*/
/**
- * Class for an OpenStack Swift based file backend.
+ * @brief Class for an OpenStack Swift based file backend.
*
* This requires the SwiftCloudFiles MediaWiki extension, which includes
* the php-cloudfiles library (https://github.com/rackspace/php-cloudfiles).
* php-cloudfiles requires the curl, fileinfo, and mb_string PHP extensions.
*
- * Status messages should avoid mentioning the Swift account name
+ * Status messages should avoid mentioning the Swift account name.
* Likewise, error suppression should be used to avoid path disclosure.
*
* @ingroup FileBackend
* @since 1.19
*/
-class SwiftFileBackend extends FileBackend {
+class SwiftFileBackend extends FileBackendStore {
/** @var CF_Authentication */
protected $auth; // Swift authentication handler
protected $authTTL; // integer seconds
protected $swiftAnonUser; // string; username to handle unauthenticated requests
- protected $maxContCacheSize = 20; // integer; max containers with entries
+ protected $maxContCacheSize = 300; // integer; max containers with entries
/** @var CF_Connection */
protected $conn; // Swift connection handle
protected $connContainers = array(); // container object cache
/**
- * @see FileBackend::__construct()
+ * @see FileBackendStore::__construct()
* Additional $config params include:
* swiftAuthUrl : Swift authentication server URL
* swiftUser : Swift user used by MediaWiki (account:username)
* swiftKey : Swift authentication key for the above user
* swiftAuthTTL : Swift authentication TTL (seconds)
* swiftAnonUser : Swift user used for end-user requests (account:username)
- * shardViaHashLevels : Map of container names to the number of hash levels
+ * shardViaHashLevels : Map of container names to sharding config with:
+ * 'base' : base of hash characters, 16 or 36
+ * 'levels' : the number of hash levels (and digits)
+ * 'repeat' : hash subdirectories are prefixed with all the
+ * parent hash directory names (e.g. "a/ab/abc")
*/
public function __construct( array $config ) {
parent::__construct( $config );
// Required settings
$this->auth = new CF_Authentication(
- $config['swiftUser'],
- $config['swiftKey'],
+ $config['swiftUser'],
+ $config['swiftKey'],
null, // account; unused
- $config['swiftAuthUrl']
+ $config['swiftAuthUrl']
);
// Optional settings
$this->authTTL = isset( $config['swiftAuthTTL'] )
? $config['swiftAuthTTL']
- : 120; // some sane number
+ : 5 * 60; // some sane number
$this->swiftAnonUser = isset( $config['swiftAnonUser'] )
? $config['swiftAnonUser']
: '';
$this->shardViaHashLevels = isset( $config['shardViaHashLevels'] )
? $config['shardViaHashLevels']
: '';
+ // Cache container info to mask latency
+ $this->memCache = wfGetMainCache();
}
/**
- * @see FileBackend::resolveContainerPath()
+ * @see FileBackendStore::resolveContainerPath()
+ * @return null
*/
protected function resolveContainerPath( $container, $relStoragePath ) {
if ( strlen( urlencode( $relStoragePath ) ) > 1024 ) {
}
/**
- * @see FileBackend::isPathUsableInternal()
+ * @see FileBackendStore::isPathUsableInternal()
+ * @return bool
*/
public function isPathUsableInternal( $storagePath ) {
list( $container, $rel ) = $this->resolveStoragePathReal( $storagePath );
}
/**
- * @see FileBackend::doCopyInternal()
+ * @see FileBackendStore::doCreateInternal()
+ * @return Status
*/
protected function doCreateInternal( array $params ) {
$status = Status::newGood();
try {
$dContObj = $this->getContainer( $dstCont );
if ( empty( $params['overwrite'] ) &&
- $this->fileExists( array( 'src' => $params['dst'], 'latest' => 1 ) ) )
+ $this->fileExists( array( 'src' => $params['dst'], 'latest' => 1 ) ) )
{
$status->fatal( 'backend-fail-alreadyexists', $params['dst'] );
return $status;
}
/**
- * @see FileBackend::doStoreInternal()
+ * @see FileBackendStore::doStoreInternal()
+ * @return Status
*/
protected function doStoreInternal( array $params ) {
$status = Status::newGood();
try {
$dContObj = $this->getContainer( $dstCont );
if ( empty( $params['overwrite'] ) &&
- $this->fileExists( array( 'src' => $params['dst'], 'latest' => 1 ) ) )
+ $this->fileExists( array( 'src' => $params['dst'], 'latest' => 1 ) ) )
{
$status->fatal( 'backend-fail-alreadyexists', $params['dst'] );
return $status;
}
/**
- * @see FileBackend::doCopyInternal()
+ * @see FileBackendStore::doCopyInternal()
+ * @return Status
*/
protected function doCopyInternal( array $params ) {
$status = Status::newGood();
$sContObj = $this->getContainer( $srcCont );
$dContObj = $this->getContainer( $dstCont );
if ( empty( $params['overwrite'] ) &&
- $this->fileExists( array( 'src' => $params['dst'], 'latest' => 1 ) ) )
+ $this->fileExists( array( 'src' => $params['dst'], 'latest' => 1 ) ) )
{
$status->fatal( 'backend-fail-alreadyexists', $params['dst'] );
return $status;
}
/**
- * @see FileBackend::doDeleteInternal()
+ * @see FileBackendStore::doDeleteInternal()
+ * @return Status
*/
protected function doDeleteInternal( array $params ) {
$status = Status::newGood();
}
/**
- * @see FileBackend::doPrepareInternal()
+ * @see FileBackendStore::doPrepareInternal()
+ * @return Status
*/
protected function doPrepareInternal( $fullCont, $dir, array $params ) {
$status = Status::newGood();
}
/**
- * @see FileBackend::doSecureInternal()
+ * @see FileBackendStore::doSecureInternal()
+ * @return Status
*/
protected function doSecureInternal( $fullCont, $dir, array $params ) {
$status = Status::newGood();
}
/**
- * @see FileBackend::doCleanInternal()
+ * @see FileBackendStore::doCleanInternal()
+ * @return Status
*/
protected function doCleanInternal( $fullCont, $dir, array $params ) {
$status = Status::newGood();
}
/**
- * @see FileBackend::doFileExists()
+ * @see FileBackendStore::doFileExists()
+ * @return array|bool|null
*/
protected function doGetFileStat( array $params ) {
list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
try {
$contObj = $this->getContainer( $srcCont );
$srcObj = $contObj->get_object( $srcRel, $this->headersFromParams( $params ) );
+ $this->addMissingMetadata( $srcObj, $params['src'] );
$stat = array(
// Convert dates like "Tue, 03 Jan 2012 22:01:04 GMT" to TS_MW
'mtime' => wfTimestamp( TS_MW, $srcObj->last_modified ),
}
/**
- * @see FileBackendBase::getFileContents()
+ * Fill in any missing object metadata and save it to Swift
+ *
+ * @param $obj CF_Object
+ * @param $path string Storage path to object
+ * @return bool Success
+ * @throws Exception cloudfiles exceptions
+ */
+ protected function addMissingMetadata( CF_Object $obj, $path ) {
+ if ( isset( $obj->metadata['Sha1base36'] ) ) {
+ return true; // nothing to do
+ }
+ $status = Status::newGood();
+ $scopeLockS = $this->getScopedFileLocks( array( $path ), LockManager::LOCK_UW, $status );
+ if ( $status->isOK() ) {
+ $tmpFile = $this->getLocalCopy( array( 'src' => $path, 'latest' => 1 ) );
+ if ( $tmpFile ) {
+ $hash = $tmpFile->getSha1Base36();
+ if ( $hash !== false ) {
+ $obj->metadata['Sha1base36'] = $hash;
+ $obj->sync_metadata(); // save to Swift
+ return true; // success
+ }
+ }
+ }
+ $obj->metadata['Sha1base36'] = false;
+ return false; // failed
+ }
+
+ /**
+ * @see FileBackend::getFileContents()
+ * @return bool|null|string
*/
public function getFileContents( array $params ) {
list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
}
/**
- * @see FileBackend::getFileListInternal()
+ * @see FileBackendStore::doDirectoryExists()
+ * @return bool|null
+ */
+ protected function doDirectoryExists( $fullCont, $dir, array $params ) {
+ try {
+ $container = $this->getContainer( $fullCont );
+ $prefix = ( $dir == '' ) ? null : "{$dir}/";
+ return ( count( $container->list_objects( 1, null, $prefix ) ) > 0 );
+ } catch ( NoSuchContainerException $e ) {
+ return false;
+ } catch ( InvalidResponseException $e ) {
+ } catch ( Exception $e ) { // some other exception?
+ $this->logException( $e, __METHOD__, array( 'cont' => $fullCont, 'dir' => $dir ) );
+ }
+
+ return null; // error
+ }
+
+ /**
+ * @see FileBackendStore::getDirectoryListInternal()
+ * @return SwiftFileBackendDirList
+ */
+ public function getDirectoryListInternal( $fullCont, $dir, array $params ) {
+ return new SwiftFileBackendDirList( $this, $fullCont, $dir, $params );
+ }
+
+ /**
+ * @see FileBackendStore::getFileListInternal()
+ * @return SwiftFileBackendFileList
*/
public function getFileListInternal( $fullCont, $dir, array $params ) {
- return new SwiftFileBackendFileList( $this, $fullCont, $dir );
+ return new SwiftFileBackendFileList( $this, $fullCont, $dir, $params );
}
/**
* Do not call this function outside of SwiftFileBackendFileList
- *
+ *
* @param $fullCont string Resolved container name
* @param $dir string Resolved storage directory with no trailing slash
- * @param $after string Storage path of file to list items after
+ * @param $after string|null Storage path of file to list items after
* @param $limit integer Max number of items to list
- * @return Array
+ * @param $params Array Includes flag for 'topOnly'
+ * @return Array List of relative paths of dirs directly under $dir
+ */
+ public function getDirListPageInternal( $fullCont, $dir, &$after, $limit, array $params ) {
+ $dirs = array();
+
+ try {
+ $container = $this->getContainer( $fullCont );
+ $prefix = ( $dir == '' ) ? null : "{$dir}/";
+ // Non-recursive: only list dirs right under $dir
+ if ( !empty( $params['topOnly'] ) ) {
+ $objects = $container->list_objects( $limit, $after, $prefix, null, '/' );
+ foreach ( $objects as $object ) { // files and dirs
+ if ( substr( $object, -1 ) === '/' ) {
+ $dirs[] = $object; // directories end in '/'
+ }
+ $after = $object; // update last item
+ }
+ // Recursive: list all dirs under $dir and its subdirs
+ } else {
+ // Get directory from last item of prior page
+ $lastDir = $this->getParentDir( $after ); // must be first page
+ $objects = $container->list_objects( $limit, $after, $prefix );
+ foreach ( $objects as $object ) { // files
+ $objectDir = $this->getParentDir( $object ); // directory of object
+ if ( $objectDir !== false ) { // file has a parent dir
+ // Swift stores paths in UTF-8, using binary sorting.
+ // See function "create_container_table" in common/db.py.
+ // If a directory is not "greater" than the last one,
+ // then it was already listed by the calling iterator.
+ if ( $objectDir > $lastDir ) {
+ $pDir = $objectDir;
+ do { // add dir and all its parent dirs
+ $dirs[] = "{$pDir}/";
+ $pDir = $this->getParentDir( $pDir );
+ } while ( $pDir !== false // sanity
+ && $pDir > $lastDir // not done already
+ && strlen( $pDir ) > strlen( $dir ) // within $dir
+ );
+ }
+ $lastDir = $objectDir;
+ }
+ $after = $object; // update last item
+ }
+ }
+ } catch ( NoSuchContainerException $e ) {
+ } catch ( InvalidResponseException $e ) {
+ } catch ( Exception $e ) { // some other exception?
+ $this->logException( $e, __METHOD__, array( 'cont' => $fullCont, 'dir' => $dir ) );
+ }
+
+ return $dirs;
+ }
+
+ protected function getParentDir( $path ) {
+ return ( strpos( $path, '/' ) !== false ) ? dirname( $path ) : false;
+ }
+
+ /**
+ * Do not call this function outside of SwiftFileBackendFileList
+ *
+ * @param $fullCont string Resolved container name
+ * @param $dir string Resolved storage directory with no trailing slash
+ * @param $after string|null Storage path of file to list items after
+ * @param $limit integer Max number of items to list
+ * @param $params Array Includes flag for 'topOnly'
+ * @return Array List of relative paths of files under $dir
*/
- public function getFileListPageInternal( $fullCont, $dir, $after, $limit ) {
+ public function getFileListPageInternal( $fullCont, $dir, &$after, $limit, array $params ) {
$files = array();
try {
$container = $this->getContainer( $fullCont );
$prefix = ( $dir == '' ) ? null : "{$dir}/";
- $files = $container->list_objects( $limit, $after, $prefix );
+ // Non-recursive: only list files right under $dir
+ if ( !empty( $params['topOnly'] ) ) { // files and dirs
+ $objects = $container->list_objects( $limit, $after, $prefix, null, '/' );
+ foreach ( $objects as $object ) {
+ if ( substr( $object, -1 ) !== '/' ) {
+ $files[] = $object; // directories end in '/'
+ }
+ }
+ // Recursive: list all files under $dir and its subdirs
+ } else { // files
+ $files = $container->list_objects( $limit, $after, $prefix );
+ }
+ $after = end( $files ); // update last item
+ reset( $files ); // reset pointer
} catch ( NoSuchContainerException $e ) {
- } catch ( NoSuchObjectException $e ) {
} catch ( InvalidResponseException $e ) {
} catch ( Exception $e ) { // some other exception?
$this->logException( $e, __METHOD__, array( 'cont' => $fullCont, 'dir' => $dir ) );
}
/**
- * @see FileBackend::doGetFileSha1base36()
+ * @see FileBackendStore::doGetFileSha1base36()
+ * @return bool
*/
- public function doGetFileSha1base36( array $params ) {
+ protected function doGetFileSha1base36( array $params ) {
$stat = $this->getFileStat( $params );
if ( $stat ) {
return $stat['sha1'];
}
/**
- * @see FileBackend::doStreamFile()
+ * @see FileBackendStore::doStreamFile()
+ * @return Status
*/
protected function doStreamFile( array $params ) {
$status = Status::newGood();
}
try {
- $output = fopen( 'php://output', 'w' );
- // FileBackend::streamFile() already checks existence
+ $output = fopen( 'php://output', 'wb' );
$obj = new CF_Object( $cont, $srcRel, false, false ); // skip HEAD request
$obj->stream( $output, $this->headersFromParams( $params ) );
} catch ( InvalidResponseException $e ) { // 404? connection problem?
}
/**
- * @see FileBackend::getLocalCopy()
+ * @see FileBackendStore::getLocalCopy()
+ * @return null|TempFSFile
*/
public function getLocalCopy( array $params ) {
list( $srcCont, $srcRel ) = $this->resolveStoragePathReal( $params['src'] );
return $tmpFile;
}
+ /**
+ * @see FileBackendStore::directoriesAreVirtual()
+ * @return bool
+ */
+ protected function directoriesAreVirtual() {
+ return true;
+ }
+
/**
* Get headers to send to Swift when reading a file based
- * on a FileBackend params array, e.g. that of getLocalCopy().
+ * on a FileBackend params array, e.g. that of getLocalCopy().
* $params is currently only checked for a 'latest' flag.
- *
+ *
* @param $params Array
- * @return Array
+ * @return Array
*/
protected function headersFromParams( array $params ) {
$hdrs = array();
$url = $creds['storage_url'] . '/' . rawurlencode( $contObj->name );
// Note: 10 second timeout consistent with php-cloudfiles
- $req = new CurlHttpRequest( $url, array( 'method' => 'POST', 'timeout' => 10 ) );
+ $req = new CurlHttpRequest( $url, array( 'method' => 'POST', 'timeout' => 10 ) );
$req->setHeader( 'X-Auth-Token', $creds['auth_token'] );
$req->setHeader( 'X-Container-Read', implode( ',', $readGrps ) );
$req->setHeader( 'X-Container-Write', implode( ',', $writeGrps ) );
/**
* Get a connection to the Swift proxy
*
- * @return CF_Connection|false
+ * @return CF_Connection|bool False on failure
* @throws InvalidResponseException
*/
protected function getConnection() {
}
/**
- * @see FileBackend::doClearCache()
+ * @see FileBackendStore::doClearCache()
*/
protected function doClearCache( array $paths = null ) {
$this->connContainers = array(); // clear container object cache
* Use $reCache if the file count or byte count is needed.
*
* @param $container string Container name
- * @param $reCache bool Refresh the process cache
+ * @param $bypassCache bool Bypass all caches and load from Swift
* @return CF_Container
+ * @throws InvalidResponseException
*/
- protected function getContainer( $container, $reCache = false ) {
+ protected function getContainer( $container, $bypassCache = false ) {
$conn = $this->getConnection(); // Swift proxy connection
- if ( $reCache ) {
- unset( $this->connContainers[$container] ); // purge cache
+ if ( $bypassCache ) { // purge cache
+ unset( $this->connContainers[$container] );
+ } elseif ( !isset( $this->connContainers[$container] ) ) {
+ $this->primeContainerCache( array( $container ) ); // check persistent cache
}
if ( !isset( $this->connContainers[$container] ) ) {
$contObj = $conn->get_container( $container );
// NoSuchContainerException not thrown: container must exist
if ( count( $this->connContainers ) >= $this->maxContCacheSize ) { // trim cache?
reset( $this->connContainers );
- $key = key( $this->connContainers );
- unset( $this->connContainers[$key] );
+ unset( $this->connContainers[key( $this->connContainers )] );
}
$this->connContainers[$container] = $contObj; // cache it
+ if ( !$bypassCache ) {
+ $this->setContainerCache( $container, // update persistent cache
+ array( 'bytes' => $contObj->bytes_used, 'count' => $contObj->object_count )
+ );
+ }
}
return $this->connContainers[$container];
}
*
* @param $container string Container name
* @return CF_Container
+ * @throws InvalidResponseException
*/
protected function createContainer( $container ) {
$conn = $this->getConnection(); // Swift proxy connection
*
* @param $container string Container name
* @return void
+ * @throws InvalidResponseException
*/
protected function deleteContainer( $container ) {
$conn = $this->getConnection(); // Swift proxy connection
unset( $this->connContainers[$container] ); // purge cache
}
+ /**
+ * @see FileBackendStore::doPrimeContainerCache()
+ * @return void
+ */
+ protected function doPrimeContainerCache( array $containerInfo ) {
+ try {
+ $conn = $this->getConnection(); // Swift proxy connection
+ foreach ( $containerInfo as $container => $info ) {
+ $this->connContainers[$container] = new CF_Container(
+ $conn->cfs_auth,
+ $conn->cfs_http,
+ $container,
+ $info['count'],
+ $info['bytes']
+ );
+ }
+ } catch ( InvalidResponseException $e ) {
+ } catch ( Exception $e ) { // some other exception?
+ $this->logException( $e, __METHOD__, array() );
+ }
+ }
+
/**
* Log an unexpected exception for this backend
- *
+ *
* @param $e Exception
* @param $func string
* @param $params Array
*/
protected function logException( Exception $e, $func, array $params ) {
wfDebugLog( 'SwiftBackend',
- get_class( $e ) . " in '{$this->name}': '{$func}' with " . serialize( $params )
+ get_class( $e ) . " in '{$func}' (given '" . FormatJson::encode( $params ) . "')" .
+ ( $e instanceof InvalidResponseException
+ ? ": {$e->getMessage()}"
+ : ""
+ )
);
}
}
/**
- * SwiftFileBackend helper class to page through object listings.
+ * SwiftFileBackend helper class to page through listings.
* Swift also has a listing limit of 10,000 objects for sanity.
* Do not use this class from places outside SwiftFileBackend.
*
* @ingroup FileBackend
*/
-class SwiftFileBackendFileList implements Iterator {
+abstract class SwiftFileBackendList implements Iterator {
/** @var Array */
protected $bufferIter = array();
protected $bufferAfter = null; // string; list items *after* this path
protected $pos = 0; // integer
+ /** @var Array */
+ protected $params = array();
/** @var SwiftFileBackend */
- protected $backend;
- protected $container; //
- protected $dir; // string storage directory
+ protected $backend;
+ protected $container; // string; container name
+ protected $dir; // string; storage directory
protected $suffixStart; // integer
const PAGE_SIZE = 5000; // file listing buffer size
* @param $backend SwiftFileBackend
* @param $fullCont string Resolved container name
* @param $dir string Resolved directory relative to container
+ * @param $params Array
*/
- public function __construct( SwiftFileBackend $backend, $fullCont, $dir ) {
+ public function __construct( SwiftFileBackend $backend, $fullCont, $dir, array $params ) {
$this->backend = $backend;
$this->container = $fullCont;
$this->dir = $dir;
} else { // dir within container
$this->suffixStart = strlen( $this->dir ) + 1; // size of "path/to/dir/"
}
+ $this->params = $params;
}
- public function current() {
- return substr( current( $this->bufferIter ), $this->suffixStart );
- }
-
+ /**
+ * @see Iterator::key()
+ * @return integer
+ */
public function key() {
return $this->pos;
}
+ /**
+ * @see Iterator::next()
+ * @return void
+ */
public function next() {
// Advance to the next file in the page
next( $this->bufferIter );
// Check if there are no files left in this page and
// advance to the next page if this page was not empty.
if ( !$this->valid() && count( $this->bufferIter ) ) {
- $this->bufferAfter = end( $this->bufferIter );
- $this->bufferIter = $this->backend->getFileListPageInternal(
- $this->container, $this->dir, $this->bufferAfter, self::PAGE_SIZE
- );
+ $this->bufferIter = $this->pageFromList(
+ $this->container, $this->dir, $this->bufferAfter, self::PAGE_SIZE, $this->params
+ ); // updates $this->bufferAfter
}
}
+ /**
+ * @see Iterator::rewind()
+ * @return void
+ */
public function rewind() {
$this->pos = 0;
$this->bufferAfter = null;
- $this->bufferIter = $this->backend->getFileListPageInternal(
- $this->container, $this->dir, $this->bufferAfter, self::PAGE_SIZE
- );
+ $this->bufferIter = $this->pageFromList(
+ $this->container, $this->dir, $this->bufferAfter, self::PAGE_SIZE, $this->params
+ ); // updates $this->bufferAfter
}
+ /**
+ * @see Iterator::valid()
+ * @return bool
+ */
public function valid() {
- return ( current( $this->bufferIter ) !== false ); // no paths can have this value
+ if ( $this->bufferIter === null ) {
+ return false; // some failure?
+ } else {
+ return ( current( $this->bufferIter ) !== false ); // no paths can have this value
+ }
+ }
+
+ /**
+ * Get the given list portion (page)
+ *
+ * @param $container string Resolved container name
+ * @param $dir string Resolved path relative to container
+ * @param $after string|null
+ * @param $limit integer
+ * @param $params Array
+ * @return Traversable|Array|null Returns null on failure
+ */
+ abstract protected function pageFromList( $container, $dir, &$after, $limit, array $params );
+}
+
+/**
+ * Iterator for listing directories
+ */
+class SwiftFileBackendDirList extends SwiftFileBackendList {
+ /**
+ * @see Iterator::current()
+ * @return string|bool String (relative path) or false
+ */
+ public function current() {
+ return substr( current( $this->bufferIter ), $this->suffixStart, -1 );
+ }
+
+ /**
+ * @see SwiftFileBackendList::pageFromList()
+ * @return Array|null
+ */
+ protected function pageFromList( $container, $dir, &$after, $limit, array $params ) {
+ return $this->backend->getDirListPageInternal( $container, $dir, $after, $limit, $params );
+ }
+}
+
+/**
+ * Iterator for listing regular files
+ */
+class SwiftFileBackendFileList extends SwiftFileBackendList {
+ /**
+ * @see Iterator::current()
+ * @return string|bool String (relative path) or false
+ */
+ public function current() {
+ return substr( current( $this->bufferIter ), $this->suffixStart );
+ }
+
+ /**
+ * @see SwiftFileBackendList::pageFromList()
+ * @return Array|null
+ */
+ protected function pageFromList( $container, $dir, &$after, $limit, array $params ) {
+ return $this->backend->getFileListPageInternal( $container, $dir, $after, $limit, $params );
}
}