MediaWiki  master
ExternalStoreDB.php
Go to the documentation of this file.
1 <?php
27 use Wikimedia\ScopedCallback;
28 
40  private $lbFactory;
41 
47  public function __construct( array $params ) {
48  parent::__construct( $params );
49  if ( !isset( $params['lbFactory'] ) || !( $params['lbFactory'] instanceof LBFactory ) ) {
50  throw new InvalidArgumentException( "LBFactory required in 'lbFactory' field." );
51  }
52  $this->lbFactory = $params['lbFactory'];
53  }
54 
65  public function fetchFromURL( $url ) {
66  list( $cluster, $id, $itemID ) = $this->parseURL( $url );
67  $ret = $this->fetchBlob( $cluster, $id, $itemID );
68 
69  if ( $itemID !== false && $ret !== false ) {
70  return $ret->getItem( $itemID );
71  }
72 
73  return $ret;
74  }
75 
86  public function batchFetchFromURLs( array $urls ) {
87  $batched = $inverseUrlMap = [];
88  foreach ( $urls as $url ) {
89  list( $cluster, $id, $itemID ) = $this->parseURL( $url );
90  $batched[$cluster][$id][] = $itemID;
91  // false $itemID gets cast to int, but should be ok
92  // since we do === from the $itemID in $batched
93  $inverseUrlMap[$cluster][$id][$itemID] = $url;
94  }
95  $ret = [];
96  foreach ( $batched as $cluster => $batchByCluster ) {
97  $res = $this->batchFetchBlobs( $cluster, $batchByCluster );
99  foreach ( $res as $id => $blob ) {
100  foreach ( $batchByCluster[$id] as $itemID ) {
101  $url = $inverseUrlMap[$cluster][$id][$itemID];
102  if ( $itemID === false ) {
103  $ret[$url] = $blob;
104  } else {
105  $ret[$url] = $blob->getItem( $itemID );
106  }
107  }
108  }
109  }
110 
111  return $ret;
112  }
113 
117  public function store( $location, $data ) {
118  $dbw = $this->getPrimary( $location );
119  $dbw->insert(
120  $this->getTable( $dbw, $location ),
121  [ 'blob_text' => $data ],
122  __METHOD__
123  );
124  $id = $dbw->insertId();
125  if ( !$id ) {
126  throw new MWException( __METHOD__ . ': no insert ID' );
127  }
128 
129  return "DB://$location/$id";
130  }
131 
135  public function isReadOnly( $location ) {
136  if ( parent::isReadOnly( $location ) ) {
137  return true;
138  }
139 
140  $lb = $this->getLoadBalancer( $location );
141  $domainId = $this->getDomainId( $lb->getServerInfo( $lb->getWriterIndex() ) );
142 
143  return ( $lb->getReadOnlyReason( $domainId ) !== false );
144  }
145 
152  private function getLoadBalancer( $cluster ) {
153  return $this->lbFactory->getExternalLB( $cluster );
154  }
155 
163  public function getReplica( $cluster ) {
164  $lb = $this->getLoadBalancer( $cluster );
165 
166  return $lb->getConnectionRef(
167  DB_REPLICA,
168  [],
169  $this->getDomainId( $lb->getServerInfo( $lb->getWriterIndex() ) ),
170  $lb::CONN_TRX_AUTOCOMMIT
171  );
172  }
173 
181  public function getPrimary( $cluster ) {
182  $lb = $this->getLoadBalancer( $cluster );
183 
184  return $lb->getMaintenanceConnectionRef(
185  DB_PRIMARY,
186  [],
187  $this->getDomainId( $lb->getServerInfo( $lb->getWriterIndex() ) ),
188  $lb::CONN_TRX_AUTOCOMMIT
189  );
190  }
191 
197  public function getMaster( $cluster ) {
198  wfDeprecated( __METHOD__, '1.37' );
199  return $this->getPrimary( $cluster );
200  }
201 
206  private function getDomainId( array $server ) {
207  if ( $this->isDbDomainExplicit ) {
208  return $this->dbDomain; // explicit foreign domain
209  }
210 
211  if ( isset( $server['dbname'] ) ) {
212  // T200471: for b/c, treat any "dbname" field as forcing which database to use.
213  // MediaWiki/LoadBalancer previously did not enforce any concept of a local DB
214  // domain, but rather assumed that the LB server configuration matched $wgDBname.
215  // This check is useful when the external storage DB for this cluster does not use
216  // the same name as the corresponding "main" DB(s) for wikis.
217  $domain = new DatabaseDomain(
218  $server['dbname'],
219  $server['schema'] ?? null,
220  $server['tablePrefix'] ?? ''
221  );
222 
223  return $domain->getId();
224  }
225 
226  return false; // local LB domain
227  }
228 
236  public function getTable( $db, $cluster = null ) {
237  if ( $cluster !== null ) {
238  $lb = $this->getLoadBalancer( $cluster );
239  $info = $lb->getServerInfo( $lb->getWriterIndex() );
240  if ( isset( $info['blobs table'] ) ) {
241  return $info['blobs table'];
242  }
243  }
244 
245  return $db->getLBInfo( 'blobs table' ) ?? 'blobs'; // b/c
246  }
247 
255  public function initializeTable( $cluster ) {
256  global $IP;
257 
258  static $supportedTypes = [ 'mysql', 'sqlite' ];
259 
260  $dbw = $this->getPrimary( $cluster );
261  if ( !in_array( $dbw->getType(), $supportedTypes, true ) ) {
262  throw new DBUnexpectedError( $dbw, "RDBMS type '{$dbw->getType()}' not supported." );
263  }
264 
265  $sqlFilePath = "$IP/maintenance/storage/blobs.sql";
266  $sql = file_get_contents( $sqlFilePath );
267  if ( $sql === false ) {
268  throw new RuntimeException( "Failed to read '$sqlFilePath'." );
269  }
270 
271  $rawTable = $this->getTable( $dbw, $cluster ); // e.g. "blobs_cluster23"
272  $encTable = $dbw->tableName( $rawTable );
273  $dbw->query(
274  str_replace(
275  [ '/*$wgDBprefix*/blobs', '/*_*/blobs' ],
276  [ $encTable, $encTable ],
277  $sql
278  ),
279  __METHOD__,
280  $dbw::QUERY_IGNORE_DBO_TRX
281  );
282  }
283 
293  private function fetchBlob( $cluster, $id, $itemID ) {
300  static $externalBlobCache = [];
301 
302  $cacheID = ( $itemID === false ) ? "$cluster/$id" : "$cluster/$id/";
303  $cacheID = "$cacheID@{$this->dbDomain}";
304 
305  if ( isset( $externalBlobCache[$cacheID] ) ) {
306  $this->logger->debug( __METHOD__ . ": cache hit on $cacheID" );
307 
308  return $externalBlobCache[$cacheID];
309  }
310 
311  $this->logger->debug( __METHOD__ . ": cache miss on $cacheID" );
312 
313  $dbr = $this->getReplica( $cluster );
314  $ret = $dbr->selectField(
315  $this->getTable( $dbr, $cluster ),
316  'blob_text',
317  [ 'blob_id' => $id ],
318  __METHOD__
319  );
320  if ( $ret === false ) {
321  // Try the primary DB
322  $this->logger->warning( __METHOD__ . ": primary DB fallback on $cacheID" );
323  $scope = $this->lbFactory->getTransactionProfiler()->silenceForScope();
324  $dbw = $this->getPrimary( $cluster );
325  $ret = $dbw->selectField(
326  $this->getTable( $dbw, $cluster ),
327  'blob_text',
328  [ 'blob_id' => $id ],
329  __METHOD__
330  );
331  ScopedCallback::consume( $scope );
332  if ( $ret === false ) {
333  $this->logger->warning( __METHOD__ . ": primary DB failed to find $cacheID" );
334  }
335  }
336  if ( $itemID !== false && $ret !== false ) {
337  // Unserialise object; caller extracts item
338  $ret = unserialize( $ret );
339  }
340 
341  $externalBlobCache = [ $cacheID => $ret ];
342 
343  return $ret;
344  }
345 
354  private function batchFetchBlobs( $cluster, array $ids ) {
355  $dbr = $this->getReplica( $cluster );
356  $res = $dbr->newSelectQueryBuilder()
357  ->select( [ 'blob_id', 'blob_text' ] )
358  ->from( $this->getTable( $dbr, $cluster ) )
359  ->where( [ 'blob_id' => array_keys( $ids ) ] )
360  ->caller( __METHOD__ )
361  ->fetchResultSet();
362 
363  $ret = [];
364  if ( $res !== false ) {
365  $this->mergeBatchResult( $ret, $ids, $res );
366  }
367  if ( $ids ) {
368  // Try the primary
369  $this->logger->info(
370  __METHOD__ . ": primary fallback on '$cluster' for: " .
371  implode( ',', array_keys( $ids ) )
372  );
373  $scope = $this->lbFactory->getTransactionProfiler()->silenceForScope();
374  $dbw = $this->getPrimary( $cluster );
375  $res = $dbw->newSelectQueryBuilder()
376  ->select( [ 'blob_id', 'blob_text' ] )
377  ->from( $this->getTable( $dbr, $cluster ) )
378  ->where( [ 'blob_id' => array_keys( $ids ) ] )
379  ->caller( __METHOD__ )
380  ->fetchResultSet();
381  ScopedCallback::consume( $scope );
382  if ( $res === false ) {
383  $this->logger->error( __METHOD__ . ": primary failed on '$cluster'" );
384  } else {
385  $this->mergeBatchResult( $ret, $ids, $res );
386  }
387  }
388  if ( $ids ) {
389  $this->logger->error(
390  __METHOD__ . ": primary on '$cluster' failed locating items: " .
391  implode( ',', array_keys( $ids ) )
392  );
393  }
394 
395  return $ret;
396  }
397 
404  private function mergeBatchResult( array &$ret, array &$ids, $res ) {
405  foreach ( $res as $row ) {
406  $id = $row->blob_id;
407  $itemIDs = $ids[$id];
408  unset( $ids[$id] ); // to track if everything is found
409  if ( count( $itemIDs ) === 1 && reset( $itemIDs ) === false ) {
410  // single result stored per blob
411  $ret[$id] = $row->blob_text;
412  } else {
413  // multi result stored per blob
414  $ret[$id] = unserialize( $row->blob_text );
415  }
416  }
417  }
418 
423  protected function parseURL( $url ) {
424  $path = explode( '/', $url );
425 
426  return [
427  $path[2], // cluster
428  $path[3], // id
429  $path[4] ?? false // itemID
430  ];
431  }
432 }
unserialize( $serialized)
wfDeprecated( $function, $version=false, $component=false, $callerOffset=2)
Logs a warning that a deprecated feature was used.
if(!defined( 'MEDIAWIKI')) if(ini_get( 'mbstring.func_overload')) if(!defined( 'MW_ENTRY_POINT')) global $IP
Environment checks.
Definition: Setup.php:91
External storage in a SQL database.
getPrimary( $cluster)
Get a primary database connection for the specified cluster.
__construct(array $params)
getReplica( $cluster)
Get a replica DB connection for the specified cluster.
getMaster( $cluster)
initializeTable( $cluster)
Create the appropriate blobs table on this cluster.
fetchFromURL( $url)
Fetch data from given external store URL.
getTable( $db, $cluster=null)
Get the 'blobs' table name for this database.
store( $location, $data)
Insert a data item into a given location.The location name The data item string|bool The URL of the s...
batchFetchFromURLs(array $urls)
Fetch multiple URLs from given external store.
isReadOnly( $location)
Check if a given location is read-only.The location name bool Whether this location is read-only 1....
Base class for external storage.
array $params
Usage context options for this instance.
string $dbDomain
Default database domain to store content under.
MediaWiki exception.
Definition: MWException.php:29
Helper class used for automatically marking an IDatabase connection as reusable (once it no longer ma...
Definition: DBConnRef.php:29
Class to handle database/schema/prefix specifications for IDatabase.
Basic database interface for live and lazy-loaded relation database handles.
Definition: IDatabase.php:39
Create and track the database connections and transactions for a given database cluster.
const DB_REPLICA
Definition: defines.php:26
const DB_PRIMARY
Definition: defines.php:28