3 * External storage in SQL database.
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License along
16 * with this program; if not, write to the Free Software Foundation, Inc.,
17 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
18 * http://www.gnu.org/copyleft/gpl.html
23 use MediaWiki\MediaWikiServices
;
24 use Wikimedia\Rdbms\ILoadBalancer
;
25 use Wikimedia\Rdbms\IDatabase
;
26 use Wikimedia\Rdbms\DBConnRef
;
27 use Wikimedia\Rdbms\MaintainableDBConnRef
;
28 use Wikimedia\Rdbms\DatabaseDomain
;
31 * DB accessible external objects.
33 * In this system, each store "location" maps to a database "cluster".
34 * The clusters must be defined in the normal LBFactory configuration.
36 * @ingroup ExternalStorage
38 class ExternalStoreDB
extends ExternalStoreMedium
{
40 * The provided URL is in the form of DB://cluster/id
41 * or DB://cluster/id/itemid for concatened storage.
44 * @return string|bool False if missing
45 * @see ExternalStoreMedium::fetchFromURL()
47 public function fetchFromURL( $url ) {
48 list( $cluster, $id, $itemID ) = $this->parseURL( $url );
49 $ret = $this->fetchBlob( $cluster, $id, $itemID );
51 if ( $itemID !== false && $ret !== false ) {
52 return $ret->getItem( $itemID );
59 * Fetch data from given external store URLs.
60 * The provided URLs are in the form of DB://cluster/id
61 * or DB://cluster/id/itemid for concatened storage.
63 * @param array $urls An array of external store URLs
64 * @return array A map from url to stored content. Failed results
65 * are not represented.
67 public function batchFetchFromURLs( array $urls ) {
68 $batched = $inverseUrlMap = [];
69 foreach ( $urls as $url ) {
70 list( $cluster, $id, $itemID ) = $this->parseURL( $url );
71 $batched[$cluster][$id][] = $itemID;
72 // false $itemID gets cast to int, but should be ok
73 // since we do === from the $itemID in $batched
74 $inverseUrlMap[$cluster][$id][$itemID] = $url;
77 foreach ( $batched as $cluster => $batchByCluster ) {
78 $res = $this->batchFetchBlobs( $cluster, $batchByCluster );
79 /** @var HistoryBlob $blob */
80 foreach ( $res as $id => $blob ) {
81 foreach ( $batchByCluster[$id] as $itemID ) {
82 $url = $inverseUrlMap[$cluster][$id][$itemID];
83 if ( $itemID === false ) {
86 $ret[$url] = $blob->getItem( $itemID );
95 public function store( $location, $data ) {
96 $dbw = $this->getMaster( $location );
97 $dbw->insert( $this->getTable( $dbw ),
98 [ 'blob_text' => $data ],
100 $id = $dbw->insertId();
102 throw new MWException( __METHOD__
. ': no insert ID' );
105 return "DB://$location/$id";
108 public function isReadOnly( $location ) {
109 return ( $this->getLoadBalancer( $location )->getReadOnlyReason() !== false );
113 * Get a LoadBalancer for the specified cluster
115 * @param string $cluster Cluster name
116 * @return ILoadBalancer
118 private function getLoadBalancer( $cluster ) {
119 $lbFactory = MediaWikiServices
::getInstance()->getDBLoadBalancerFactory();
120 return $lbFactory->getExternalLB( $cluster );
124 * Get a replica DB connection for the specified cluster
126 * @param string $cluster Cluster name
129 public function getSlave( $cluster ) {
130 global $wgDefaultExternalStore;
132 $lb = $this->getLoadBalancer( $cluster );
133 $domainId = $this->getDomainId( $lb->getServerInfo( $lb->getWriterIndex() ) );
135 if ( !in_array( "DB://" . $cluster, (array)$wgDefaultExternalStore ) ) {
136 wfDebug( "read only external store\n" );
137 $lb->allowLagged( true );
139 wfDebug( "writable external store\n" );
142 $db = $lb->getConnectionRef( DB_REPLICA
, [], $domainId );
143 $db->clearFlag( DBO_TRX
); // sanity
149 * Get a master database connection for the specified cluster
151 * @param string $cluster Cluster name
152 * @return MaintainableDBConnRef
154 public function getMaster( $cluster ) {
155 $lb = $this->getLoadBalancer( $cluster );
156 $domainId = $this->getDomainId( $lb->getServerInfo( $lb->getWriterIndex() ) );
158 $db = $lb->getMaintenanceConnectionRef( DB_MASTER
, [], $domainId );
159 $db->clearFlag( DBO_TRX
); // sanity
165 * @param array $server Master DB server configuration array for LoadBalancer
166 * @return string|bool Database domain ID or false
168 private function getDomainId( array $server ) {
169 if ( isset( $this->params
['wiki'] ) ) {
170 return $this->params
['wiki']; // explicit domain
173 if ( isset( $server['dbname'] ) ) {
174 // T200471: for b/c, treat any "dbname" field as forcing which database to use.
175 // MediaWiki/LoadBalancer previously did not enforce any concept of a local DB
176 // domain, but rather assumed that the LB server configuration matched $wgDBname.
177 // This check is useful when the external storage DB for this cluster does not use
178 // the same name as the corresponding "main" DB(s) for wikis.
179 $domain = new DatabaseDomain(
181 $server['schema'] ??
null,
182 $server['tablePrefix'] ??
''
185 return $domain->getId();
188 return false; // local LB domain
192 * Get the 'blobs' table name for this database
194 * @param IDatabase $db
195 * @return string Table name ('blobs' by default)
197 public function getTable( $db ) {
198 $table = $db->getLBInfo( 'blobs table' );
199 if ( is_null( $table ) ) {
207 * Fetch a blob item out of the database; a cache of the last-loaded
208 * blob will be kept so that multiple loads out of a multi-item blob
209 * can avoid redundant database access and decompression.
210 * @param string $cluster
212 * @param string $itemID
213 * @return HistoryBlob|bool Returns false if missing
215 private function fetchBlob( $cluster, $id, $itemID ) {
217 * One-step cache variable to hold base blobs; operations that
218 * pull multiple revisions may often pull multiple times from
219 * the same blob. By keeping the last-used one open, we avoid
220 * redundant unserialization and decompression overhead.
222 static $externalBlobCache = [];
224 $cacheID = ( $itemID === false ) ?
"$cluster/$id" : "$cluster/$id/";
225 if ( isset( $externalBlobCache[$cacheID] ) ) {
226 wfDebugLog( 'ExternalStoreDB-cache',
227 "ExternalStoreDB::fetchBlob cache hit on $cacheID" );
229 return $externalBlobCache[$cacheID];
232 wfDebugLog( 'ExternalStoreDB-cache',
233 "ExternalStoreDB::fetchBlob cache miss on $cacheID" );
235 $dbr = $this->getSlave( $cluster );
236 $ret = $dbr->selectField( $this->getTable( $dbr ),
237 'blob_text', [ 'blob_id' => $id ], __METHOD__
);
238 if ( $ret === false ) {
239 wfDebugLog( 'ExternalStoreDB',
240 "ExternalStoreDB::fetchBlob master fallback on $cacheID" );
242 $dbw = $this->getMaster( $cluster );
243 $ret = $dbw->selectField( $this->getTable( $dbw ),
244 'blob_text', [ 'blob_id' => $id ], __METHOD__
);
245 if ( $ret === false ) {
246 wfDebugLog( 'ExternalStoreDB',
247 "ExternalStoreDB::fetchBlob master failed to find $cacheID" );
250 if ( $itemID !== false && $ret !== false ) {
251 // Unserialise object; caller extracts item
252 $ret = unserialize( $ret );
255 $externalBlobCache = [ $cacheID => $ret ];
261 * Fetch multiple blob items out of the database
263 * @param string $cluster A cluster name valid for use with LBFactory
264 * @param array $ids A map from the blob_id's to look for to the requested itemIDs in the blobs
265 * @return array A map from the blob_id's requested to their content.
266 * Unlocated ids are not represented
268 private function batchFetchBlobs( $cluster, array $ids ) {
269 $dbr = $this->getSlave( $cluster );
270 $res = $dbr->select( $this->getTable( $dbr ),
271 [ 'blob_id', 'blob_text' ], [ 'blob_id' => array_keys( $ids ) ], __METHOD__
);
273 if ( $res !== false ) {
274 $this->mergeBatchResult( $ret, $ids, $res );
277 wfDebugLog( __CLASS__
, __METHOD__
.
278 " master fallback on '$cluster' for: " .
279 implode( ',', array_keys( $ids ) ) );
281 $dbw = $this->getMaster( $cluster );
282 $res = $dbw->select( $this->getTable( $dbr ),
283 [ 'blob_id', 'blob_text' ],
284 [ 'blob_id' => array_keys( $ids ) ],
286 if ( $res === false ) {
287 wfDebugLog( __CLASS__
, __METHOD__
. " master failed on '$cluster'" );
289 $this->mergeBatchResult( $ret, $ids, $res );
293 wfDebugLog( __CLASS__
, __METHOD__
.
294 " master on '$cluster' failed locating items: " .
295 implode( ',', array_keys( $ids ) ) );
302 * Helper function for self::batchFetchBlobs for merging master/replica DB results
303 * @param array &$ret Current self::batchFetchBlobs return value
304 * @param array &$ids Map from blob_id to requested itemIDs
305 * @param mixed $res DB result from Database::select
307 private function mergeBatchResult( array &$ret, array &$ids, $res ) {
308 foreach ( $res as $row ) {
310 $itemIDs = $ids[$id];
311 unset( $ids[$id] ); // to track if everything is found
312 if ( count( $itemIDs ) === 1 && reset( $itemIDs ) === false ) {
313 // single result stored per blob
314 $ret[$id] = $row->blob_text
;
316 // multi result stored per blob
317 $ret[$id] = unserialize( $row->blob_text
);
326 protected function parseURL( $url ) {
327 $path = explode( '/', $url );
332 isset( $path[4] ) ?
$path[4] : false // itemID