includes/poolcounter/PoolCounterRedis.php

   1 <?php
   2 /**
   3  * This program is free software; you can redistribute it and/or modify
   4  * it under the terms of the GNU General Public License as published by
   5  * the Free Software Foundation; either version 2 of the License, or
   6  * (at your option) any later version.
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  11  * GNU General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public License along
  14  * with this program; if not, write to the Free Software Foundation, Inc.,
  15  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  16  * http://www.gnu.org/copyleft/gpl.html
  17  *
  18  * @file
  19  * @author Aaron Schulz
  20  */
  21
  22 /**
  23  * Version of PoolCounter that uses Redis
  24  *
  25  * There are four main redis keys used to track each pool counter key:
  26  *   - poolcounter:l-slots-*     : A list of available slot IDs for a pool.
  27  *   - poolcounter:z-renewtime-* : A sorted set of (slot ID, UNIX timestamp as score)
  28  *                                 used for tracking the next time a slot should be
  29  *                                 released. This is -1 when a slot is created, and is
  30  *                                 set when released (expired), locked, and unlocked.
  31  *   - poolcounter:z-wait-*      : A sorted set of (slot ID, UNIX timestamp as score)
  32  *                                 used for tracking waiting processes (and wait time).
  33  *   - poolcounter:l-wakeup-*    : A list pushed to for the sake of waking up processes
  34  *                                 when a any process in the pool finishes (lasts for 1ms).
  35  * For a given pool key, all the redis keys start off non-existing and are deleted if not
  36  * used for a while to prevent garbage from building up on the server. They are atomically
  37  * re-initialized as needed. The "z-renewtime" key is used for detecting sessions which got
  38  * slots but then disappeared. Stale entries from there have their timestamp updated and the
  39  * corresponding slots freed up. The "z-wait" key is used for detecting processes registered
  40  * as waiting but that disappeared. Stale entries from there are deleted and the corresponding
  41  * slots are freed up. The worker count is included in all the redis key names as it does not
  42  * vary within each $wgPoolCounterConf type and doing so handles configuration changes.
  43  *
  44  * This class requires Redis 2.6 as it makes use Lua scripts for fast atomic operations.
  45  * Also this should be on a server plenty of RAM for the working set to avoid evictions.
  46  * Evictions could temporarily allow wait queues to double in size or temporarily cause
  47  * pools to appear as full when they are not. Using volatile-ttl and bumping memory-samples
  48  * and redis.conf can be helpful otherwise.
  49  *
  50  * @ingroup Redis
  51  * @since 1.23
  52  */
  53 class PoolCounterRedis extends PoolCounter {
  54         /** @var HashRing */
  55         protected $ring;
  56         /** @var RedisConnectionPool */
  57         protected $pool;
  58         /** @var array (server label => host) map */
  59         protected $serversByLabel;
  60         /** @var string SHA-1 of the key */
  61         protected $keySha1;
  62         /** @var int TTL for locks to expire (work should finish in this time) */
  63         protected $lockTTL;
  64
  65         /** @var RedisConnRef */
  66         protected $conn;
  67         /** @var string Pool slot value */
  68         protected $slot;
  69         /** @var int AWAKE_* constant */
  70         protected $onRelease;
  71         /** @var string Unique string to identify this process */
  72         protected $session;
  73         /** @var int UNIX timestamp */
  74         protected $slotTime;
  75
  76         const AWAKE_ONE = 1; // wake-up if when a slot can be taken from an existing process
  77         const AWAKE_ALL = 2; // wake-up if an existing process finishes and wake up such others
  78
  79         /** @var array List of active PoolCounterRedis objects in this script */
  80         protected static $active = null;
  81
  82         function __construct( $conf, $type, $key ) {
  83                 parent::__construct( $conf, $type, $key );
  84
  85                 $this->serversByLabel = $conf['servers'];
  86                 $this->ring = new HashRing( array_fill_keys( array_keys( $conf['servers'] ), 100 ) );
  87
  88                 $conf['redisConfig']['serializer'] = 'none'; // for use with Lua
  89                 $this->pool = RedisConnectionPool::singleton( $conf['redisConfig'] );
  90
  91                 $this->keySha1 = sha1( $this->key );
  92                 $met = ini_get( 'max_execution_time' ); // usually 0 in CLI mode
  93                 $this->lockTTL = $met ? 2*$met : 3600;
  94
  95                 if ( self::$active === null ) {
  96                         self::$active = array();
  97                         register_shutdown_function( array( __CLASS__, 'releaseAll' ) );
  98                 }
  99         }
 100
 101         /**
 102          * @return Status Uses RediConnRef as value on success
 103          */
 104         protected function getConnection() {
 105                 if ( !isset( $this->conn ) ) {
 106                         $conn = false;
 107                         $servers = $this->ring->getLocations( $this->key, 3 );
 108                         ArrayUtils::consistentHashSort( $servers, $this->key );
 109                         foreach ( $servers as $server ) {
 110                                 $conn = $this->pool->getConnection( $this->serversByLabel[$server] );
 111                                 if ( $conn ) {
 112                                         break;
 113                                 }
 114                         }
 115                         if ( !$conn ) {
 116                                 return Status::newFatal( 'pool-servererror', implode( ', ', $servers ) );
 117                         }
 118                         $this->conn = $conn;
 119                 }
 120                 return Status::newGood( $this->conn );
 121         }
 122
 123         function acquireForMe() {
 124                 $section = new ProfileSection( __METHOD__ );
 125
 126                 return $this->waitForSlotOrNotif( self::AWAKE_ONE );
 127         }
 128
 129         function acquireForAnyone() {
 130                 $section = new ProfileSection( __METHOD__ );
 131
 132                 return $this->waitForSlotOrNotif( self::AWAKE_ALL );
 133         }
 134
 135         function release() {
 136                 $section = new ProfileSection( __METHOD__ );
 137
 138                 if ( $this->slot === null ) {
 139                         return Status::newGood( PoolCounter::NOT_LOCKED ); // not locked
 140                 }
 141
 142                 $status = $this->getConnection();
 143                 if ( !$status->isOK() ) {
 144                         return $status;
 145                 }
 146                 $conn = $status->value;
 147
 148                 static $script =
 149 <<<LUA
 150                 local kSlots,kSlotsNextRelease,kWakeup,kWaiting = unpack(KEYS)
 151                 local rMaxWorkers,rExpiry,rSlot,rSlotTime,rAwakeAll,rTime = unpack(ARGV)
 152                 -- Add the slots back to the list (if rSlot is "w" then it is not a slot).
 153                 -- Treat the list as expired if the "next release" time sorted-set is missing.
 154                 if rSlot ~= 'w' and redis.call('exists',kSlotsNextRelease) == 1 then
 155                         if 1*redis.call('zScore',kSlotsNextRelease,rSlot) ~= (rSlotTime + rExpiry) then
 156                                 -- Slot lock expired and was released already
 157                         elseif redis.call('lLen',kSlots) >= (1*rMaxWorkers - 1) then
 158                                 -- Clear list to save space; it will re-init as needed
 159                                 redis.call('del',kSlots,kSlotsNextRelease)
 160                         else
 161                                 -- Add slot back to pool and update the "next release" time
 162                                 redis.call('rPush',kSlots,rSlot)
 163                                 redis.call('zAdd',kSlotsNextRelease,rTime + 30,rSlot)
 164                                 -- Always keep renewing the expiry on use
 165                                 redis.call('expireAt',kSlots,math.ceil(rTime + rExpiry))
 166                                 redis.call('expireAt',kSlotsNextRelease,math.ceil(rTime + rExpiry))
 167                         end
 168                 end
 169                 -- Update an ephemeral list to wake up other clients that can
 170                 -- reuse any cached work from this process. Only do this if no
 171                 -- slots are currently free (e.g. clients could be waiting).
 172                 if 1*rAwakeAll == 1 then
 173                         local count = redis.call('zCard',kWaiting)
 174                         for i = 1,count do
 175                                 redis.call('rPush',kWakeup,'w')
 176                         end
 177                         redis.call('pexpire',kWakeup,1)
 178                 end
 179                 return 1
 180 LUA;
 181                 try {
 182                         $res = $conn->luaEval( $script,
 183                                 array(
 184                                         $this->getSlotListKey(),
 185                                         $this->getSlotRTimeSetKey(),
 186                                         $this->getWakeupListKey(),
 187                                         $this->getWaitSetKey(),
 188                                         $this->workers,
 189                                         $this->lockTTL,
 190                                         $this->slot,
 191                                         $this->slotTime, // used for CAS-style sanity check
 192                                         ( $this->onRelease === self::AWAKE_ALL ) ? 1 : 0,
 193                                         microtime( true )
 194                                 ),
 195                                 4 # number of first argument(s) that are keys
 196                         );
 197                 } catch ( RedisException $e ) {
 198                         return Status::newFatal( 'pool-error-unknown', $e->getMessage() );
 199                 }
 200
 201                 $this->slot = null;
 202                 $this->slotTime = null;
 203                 $this->onRelease = null;
 204                 unset( self::$active[$this->session] );
 205
 206                 return Status::newGood( PoolCounter::RELEASED );
 207         }
 208
 209         /**
 210          * @param int $doWakeup AWAKE_* constant
 211          * @return Status
 212          */
 213         protected function waitForSlotOrNotif( $doWakeup ) {
 214                 if ( $this->slot !== null ) {
 215                         return Status::newGood( PoolCounter::LOCK_HELD ); // already acquired
 216                 }
 217
 218                 $status = $this->getConnection();
 219                 if ( !$status->isOK() ) {
 220                         return $status;
 221                 }
 222                 $conn = $status->value;
 223
 224                 $now = microtime( true );
 225                 try {
 226                         $slot = $this->initAndPopPoolSlotList( $conn, $now );
 227                         if ( ctype_digit( $slot ) ) {
 228                                 // Pool slot acquired by this process
 229                                 $slotTime = $now;
 230                         } elseif ( $slot === 'QUEUE_FULL' ) {
 231                                 // Too many processes are waiting for pooled processes to finish
 232                                 return Status::newGood( PoolCounter::QUEUE_FULL );
 233                         } elseif ( $slot === 'QUEUE_WAIT' ) {
 234                                 // This process is now registered as waiting
 235                                 $keys = ( $doWakeup == self::AWAKE_ALL )
 236                                         // Wait for an open slot or wake-up signal (preferring the later)
 237                                         ? array( $this->getWakeupListKey(), $this->getSlotListKey() )
 238                                         // Just wait for an actual pool slot
 239                                         : array( $this->getSlotListKey() );
 240
 241                                 $res = $conn->blPop( $keys, $this->timeout );
 242                                 if ( $res === array() ) {
 243                                         $conn->zRem( $this->getWaitSetKey(), $this->session ); // no longer waiting
 244                                         return Status::newGood( PoolCounter::TIMEOUT );
 245                                 }
 246
 247                                 $slot = $res[1]; // pool slot or "w" for wake-up notifications
 248                                 $slotTime = microtime( true ); // last microtime() was a few RTTs ago
 249                                 // Unregister this process as waiting and bump slot "next release" time
 250                                 $this->registerAcquisitionTime( $conn, $slot, $slotTime );
 251                         } else {
 252                                 return Status::newFatal( 'pool-error-unknown', "Server gave slot '$slot'." );
 253                         }
 254                 } catch ( RedisException $e ) {
 255                         return Status::newFatal( 'pool-error-unknown', $e->getMessage() );
 256                 }
 257
 258                 if ( $slot !== 'w' ) {
 259                         $this->slot = $slot;
 260                         $this->slotTime = $slotTime;
 261                         $this->onRelease = $doWakeup;
 262                         self::$active[$this->session] = $this;
 263                 }
 264
 265                 return Status::newGood( $slot === 'w' ? PoolCounter::DONE : PoolCounter::LOCKED );
 266         }
 267
 268         /**
 269          * @param RedisConnRef $conn
 270          * @param float $now UNIX timestamp
 271          * @return string|bool False on failure
 272          */
 273         protected function initAndPopPoolSlotList( RedisConnRef $conn, $now ) {
 274                 static $script =
 275 <<<LUA
 276                 local kSlots,kSlotsNextRelease,kSlotWaits = unpack(KEYS)
 277                 local rMaxWorkers,rMaxQueue,rTimeout,rExpiry,rSess,rTime = unpack(ARGV)
 278                 -- Initialize if the "next release" time sorted-set is empty. The slot key
 279                 -- itself is empty if all slots are busy or when nothing is initialized.
 280                 -- If the list is empty but the set is not, then it is the later case.
 281                 -- For sanity, if the list exists but not the set, then reset everything.
 282                 if redis.call('exists',kSlotsNextRelease) == 0 then
 283                         redis.call('del',kSlots)
 284                         for i = 1,1*rMaxWorkers do
 285                                 redis.call('rPush',kSlots,i)
 286                                 redis.call('zAdd',kSlotsNextRelease,-1,i)
 287                         end
 288                 -- Otherwise do maintenance to clean up after network partitions
 289                 else
 290                         -- Find stale slot locks and add free them (avoid duplicates for sanity)
 291                         local staleLocks = redis.call('zRangeByScore',kSlotsNextRelease,0,rTime)
 292                         for k,slot in ipairs(staleLocks) do
 293                                 redis.call('lRem',kSlots,0,slot)
 294                                 redis.call('rPush',kSlots,slot)
 295                                 redis.call('zAdd',kSlotsNextRelease,rTime + 30,slot)
 296                         end
 297                         -- Find stale wait slot entries and remove them
 298                         redis.call('zRemRangeByScore',kSlotWaits,0,rTime - 2*rTimeout)
 299                 end
 300                 local slot
 301                 -- Try to acquire a slot if possible now
 302                 if redis.call('lLen',kSlots) > 0 then
 303                         slot = redis.call('lPop',kSlots)
 304                         -- Update the slot "next release" time
 305                         redis.call('zAdd',kSlotsNextRelease,rTime + rExpiry,slot)
 306                 elseif redis.call('zCard',kSlotWaits) >= 1*rMaxQueue then
 307                         slot = 'QUEUE_FULL'
 308                 else
 309                         slot = 'QUEUE_WAIT'
 310                         -- Register this process as waiting
 311                         redis.call('zAdd',kSlotWaits,rTime,rSess)
 312                         redis.call('expireAt',kSlotWaits,math.ceil(rTime + 2*rTimeout))
 313                 end
 314                 -- Always keep renewing the expiry on use
 315                 redis.call('expireAt',kSlots,math.ceil(rTime + rExpiry))
 316                 redis.call('expireAt',kSlotsNextRelease,math.ceil(rTime + rExpiry))
 317                 return slot
 318 LUA;
 319                 return $conn->luaEval( $script,
 320                         array(
 321                                 $this->getSlotListKey(),
 322                                 $this->getSlotRTimeSetKey(),
 323                                 $this->getWaitSetKey(),
 324                                 $this->workers,
 325                                 $this->maxqueue,
 326                                 $this->timeout,
 327                                 $this->lockTTL,
 328                                 $this->session,
 329                                 $now
 330                         ),
 331                         3 # number of first argument(s) that are keys
 332                 );
 333         }
 334
 335         /**
 336          * @param RedisConnRef $conn
 337          * @param string $slot
 338          * @param float $now
 339          * @return int|bool False on failure
 340          */
 341         protected function registerAcquisitionTime( RedisConnRef $conn, $slot, $now ) {
 342                 static $script =
 343 <<<LUA
 344                 local kSlots,kSlotsNextRelease,kSlotWaits = unpack(KEYS)
 345                 local rSlot,rExpiry,rSess,rTime = unpack(ARGV)
 346                 -- If rSlot is 'w' then the client was told to wake up but got no slot
 347                 if rSlot ~= 'w' then
 348                         -- Update the slot "next release" time
 349                         redis.call('zAdd',kSlotsNextRelease,rTime + rExpiry,rSlot)
 350                         -- Always keep renewing the expiry on use
 351                         redis.call('expireAt',kSlots,math.ceil(rTime + rExpiry))
 352                         redis.call('expireAt',kSlotsNextRelease,math.ceil(rTime + rExpiry))
 353                 end
 354                 -- Unregister this process as waiting
 355                 redis.call('zRem',kSlotWaits,rSess)
 356                 return 1
 357 LUA;
 358                 return $conn->luaEval( $script,
 359                         array(
 360                                 $this->getSlotListKey(),
 361                                 $this->getSlotRTimeSetKey(),
 362                                 $this->getWaitSetKey(),
 363                                 $slot,
 364                                 $this->lockTTL,
 365                                 $this->session,
 366                                 $now
 367                         ),
 368                         3 # number of first argument(s) that are keys
 369                 );
 370         }
 371
 372         /**
 373          * @return string
 374          */
 375         protected function getSlotListKey() {
 376                 return "poolcounter:l-slots-{$this->keySha1}-{$this->workers}";
 377         }
 378
 379         /**
 380          * @return string
 381          */
 382         protected function getSlotRTimeSetKey() {
 383                 return "poolcounter:z-renewtime-{$this->keySha1}-{$this->workers}";
 384         }
 385
 386         /**
 387          * @return string
 388          */
 389         protected function getWaitSetKey() {
 390                 return "poolcounter:z-wait-{$this->keySha1}-{$this->workers}";
 391         }
 392
 393         /**
 394          * @return string
 395          */
 396         protected function getWakeupListKey() {
 397                 return "poolcounter:l-wakeup-{$this->keySha1}-{$this->workers}";
 398         }
 399
 400         /**
 401          * Try to make sure that locks get released (even with exceptions and fatals)
 402          */
 403         public static function releaseAll() {
 404                 foreach ( self::$active as $poolCounter ) {
 405                         try {
 406                                 if ( $poolCounter->slot !== null ) {
 407                                         $poolCounter->release();
 408                                 }
 409                         } catch ( Exception $e ) {
 410                         }
 411                 }
 412         }
 413 }