // unprefix
400 list( $value, $curTTL ) = $this->unwrap( $wrappedValues[$vKey], $now );
401 if ( $value !== false ) {
402 $result[$key] = $value;
403 // Force dependent keys to be seen as stale for a while after purging
404 // to reduce race conditions involving stale data getting cached
405 $purgeValues = $purgeValuesForAll;
406 if ( isset( $purgeValuesByKey[$key] ) ) {
407 $purgeValues = array_merge( $purgeValues, $purgeValuesByKey[$key] );
409 foreach ( $purgeValues as $purge ) {
410 $safeTimestamp = $purge[self
::FLD_TIME
] +
$purge[self
::FLD_HOLDOFF
];
411 if ( $safeTimestamp >= $wrappedValues[$vKey][self
::FLD_TIME
] ) {
412 // How long ago this value was invalidated by *this* check key
413 $ago = min( $purge[self
::FLD_TIME
] - $now, self
::TINY_NEGATIVE
);
414 // How long ago this value was invalidated by *any* known check key
415 $curTTL = min( $curTTL, $ago );
419 $curTTLs[$key] = $curTTL;
420 $asOfs[$key] = ( $value !== false ) ?
$wrappedValues[$vKey][self
::FLD_TIME
] : null;
428 * @param array $timeKeys List of prefixed time check keys
429 * @param array $wrappedValues
431 * @return array List of purge value arrays
433 private function processCheckKeys( array $timeKeys, array $wrappedValues, $now ) {
435 foreach ( $timeKeys as $timeKey ) {
436 $purge = isset( $wrappedValues[$timeKey] )
437 ?
$this->parsePurgeValue( $wrappedValues[$timeKey] )
439 if ( $purge === false ) {
440 // Key is not set or malformed; regenerate
441 $newVal = $this->makePurgeValue( $now, self
::HOLDOFF_TTL
);
442 $this->cache
->add( $timeKey, $newVal, self
::CHECK_KEY_TTL
);
443 $purge = $this->parsePurgeValue( $newVal );
445 $purgeValues[] = $purge;
451 * Set the value of a key in cache
453 * Simply calling this method when source data changes is not valid because
454 * the changes do not replicate to the other WAN sites. In that case, delete()
455 * should be used instead. This method is intended for use on cache misses.
457 * If the data was read from a snapshot-isolated transactions (e.g. the default
458 * REPEATABLE-READ in innoDB), use 'since' to avoid the following race condition:
460 * - b) T2 updates a row, calls delete(), and commits
461 * - c) The HOLDOFF_TTL passes, expiring the delete() tombstone
462 * - d) T1 reads the row and calls set() due to a cache miss
463 * - e) Stale value is stuck in cache
465 * Setting 'lag' and 'since' help avoids keys getting stuck in stale states.
467 * Be aware that this does not update the process cache for getWithSetCallback()
468 * callers. Keys accessed via that method are not generally meant to also be set
469 * using this primitive method.
471 * Do not use this method on versioned keys accessed via getWithSetCallback().
475 * $dbr = wfGetDB( DB_REPLICA );
476 * $setOpts = Database::getCacheSetOptions( $dbr );
477 * // Fetch the row from the DB
478 * $row = $dbr->selectRow( ... );
479 * $key = $cache->makeKey( 'building', $buildingId );
480 * $cache->set( $key, $row, $cache::TTL_DAY, $setOpts );
483 * @param string $key Cache key
484 * @param mixed $value
485 * @param int $ttl Seconds to live. Special values are:
486 * - WANObjectCache::TTL_INDEFINITE: Cache forever
487 * @param array $opts Options map:
488 * - lag : Seconds of replica DB lag. Typically, this is either the replica DB lag
489 * before the data was read or, if applicable, the replica DB lag before
490 * the snapshot-isolated transaction the data was read from started.
491 * Use false to indicate that replication is not running.
493 * - since : UNIX timestamp of the data in $value. Typically, this is either
494 * the current time the data was read or (if applicable) the time when
495 * the snapshot-isolated transaction the data was read from started.
497 * - pending : Whether this data is possibly from an uncommitted write transaction.
498 * Generally, other threads should not see values from the future and
499 * they certainly should not see ones that ended up getting rolled back.
501 * - lockTSE : if excessive replication/snapshot lag is detected, then store the value
502 * with this TTL and flag it as stale. This is only useful if the reads for this key
503 * use getWithSetCallback() with "lockTSE" set. Note that if "staleTTL" is set
504 * then it will still add on to this TTL in the excessive lag scenario.
505 * Default: WANObjectCache::TSE_NONE
506 * - staleTTL : Seconds to keep the key around if it is stale. The get()/getMulti()
507 * methods return such stale values with a $curTTL of 0, and getWithSetCallback()
508 * will call the regeneration callback in such cases, passing in the old value
509 * and its as-of time to the callback. This is useful if adaptiveTTL() is used
510 * on the old value's as-of time when it is verified as still being correct.
511 * Default: WANObjectCache::STALE_TTL_NONE.
512 * @note Options added in 1.28: staleTTL
513 * @return bool Success
515 final public function set( $key, $value, $ttl = 0, array $opts = [] ) {
516 $now = $this->getCurrentTime();
517 $lockTSE = $opts['lockTSE'] ?? self
::TSE_NONE
;
518 $staleTTL = $opts['staleTTL'] ?? self
::STALE_TTL_NONE
;
519 $age = isset( $opts['since'] ) ?
max( 0, $now - $opts['since'] ) : 0;
520 $lag = $opts['lag'] ??
0;
522 // Do not cache potentially uncommitted data as it might get rolled back
523 if ( !empty( $opts['pending'] ) ) {
524 $this->logger
->info( 'Rejected set() for {cachekey} due to pending writes.',
525 [ 'cachekey' => $key ] );
527 return true; // no-op the write for being unsafe
530 $wrapExtra = []; // additional wrapped value fields
531 // Check if there's a risk of writing stale data after the purge tombstone expired
532 if ( $lag === false ||
( $lag +
$age ) > self
::MAX_READ_LAG
) {
533 // Case A: read lag with "lockTSE"; save but record value as stale
534 if ( $lockTSE >= 0 ) {
535 $ttl = max( 1, (int)$lockTSE ); // set() expects seconds
536 $wrapExtra[self
::FLD_FLAGS
] = self
::FLG_STALE
; // mark as stale
537 // Case B: any long-running transaction; ignore this set()
538 } elseif ( $age > self
::MAX_READ_LAG
) {
539 $this->logger
->info( 'Rejected set() for {cachekey} due to snapshot lag.',
540 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ] );
542 return true; // no-op the write for being unsafe
543 // Case C: high replication lag; lower TTL instead of ignoring all set()s
544 } elseif ( $lag === false ||
$lag > self
::MAX_READ_LAG
) {
545 $ttl = $ttl ?
min( $ttl, self
::TTL_LAGGED
) : self
::TTL_LAGGED
;
546 $this->logger
->warning( 'Lowered set() TTL for {cachekey} due to replication lag.',
547 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ] );
548 // Case D: medium length request with medium replication lag; ignore this set()
550 $this->logger
->info( 'Rejected set() for {cachekey} due to high read lag.',
551 [ 'cachekey' => $key, 'lag' => $lag, 'age' => $age ] );
553 return true; // no-op the write for being unsafe
557 // Wrap that value with time/TTL/version metadata
558 $wrapped = $this->wrap( $value, $ttl, $now ) +
$wrapExtra;
560 $func = function ( $cache, $key, $cWrapped ) use ( $wrapped ) {
561 return ( is_string( $cWrapped ) )
562 ?
false // key is tombstoned; do nothing
566 return $this->cache
->merge( self
::VALUE_KEY_PREFIX
. $key, $func, $ttl +
$staleTTL, 1 );
570 * Purge a key from all datacenters
572 * This should only be called when the underlying data (being cached)
573 * changes in a significant way. This deletes the key and starts a hold-off
574 * period where the key cannot be written to for a few seconds (HOLDOFF_TTL).
575 * This is done to avoid the following race condition:
576 * - a) Some DB data changes and delete() is called on a corresponding key
577 * - b) A request refills the key with a stale value from a lagged DB
578 * - c) The stale value is stuck there until the key is expired/evicted
580 * This is implemented by storing a special "tombstone" value at the cache
581 * key that this class recognizes; get() calls will return false for the key
582 * and any set() calls will refuse to replace tombstone values at the key.
583 * For this to always avoid stale value writes, the following must hold:
584 * - a) Replication lag is bounded to being less than HOLDOFF_TTL; or
585 * - b) If lag is higher, the DB will have gone into read-only mode already
587 * Note that set() can also be lag-aware and lower the TTL if it's high.
589 * Be aware that this does not clear the process cache. Even if it did, callbacks
590 * used by getWithSetCallback() might still return stale data in the case of either
591 * uncommitted or not-yet-replicated changes (callback generally use replica DBs).
593 * When using potentially long-running ACID transactions, a good pattern is
594 * to use a pre-commit hook to issue the delete. This means that immediately
595 * after commit, callers will see the tombstone in cache upon purge relay.
596 * It also avoids the following race condition:
597 * - a) T1 begins, changes a row, and calls delete()
598 * - b) The HOLDOFF_TTL passes, expiring the delete() tombstone
599 * - c) T2 starts, reads the row and calls set() due to a cache miss
600 * - d) T1 finally commits
601 * - e) Stale value is stuck in cache
605 * $dbw->startAtomic( __METHOD__ ); // start of request
606 * ... <execute some stuff> ...
607 * // Update the row in the DB
608 * $dbw->update( ... );
609 * $key = $cache->makeKey( 'homes', $homeId );
610 * // Purge the corresponding cache entry just before committing
611 * $dbw->onTransactionPreCommitOrIdle( function() use ( $cache, $key ) {
612 * $cache->delete( $key );
614 * ... <execute some stuff> ...
615 * $dbw->endAtomic( __METHOD__ ); // end of request
618 * The $ttl parameter can be used when purging values that have not actually changed
619 * recently. For example, a cleanup script to purge cache entries does not really need
620 * a hold-off period, so it can use HOLDOFF_NONE. Likewise for user-requested purge.
621 * Note that $ttl limits the effective range of 'lockTSE' for getWithSetCallback().
623 * If called twice on the same key, then the last hold-off TTL takes precedence. For
624 * idempotence, the $ttl should not vary for different delete() calls on the same key.
626 * @param string $key Cache key
627 * @param int $ttl Tombstone TTL; Default: WANObjectCache::HOLDOFF_TTL
628 * @return bool True if the item was purged or not found, false on failure
630 final public function delete( $key, $ttl = self
::HOLDOFF_TTL
) {
631 $key = self
::VALUE_KEY_PREFIX
. $key;
634 // Publish the purge to all datacenters
635 $ok = $this->relayDelete( $key );
637 // Publish the purge to all datacenters
638 $ok = $this->relayPurge( $key, $ttl, self
::HOLDOFF_NONE
);
645 * Fetch the value of a timestamp "check" key
647 * The key will be *initialized* to the current time if not set,
648 * so only call this method if this behavior is actually desired
650 * The timestamp can be used to check whether a cached value is valid.
651 * Callers should not assume that this returns the same timestamp in
652 * all datacenters due to relay delays.
654 * The level of staleness can roughly be estimated from this key, but
655 * if the key was evicted from cache, such calculations may show the
656 * time since expiry as ~0 seconds.
658 * Note that "check" keys won't collide with other regular keys.
661 * @return float UNIX timestamp
663 final public function getCheckKeyTime( $key ) {
664 return $this->getMultiCheckKeyTime( [ $key ] )[$key];
668 * Fetch the values of each timestamp "check" key
670 * This works like getCheckKeyTime() except it takes a list of keys
671 * and returns a map of timestamps instead of just that of one key
673 * This might be useful if both:
674 * - a) a class of entities each depend on hundreds of other entities
675 * - b) these other entities are depended upon by millions of entities
677 * The later entities can each use a "check" key to invalidate their dependee entities.
678 * However, it is expensive for the former entities to verify against all of the relevant
679 * "check" keys during each getWithSetCallback() call. A less expensive approach is to do
680 * these verifications only after a "time-till-verify" (TTV) has passed. This is a middle
681 * ground between using blind TTLs and using constant verification. The adaptiveTTL() method
682 * can be used to dynamically adjust the TTV. Also, the initial TTV can make use of the
683 * last-modified times of the dependant entities (either from the DB or the "check" keys).
687 * $value = $cache->getWithSetCallback(
688 * $cache->makeGlobalKey( 'wikibase-item', $id ),
689 * self::INITIAL_TTV, // initial time-till-verify
690 * function ( $oldValue, &$ttv, &$setOpts, $oldAsOf ) use ( $checkKeys, $cache ) {
691 * $now = microtime( true );
692 * // Use $oldValue if it passes max ultimate age and "check" key comparisons
694 * $oldAsOf > max( $cache->getMultiCheckKeyTime( $checkKeys ) ) &&
695 * ( $now - $oldValue['ctime'] ) <= self::MAX_CACHE_AGE
697 * // Increase time-till-verify by 50% of last time to reduce overhead
698 * $ttv = $cache->adaptiveTTL( $oldAsOf, self::MAX_TTV, self::MIN_TTV, 1.5 );
699 * // Unlike $oldAsOf, "ctime" is the ultimate age of the cached data
703 * $mtimes = []; // dependency last-modified times; passed by reference
704 * $value = [ 'data' => $this->fetchEntityData( $mtimes ), 'ctime' => $now ];
705 * // Guess time-till-change among the dependencies, e.g. 1/(total change rate)
706 * $ttc = 1 / array_sum( array_map(
707 * function ( $mtime ) use ( $now ) {
708 * return 1 / ( $mtime ? ( $now - $mtime ) : 900 );
712 * // The time-to-verify should not be overly pessimistic nor optimistic
713 * $ttv = min( max( $ttc, self::MIN_TTV ), self::MAX_TTV );
717 * [ 'staleTTL' => $cache::TTL_DAY ] // keep around to verify and re-save
721 * @see WANObjectCache::getCheckKeyTime()
722 * @see WANObjectCache::getWithSetCallback()
725 * @return float[] Map of (key => UNIX timestamp)
728 final public function getMultiCheckKeyTime( array $keys ) {
730 foreach ( $keys as $key ) {
731 $rawKeys[$key] = self
::TIME_KEY_PREFIX
. $key;
734 $rawValues = $this->cache
->getMulti( $rawKeys );
735 $rawValues +
= array_fill_keys( $rawKeys, false );
738 foreach ( $rawKeys as $key => $rawKey ) {
739 $purge = $this->parsePurgeValue( $rawValues[$rawKey] );
740 if ( $purge !== false ) {
741 $time = $purge[self
::FLD_TIME
];
743 // Casting assures identical floats for the next getCheckKeyTime() calls
744 $now = (string)$this->getCurrentTime();
747 $this->makePurgeValue( $now, self
::HOLDOFF_TTL
),
753 $times[$key] = $time;
760 * Purge a "check" key from all datacenters, invalidating keys that use it
762 * This should only be called when the underlying data (being cached)
763 * changes in a significant way, and it is impractical to call delete()
764 * on all keys that should be changed. When get() is called on those
765 * keys, the relevant "check" keys must be supplied for this to work.
767 * The "check" key essentially represents a last-modified time of an entity.
768 * When the key is touched, the timestamp will be updated to the current time.
769 * Keys using the "check" key via get(), getMulti(), or getWithSetCallback() will
770 * be invalidated. This approach is useful if many keys depend on a single entity.
772 * The timestamp of the "check" key is treated as being HOLDOFF_TTL seconds in the
773 * future by get*() methods in order to avoid race conditions where keys are updated
774 * with stale values (e.g. from a lagged replica DB). A high TTL is set on the "check"
775 * key, making it possible to know the timestamp of the last change to the corresponding
776 * entities in most cases. This might use more cache space than resetCheckKey().
778 * When a few important keys get a large number of hits, a high cache time is usually
779 * desired as well as "lockTSE" logic. The resetCheckKey() method is less appropriate
780 * in such cases since the "time since expiry" cannot be inferred, causing any get()
781 * after the reset to treat the key as being "hot", resulting in more stale value usage.
783 * Note that "check" keys won't collide with other regular keys.
785 * @see WANObjectCache::get()
786 * @see WANObjectCache::getWithSetCallback()
787 * @see WANObjectCache::resetCheckKey()
789 * @param string $key Cache key
790 * @param int $holdoff HOLDOFF_TTL or HOLDOFF_NONE constant
791 * @return bool True if the item was purged or not found, false on failure
793 final public function touchCheckKey( $key, $holdoff = self
::HOLDOFF_TTL
) {
794 // Publish the purge to all datacenters
795 return $this->relayPurge( self
::TIME_KEY_PREFIX
. $key, self
::CHECK_KEY_TTL
, $holdoff );
799 * Delete a "check" key from all datacenters, invalidating keys that use it
801 * This is similar to touchCheckKey() in that keys using it via get(), getMulti(),
802 * or getWithSetCallback() will be invalidated. The differences are:
803 * - a) The "check" key will be deleted from all caches and lazily
804 * re-initialized when accessed (rather than set everywhere)
805 * - b) Thus, dependent keys will be known to be stale, but not
806 * for how long (they are treated as "just" purged), which
807 * effects any lockTSE logic in getWithSetCallback()
808 * - c) Since "check" keys are initialized only on the server the key hashes
809 * to, any temporary ejection of that server will cause the value to be
810 * seen as purged as a new server will initialize the "check" key.
812 * The advantage here is that the "check" keys, which have high TTLs, will only
813 * be created when a get*() method actually uses that key. This is better when
814 * a large number of "check" keys are invalided in a short period of time.
816 * Note that "check" keys won't collide with other regular keys.
818 * @see WANObjectCache::get()
819 * @see WANObjectCache::getWithSetCallback()
820 * @see WANObjectCache::touchCheckKey()
822 * @param string $key Cache key
823 * @return bool True if the item was purged or not found, false on failure
825 final public function resetCheckKey( $key ) {
826 // Publish the purge to all datacenters
827 return $this->relayDelete( self
::TIME_KEY_PREFIX
. $key );
831 * Method to fetch/regenerate cache keys
833 * On cache miss, the key will be set to the callback result via set()
834 * (unless the callback returns false) and that result will be returned.
835 * The arguments supplied to the callback are:
836 * - $oldValue : current cache value or false if not present
837 * - &$ttl : a reference to the TTL which can be altered
838 * - &$setOpts : a reference to options for set() which can be altered
839 * - $oldAsOf : generation UNIX timestamp of $oldValue or null if not present (since 1.28)
841 * It is strongly recommended to set the 'lag' and 'since' fields to avoid race conditions
842 * that can cause stale values to get stuck at keys. Usually, callbacks ignore the current
843 * value, but it can be used to maintain "most recent X" values that come from time or
844 * sequence based source data, provided that the "as of" id/time is tracked. Note that
845 * preemptive regeneration and $checkKeys can result in a non-false current value.
847 * Usage of $checkKeys is similar to get() and getMulti(). However, rather than the caller
848 * having to inspect a "current time left" variable (e.g. $curTTL, $curTTLs), a cache
849 * regeneration will automatically be triggered using the callback.
851 * The $ttl argument and "hotTTR" option (in $opts) use time-dependant randomization
852 * to avoid stampedes. Keys that are slow to regenerate and either heavily used
853 * or subject to explicit (unpredictable) purges, may need additional mechanisms.
854 * The simplest way to avoid stampedes for such keys is to use 'lockTSE' (in $opts).
855 * If explicit purges are needed, also:
856 * - a) Pass $key into $checkKeys
857 * - b) Use touchCheckKey( $key ) instead of delete( $key )
859 * Example usage (typical key):
861 * $catInfo = $cache->getWithSetCallback(
862 * // Key to store the cached value under
863 * $cache->makeKey( 'cat-attributes', $catId ),
864 * // Time-to-live (in seconds)
865 * $cache::TTL_MINUTE,
866 * // Function that derives the new key value
867 * function ( $oldValue, &$ttl, array &$setOpts ) {
868 * $dbr = wfGetDB( DB_REPLICA );
869 * // Account for any snapshot/replica DB lag
870 * $setOpts += Database::getCacheSetOptions( $dbr );
872 * return $dbr->selectRow( ... );
877 * Example usage (key that is expensive and hot):
879 * $catConfig = $cache->getWithSetCallback(
880 * // Key to store the cached value under
881 * $cache->makeKey( 'site-cat-config' ),
882 * // Time-to-live (in seconds)
884 * // Function that derives the new key value
885 * function ( $oldValue, &$ttl, array &$setOpts ) {
886 * $dbr = wfGetDB( DB_REPLICA );
887 * // Account for any snapshot/replica DB lag
888 * $setOpts += Database::getCacheSetOptions( $dbr );
890 * return CatConfig::newFromRow( $dbr->selectRow( ... ) );
893 * // Calling touchCheckKey() on this key invalidates the cache
894 * 'checkKeys' => [ $cache->makeKey( 'site-cat-config' ) ],
895 * // Try to only let one datacenter thread manage cache updates at a time
897 * // Avoid querying cache servers multiple times in a web request
898 * 'pcTTL' => $cache::TTL_PROC_LONG
903 * Example usage (key with dynamic dependencies):
905 * $catState = $cache->getWithSetCallback(
906 * // Key to store the cached value under
907 * $cache->makeKey( 'cat-state', $cat->getId() ),
908 * // Time-to-live (seconds)
910 * // Function that derives the new key value
911 * function ( $oldValue, &$ttl, array &$setOpts ) {
912 * // Determine new value from the DB
913 * $dbr = wfGetDB( DB_REPLICA );
914 * // Account for any snapshot/replica DB lag
915 * $setOpts += Database::getCacheSetOptions( $dbr );
917 * return CatState::newFromResults( $dbr->select( ... ) );
920 * // The "check" keys that represent things the value depends on;
921 * // Calling touchCheckKey() on any of them invalidates the cache
923 * $cache->makeKey( 'sustenance-bowls', $cat->getRoomId() ),
924 * $cache->makeKey( 'people-present', $cat->getHouseId() ),
925 * $cache->makeKey( 'cat-laws', $cat->getCityId() ),
931 * Example usage (key that is expensive with too many DB dependencies for "check keys"):
933 * $catToys = $cache->getWithSetCallback(
934 * // Key to store the cached value under
935 * $cache->makeKey( 'cat-toys', $catId ),
936 * // Time-to-live (seconds)
938 * // Function that derives the new key value
939 * function ( $oldValue, &$ttl, array &$setOpts ) {
940 * // Determine new value from the DB
941 * $dbr = wfGetDB( DB_REPLICA );
942 * // Account for any snapshot/replica DB lag
943 * $setOpts += Database::getCacheSetOptions( $dbr );
945 * return CatToys::newFromResults( $dbr->select( ... ) );
948 * // Get the highest timestamp of any of the cat's toys
949 * 'touchedCallback' => function ( $value ) use ( $catId ) {
950 * $dbr = wfGetDB( DB_REPLICA );
951 * $ts = $dbr->selectField( 'cat_toys', 'MAX(ct_touched)', ... );
953 * return wfTimestampOrNull( TS_UNIX, $ts );
955 * // Avoid DB queries for repeated access
956 * 'pcTTL' => $cache::TTL_PROC_SHORT
961 * Example usage (hot key holding most recent 100 events):
963 * $lastCatActions = $cache->getWithSetCallback(
964 * // Key to store the cached value under
965 * $cache->makeKey( 'cat-last-actions', 100 ),
966 * // Time-to-live (in seconds)
968 * // Function that derives the new key value
969 * function ( $oldValue, &$ttl, array &$setOpts ) {
970 * $dbr = wfGetDB( DB_REPLICA );
971 * // Account for any snapshot/replica DB lag
972 * $setOpts += Database::getCacheSetOptions( $dbr );
974 * // Start off with the last cached list
975 * $list = $oldValue ?: [];
976 * // Fetch the last 100 relevant rows in descending order;
977 * // only fetch rows newer than $list[0] to reduce scanning
978 * $rows = iterator_to_array( $dbr->select( ... ) );
979 * // Merge them and get the new "last 100" rows
980 * return array_slice( array_merge( $new, $list ), 0, 100 );
983 * // Try to only let one datacenter thread manage cache updates at a time
985 * // Use a magic value when no cache value is ready rather than stampeding
986 * 'busyValue' => 'computing'
991 * Example usage (key holding an LRU subkey:value map; this can avoid flooding cache with
992 * keys for an unlimited set of (constraint,situation) pairs, thereby avoiding elevated
993 * cache evictions and wasted memory):
995 * $catSituationTolerabilityCache = $this->cache->getWithSetCallback(
996 * // Group by constraint ID/hash, cat family ID/hash, or something else useful
997 * $this->cache->makeKey( 'cat-situation-tolerability-checks', $groupKey ),
998 * WANObjectCache::TTL_DAY, // rarely used groups should fade away
999 * // The $scenarioKey format is $constraintId:<ID/hash of $situation>
1000 * function ( $cacheMap ) use ( $scenarioKey, $constraintId, $situation ) {
1001 * $lruCache = MapCacheLRU::newFromArray( $cacheMap ?: [], self::CACHE_SIZE );
1002 * $result = $lruCache->get( $scenarioKey ); // triggers LRU bump if present
1003 * if ( $result === null || $this->isScenarioResultExpired( $result ) ) {
1004 * $result = $this->checkScenarioTolerability( $constraintId, $situation );
1005 * $lruCache->set( $scenarioKey, $result, 3 / 8 );
1007 * // Save the new LRU cache map and reset the map's TTL
1008 * return $lruCache->toArray();
1011 * // Once map is > 1 sec old, consider refreshing
1013 * // Update within 5 seconds after "ageNew" given a 1hz cache check rate
1015 * // Avoid querying cache servers multiple times in a request; this also means
1016 * // that a request can only alter the value of any given constraint key once
1017 * 'pcTTL' => WANObjectCache::TTL_PROC_LONG
1020 * $tolerability = isset( $catSituationTolerabilityCache[$scenarioKey] )
1021 * ? $catSituationTolerabilityCache[$scenarioKey]
1022 * : $this->checkScenarioTolerability( $constraintId, $situation );
1025 * @see WANObjectCache::get()
1026 * @see WANObjectCache::set()
1028 * @param string $key Cache key made from makeKey() or makeGlobalKey()
1029 * @param int $ttl Seconds to live for key updates. Special values are:
1030 * - WANObjectCache::TTL_INDEFINITE: Cache forever (subject to LRU-style evictions)
1031 * - WANObjectCache::TTL_UNCACHEABLE: Do not cache (if the key exists, it is not deleted)
1032 * @param callable $callback Value generation function
1033 * @param array $opts Options map:
1034 * - checkKeys: List of "check" keys. The key at $key will be seen as stale when either
1035 * touchCheckKey() or resetCheckKey() is called on any of the keys in this list. This
1036 * is useful if thousands or millions of keys depend on the same entity. The entity can
1037 * simply have its "check" key updated whenever the entity is modified.
1039 * - graceTTL: If the key is invalidated (by "checkKeys") less than this many seconds ago,
1040 * consider reusing the stale value. The odds of a refresh becomes more likely over time,
1041 * becoming certain once the grace period is reached. This can reduce traffic spikes
1042 * when millions of keys are compared to the same "check" key and touchCheckKey() or
1043 * resetCheckKey() is called on that "check" key. This option is not useful for the
1044 * case of the key simply expiring on account of its TTL (use "lowTTL" instead).
1045 * Default: WANObjectCache::GRACE_TTL_NONE.
1046 * - lockTSE: If the key is tombstoned or invalidated (by "checkKeys") less than this many
1047 * seconds ago, try to have a single thread handle cache regeneration at any given time.
1048 * Other threads will try to use stale values if possible. If, on miss, the time since
1049 * expiration is low, the assumption is that the key is hot and that a stampede is worth
1050 * avoiding. Setting this above WANObjectCache::HOLDOFF_TTL makes no difference. The
1051 * higher this is set, the higher the worst-case staleness can be. This option does not
1052 * by itself handle the case of the key simply expiring on account of its TTL, so make
1053 * sure that "lowTTL" is not disabled when using this option.
1054 * Use WANObjectCache::TSE_NONE to disable this logic.
1055 * Default: WANObjectCache::TSE_NONE.
1056 * - busyValue: If no value exists and another thread is currently regenerating it, use this
1057 * as a fallback value (or a callback to generate such a value). This assures that cache
1058 * stampedes cannot happen if the value falls out of cache. This can be used as insurance
1059 * against cache regeneration becoming very slow for some reason (greater than the TTL).
1061 * - pcTTL: Process cache the value in this PHP instance for this many seconds. This avoids
1062 * network I/O when a key is read several times. This will not cache when the callback
1063 * returns false, however. Note that any purges will not be seen while process cached;
1064 * since the callback should use replica DBs and they may be lagged or have snapshot
1065 * isolation anyway, this should not typically matter.
1066 * Default: WANObjectCache::TTL_UNCACHEABLE.
1067 * - pcGroup: Process cache group to use instead of the primary one. If set, this must be
1068 * of the format ALPHANUMERIC_NAME:MAX_KEY_SIZE, e.g. "mydata:10". Use this for storing
1069 * large values, small yet numerous values, or some values with a high cost of eviction.
1070 * It is generally preferable to use a class constant when setting this value.
1071 * This has no effect unless pcTTL is used.
1072 * Default: WANObjectCache::PC_PRIMARY.
1073 * - version: Integer version number. This allows for callers to make breaking changes to
1074 * how values are stored while maintaining compatability and correct cache purges. New
1075 * versions are stored alongside older versions concurrently. Avoid storing class objects
1076 * however, as this reduces compatibility (due to serialization).
1078 * - minAsOf: Reject values if they were generated before this UNIX timestamp.
1079 * This is useful if the source of a key is suspected of having possibly changed
1080 * recently, and the caller wants any such changes to be reflected.
1081 * Default: WANObjectCache::MIN_TIMESTAMP_NONE.
1082 * - hotTTR: Expected time-till-refresh (TTR) in seconds for keys that average ~1 hit per
1083 * second (e.g. 1Hz). Keys with a hit rate higher than 1Hz will refresh sooner than this
1084 * TTR and vise versa. Such refreshes won't happen until keys are "ageNew" seconds old.
1085 * This uses randomization to avoid triggering cache stampedes. The TTR is useful at
1086 * reducing the impact of missed cache purges, since the effect of a heavily referenced
1087 * key being stale is worse than that of a rarely referenced key. Unlike simply lowering
1088 * $ttl, seldomly used keys are largely unaffected by this option, which makes it
1089 * possible to have a high hit rate for the "long-tail" of less-used keys.
1090 * Default: WANObjectCache::HOT_TTR.
1091 * - lowTTL: Consider pre-emptive updates when the current TTL (seconds) of the key is less
1092 * than this. It becomes more likely over time, becoming certain once the key is expired.
1093 * This helps avoid cache stampedes that might be triggered due to the key expiring.
1094 * Default: WANObjectCache::LOW_TTL.
1095 * - ageNew: Consider popularity refreshes only once a key reaches this age in seconds.
1096 * Default: WANObjectCache::AGE_NEW.
1097 * - staleTTL: Seconds to keep the key around if it is stale. This means that on cache
1098 * miss the callback may get $oldValue/$oldAsOf values for keys that have already been
1099 * expired for this specified time. This is useful if adaptiveTTL() is used on the old
1100 * value's as-of time when it is verified as still being correct.
1101 * Default: WANObjectCache::STALE_TTL_NONE
1102 * - touchedCallback: A callback that takes the current value and returns a UNIX timestamp
1103 * indicating the last time a dynamic dependency changed. Null can be returned if there
1104 * are no relevant dependency changes to check. This can be used to check against things
1105 * like last-modified times of files or DB timestamp fields. This should generally not be
1106 * used for small and easily queried values in a DB if the callback itself ends up doing
1107 * a similarly expensive DB query to check a timestamp. Usages of this option makes the
1108 * most sense for values that are moderately to highly expensive to regenerate and easy
1109 * to query for dependency timestamps. The use of "pcTTL" reduces timestamp queries.
1111 * @return mixed Value found or written to the key
1112 * @note Options added in 1.28: version, busyValue, hotTTR, ageNew, pcGroup, minAsOf
1113 * @note Options added in 1.31: staleTTL, graceTTL
1114 * @note Options added in 1.33: touchedCallback
1115 * @note Callable type hints are not used to avoid class-autoloading
1117 final public function getWithSetCallback( $key, $ttl, $callback, array $opts = [] ) {
1118 $pcTTL = $opts['pcTTL'] ?? self
::TTL_UNCACHEABLE
;
1120 // Try the process cache if enabled and the cache callback is not within a cache callback.
1121 // Process cache use in nested callbacks is not lag-safe with regard to HOLDOFF_TTL since
1122 // the in-memory value is further lagged than the shared one since it uses a blind TTL.
1123 if ( $pcTTL >= 0 && $this->callbackDepth
== 0 ) {
1124 $group = $opts['pcGroup'] ?? self
::PC_PRIMARY
;
1125 $procCache = $this->getProcessCache( $group );
1126 $value = $procCache->has( $key, $pcTTL ) ?
$procCache->get( $key ) : false;
1132 if ( $value === false ) {
1133 // Fetch the value over the network
1134 if ( isset( $opts['version'] ) ) {
1135 $version = $opts['version'];
1137 $cur = $this->doGetWithSetCallback(
1140 function ( $oldValue, &$ttl, &$setOpts, $oldAsOf )
1141 use ( $callback, $version ) {
1142 if ( is_array( $oldValue )
1143 && array_key_exists( self
::VFLD_DATA
, $oldValue )
1144 && array_key_exists( self
::VFLD_VERSION
, $oldValue )
1145 && $oldValue[self
::VFLD_VERSION
] === $version
1147 $oldData = $oldValue[self
::VFLD_DATA
];
1149 // VFLD_DATA is not set if an old, unversioned, key is present
1155 self
::VFLD_DATA
=> $callback( $oldData, $ttl, $setOpts, $oldAsOf ),
1156 self
::VFLD_VERSION
=> $version
1162 if ( $cur[self
::VFLD_VERSION
] === $version ) {
1163 // Value created or existed before with version; use it
1164 $value = $cur[self
::VFLD_DATA
];
1166 // Value existed before with a different version; use variant key.
1167 // Reflect purges to $key by requiring that this key value be newer.
1168 $value = $this->doGetWithSetCallback(
1169 $this->makeGlobalKey( 'WANCache-key-variant', md5( $key ), $version ),
1172 // Regenerate value if not newer than $key
1173 [ 'version' => null, 'minAsOf' => $asOf ] +
$opts
1177 $value = $this->doGetWithSetCallback( $key, $ttl, $callback, $opts );
1180 // Update the process cache if enabled
1181 if ( $procCache && $value !== false ) {
1182 $procCache->set( $key, $value );
1190 * Do the actual I/O for getWithSetCallback() when needed
1192 * @see WANObjectCache::getWithSetCallback()
1194 * @param string $key
1196 * @param callable $callback
1197 * @param array $opts Options map for getWithSetCallback()
1198 * @param float|null &$asOf Cache generation timestamp of returned value [returned]
1200 * @note Callable type hints are not used to avoid class-autoloading
1202 protected function doGetWithSetCallback( $key, $ttl, $callback, array $opts, &$asOf = null ) {
1203 $lowTTL = $opts['lowTTL'] ??
min( self
::LOW_TTL
, $ttl );
1204 $lockTSE = $opts['lockTSE'] ?? self
::TSE_NONE
;
1205 $staleTTL = $opts['staleTTL'] ?? self
::STALE_TTL_NONE
;
1206 $graceTTL = $opts['graceTTL'] ?? self
::GRACE_TTL_NONE
;
1207 $checkKeys = $opts['checkKeys'] ??
[];
1208 $busyValue = $opts['busyValue'] ??
null;
1209 $popWindow = $opts['hotTTR'] ?? self
::HOT_TTR
;
1210 $ageNew = $opts['ageNew'] ?? self
::AGE_NEW
;
1211 $minTime = $opts['minAsOf'] ?? self
::MIN_TIMESTAMP_NONE
;
1212 $versioned = isset( $opts['version'] );
1213 $touchedCallback = $opts['touchedCallback'] ??
null;
1215 // Get a collection name to describe this class of key
1216 $kClass = $this->determineKeyClass( $key );
1218 // Get the current key value
1220 $cValue = $this->get( $key, $curTTL, $checkKeys, $asOf ); // current value
1221 $value = $cValue; // return value
1223 // Apply additional dynamic expiration logic if supplied
1224 $curTTL = $this->applyTouchedCallback( $value, $asOf, $curTTL, $touchedCallback );
1226 $preCallbackTime = $this->getCurrentTime();
1227 // Determine if a cached value regeneration is needed or desired
1229 $this->isValid( $value, $versioned, $asOf, $minTime ) &&
1230 $this->isAliveOrInGracePeriod( $curTTL, $graceTTL )
1232 $preemptiveRefresh = (
1233 $this->worthRefreshExpiring( $curTTL, $lowTTL ) ||
1234 $this->worthRefreshPopular( $asOf, $ageNew, $popWindow, $preCallbackTime )
1237 if ( !$preemptiveRefresh ) {
1238 $this->stats
->increment( "wanobjectcache.$kClass.hit.good" );
1241 } elseif ( $this->asyncHandler
) {
1242 // Update the cache value later, such during post-send of an HTTP request
1243 $func = $this->asyncHandler
;
1244 $func( function () use ( $key, $ttl, $callback, $opts, $asOf ) {
1245 $opts['minAsOf'] = INF
; // force a refresh
1246 $this->doGetWithSetCallback( $key, $ttl, $callback, $opts, $asOf );
1248 $this->stats
->increment( "wanobjectcache.$kClass.hit.refresh" );
1254 // Only a tombstoned key yields no value yet has a (negative) "current time left"
1255 $isKeyTombstoned = ( $curTTL !== null && $value === false );
1256 // Decide if only one thread should handle regeneration at a time
1258 // Note that since tombstones no-op set(), $lockTSE and $curTTL cannot be used to
1259 // deduce the key hotness because $curTTL will always keep increasing until the
1260 // tombstone expires or is overwritten by a new tombstone. Also, even if $lockTSE
1261 // is not set, constant regeneration of a key for the tombstone lifetime might be
1262 // very expensive. Assume tombstoned keys are possibly hot in order to reduce
1263 // the risk of high regeneration load after the delete() method is called.
1265 // Assume a key is hot if requested soon ($lockTSE seconds) after invalidation.
1266 // This avoids stampedes when timestamps from $checkKeys/$touchedCallback bump.
1267 ( $curTTL !== null && $curTTL <= 0 && abs( $curTTL ) <= $lockTSE ) ||
1268 // Assume a key is hot if there is no value and a busy fallback is given.
1269 // This avoids stampedes on eviction or preemptive regeneration taking too long.
1270 ( $busyValue !== null && $value === false );
1272 $lockAcquired = false;
1274 // Acquire a datacenter-local non-blocking lock
1275 if ( $this->cache
->add( self
::MUTEX_KEY_PREFIX
. $key, 1, self
::LOCK_TTL
) ) {
1276 // Lock acquired; this thread will recompute the value and update cache
1277 $lockAcquired = true;
1278 } elseif ( $this->isValid( $value, $versioned, $asOf, $minTime ) ) {
1279 // Lock not acquired and a stale value exists; use the stale value
1280 $this->stats
->increment( "wanobjectcache.$kClass.hit.stale" );
1284 // Lock not acquired and no stale value exists
1285 if ( $isKeyTombstoned ) {
1286 // Use the INTERIM value from the last thread that regenerated it if possible
1287 $value = $this->getInterimValue( $key, $versioned, $minTime, $asOf );
1288 if ( $value !== false ) {
1289 $this->stats
->increment( "wanobjectcache.$kClass.hit.volatile" );
1295 if ( $busyValue !== null ) {
1296 // Use the busy fallback value if nothing else
1297 $miss = is_infinite( $minTime ) ?
'renew' : 'miss';
1298 $this->stats
->increment( "wanobjectcache.$kClass.$miss.busy" );
1300 return is_callable( $busyValue ) ?
$busyValue() : $busyValue;
1305 if ( !is_callable( $callback ) ) {
1306 throw new InvalidArgumentException( "Invalid cache miss callback provided." );
1309 // Generate the new value from the callback...
1311 ++
$this->callbackDepth
;
1313 $value = call_user_func_array( $callback, [ $cValue, &$ttl, &$setOpts, $asOf ] );
1315 --$this->callbackDepth
;
1317 $valueIsCacheable = ( $value !== false && $ttl >= 0 );
1319 if ( $valueIsCacheable ) {
1320 if ( $isKeyTombstoned ) {
1321 // When delete() is called, writes are write-holed by the tombstone,
1322 // so use a special INTERIM key to pass the new value among threads.
1323 $tempTTL = max( self
::INTERIM_KEY_TTL
, (int)$lockTSE ); // set() expects seconds
1324 $newAsOf = $this->getCurrentTime();
1325 $wrapped = $this->wrap( $value, $tempTTL, $newAsOf );
1326 // Avoid using set() to avoid pointless mcrouter broadcasting
1327 $this->setInterimValue( $key, $wrapped, $tempTTL );
1328 } elseif ( !$useMutex ||
$lockAcquired ) {
1329 // Save the value unless a lock-winning thread is already expected to do that
1330 $setOpts['lockTSE'] = $lockTSE;
1331 $setOpts['staleTTL'] = $staleTTL;
1332 // Use best known "since" timestamp if not provided
1333 $setOpts +
= [ 'since' => $preCallbackTime ];
1334 // Update the cache; this will fail if the key is tombstoned
1335 $this->set( $key, $value, $ttl, $setOpts );
1339 if ( $lockAcquired ) {
1340 // Avoid using delete() to avoid pointless mcrouter broadcasting
1341 $this->cache
->changeTTL( self
::MUTEX_KEY_PREFIX
. $key, (int)$preCallbackTime - 60 );
1344 $miss = is_infinite( $minTime ) ?
'renew' : 'miss';
1345 $this->stats
->increment( "wanobjectcache.$kClass.$miss.compute" );
1351 * @param mixed $value
1352 * @param float $asOf
1353 * @param float $curTTL
1354 * @param callable|null $callback
1357 protected function applyTouchedCallback( $value, $asOf, $curTTL, $callback ) {
1358 if ( $callback === null ) {
1362 if ( !is_callable( $callback ) ) {
1363 throw new InvalidArgumentException( "Invalid expiration callback provided." );
1366 if ( $value !== false ) {
1367 $touched = $callback( $value );
1368 if ( $touched !== null && $touched >= $asOf ) {
1369 $curTTL = min( $curTTL, self
::TINY_NEGATIVE
, $asOf - $touched );
1377 * @param string $key
1378 * @param bool $versioned
1379 * @param float $minTime
1380 * @param mixed &$asOf
1383 protected function getInterimValue( $key, $versioned, $minTime, &$asOf ) {
1384 if ( !$this->useInterimHoldOffCaching
) {
1385 return false; // disabled
1388 $wrapped = $this->cache
->get( self
::INTERIM_KEY_PREFIX
. $key );
1389 list( $value ) = $this->unwrap( $wrapped, $this->getCurrentTime() );
1390 if ( $this->isValid( $value, $versioned, $asOf, $minTime ) ) {
1391 $asOf = $wrapped[self
::FLD_TIME
];
1400 * @param string $key
1401 * @param array $wrapped
1402 * @param int $tempTTL
1404 protected function setInterimValue( $key, $wrapped, $tempTTL ) {
1405 $this->cache
->merge(
1406 self
::INTERIM_KEY_PREFIX
. $key,
1407 function () use ( $wrapped ) {
1416 * Method to fetch multiple cache keys at once with regeneration
1418 * This works the same as getWithSetCallback() except:
1419 * - a) The $keys argument expects the result of WANObjectCache::makeMultiKeys()
1420 * - b) The $callback argument expects a callback taking the following arguments:
1421 * - $id: ID of an entity to query
1422 * - $oldValue : the prior cache value or false if none was present
1423 * - &$ttl : a reference to the new value TTL in seconds
1424 * - &$setOpts : a reference to options for set() which can be altered
1425 * - $oldAsOf : generation UNIX timestamp of $oldValue or null if not present
1426 * Aside from the additional $id argument, the other arguments function the same
1427 * way they do in getWithSetCallback().
1428 * - c) The return value is a map of (cache key => value) in the order of $keyedIds
1430 * @see WANObjectCache::getWithSetCallback()
1431 * @see WANObjectCache::getMultiWithUnionSetCallback()
1435 * $rows = $cache->getMultiWithSetCallback(
1436 * // Map of cache keys to entity IDs
1437 * $cache->makeMultiKeys(
1438 * $this->fileVersionIds(),
1439 * function ( $id, WANObjectCache $cache ) {
1440 * return $cache->makeKey( 'file-version', $id );
1443 * // Time-to-live (in seconds)
1445 * // Function that derives the new key value
1446 * function ( $id, $oldValue, &$ttl, array &$setOpts ) {
1447 * $dbr = wfGetDB( DB_REPLICA );
1448 * // Account for any snapshot/replica DB lag
1449 * $setOpts += Database::getCacheSetOptions( $dbr );
1451 * // Load the row for this file
1452 * $queryInfo = File::getQueryInfo();
1453 * $row = $dbr->selectRow(
1454 * $queryInfo['tables'],
1455 * $queryInfo['fields'],
1459 * $queryInfo['joins']
1462 * return $row ? (array)$row : false;
1465 * // Process cache for 30 seconds
1467 * // Use a dedicated 500 item cache (initialized on-the-fly)
1468 * 'pcGroup' => 'file-versions:500'
1471 * $files = array_map( [ __CLASS__, 'newFromRow' ], $rows );
1474 * @param ArrayIterator $keyedIds Result of WANObjectCache::makeMultiKeys()
1475 * @param int $ttl Seconds to live for key updates
1476 * @param callable $callback Callback the yields entity regeneration callbacks
1477 * @param array $opts Options map
1478 * @return array Map of (cache key => value) in the same order as $keyedIds
1481 final public function getMultiWithSetCallback(
1482 ArrayIterator
$keyedIds, $ttl, callable
$callback, array $opts = []
1484 $valueKeys = array_keys( $keyedIds->getArrayCopy() );
1485 $checkKeys = $opts['checkKeys'] ??
[];
1486 $pcTTL = $opts['pcTTL'] ?? self
::TTL_UNCACHEABLE
;
1488 // Load required keys into process cache in one go
1489 $this->warmupCache
= $this->getRawKeysForWarmup(
1490 $this->getNonProcessCachedKeys( $valueKeys, $opts, $pcTTL ),
1493 $this->warmupKeyMisses
= 0;
1495 // Wrap $callback to match the getWithSetCallback() format while passing $id to $callback
1496 $id = null; // current entity ID
1497 $func = function ( $oldValue, &$ttl, &$setOpts, $oldAsOf ) use ( $callback, &$id ) {
1498 return $callback( $id, $oldValue, $ttl, $setOpts, $oldAsOf );
1502 foreach ( $keyedIds as $key => $id ) { // preserve order
1503 $values[$key] = $this->getWithSetCallback( $key, $ttl, $func, $opts );
1506 $this->warmupCache
= [];
1512 * Method to fetch/regenerate multiple cache keys at once
1514 * This works the same as getWithSetCallback() except:
1515 * - a) The $keys argument expects the result of WANObjectCache::makeMultiKeys()
1516 * - b) The $callback argument expects a callback returning a map of (ID => new value)
1517 * for all entity IDs in $ids and it takes the following arguments:
1518 * - $ids: a list of entity IDs that require cache regeneration
1519 * - &$ttls: a reference to the (entity ID => new TTL) map
1520 * - &$setOpts: a reference to options for set() which can be altered
1521 * - c) The return value is a map of (cache key => value) in the order of $keyedIds
1522 * - d) The "lockTSE" and "busyValue" options are ignored
1524 * @see WANObjectCache::getWithSetCallback()
1525 * @see WANObjectCache::getMultiWithSetCallback()
1529 * $rows = $cache->getMultiWithUnionSetCallback(
1530 * // Map of cache keys to entity IDs
1531 * $cache->makeMultiKeys(
1532 * $this->fileVersionIds(),
1533 * function ( $id, WANObjectCache $cache ) {
1534 * return $cache->makeKey( 'file-version', $id );
1537 * // Time-to-live (in seconds)
1539 * // Function that derives the new key value
1540 * function ( array $ids, array &$ttls, array &$setOpts ) {
1541 * $dbr = wfGetDB( DB_REPLICA );
1542 * // Account for any snapshot/replica DB lag
1543 * $setOpts += Database::getCacheSetOptions( $dbr );
1545 * // Load the rows for these files
1547 * $queryInfo = File::getQueryInfo();
1548 * $res = $dbr->select(
1549 * $queryInfo['tables'],
1550 * $queryInfo['fields'],
1554 * $queryInfo['joins']
1556 * foreach ( $res as $row ) {
1557 * $rows[$row->id] = $row;
1558 * $mtime = wfTimestamp( TS_UNIX, $row->timestamp );
1559 * $ttls[$row->id] = $this->adaptiveTTL( $mtime, $ttls[$row->id] );
1566 * $files = array_map( [ __CLASS__, 'newFromRow' ], $rows );
1569 * @param ArrayIterator $keyedIds Result of WANObjectCache::makeMultiKeys()
1570 * @param int $ttl Seconds to live for key updates
1571 * @param callable $callback Callback the yields entity regeneration callbacks
1572 * @param array $opts Options map
1573 * @return array Map of (cache key => value) in the same order as $keyedIds
1576 final public function getMultiWithUnionSetCallback(
1577 ArrayIterator
$keyedIds, $ttl, callable
$callback, array $opts = []
1579 $idsByValueKey = $keyedIds->getArrayCopy();
1580 $valueKeys = array_keys( $idsByValueKey );
1581 $checkKeys = $opts['checkKeys'] ??
[];
1582 $pcTTL = $opts['pcTTL'] ?? self
::TTL_UNCACHEABLE
;
1583 unset( $opts['lockTSE'] ); // incompatible
1584 unset( $opts['busyValue'] ); // incompatible
1586 // Load required keys into process cache in one go
1587 $keysGet = $this->getNonProcessCachedKeys( $valueKeys, $opts, $pcTTL );
1588 $this->warmupCache
= $this->getRawKeysForWarmup( $keysGet, $checkKeys );
1589 $this->warmupKeyMisses
= 0;
1591 // IDs of entities known to be in need of regeneration
1594 // Find out which keys are missing/deleted/stale
1597 $curByKey = $this->getMulti( $keysGet, $curTTLs, $checkKeys, $asOfs );
1598 foreach ( $keysGet as $key ) {
1599 if ( !array_key_exists( $key, $curByKey ) ||
$curTTLs[$key] < 0 ) {
1600 $idsRegen[] = $idsByValueKey[$key];
1604 // Run the callback to populate the regeneration value map for all required IDs
1606 $newTTLsById = array_fill_keys( $idsRegen, $ttl );
1607 $newValsById = $idsRegen ?
$callback( $idsRegen, $newTTLsById, $newSetOpts ) : [];
1609 // Wrap $callback to match the getWithSetCallback() format while passing $id to $callback
1610 $id = null; // current entity ID
1611 $func = function ( $oldValue, &$ttl, &$setOpts, $oldAsOf )
1612 use ( $callback, &$id, $newValsById, $newTTLsById, $newSetOpts )
1614 if ( array_key_exists( $id, $newValsById ) ) {
1615 // Value was already regerated as expected, so use the value in $newValsById
1616 $newValue = $newValsById[$id];
1617 $ttl = $newTTLsById[$id];
1618 $setOpts = $newSetOpts;
1620 // Pre-emptive/popularity refresh and version mismatch cases are not detected
1621 // above and thus $newValsById has no entry. Run $callback on this single entity.
1622 $ttls = [ $id => $ttl ];
1623 $newValue = $callback( [ $id ], $ttls, $setOpts )[$id];
1630 // Run the cache-aside logic using warmupCache instead of persistent cache queries
1632 foreach ( $idsByValueKey as $key => $id ) { // preserve order
1633 $values[$key] = $this->getWithSetCallback( $key, $ttl, $func, $opts );
1636 $this->warmupCache
= [];
1642 * Set a key to soon expire in the local cluster if it pre-dates $purgeTimestamp
1644 * This sets stale keys' time-to-live at HOLDOFF_TTL seconds, which both avoids
1645 * broadcasting in mcrouter setups and also avoids races with new tombstones.
1647 * @param string $key Cache key
1648 * @param int $purgeTimestamp UNIX timestamp of purge
1649 * @param bool &$isStale Whether the key is stale
1650 * @return bool Success
1653 final public function reap( $key, $purgeTimestamp, &$isStale = false ) {
1654 $minAsOf = $purgeTimestamp + self
::HOLDOFF_TTL
;
1655 $wrapped = $this->cache
->get( self
::VALUE_KEY_PREFIX
. $key );
1656 if ( is_array( $wrapped ) && $wrapped[self
::FLD_TIME
] < $minAsOf ) {
1658 $this->logger
->warning( "Reaping stale value key '$key'." );
1659 $ttlReap = self
::HOLDOFF_TTL
; // avoids races with tombstone creation
1660 $ok = $this->cache
->changeTTL( self
::VALUE_KEY_PREFIX
. $key, $ttlReap );
1662 $this->logger
->error( "Could not complete reap of key '$key'." );
1674 * Set a "check" key to soon expire in the local cluster if it pre-dates $purgeTimestamp
1676 * @param string $key Cache key
1677 * @param int $purgeTimestamp UNIX timestamp of purge
1678 * @param bool &$isStale Whether the key is stale
1679 * @return bool Success
1682 final public function reapCheckKey( $key, $purgeTimestamp, &$isStale = false ) {
1683 $purge = $this->parsePurgeValue( $this->cache
->get( self
::TIME_KEY_PREFIX
. $key ) );
1684 if ( $purge && $purge[self
::FLD_TIME
] < $purgeTimestamp ) {
1686 $this->logger
->warning( "Reaping stale check key '$key'." );
1687 $ok = $this->cache
->changeTTL( self
::TIME_KEY_PREFIX
. $key, self
::TTL_SECOND
);
1689 $this->logger
->error( "Could not complete reap of check key '$key'." );
1701 * @see BagOStuff::makeKey()
1702 * @param string $class Key class
1703 * @param string|null $component [optional] Key component (starting with a key collection name)
1704 * @return string Colon-delimited list of $keyspace followed by escaped components of $args
1707 public function makeKey( $class, $component = null ) {
1708 return $this->cache
->makeKey( ...func_get_args() );
1712 * @see BagOStuff::makeGlobalKey()
1713 * @param string $class Key class
1714 * @param string|null $component [optional] Key component (starting with a key collection name)
1715 * @return string Colon-delimited list of $keyspace followed by escaped components of $args
1718 public function makeGlobalKey( $class, $component = null ) {
1719 return $this->cache
->makeGlobalKey( ...func_get_args() );
1723 * @param array $entities List of entity IDs
1724 * @param callable $keyFunc Callback yielding a key from (entity ID, this WANObjectCache)
1725 * @return ArrayIterator Iterator yielding (cache key => entity ID) in $entities order
1728 final public function makeMultiKeys( array $entities, callable
$keyFunc ) {
1730 foreach ( $entities as $entity ) {
1731 $map[$keyFunc( $entity, $this )] = $entity;
1734 return new ArrayIterator( $map );
1738 * Get the "last error" registered; clearLastError() should be called manually
1739 * @return int ERR_* class constant for the "last error" registry
1741 final public function getLastError() {
1742 $code = $this->cache
->getLastError();
1744 case BagOStuff
::ERR_NONE
:
1745 return self
::ERR_NONE
;
1746 case BagOStuff
::ERR_NO_RESPONSE
:
1747 return self
::ERR_NO_RESPONSE
;
1748 case BagOStuff
::ERR_UNREACHABLE
:
1749 return self
::ERR_UNREACHABLE
;
1751 return self
::ERR_UNEXPECTED
;
1756 * Clear the "last error" registry
1758 final public function clearLastError() {
1759 $this->cache
->clearLastError();
1763 * Clear the in-process caches; useful for testing
1767 public function clearProcessCache() {
1768 $this->processCaches
= [];
1772 * Enable or disable the use of brief caching for tombstoned keys
1774 * When a key is purged via delete(), there normally is a period where caching
1775 * is hold-off limited to an extremely short time. This method will disable that
1776 * caching, forcing the callback to run for any of:
1777 * - WANObjectCache::getWithSetCallback()
1778 * - WANObjectCache::getMultiWithSetCallback()
1779 * - WANObjectCache::getMultiWithUnionSetCallback()
1781 * This is useful when both:
1782 * - a) the database used by the callback is known to be up-to-date enough
1783 * for some particular purpose (e.g. replica DB has applied transaction X)
1784 * - b) the caller needs to exploit that fact, and therefore needs to avoid the
1785 * use of inherently volatile and possibly stale interim keys
1787 * @see WANObjectCache::delete()
1788 * @param bool $enabled Whether to enable interim caching
1791 final public function useInterimHoldOffCaching( $enabled ) {
1792 $this->useInterimHoldOffCaching
= $enabled;
1796 * @param int $flag ATTR_* class constant
1797 * @return int QOS_* class constant
1800 public function getQoS( $flag ) {
1801 return $this->cache
->getQoS( $flag );
1805 * Get a TTL that is higher for objects that have not changed recently
1807 * This is useful for keys that get explicit purges and DB or purge relay
1808 * lag is a potential concern (especially how it interacts with CDN cache)
1812 * // Last-modified time of page
1813 * $mtime = wfTimestamp( TS_UNIX, $page->getTimestamp() );
1814 * // Get adjusted TTL. If $mtime is 3600 seconds ago and $minTTL/$factor left at
1815 * // defaults, then $ttl is 3600 * .2 = 720. If $minTTL was greater than 720, then
1816 * // $ttl would be $minTTL. If $maxTTL was smaller than 720, $ttl would be $maxTTL.
1817 * $ttl = $cache->adaptiveTTL( $mtime, $cache::TTL_DAY );
1820 * Another use case is when there are no applicable "last modified" fields in the DB,
1821 * and there are too many dependencies for explicit purges to be viable, and the rate of
1822 * change to relevant content is unstable, and it is highly valued to have the cached value
1823 * be as up-to-date as possible.
1827 * $query = "<some complex query>";
1828 * $idListFromComplexQuery = $cache->getWithSetCallback(
1829 * $cache->makeKey( 'complex-graph-query', $hashOfQuery ),
1830 * GraphQueryClass::STARTING_TTL,
1831 * function ( $oldValue, &$ttl, array &$setOpts, $oldAsOf ) use ( $query, $cache ) {
1832 * $gdb = $this->getReplicaGraphDbConnection();
1833 * // Account for any snapshot/replica DB lag
1834 * $setOpts += GraphDatabase::getCacheSetOptions( $gdb );
1836 * $newList = iterator_to_array( $gdb->query( $query ) );
1837 * sort( $newList, SORT_NUMERIC ); // normalize
1839 * $minTTL = GraphQueryClass::MIN_TTL;
1840 * $maxTTL = GraphQueryClass::MAX_TTL;
1841 * if ( $oldValue !== false ) {
1842 * // Note that $oldAsOf is the last time this callback ran
1843 * $ttl = ( $newList === $oldValue )
1844 * // No change: cache for 150% of the age of $oldValue
1845 * ? $cache->adaptiveTTL( $oldAsOf, $maxTTL, $minTTL, 1.5 )
1846 * // Changed: cache for 50% of the age of $oldValue
1847 * : $cache->adaptiveTTL( $oldAsOf, $maxTTL, $minTTL, .5 );
1853 * // Keep stale values around for doing comparisons for TTL calculations.
1854 * // High values improve long-tail keys hit-rates, though might waste space.
1855 * 'staleTTL' => GraphQueryClass::GRACE_TTL
1860 * @param int|float $mtime UNIX timestamp
1861 * @param int $maxTTL Maximum TTL (seconds)
1862 * @param int $minTTL Minimum TTL (seconds); Default: 30
1863 * @param float $factor Value in the range (0,1); Default: .2
1864 * @return int Adaptive TTL
1867 public function adaptiveTTL( $mtime, $maxTTL, $minTTL = 30, $factor = 0.2 ) {
1868 if ( is_float( $mtime ) ||
ctype_digit( $mtime ) ) {
1869 $mtime = (int)$mtime; // handle fractional seconds and string integers
1872 if ( !is_int( $mtime ) ||
$mtime <= 0 ) {
1873 return $minTTL; // no last-modified time provided
1876 $age = $this->getCurrentTime() - $mtime;
1878 return (int)min( $maxTTL, max( $minTTL, $factor * $age ) );
1882 * @return int Number of warmup key cache misses last round
1885 final public function getWarmupKeyMisses() {
1886 return $this->warmupKeyMisses
;
1890 * Do the actual async bus purge of a key
1892 * This must set the key to "PURGED:<UNIX timestamp>:<holdoff>"
1894 * @param string $key Cache key
1895 * @param int $ttl How long to keep the tombstone [seconds]
1896 * @param int $holdoff HOLDOFF_* constant controlling how long to ignore sets for this key
1897 * @return bool Success
1899 protected function relayPurge( $key, $ttl, $holdoff ) {
1900 if ( $this->mcrouterAware
) {
1901 // See https://github.com/facebook/mcrouter/wiki/Multi-cluster-broadcast-setup
1902 // Wildcards select all matching routes, e.g. the WAN cluster on all DCs
1903 $ok = $this->cache
->set(
1904 "/*/{$this->cluster}/{$key}",
1905 $this->makePurgeValue( $this->getCurrentTime(), self
::HOLDOFF_NONE
),
1909 // This handles the mcrouter and the single-DC case
1910 $ok = $this->cache
->set(
1912 $this->makePurgeValue( $this->getCurrentTime(), self
::HOLDOFF_NONE
),
1921 * Do the actual async bus delete of a key
1923 * @param string $key Cache key
1924 * @return bool Success
1926 protected function relayDelete( $key ) {
1927 if ( $this->mcrouterAware
) {
1928 // See https://github.com/facebook/mcrouter/wiki/Multi-cluster-broadcast-setup
1929 // Wildcards select all matching routes, e.g. the WAN cluster on all DCs
1930 $ok = $this->cache
->delete( "/*/{$this->cluster}/{$key}" );
1932 // Some other proxy handles broadcasting or there is only one datacenter
1933 $ok = $this->cache
->delete( $key );
1940 * Check if a key is fresh or in the grace window and thus due for randomized reuse
1942 * If $curTTL > 0 (e.g. not expired) this returns true. Otherwise, the chance of returning
1943 * true decrease steadily from 100% to 0% as the |$curTTL| moves from 0 to $graceTTL seconds.
1944 * This handles widely varying levels of cache access traffic.
1946 * If $curTTL <= -$graceTTL (e.g. already expired), then this returns false.
1948 * @param float $curTTL Approximate TTL left on the key if present
1949 * @param int $graceTTL Consider using stale values if $curTTL is greater than this
1952 protected function isAliveOrInGracePeriod( $curTTL, $graceTTL ) {
1953 if ( $curTTL > 0 ) {
1955 } elseif ( $graceTTL <= 0 ) {
1959 $ageStale = abs( $curTTL ); // seconds of staleness
1960 $curGTTL = ( $graceTTL - $ageStale ); // current grace-time-to-live
1961 if ( $curGTTL <= 0 ) {
1962 return false; // already out of grace period
1965 // Chance of using a stale value is the complement of the chance of refreshing it
1966 return !$this->worthRefreshExpiring( $curGTTL, $graceTTL );
1970 * Check if a key is nearing expiration and thus due for randomized regeneration
1972 * This returns false if $curTTL >= $lowTTL. Otherwise, the chance of returning true
1973 * increases steadily from 0% to 100% as the $curTTL moves from $lowTTL to 0 seconds.
1974 * This handles widely varying levels of cache access traffic.
1976 * If $curTTL <= 0 (e.g. already expired), then this returns false.
1978 * @param float $curTTL Approximate TTL left on the key if present
1979 * @param float $lowTTL Consider a refresh when $curTTL is less than this
1982 protected function worthRefreshExpiring( $curTTL, $lowTTL ) {
1983 if ( $lowTTL <= 0 ) {
1985 } elseif ( $curTTL >= $lowTTL ) {
1987 } elseif ( $curTTL <= 0 ) {
1991 $chance = ( 1 - $curTTL / $lowTTL );
1993 return mt_rand( 1, 1e9
) <= 1e9
* $chance;
1997 * Check if a key is due for randomized regeneration due to its popularity
1999 * This is used so that popular keys can preemptively refresh themselves for higher
2000 * consistency (especially in the case of purge loss/delay). Unpopular keys can remain
2001 * in cache with their high nominal TTL. This means popular keys keep good consistency,
2002 * whether the data changes frequently or not, and long-tail keys get to stay in cache
2003 * and get hits too. Similar to worthRefreshExpiring(), randomization is used.
2005 * @param float $asOf UNIX timestamp of the value
2006 * @param int $ageNew Age of key when this might recommend refreshing (seconds)
2007 * @param int $timeTillRefresh Age of key when it should be refreshed if popular (seconds)
2008 * @param float $now The current UNIX timestamp
2011 protected function worthRefreshPopular( $asOf, $ageNew, $timeTillRefresh, $now ) {
2012 if ( $ageNew < 0 ||
$timeTillRefresh <= 0 ) {
2016 $age = $now - $asOf;
2017 $timeOld = $age - $ageNew;
2018 if ( $timeOld <= 0 ) {
2022 // Lifecycle is: new, ramp-up refresh chance, full refresh chance.
2023 // Note that the "expected # of refreshes" for the ramp-up time range is half of what it
2024 // would be if P(refresh) was at its full value during that time range.
2025 $refreshWindowSec = max( $timeTillRefresh - $ageNew - self
::RAMPUP_TTL
/ 2, 1 );
2026 // P(refresh) * (# hits in $refreshWindowSec) = (expected # of refreshes)
2027 // P(refresh) * ($refreshWindowSec * $popularHitsPerSec) = 1
2028 // P(refresh) = 1/($refreshWindowSec * $popularHitsPerSec)
2029 $chance = 1 / ( self
::HIT_RATE_HIGH
* $refreshWindowSec );
2031 // Ramp up $chance from 0 to its nominal value over RAMPUP_TTL seconds to avoid stampedes
2032 $chance *= ( $timeOld <= self
::RAMPUP_TTL
) ?
$timeOld / self
::RAMPUP_TTL
: 1;
2034 return mt_rand( 1, 1e9
) <= 1e9
* $chance;
2038 * Check if $value is not false, versioned (if needed), and not older than $minTime (if set)
2040 * @param array|bool $value
2041 * @param bool $versioned
2042 * @param float $asOf The time $value was generated
2043 * @param float $minTime The last time the main value was generated (0.0 if unknown)
2046 protected function isValid( $value, $versioned, $asOf, $minTime ) {
2047 if ( $value === false ) {
2049 } elseif ( $versioned && !isset( $value[self
::VFLD_VERSION
] ) ) {
2051 } elseif ( $minTime > 0 && $asOf < $minTime ) {
2059 * Do not use this method outside WANObjectCache
2061 * @param mixed $value
2062 * @param int $ttl [0=forever]
2063 * @param float $now Unix Current timestamp just before calling set()
2066 protected function wrap( $value, $ttl, $now ) {
2068 self
::FLD_VERSION
=> self
::VERSION
,
2069 self
::FLD_VALUE
=> $value,
2070 self
::FLD_TTL
=> $ttl,
2071 self
::FLD_TIME
=> $now
2076 * Do not use this method outside WANObjectCache
2078 * @param array|string|bool $wrapped
2079 * @param float $now Unix Current timestamp (preferrably pre-query)
2080 * @return array (mixed; false if absent/tombstoned/malformed, current time left)
2082 protected function unwrap( $wrapped, $now ) {
2083 // Check if the value is a tombstone
2084 $purge = $this->parsePurgeValue( $wrapped );
2085 if ( $purge !== false ) {
2086 // Purged values should always have a negative current $ttl
2087 $curTTL = min( $purge[self
::FLD_TIME
] - $now, self
::TINY_NEGATIVE
);
2088 return [ false, $curTTL ];
2091 if ( !is_array( $wrapped ) // not found
2092 ||
!isset( $wrapped[self
::FLD_VERSION
] ) // wrong format
2093 ||
$wrapped[self
::FLD_VERSION
] !== self
::VERSION
// wrong version
2095 return [ false, null ];
2098 $flags = $wrapped[self
::FLD_FLAGS
] ??
0;
2099 if ( ( $flags & self
::FLG_STALE
) == self
::FLG_STALE
) {
2100 // Treat as expired, with the cache time as the expiration
2101 $age = $now - $wrapped[self
::FLD_TIME
];
2102 $curTTL = min( -$age, self
::TINY_NEGATIVE
);
2103 } elseif ( $wrapped[self
::FLD_TTL
] > 0 ) {
2104 // Get the approximate time left on the key
2105 $age = $now - $wrapped[self
::FLD_TIME
];
2106 $curTTL = max( $wrapped[self
::FLD_TTL
] - $age, 0.0 );
2108 // Key had no TTL, so the time left is unbounded
2112 if ( $wrapped[self
::FLD_TIME
] < $this->epoch
) {
2113 // Values this old are ignored
2114 return [ false, null ];
2117 return [ $wrapped[self
::FLD_VALUE
], $curTTL ];
2121 * @param array $keys
2122 * @param string $prefix
2125 protected static function prefixCacheKeys( array $keys, $prefix ) {
2127 foreach ( $keys as $key ) {
2128 $res[] = $prefix . $key;
2135 * @param string $key String of the format <scope>:<class>[:<class or variable>]...
2138 protected function determineKeyClass( $key ) {
2139 $parts = explode( ':', $key );
2141 return $parts[1] ??
$parts[0]; // sanity
2145 * @param string|array|bool $value Possible string of the form "PURGED:<timestamp>:<holdoff>"
2146 * @return array|bool Array containing a UNIX timestamp (float) and holdoff period (integer),
2147 * or false if value isn't a valid purge value
2149 protected function parsePurgeValue( $value ) {
2150 if ( !is_string( $value ) ) {
2154 $segments = explode( ':', $value, 3 );
2155 if ( !isset( $segments[0] ) ||
!isset( $segments[1] )
2156 ||
"{$segments[0]}:" !== self
::PURGE_VAL_PREFIX
2161 if ( !isset( $segments[2] ) ) {
2162 // Back-compat with old purge values without holdoff
2163 $segments[2] = self
::HOLDOFF_TTL
;
2166 if ( $segments[1] < $this->epoch
) {
2167 // Values this old are ignored
2172 self
::FLD_TIME
=> (float)$segments[1],
2173 self
::FLD_HOLDOFF
=> (int)$segments[2],
2178 * @param float $timestamp
2179 * @param int $holdoff In seconds
2180 * @return string Wrapped purge value
2182 protected function makePurgeValue( $timestamp, $holdoff ) {
2183 return self
::PURGE_VAL_PREFIX
. (float)$timestamp . ':' . (int)$holdoff;
2187 * @param string $group
2188 * @return MapCacheLRU
2190 protected function getProcessCache( $group ) {
2191 if ( !isset( $this->processCaches
[$group] ) ) {
2192 list( , $n ) = explode( ':', $group );
2193 $this->processCaches
[$group] = new MapCacheLRU( (int)$n );
2196 return $this->processCaches
[$group];
2200 * @param array $keys
2201 * @param array $opts
2203 * @return array List of keys
2205 private function getNonProcessCachedKeys( array $keys, array $opts, $pcTTL ) {
2207 if ( isset( $opts['pcTTL'] ) && $opts['pcTTL'] > 0 && $this->callbackDepth
== 0 ) {
2208 $pcGroup = $opts['pcGroup'] ?? self
::PC_PRIMARY
;
2209 $procCache = $this->getProcessCache( $pcGroup );
2210 foreach ( $keys as $key ) {
2211 if ( $procCache->has( $key, $pcTTL ) ) {
2212 $keysFound[] = $key;
2217 return array_diff( $keys, $keysFound );
2221 * @param array $keys
2222 * @param array $checkKeys
2223 * @return array Map of (cache key => mixed)
2225 private function getRawKeysForWarmup( array $keys, array $checkKeys ) {
2231 // Get all the value keys to fetch...
2232 foreach ( $keys as $key ) {
2233 $keysWarmUp[] = self
::VALUE_KEY_PREFIX
. $key;
2235 // Get all the check keys to fetch...
2236 foreach ( $checkKeys as $i => $checkKeyOrKeys ) {
2237 if ( is_int( $i ) ) {
2238 // Single check key that applies to all value keys
2239 $keysWarmUp[] = self
::TIME_KEY_PREFIX
. $checkKeyOrKeys;
2241 // List of check keys that apply to value key $i
2242 $keysWarmUp = array_merge(
2244 self
::prefixCacheKeys( $checkKeyOrKeys, self
::TIME_KEY_PREFIX
)
2249 $warmupCache = $this->cache
->getMulti( $keysWarmUp );
2250 $warmupCache +
= array_fill_keys( $keysWarmUp, false );
2252 return $warmupCache;
2256 * @return float UNIX timestamp
2257 * @codeCoverageIgnore
2259 protected function getCurrentTime() {
2260 if ( $this->wallClockOverride
) {
2261 return $this->wallClockOverride
;
2264 $clockTime = (float)time(); // call this first
2265 // microtime() uses an initial gettimeofday() call added to usage clocks.
2266 // This can severely drift from time() and the microtime() value of other threads
2267 // due to undercounting of the amount of time elapsed. Instead of seeing the current
2268 // time as being in the past, use the value of time(). This avoids setting cache values
2269 // that will immediately be seen as expired and possibly cause stampedes.
2270 return max( microtime( true ), $clockTime );
2274 * @param float|null &$time Mock UNIX timestamp for testing
2275 * @codeCoverageIgnore
2277 public function setMockTime( &$time ) {
2278 $this->wallClockOverride
=& $time;
2279 $this->cache
->setMockTime( $time );