Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
69.21% |
245 / 354 |
|
70.73% |
29 / 41 |
CRAP | |
0.00% |
0 / 1 |
| FileBackendMultiWrite | |
69.41% |
245 / 353 |
|
70.73% |
29 / 41 |
605.57 | |
0.00% |
0 / 1 |
| __construct | |
85.71% |
24 / 28 |
|
0.00% |
0 / 1 |
10.29 | |||
| doOperationsInternal | |
61.54% |
32 / 52 |
|
0.00% |
0 / 1 |
22.62 | |||
| consistencyCheck | |
79.07% |
34 / 43 |
|
0.00% |
0 / 1 |
22.31 | |||
| accessibilityCheck | |
77.78% |
7 / 9 |
|
0.00% |
0 / 1 |
5.27 | |||
| resyncFiles | |
0.00% |
0 / 51 |
|
0.00% |
0 / 1 |
272 | |||
| fileStoragePathsForOps | |
90.91% |
10 / 11 |
|
0.00% |
0 / 1 |
7.04 | |||
| substOpBatchPaths | |
100.00% |
8 / 8 |
|
100.00% |
1 / 1 |
4 | |||
| substOpPaths | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| substPaths | |
100.00% |
5 / 5 |
|
100.00% |
1 / 1 |
1 | |||
| unsubstPaths | |
100.00% |
5 / 5 |
|
100.00% |
1 / 1 |
1 | |||
| hasVolatileSources | |
75.00% |
3 / 4 |
|
0.00% |
0 / 1 |
4.25 | |||
| doQuickOperationsInternal | |
100.00% |
19 / 19 |
|
100.00% |
1 / 1 |
5 | |||
| doPrepare | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| doSecure | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| doPublish | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| doClean | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| doDirectoryOp | |
68.75% |
11 / 16 |
|
0.00% |
0 / 1 |
4.49 | |||
| concatenate | |
100.00% |
5 / 5 |
|
100.00% |
1 / 1 |
1 | |||
| fileExists | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| getFileTimestamp | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| getFileSize | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| getFileStat | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| getFileXAttributes | |
0.00% |
0 / 3 |
|
0.00% |
0 / 1 |
2 | |||
| getFileContentsMulti | |
100.00% |
7 / 7 |
|
100.00% |
1 / 1 |
2 | |||
| getFileSha1Base36 | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| getFileProps | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| streamFile | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| getLocalReferenceMulti | |
100.00% |
7 / 7 |
|
100.00% |
1 / 1 |
2 | |||
| getLocalCopyMulti | |
100.00% |
7 / 7 |
|
100.00% |
1 / 1 |
2 | |||
| getFileHttpUrl | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| addShellboxInputFile | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| directoryExists | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| getDirectoryList | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| getFileList | |
75.00% |
3 / 4 |
|
0.00% |
0 / 1 |
3.14 | |||
| getFileListForWrite | |
0.00% |
0 / 8 |
|
0.00% |
0 / 1 |
20 | |||
| getFeatures | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| clearCache | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
3 | |||
| preloadCache | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| preloadFileStat | |
0.00% |
0 / 3 |
|
0.00% |
0 / 1 |
2 | |||
| getScopedLocksForOps | |
100.00% |
14 / 14 |
|
100.00% |
1 / 1 |
1 | |||
| getReadIndexFromParams | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
2 | |||
| 1 | <?php |
| 2 | /** |
| 3 | * Proxy backend that mirrors writes to several internal backends. |
| 4 | * |
| 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. |
| 9 | * |
| 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. |
| 14 | * |
| 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 |
| 19 | * |
| 20 | * @file |
| 21 | * @ingroup FileBackend |
| 22 | */ |
| 23 | |
| 24 | namespace Wikimedia\FileBackend; |
| 25 | |
| 26 | use InvalidArgumentException; |
| 27 | use LockManager; |
| 28 | use LogicException; |
| 29 | use Shellbox\Command\BoxedCommand; |
| 30 | use StatusValue; |
| 31 | use StringUtils; |
| 32 | use Wikimedia\Timestamp\ConvertibleTimestamp; |
| 33 | |
| 34 | /** |
| 35 | * @brief Proxy backend that mirrors writes to several internal backends. |
| 36 | * |
| 37 | * This class defines a multi-write backend. Multiple backends can be |
| 38 | * registered to this proxy backend and it will act as a single backend. |
| 39 | * Use this when all access to those backends is through this proxy backend. |
| 40 | * At least one of the backends must be declared the "master" backend. |
| 41 | * |
| 42 | * Only use this class when transitioning from one storage system to another. |
| 43 | * |
| 44 | * Read operations are only done on the 'master' backend for consistency. |
| 45 | * Except on getting list of thumbnails for write operations. |
| 46 | * Write operations are performed on all backends, starting with the master. |
| 47 | * This makes a best-effort to have transactional semantics, but since requests |
| 48 | * may sometimes fail, the use of "autoResync" or background scripts to fix |
| 49 | * inconsistencies is important. |
| 50 | * |
| 51 | * @ingroup FileBackend |
| 52 | * @since 1.19 |
| 53 | */ |
| 54 | class FileBackendMultiWrite extends FileBackend { |
| 55 | /** @var FileBackendStore[] Prioritized list of FileBackendStore objects */ |
| 56 | protected $backends = []; |
| 57 | |
| 58 | /** @var int Index of master backend */ |
| 59 | protected $masterIndex = -1; |
| 60 | /** @var int Index of read affinity backend */ |
| 61 | protected $readIndex = -1; |
| 62 | |
| 63 | /** @var int Bitfield */ |
| 64 | protected $syncChecks = 0; |
| 65 | /** @var string|bool */ |
| 66 | protected $autoResync = false; |
| 67 | |
| 68 | /** @var bool */ |
| 69 | protected $asyncWrites = false; |
| 70 | |
| 71 | /** @var int Compare file sizes among backends */ |
| 72 | private const CHECK_SIZE = 1; |
| 73 | /** @var int Compare file mtimes among backends */ |
| 74 | private const CHECK_TIME = 2; |
| 75 | /** @var int Compare file hashes among backends */ |
| 76 | private const CHECK_SHA1 = 4; |
| 77 | |
| 78 | /** |
| 79 | * Construct a proxy backend that consists of several internal backends. |
| 80 | * Locking and read-only checks are handled by the proxy backend. |
| 81 | * |
| 82 | * Additional $config params include: |
| 83 | * - backends : Array of backend config and multi-backend settings. |
| 84 | * Each value is the config used in the constructor of a |
| 85 | * FileBackendStore class, but with these additional settings: |
| 86 | * - class : The name of the backend class |
| 87 | * - isMultiMaster : This must be set for one backend. |
| 88 | * - readAffinity : Use this for reads without 'latest' set. |
| 89 | * - syncChecks : Integer bitfield of internal backend sync checks to perform. |
| 90 | * Possible bits include the FileBackendMultiWrite::CHECK_* constants. |
| 91 | * There are constants for SIZE, TIME, and SHA1. |
| 92 | * The checks are done before allowing any file operations. |
| 93 | * - autoResync : Automatically resync the clone backends to the master backend |
| 94 | * when pre-operation sync checks fail. This should only be used |
| 95 | * if the master backend is stable and not missing any files. |
| 96 | * Use "conservative" to limit resyncing to copying newer master |
| 97 | * backend files over older (or non-existing) clone backend files. |
| 98 | * Cases that cannot be handled will result in operation abortion. |
| 99 | * - replication : Set to 'async' to defer file operations on the non-master backends. |
| 100 | * This will apply such updates post-send for web requests. Note that |
| 101 | * any checks from "syncChecks" are still synchronous. |
| 102 | */ |
| 103 | public function __construct( array $config ) { |
| 104 | parent::__construct( $config ); |
| 105 | $this->syncChecks = $config['syncChecks'] ?? self::CHECK_SIZE; |
| 106 | $this->autoResync = $config['autoResync'] ?? false; |
| 107 | $this->asyncWrites = isset( $config['replication'] ) && $config['replication'] === 'async'; |
| 108 | // Construct backends here rather than via registration |
| 109 | // to keep these backends hidden from outside the proxy. |
| 110 | $namesUsed = []; |
| 111 | foreach ( $config['backends'] as $index => $beConfig ) { |
| 112 | $name = $beConfig['name']; |
| 113 | if ( isset( $namesUsed[$name] ) ) { // don't break FileOp predicates |
| 114 | throw new LogicException( "Two or more backends defined with the name $name." ); |
| 115 | } |
| 116 | $namesUsed[$name] = 1; |
| 117 | // Alter certain sub-backend settings |
| 118 | unset( $beConfig['readOnly'] ); // use proxy backend setting |
| 119 | unset( $beConfig['lockManager'] ); // lock under proxy backend |
| 120 | $beConfig['domainId'] = $this->domainId; // use the proxy backend wiki ID |
| 121 | $beConfig['logger'] = $this->logger; // use the proxy backend logger |
| 122 | if ( !empty( $beConfig['isMultiMaster'] ) ) { |
| 123 | if ( $this->masterIndex >= 0 ) { |
| 124 | throw new LogicException( 'More than one master backend defined.' ); |
| 125 | } |
| 126 | $this->masterIndex = $index; // this is the "master" |
| 127 | } |
| 128 | if ( !empty( $beConfig['readAffinity'] ) ) { |
| 129 | $this->readIndex = $index; // prefer this for reads |
| 130 | } |
| 131 | // Create sub-backend object |
| 132 | if ( !isset( $beConfig['class'] ) ) { |
| 133 | throw new InvalidArgumentException( 'No class given for a backend config.' ); |
| 134 | } |
| 135 | $class = $beConfig['class']; |
| 136 | $this->backends[$index] = new $class( $beConfig ); |
| 137 | } |
| 138 | if ( $this->masterIndex < 0 ) { // need backends and must have a master |
| 139 | throw new LogicException( 'No master backend defined.' ); |
| 140 | } |
| 141 | if ( $this->readIndex < 0 ) { |
| 142 | $this->readIndex = $this->masterIndex; // default |
| 143 | } |
| 144 | } |
| 145 | |
| 146 | final protected function doOperationsInternal( array $ops, array $opts ) { |
| 147 | $status = $this->newStatus(); |
| 148 | |
| 149 | $fname = __METHOD__; |
| 150 | $mbe = $this->backends[$this->masterIndex]; // convenience |
| 151 | |
| 152 | // Acquire any locks as needed |
| 153 | $scopeLock = null; |
| 154 | if ( empty( $opts['nonLocking'] ) ) { |
| 155 | $scopeLock = $this->getScopedLocksForOps( $ops, $status ); |
| 156 | if ( !$status->isOK() ) { |
| 157 | return $status; // abort |
| 158 | } |
| 159 | } |
| 160 | // Get the list of paths to read/write |
| 161 | $relevantPaths = $this->fileStoragePathsForOps( $ops ); |
| 162 | // Clear any cache entries (after locks acquired) |
| 163 | $this->clearCache( $relevantPaths ); |
| 164 | $opts['preserveCache'] = true; // only locked files are cached |
| 165 | // Check if the paths are valid and accessible on all backends |
| 166 | $status->merge( $this->accessibilityCheck( $relevantPaths ) ); |
| 167 | if ( !$status->isOK() ) { |
| 168 | return $status; // abort |
| 169 | } |
| 170 | // Do a consistency check to see if the backends are consistent |
| 171 | $syncStatus = $this->consistencyCheck( $relevantPaths ); |
| 172 | if ( !$syncStatus->isOK() ) { |
| 173 | $this->logger->error( |
| 174 | "$fname: failed sync check: " . implode( ', ', $relevantPaths ) |
| 175 | ); |
| 176 | // Try to resync the clone backends to the master on the spot |
| 177 | if ( |
| 178 | $this->autoResync === false || |
| 179 | !$this->resyncFiles( $relevantPaths, $this->autoResync )->isOK() |
| 180 | ) { |
| 181 | $status->merge( $syncStatus ); |
| 182 | |
| 183 | return $status; // abort |
| 184 | } |
| 185 | } |
| 186 | // Actually attempt the operation batch on the master backend |
| 187 | $realOps = $this->substOpBatchPaths( $ops, $mbe ); |
| 188 | $masterStatus = $mbe->doOperations( $realOps, $opts ); |
| 189 | $status->merge( $masterStatus ); |
| 190 | // Propagate the operations to the clone backends if there were no unexpected errors |
| 191 | // and everything didn't fail due to predicted errors. If $ops only had one operation, |
| 192 | // this might avoid backend sync inconsistencies. |
| 193 | if ( $masterStatus->isOK() && $masterStatus->successCount > 0 ) { |
| 194 | foreach ( $this->backends as $index => $backend ) { |
| 195 | if ( $index === $this->masterIndex ) { |
| 196 | continue; // done already |
| 197 | } |
| 198 | |
| 199 | $realOps = $this->substOpBatchPaths( $ops, $backend ); |
| 200 | if ( $this->asyncWrites && !$this->hasVolatileSources( $ops ) ) { |
| 201 | // Bind $scopeLock to the callback to preserve locks |
| 202 | $this->callNowOrLater( |
| 203 | function () use ( |
| 204 | $backend, $realOps, $opts, $scopeLock, $relevantPaths, $fname |
| 205 | ) { |
| 206 | $this->logger->debug( |
| 207 | "$fname: '{$backend->getName()}' async replication; paths: " . |
| 208 | implode( ', ', $relevantPaths ) |
| 209 | ); |
| 210 | $backend->doOperations( $realOps, $opts ); |
| 211 | } |
| 212 | ); |
| 213 | } else { |
| 214 | $this->logger->debug( |
| 215 | "$fname: '{$backend->getName()}' sync replication; paths: " . |
| 216 | implode( ', ', $relevantPaths ) |
| 217 | ); |
| 218 | $status->merge( $backend->doOperations( $realOps, $opts ) ); |
| 219 | } |
| 220 | } |
| 221 | } |
| 222 | // Make 'success', 'successCount', and 'failCount' fields reflect |
| 223 | // the overall operation, rather than all the batches for each backend. |
| 224 | // Do this by only using success values from the master backend's batch. |
| 225 | $status->success = $masterStatus->success; |
| 226 | $status->successCount = $masterStatus->successCount; |
| 227 | $status->failCount = $masterStatus->failCount; |
| 228 | |
| 229 | return $status; |
| 230 | } |
| 231 | |
| 232 | /** |
| 233 | * Check that a set of files are consistent across all internal backends |
| 234 | * |
| 235 | * This method should only be called if the files are locked or the backend |
| 236 | * is in read-only mode |
| 237 | * |
| 238 | * @param string[] $paths List of storage paths |
| 239 | * @return StatusValue |
| 240 | */ |
| 241 | public function consistencyCheck( array $paths ) { |
| 242 | $status = $this->newStatus(); |
| 243 | if ( $this->syncChecks == 0 || count( $this->backends ) <= 1 ) { |
| 244 | return $status; // skip checks |
| 245 | } |
| 246 | |
| 247 | // Preload all of the stat info in as few round trips as possible |
| 248 | foreach ( $this->backends as $backend ) { |
| 249 | $realPaths = $this->substPaths( $paths, $backend ); |
| 250 | $backend->preloadFileStat( [ 'srcs' => $realPaths, 'latest' => true ] ); |
| 251 | } |
| 252 | |
| 253 | foreach ( $paths as $path ) { |
| 254 | $params = [ 'src' => $path, 'latest' => true ]; |
| 255 | // Get the state of the file on the master backend |
| 256 | $masterBackend = $this->backends[$this->masterIndex]; |
| 257 | $masterParams = $this->substOpPaths( $params, $masterBackend ); |
| 258 | $masterStat = $masterBackend->getFileStat( $masterParams ); |
| 259 | if ( $masterStat === self::STAT_ERROR ) { |
| 260 | $status->fatal( 'backend-fail-stat', $path ); |
| 261 | continue; |
| 262 | } |
| 263 | if ( $this->syncChecks & self::CHECK_SHA1 ) { |
| 264 | $masterSha1 = $masterBackend->getFileSha1Base36( $masterParams ); |
| 265 | if ( ( $masterSha1 !== false ) !== (bool)$masterStat ) { |
| 266 | $status->fatal( 'backend-fail-hash', $path ); |
| 267 | continue; |
| 268 | } |
| 269 | } else { |
| 270 | $masterSha1 = null; // unused |
| 271 | } |
| 272 | |
| 273 | // Check if all clone backends agree with the master... |
| 274 | foreach ( $this->backends as $index => $cloneBackend ) { |
| 275 | if ( $index === $this->masterIndex ) { |
| 276 | continue; // master |
| 277 | } |
| 278 | |
| 279 | // Get the state of the file on the clone backend |
| 280 | $cloneParams = $this->substOpPaths( $params, $cloneBackend ); |
| 281 | $cloneStat = $cloneBackend->getFileStat( $cloneParams ); |
| 282 | |
| 283 | if ( $masterStat ) { |
| 284 | // File exists in the master backend |
| 285 | if ( !$cloneStat ) { |
| 286 | // File is missing from the clone backend |
| 287 | $status->fatal( 'backend-fail-synced', $path ); |
| 288 | } elseif ( |
| 289 | ( $this->syncChecks & self::CHECK_SIZE ) && |
| 290 | $cloneStat['size'] !== $masterStat['size'] |
| 291 | ) { |
| 292 | // File in the clone backend is different |
| 293 | $status->fatal( 'backend-fail-synced', $path ); |
| 294 | } elseif ( |
| 295 | ( $this->syncChecks & self::CHECK_TIME ) && |
| 296 | abs( |
| 297 | (int)ConvertibleTimestamp::convert( TS_UNIX, $masterStat['mtime'] ) - |
| 298 | (int)ConvertibleTimestamp::convert( TS_UNIX, $cloneStat['mtime'] ) |
| 299 | ) > 30 |
| 300 | ) { |
| 301 | // File in the clone backend is significantly newer or older |
| 302 | $status->fatal( 'backend-fail-synced', $path ); |
| 303 | } elseif ( |
| 304 | ( $this->syncChecks & self::CHECK_SHA1 ) && |
| 305 | $cloneBackend->getFileSha1Base36( $cloneParams ) !== $masterSha1 |
| 306 | ) { |
| 307 | // File in the clone backend is different |
| 308 | $status->fatal( 'backend-fail-synced', $path ); |
| 309 | } |
| 310 | } else { |
| 311 | // File does not exist in the master backend |
| 312 | if ( $cloneStat ) { |
| 313 | // Stray file exists in the clone backend |
| 314 | $status->fatal( 'backend-fail-synced', $path ); |
| 315 | } |
| 316 | } |
| 317 | } |
| 318 | } |
| 319 | |
| 320 | return $status; |
| 321 | } |
| 322 | |
| 323 | /** |
| 324 | * Check that a set of file paths are usable across all internal backends |
| 325 | * |
| 326 | * @param string[] $paths List of storage paths |
| 327 | * @return StatusValue |
| 328 | */ |
| 329 | public function accessibilityCheck( array $paths ) { |
| 330 | $status = $this->newStatus(); |
| 331 | if ( count( $this->backends ) <= 1 ) { |
| 332 | return $status; // skip checks |
| 333 | } |
| 334 | |
| 335 | foreach ( $paths as $path ) { |
| 336 | foreach ( $this->backends as $backend ) { |
| 337 | $realPath = $this->substPaths( $path, $backend ); |
| 338 | if ( !$backend->isPathUsableInternal( $realPath ) ) { |
| 339 | $status->fatal( 'backend-fail-usable', $path ); |
| 340 | } |
| 341 | } |
| 342 | } |
| 343 | |
| 344 | return $status; |
| 345 | } |
| 346 | |
| 347 | /** |
| 348 | * Check that a set of files are consistent across all internal backends |
| 349 | * and re-synchronize those files against the "multi master" if needed. |
| 350 | * |
| 351 | * This method should only be called if the files are locked |
| 352 | * |
| 353 | * @param string[] $paths List of storage paths |
| 354 | * @param string|bool $resyncMode False, True, or "conservative"; see __construct() |
| 355 | * @return StatusValue |
| 356 | */ |
| 357 | public function resyncFiles( array $paths, $resyncMode = true ) { |
| 358 | $status = $this->newStatus(); |
| 359 | |
| 360 | $fname = __METHOD__; |
| 361 | foreach ( $paths as $path ) { |
| 362 | $params = [ 'src' => $path, 'latest' => true ]; |
| 363 | // Get the state of the file on the master backend |
| 364 | $masterBackend = $this->backends[$this->masterIndex]; |
| 365 | $masterParams = $this->substOpPaths( $params, $masterBackend ); |
| 366 | $masterPath = $masterParams['src']; |
| 367 | $masterStat = $masterBackend->getFileStat( $masterParams ); |
| 368 | if ( $masterStat === self::STAT_ERROR ) { |
| 369 | $status->fatal( 'backend-fail-stat', $path ); |
| 370 | $this->logger->error( "$fname: file '$masterPath' is not available" ); |
| 371 | continue; |
| 372 | } |
| 373 | $masterSha1 = $masterBackend->getFileSha1Base36( $masterParams ); |
| 374 | if ( ( $masterSha1 !== false ) !== (bool)$masterStat ) { |
| 375 | $status->fatal( 'backend-fail-hash', $path ); |
| 376 | $this->logger->error( "$fname: file '$masterPath' hash does not match stat" ); |
| 377 | continue; |
| 378 | } |
| 379 | |
| 380 | // Check of all clone backends agree with the master... |
| 381 | foreach ( $this->backends as $index => $cloneBackend ) { |
| 382 | if ( $index === $this->masterIndex ) { |
| 383 | continue; // master |
| 384 | } |
| 385 | |
| 386 | // Get the state of the file on the clone backend |
| 387 | $cloneParams = $this->substOpPaths( $params, $cloneBackend ); |
| 388 | $clonePath = $cloneParams['src']; |
| 389 | $cloneStat = $cloneBackend->getFileStat( $cloneParams ); |
| 390 | if ( $cloneStat === self::STAT_ERROR ) { |
| 391 | $status->fatal( 'backend-fail-stat', $path ); |
| 392 | $this->logger->error( "$fname: file '$clonePath' is not available" ); |
| 393 | continue; |
| 394 | } |
| 395 | $cloneSha1 = $cloneBackend->getFileSha1Base36( $cloneParams ); |
| 396 | if ( ( $cloneSha1 !== false ) !== (bool)$cloneStat ) { |
| 397 | $status->fatal( 'backend-fail-hash', $path ); |
| 398 | $this->logger->error( "$fname: file '$clonePath' hash does not match stat" ); |
| 399 | continue; |
| 400 | } |
| 401 | |
| 402 | if ( $masterSha1 === $cloneSha1 ) { |
| 403 | // File is either the same in both backends or absent from both backends |
| 404 | $this->logger->debug( "$fname: file '$clonePath' matches '$masterPath'" ); |
| 405 | } elseif ( $masterSha1 !== false ) { |
| 406 | // File is either missing from or different in the clone backend |
| 407 | if ( |
| 408 | $resyncMode === 'conservative' && |
| 409 | $cloneStat && |
| 410 | // @phan-suppress-next-line PhanTypeArraySuspiciousNullable |
| 411 | $cloneStat['mtime'] > $masterStat['mtime'] |
| 412 | ) { |
| 413 | // Do not replace files with older ones; reduces the risk of data loss |
| 414 | $status->fatal( 'backend-fail-synced', $path ); |
| 415 | } else { |
| 416 | // Copy the master backend file to the clone backend in overwrite mode |
| 417 | $fsFile = $masterBackend->getLocalReference( $masterParams ); |
| 418 | $status->merge( $cloneBackend->quickStore( [ |
| 419 | 'src' => $fsFile, |
| 420 | 'dst' => $clonePath |
| 421 | ] ) ); |
| 422 | } |
| 423 | } elseif ( $masterStat === false ) { |
| 424 | // Stray file exists in the clone backend |
| 425 | if ( $resyncMode === 'conservative' ) { |
| 426 | // Do not delete stray files; reduces the risk of data loss |
| 427 | $status->fatal( 'backend-fail-synced', $path ); |
| 428 | $this->logger->error( "$fname: not allowed to delete file '$clonePath'" ); |
| 429 | } else { |
| 430 | // Delete the stay file from the clone backend |
| 431 | $status->merge( $cloneBackend->quickDelete( [ 'src' => $clonePath ] ) ); |
| 432 | } |
| 433 | } |
| 434 | } |
| 435 | } |
| 436 | |
| 437 | if ( !$status->isOK() ) { |
| 438 | $this->logger->error( "$fname: failed to resync: " . implode( ', ', $paths ) ); |
| 439 | } |
| 440 | |
| 441 | return $status; |
| 442 | } |
| 443 | |
| 444 | /** |
| 445 | * Get a list of file storage paths to read or write for a list of operations |
| 446 | * |
| 447 | * @param array $ops Same format as doOperations() |
| 448 | * @return array List of storage paths to files (does not include directories) |
| 449 | */ |
| 450 | protected function fileStoragePathsForOps( array $ops ) { |
| 451 | $paths = []; |
| 452 | foreach ( $ops as $op ) { |
| 453 | if ( isset( $op['src'] ) ) { |
| 454 | // For things like copy/move/delete with "ignoreMissingSource" and there |
| 455 | // is no source file, nothing should happen and there should be no errors. |
| 456 | if ( empty( $op['ignoreMissingSource'] ) |
| 457 | || $this->fileExists( [ 'src' => $op['src'] ] ) |
| 458 | ) { |
| 459 | $paths[] = $op['src']; |
| 460 | } |
| 461 | } |
| 462 | if ( isset( $op['srcs'] ) ) { |
| 463 | $paths = array_merge( $paths, $op['srcs'] ); |
| 464 | } |
| 465 | if ( isset( $op['dst'] ) ) { |
| 466 | $paths[] = $op['dst']; |
| 467 | } |
| 468 | } |
| 469 | |
| 470 | return array_values( array_unique( array_filter( $paths, [ FileBackend::class, 'isStoragePath' ] ) ) ); |
| 471 | } |
| 472 | |
| 473 | /** |
| 474 | * Substitute the backend name in storage path parameters |
| 475 | * for a set of operations with that of a given internal backend. |
| 476 | * |
| 477 | * @param array $ops List of file operation arrays |
| 478 | * @param FileBackendStore $backend |
| 479 | * @return array |
| 480 | */ |
| 481 | protected function substOpBatchPaths( array $ops, FileBackendStore $backend ) { |
| 482 | $newOps = []; // operations |
| 483 | foreach ( $ops as $op ) { |
| 484 | $newOp = $op; // operation |
| 485 | foreach ( [ 'src', 'srcs', 'dst', 'dir' ] as $par ) { |
| 486 | if ( isset( $newOp[$par] ) ) { // string or array |
| 487 | $newOp[$par] = $this->substPaths( $newOp[$par], $backend ); |
| 488 | } |
| 489 | } |
| 490 | $newOps[] = $newOp; |
| 491 | } |
| 492 | |
| 493 | return $newOps; |
| 494 | } |
| 495 | |
| 496 | /** |
| 497 | * Same as substOpBatchPaths() but for a single operation |
| 498 | * |
| 499 | * @param array $ops File operation array |
| 500 | * @param FileBackendStore $backend |
| 501 | * @return array |
| 502 | */ |
| 503 | protected function substOpPaths( array $ops, FileBackendStore $backend ) { |
| 504 | $newOps = $this->substOpBatchPaths( [ $ops ], $backend ); |
| 505 | |
| 506 | return $newOps[0]; |
| 507 | } |
| 508 | |
| 509 | /** |
| 510 | * Substitute the backend of storage paths with an internal backend's name |
| 511 | * |
| 512 | * @param string[]|string $paths List of paths or single string path |
| 513 | * @param FileBackendStore $backend |
| 514 | * @return string[]|string |
| 515 | */ |
| 516 | protected function substPaths( $paths, FileBackendStore $backend ) { |
| 517 | return preg_replace( |
| 518 | '!^mwstore://' . preg_quote( $this->name, '!' ) . '/!', |
| 519 | StringUtils::escapeRegexReplacement( "mwstore://{$backend->getName()}/" ), |
| 520 | $paths // string or array |
| 521 | ); |
| 522 | } |
| 523 | |
| 524 | /** |
| 525 | * Substitute the backend of internal storage paths with the proxy backend's name |
| 526 | * |
| 527 | * @param string[]|string $paths List of paths or single string path |
| 528 | * @param FileBackendStore $backend internal storage backend |
| 529 | * @return string[]|string |
| 530 | */ |
| 531 | protected function unsubstPaths( $paths, FileBackendStore $backend ) { |
| 532 | return preg_replace( |
| 533 | '!^mwstore://' . preg_quote( $backend->getName(), '!' ) . '/!', |
| 534 | StringUtils::escapeRegexReplacement( "mwstore://{$this->name}/" ), |
| 535 | $paths // string or array |
| 536 | ); |
| 537 | } |
| 538 | |
| 539 | /** |
| 540 | * @param array[] $ops File operations for FileBackend::doOperations() |
| 541 | * @return bool Whether there are file path sources with outside lifetime/ownership |
| 542 | */ |
| 543 | protected function hasVolatileSources( array $ops ) { |
| 544 | foreach ( $ops as $op ) { |
| 545 | if ( $op['op'] === 'store' && !isset( $op['srcRef'] ) ) { |
| 546 | return true; // source file might be deleted anytime after do*Operations() |
| 547 | } |
| 548 | } |
| 549 | |
| 550 | return false; |
| 551 | } |
| 552 | |
| 553 | protected function doQuickOperationsInternal( array $ops, array $opts ) { |
| 554 | $status = $this->newStatus(); |
| 555 | // Do the operations on the master backend; setting StatusValue fields |
| 556 | $realOps = $this->substOpBatchPaths( $ops, $this->backends[$this->masterIndex] ); |
| 557 | $masterStatus = $this->backends[$this->masterIndex]->doQuickOperations( $realOps ); |
| 558 | $status->merge( $masterStatus ); |
| 559 | // Propagate the operations to the clone backends... |
| 560 | foreach ( $this->backends as $index => $backend ) { |
| 561 | if ( $index === $this->masterIndex ) { |
| 562 | continue; // done already |
| 563 | } |
| 564 | |
| 565 | $realOps = $this->substOpBatchPaths( $ops, $backend ); |
| 566 | if ( $this->asyncWrites && !$this->hasVolatileSources( $ops ) ) { |
| 567 | $this->callNowOrLater( |
| 568 | static function () use ( $backend, $realOps ) { |
| 569 | $backend->doQuickOperations( $realOps ); |
| 570 | } |
| 571 | ); |
| 572 | } else { |
| 573 | $status->merge( $backend->doQuickOperations( $realOps ) ); |
| 574 | } |
| 575 | } |
| 576 | // Make 'success', 'successCount', and 'failCount' fields reflect |
| 577 | // the overall operation, rather than all the batches for each backend. |
| 578 | // Do this by only using success values from the master backend's batch. |
| 579 | $status->success = $masterStatus->success; |
| 580 | $status->successCount = $masterStatus->successCount; |
| 581 | $status->failCount = $masterStatus->failCount; |
| 582 | |
| 583 | return $status; |
| 584 | } |
| 585 | |
| 586 | protected function doPrepare( array $params ) { |
| 587 | return $this->doDirectoryOp( 'prepare', $params ); |
| 588 | } |
| 589 | |
| 590 | protected function doSecure( array $params ) { |
| 591 | return $this->doDirectoryOp( 'secure', $params ); |
| 592 | } |
| 593 | |
| 594 | protected function doPublish( array $params ) { |
| 595 | return $this->doDirectoryOp( 'publish', $params ); |
| 596 | } |
| 597 | |
| 598 | protected function doClean( array $params ) { |
| 599 | return $this->doDirectoryOp( 'clean', $params ); |
| 600 | } |
| 601 | |
| 602 | /** |
| 603 | * @param string $method One of (doPrepare,doSecure,doPublish,doClean) |
| 604 | * @param array $params Method arguments |
| 605 | * @return StatusValue |
| 606 | */ |
| 607 | protected function doDirectoryOp( $method, array $params ) { |
| 608 | $status = $this->newStatus(); |
| 609 | |
| 610 | $realParams = $this->substOpPaths( $params, $this->backends[$this->masterIndex] ); |
| 611 | $masterStatus = $this->backends[$this->masterIndex]->$method( $realParams ); |
| 612 | $status->merge( $masterStatus ); |
| 613 | |
| 614 | foreach ( $this->backends as $index => $backend ) { |
| 615 | if ( $index === $this->masterIndex ) { |
| 616 | continue; // already done |
| 617 | } |
| 618 | |
| 619 | $realParams = $this->substOpPaths( $params, $backend ); |
| 620 | if ( $this->asyncWrites ) { |
| 621 | $this->callNowOrLater( |
| 622 | static function () use ( $backend, $method, $realParams ) { |
| 623 | $backend->$method( $realParams ); |
| 624 | } |
| 625 | ); |
| 626 | } else { |
| 627 | $status->merge( $backend->$method( $realParams ) ); |
| 628 | } |
| 629 | } |
| 630 | |
| 631 | return $status; |
| 632 | } |
| 633 | |
| 634 | public function concatenate( array $params ) { |
| 635 | $status = $this->newStatus(); |
| 636 | // We are writing to an FS file, so we don't need to do this per-backend |
| 637 | $index = $this->getReadIndexFromParams( $params ); |
| 638 | $realParams = $this->substOpPaths( $params, $this->backends[$index] ); |
| 639 | |
| 640 | $status->merge( $this->backends[$index]->concatenate( $realParams ) ); |
| 641 | |
| 642 | return $status; |
| 643 | } |
| 644 | |
| 645 | public function fileExists( array $params ) { |
| 646 | $index = $this->getReadIndexFromParams( $params ); |
| 647 | $realParams = $this->substOpPaths( $params, $this->backends[$index] ); |
| 648 | |
| 649 | return $this->backends[$index]->fileExists( $realParams ); |
| 650 | } |
| 651 | |
| 652 | public function getFileTimestamp( array $params ) { |
| 653 | $index = $this->getReadIndexFromParams( $params ); |
| 654 | $realParams = $this->substOpPaths( $params, $this->backends[$index] ); |
| 655 | |
| 656 | return $this->backends[$index]->getFileTimestamp( $realParams ); |
| 657 | } |
| 658 | |
| 659 | public function getFileSize( array $params ) { |
| 660 | $index = $this->getReadIndexFromParams( $params ); |
| 661 | $realParams = $this->substOpPaths( $params, $this->backends[$index] ); |
| 662 | |
| 663 | return $this->backends[$index]->getFileSize( $realParams ); |
| 664 | } |
| 665 | |
| 666 | public function getFileStat( array $params ) { |
| 667 | $index = $this->getReadIndexFromParams( $params ); |
| 668 | $realParams = $this->substOpPaths( $params, $this->backends[$index] ); |
| 669 | |
| 670 | return $this->backends[$index]->getFileStat( $realParams ); |
| 671 | } |
| 672 | |
| 673 | public function getFileXAttributes( array $params ) { |
| 674 | $index = $this->getReadIndexFromParams( $params ); |
| 675 | $realParams = $this->substOpPaths( $params, $this->backends[$index] ); |
| 676 | |
| 677 | return $this->backends[$index]->getFileXAttributes( $realParams ); |
| 678 | } |
| 679 | |
| 680 | public function getFileContentsMulti( array $params ) { |
| 681 | $index = $this->getReadIndexFromParams( $params ); |
| 682 | $realParams = $this->substOpPaths( $params, $this->backends[$index] ); |
| 683 | |
| 684 | $contentsM = $this->backends[$index]->getFileContentsMulti( $realParams ); |
| 685 | |
| 686 | $contents = []; // (path => FSFile) mapping using the proxy backend's name |
| 687 | foreach ( $contentsM as $path => $data ) { |
| 688 | $contents[$this->unsubstPaths( $path, $this->backends[$index] )] = $data; |
| 689 | } |
| 690 | |
| 691 | return $contents; |
| 692 | } |
| 693 | |
| 694 | public function getFileSha1Base36( array $params ) { |
| 695 | $index = $this->getReadIndexFromParams( $params ); |
| 696 | $realParams = $this->substOpPaths( $params, $this->backends[$index] ); |
| 697 | |
| 698 | return $this->backends[$index]->getFileSha1Base36( $realParams ); |
| 699 | } |
| 700 | |
| 701 | public function getFileProps( array $params ) { |
| 702 | $index = $this->getReadIndexFromParams( $params ); |
| 703 | $realParams = $this->substOpPaths( $params, $this->backends[$index] ); |
| 704 | |
| 705 | return $this->backends[$index]->getFileProps( $realParams ); |
| 706 | } |
| 707 | |
| 708 | public function streamFile( array $params ) { |
| 709 | $index = $this->getReadIndexFromParams( $params ); |
| 710 | $realParams = $this->substOpPaths( $params, $this->backends[$index] ); |
| 711 | |
| 712 | return $this->backends[$index]->streamFile( $realParams ); |
| 713 | } |
| 714 | |
| 715 | public function getLocalReferenceMulti( array $params ) { |
| 716 | $index = $this->getReadIndexFromParams( $params ); |
| 717 | $realParams = $this->substOpPaths( $params, $this->backends[$index] ); |
| 718 | |
| 719 | $fsFilesM = $this->backends[$index]->getLocalReferenceMulti( $realParams ); |
| 720 | |
| 721 | $fsFiles = []; // (path => FSFile) mapping using the proxy backend's name |
| 722 | foreach ( $fsFilesM as $path => $fsFile ) { |
| 723 | $fsFiles[$this->unsubstPaths( $path, $this->backends[$index] )] = $fsFile; |
| 724 | } |
| 725 | |
| 726 | return $fsFiles; |
| 727 | } |
| 728 | |
| 729 | public function getLocalCopyMulti( array $params ) { |
| 730 | $index = $this->getReadIndexFromParams( $params ); |
| 731 | $realParams = $this->substOpPaths( $params, $this->backends[$index] ); |
| 732 | |
| 733 | $tempFilesM = $this->backends[$index]->getLocalCopyMulti( $realParams ); |
| 734 | |
| 735 | $tempFiles = []; // (path => TempFSFile) mapping using the proxy backend's name |
| 736 | foreach ( $tempFilesM as $path => $tempFile ) { |
| 737 | $tempFiles[$this->unsubstPaths( $path, $this->backends[$index] )] = $tempFile; |
| 738 | } |
| 739 | |
| 740 | return $tempFiles; |
| 741 | } |
| 742 | |
| 743 | public function getFileHttpUrl( array $params ) { |
| 744 | $index = $this->getReadIndexFromParams( $params ); |
| 745 | $realParams = $this->substOpPaths( $params, $this->backends[$index] ); |
| 746 | |
| 747 | return $this->backends[$index]->getFileHttpUrl( $realParams ); |
| 748 | } |
| 749 | |
| 750 | public function addShellboxInputFile( BoxedCommand $command, string $boxedName, |
| 751 | array $params |
| 752 | ) { |
| 753 | $index = $this->getReadIndexFromParams( $params ); |
| 754 | $realParams = $this->substOpPaths( $params, $this->backends[$index] ); |
| 755 | return $this->backends[$index]->addShellboxInputFile( $command, $boxedName, $realParams ); |
| 756 | } |
| 757 | |
| 758 | public function directoryExists( array $params ) { |
| 759 | $realParams = $this->substOpPaths( $params, $this->backends[$this->masterIndex] ); |
| 760 | |
| 761 | return $this->backends[$this->masterIndex]->directoryExists( $realParams ); |
| 762 | } |
| 763 | |
| 764 | public function getDirectoryList( array $params ) { |
| 765 | $realParams = $this->substOpPaths( $params, $this->backends[$this->masterIndex] ); |
| 766 | |
| 767 | return $this->backends[$this->masterIndex]->getDirectoryList( $realParams ); |
| 768 | } |
| 769 | |
| 770 | public function getFileList( array $params ) { |
| 771 | if ( isset( $params['forWrite'] ) && $params['forWrite'] ) { |
| 772 | return $this->getFileListForWrite( $params ); |
| 773 | } |
| 774 | |
| 775 | $realParams = $this->substOpPaths( $params, $this->backends[$this->masterIndex] ); |
| 776 | return $this->backends[$this->masterIndex]->getFileList( $realParams ); |
| 777 | } |
| 778 | |
| 779 | private function getFileListForWrite( $params ) { |
| 780 | $files = []; |
| 781 | // Get the list of thumbnails from all backends to allow |
| 782 | // deleting all of them. Otherwise, old thumbnails existing on |
| 783 | // one backend only won't get updated in reupload (T331138). |
| 784 | foreach ( $this->backends as $backend ) { |
| 785 | $realParams = $this->substOpPaths( $params, $backend ); |
| 786 | $iterator = $backend->getFileList( $realParams ); |
| 787 | if ( $iterator !== null ) { |
| 788 | foreach ( $iterator as $file ) { |
| 789 | $files[] = $file; |
| 790 | } |
| 791 | } |
| 792 | } |
| 793 | |
| 794 | return array_unique( $files ); |
| 795 | } |
| 796 | |
| 797 | public function getFeatures() { |
| 798 | return $this->backends[$this->masterIndex]->getFeatures(); |
| 799 | } |
| 800 | |
| 801 | public function clearCache( ?array $paths = null ) { |
| 802 | foreach ( $this->backends as $backend ) { |
| 803 | $realPaths = is_array( $paths ) ? $this->substPaths( $paths, $backend ) : null; |
| 804 | $backend->clearCache( $realPaths ); |
| 805 | } |
| 806 | } |
| 807 | |
| 808 | public function preloadCache( array $paths ) { |
| 809 | $realPaths = $this->substPaths( $paths, $this->backends[$this->readIndex] ); |
| 810 | $this->backends[$this->readIndex]->preloadCache( $realPaths ); |
| 811 | } |
| 812 | |
| 813 | public function preloadFileStat( array $params ) { |
| 814 | $index = $this->getReadIndexFromParams( $params ); |
| 815 | $realParams = $this->substOpPaths( $params, $this->backends[$index] ); |
| 816 | |
| 817 | return $this->backends[$index]->preloadFileStat( $realParams ); |
| 818 | } |
| 819 | |
| 820 | public function getScopedLocksForOps( array $ops, StatusValue $status ) { |
| 821 | $realOps = $this->substOpBatchPaths( $ops, $this->backends[$this->masterIndex] ); |
| 822 | $fileOps = $this->backends[$this->masterIndex]->getOperationsInternal( $realOps ); |
| 823 | // Get the paths to lock from the master backend |
| 824 | $paths = $this->backends[$this->masterIndex]->getPathsToLockForOpsInternal( $fileOps ); |
| 825 | // Get the paths under the proxy backend's name |
| 826 | $pbPaths = [ |
| 827 | LockManager::LOCK_UW => $this->unsubstPaths( |
| 828 | $paths[LockManager::LOCK_UW], |
| 829 | $this->backends[$this->masterIndex] |
| 830 | ), |
| 831 | LockManager::LOCK_EX => $this->unsubstPaths( |
| 832 | $paths[LockManager::LOCK_EX], |
| 833 | $this->backends[$this->masterIndex] |
| 834 | ) |
| 835 | ]; |
| 836 | |
| 837 | // Actually acquire the locks |
| 838 | return $this->getScopedFileLocks( $pbPaths, 'mixed', $status ); |
| 839 | } |
| 840 | |
| 841 | /** |
| 842 | * @param array $params |
| 843 | * @return int The master or read affinity backend index, based on $params['latest'] |
| 844 | */ |
| 845 | protected function getReadIndexFromParams( array $params ) { |
| 846 | return !empty( $params['latest'] ) ? $this->masterIndex : $this->readIndex; |
| 847 | } |
| 848 | } |
| 849 | |
| 850 | /** @deprecated class alias since 1.43 */ |
| 851 | class_alias( FileBackendMultiWrite::class, 'FileBackendMultiWrite' ); |