includes/libs/rdbms/loadmonitor/LoadMonitor.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  * @ingroup Database
  20  */
  21
  22 namespace Wikimedia\Rdbms;
  23
  24 use Psr\Log\LoggerInterface;
  25 use Psr\Log\NullLogger;
  26 use Wikimedia\ScopedCallback;
  27 use IDatabase;
  28 use BagOStuff;
  29
  30 /**
  31  * Basic DB load monitor with no external dependencies
  32  * Uses memcached to cache the replication lag for a short time
  33  *
  34  * @ingroup Database
  35  */
  36 class LoadMonitor implements ILoadMonitor {
  37         /** @var ILoadBalancer */
  38         protected $parent;
  39         /** @var BagOStuff */
  40         protected $srvCache;
  41         /** @var BagOStuff */
  42         protected $mainCache;
  43         /** @var LoggerInterface */
  44         protected $replLogger;
  45
  46         /** @var float Moving average ratio (e.g. 0.1 for 10% weight to new weight) */
  47         private $movingAveRatio;
  48
  49         const VERSION = 1; // cache key version
  50
  51         public function __construct(
  52                 ILoadBalancer $lb, BagOStuff $srvCache, BagOStuff $cache, array $options = []
  53         ) {
  54                 $this->parent = $lb;
  55                 $this->srvCache = $srvCache;
  56                 $this->mainCache = $cache;
  57                 $this->replLogger = new NullLogger();
  58
  59                 $this->movingAveRatio = isset( $options['movingAveRatio'] )
  60                         ? $options['movingAveRatio']
  61                         : 0.1;
  62         }
  63
  64         public function setLogger( LoggerInterface $logger ) {
  65                 $this->replLogger = $logger;
  66         }
  67
  68         public function scaleLoads( array &$weightByServer, $domain ) {
  69                 $serverIndexes = array_keys( $weightByServer );
  70                 $states = $this->getServerStates( $serverIndexes, $domain );
  71                 $coefficientsByServer = $states['weightScales'];
  72                 foreach ( $weightByServer as $i => $weight ) {
  73                         if ( isset( $coefficientsByServer[$i] ) ) {
  74                                 $weightByServer[$i] = $weight * $coefficientsByServer[$i];
  75                         } else { // server recently added to config?
  76                                 $host = $this->parent->getServerName( $i );
  77                                 $this->replLogger->error( __METHOD__ . ": host $host not in cache" );
  78                         }
  79                 }
  80         }
  81
  82         public function getLagTimes( array $serverIndexes, $domain ) {
  83                 $states = $this->getServerStates( $serverIndexes, $domain );
  84
  85                 return $states['lagTimes'];
  86         }
  87
  88         protected function getServerStates( array $serverIndexes, $domain ) {
  89                 $writerIndex = $this->parent->getWriterIndex();
  90                 if ( count( $serverIndexes ) == 1 && reset( $serverIndexes ) == $writerIndex ) {
  91                         # Single server only, just return zero without caching
  92                         return [
  93                                 'lagTimes' => [ $writerIndex => 0 ],
  94                                 'weightScales' => [ $writerIndex => 1.0 ]
  95                         ];
  96                 }
  97
  98                 $key = $this->getCacheKey( $serverIndexes );
  99                 # Randomize TTLs to reduce stampedes (4.0 - 5.0 sec)
 100                 $ttl = mt_rand( 4e6, 5e6 ) / 1e6;
 101                 # Keep keys around longer as fallbacks
 102                 $staleTTL = 60;
 103
 104                 # (a) Check the local APC cache
 105                 $value = $this->srvCache->get( $key );
 106                 if ( $value && $value['timestamp'] > ( microtime( true ) - $ttl ) ) {
 107                         $this->replLogger->debug( __METHOD__ . ": got lag times ($key) from local cache" );
 108                         return $value; // cache hit
 109                 }
 110                 $staleValue = $value ?: false;
 111
 112                 # (b) Check the shared cache and backfill APC
 113                 $value = $this->mainCache->get( $key );
 114                 if ( $value && $value['timestamp'] > ( microtime( true ) - $ttl ) ) {
 115                         $this->srvCache->set( $key, $value, $staleTTL );
 116                         $this->replLogger->debug( __METHOD__ . ": got lag times ($key) from main cache" );
 117
 118                         return $value; // cache hit
 119                 }
 120                 $staleValue = $value ?: $staleValue;
 121
 122                 # (c) Cache key missing or expired; regenerate and backfill
 123                 if ( $this->mainCache->lock( $key, 0, 10 ) ) {
 124                         # Let this process alone update the cache value
 125                         $cache = $this->mainCache;
 126                         /** @noinspection PhpUnusedLocalVariableInspection */
 127                         $unlocker = new ScopedCallback( function () use ( $cache, $key ) {
 128                                 $cache->unlock( $key );
 129                         } );
 130                 } elseif ( $staleValue ) {
 131                         # Could not acquire lock but an old cache exists, so use it
 132                         return $staleValue;
 133                 }
 134
 135                 $lagTimes = [];
 136                 $weightScales = [];
 137                 $movAveRatio = $this->movingAveRatio;
 138                 foreach ( $serverIndexes as $i ) {
 139                         if ( $i == $this->parent->getWriterIndex() ) {
 140                                 $lagTimes[$i] = 0; // master always has no lag
 141                                 $weightScales[$i] = 1.0; // nominal weight
 142                                 continue;
 143                         }
 144
 145                         $conn = $this->parent->getAnyOpenConnection( $i );
 146                         if ( $conn ) {
 147                                 $close = false; // already open
 148                         } else {
 149                                 $conn = $this->parent->openConnection( $i, $domain );
 150                                 $close = true; // new connection
 151                         }
 152
 153                         $lastWeight = isset( $staleValue['weightScales'][$i] )
 154                                 ? $staleValue['weightScales'][$i]
 155                                 : 1.0;
 156                         $coefficient = $this->getWeightScale( $i, $conn ?: null );
 157                         $newWeight = $movAveRatio * $coefficient + ( 1 - $movAveRatio ) * $lastWeight;
 158
 159                         // Scale from 10% to 100% of nominal weight
 160                         $weightScales[$i] = max( $newWeight, .10 );
 161
 162                         if ( !$conn ) {
 163                                 $lagTimes[$i] = false;
 164                                 $host = $this->parent->getServerName( $i );
 165                                 $this->replLogger->error( __METHOD__ . ": host $host is unreachable" );
 166                                 continue;
 167                         }
 168
 169                         if ( $conn->getLBInfo( 'is static' ) ) {
 170                                 $lagTimes[$i] = 0;
 171                         } else {
 172                                 $lagTimes[$i] = $conn->getLag();
 173                                 if ( $lagTimes[$i] === false ) {
 174                                         $host = $this->parent->getServerName( $i );
 175                                         $this->replLogger->error( __METHOD__ . ": host $host is not replicating?" );
 176                                 }
 177                         }
 178
 179                         if ( $close ) {
 180                                 # Close the connection to avoid sleeper connections piling up.
 181                                 # Note that the caller will pick one of these DBs and reconnect,
 182                                 # which is slightly inefficient, but this only matters for the lag
 183                                 # time cache miss cache, which is far less common that cache hits.
 184                                 $this->parent->closeConnection( $conn );
 185                         }
 186                 }
 187
 188                 # Add a timestamp key so we know when it was cached
 189                 $value = [
 190                         'lagTimes' => $lagTimes,
 191                         'weightScales' => $weightScales,
 192                         'timestamp' => microtime( true )
 193                 ];
 194                 $this->mainCache->set( $key, $value, $staleTTL );
 195                 $this->srvCache->set( $key, $value, $staleTTL );
 196                 $this->replLogger->info( __METHOD__ . ": re-calculated lag times ($key)" );
 197
 198                 return $value;
 199         }
 200
 201         /**
 202          * @param integer $index Server index
 203          * @param IDatabase|null $conn Connection handle or null on connection failure
 204          * @return float
 205          */
 206         protected function getWeightScale( $index, IDatabase $conn = null ) {
 207                 return $conn ? 1.0 : 0.0;
 208         }
 209
 210         private function getCacheKey( array $serverIndexes ) {
 211                 sort( $serverIndexes );
 212                 // Lag is per-server, not per-DB, so key on the master DB name
 213                 return $this->srvCache->makeGlobalKey(
 214                         'lag-times',
 215                         self::VERSION,
 216                         $this->parent->getServerName( $this->parent->getWriterIndex() ),
 217                         implode( '-', $serverIndexes )
 218                 );
 219         }
 220 }