def RunCommand(self): """Command entry point for the rewrite command.""" self.continue_on_error = self.parallel_operations self.csek_hash_to_keywrapper = {} self.dest_storage_class = None self.no_preserve_acl = False self.read_args_from_stdin = False self.supported_transformation_flags = ['-k', '-s'] self.transform_types = set() self.op_failure_count = 0 self.boto_file_encryption_keywrapper = GetEncryptionKeyWrapper(config) self.boto_file_encryption_sha256 = ( self.boto_file_encryption_keywrapper.crypto_key_sha256 if self.boto_file_encryption_keywrapper else None) if self.sub_opts: for o, a in self.sub_opts: if o == '-f': self.continue_on_error = True elif o == '-k': self.transform_types.add(_TransformTypes.CRYPTO_KEY) elif o == '-I': self.read_args_from_stdin = True elif o == '-O': self.no_preserve_acl = True elif o == '-r' or o == '-R': self.recursion_requested = True self.all_versions = True elif o == '-s': self.transform_types.add(_TransformTypes.STORAGE_CLASS) self.dest_storage_class = NormalizeStorageClass(a) if self.read_args_from_stdin: if self.args: raise CommandException('No arguments allowed with the -I flag.') url_strs = StdinIterator() else: if not self.args: raise CommandException('The rewrite command (without -I) expects at ' 'least one URL.') url_strs = self.args if not self.transform_types: raise CommandException( 'rewrite command requires at least one transformation flag. ' 'Currently supported transformation flags: %s' % self.supported_transformation_flags) self.preconditions = PreconditionsFromHeaders(self.headers or {}) url_strs_generator = GenerationCheckGenerator(url_strs) # Convert recursive flag to flat wildcard to avoid performing multiple # listings. if self.recursion_requested: url_strs_generator = ConvertRecursiveToFlatWildcard(url_strs_generator) # Expand the source argument(s). name_expansion_iterator = NameExpansionIterator( self.command_name, self.debug, self.logger, self.gsutil_api, url_strs_generator, self.recursion_requested, project_id=self.project_id, continue_on_error=self.continue_on_error or self.parallel_operations, bucket_listing_fields=['name', 'size']) seek_ahead_iterator = None # Cannot seek ahead with stdin args, since we can only iterate them # once without buffering in memory. if not self.read_args_from_stdin: # Perform the same recursive-to-flat conversion on original url_strs so # that it is as true to the original iterator as possible. seek_ahead_url_strs = ConvertRecursiveToFlatWildcard(url_strs) seek_ahead_iterator = SeekAheadNameExpansionIterator( self.command_name, self.debug, self.GetSeekAheadGsutilApi(), seek_ahead_url_strs, self.recursion_requested, all_versions=self.all_versions, project_id=self.project_id) # Rather than have each worker repeatedly calculate the sha256 hash for each # decryption_key in the boto config, do this once now and cache the results. for i in range(0, MAX_DECRYPTION_KEYS): key_number = i + 1 keywrapper = CryptoKeyWrapperFromKey( config.get('GSUtil', 'decryption_key%s' % str(key_number), None)) if keywrapper is None: # Stop at first attribute absence in lexicographical iteration. break if keywrapper.crypto_type == CryptoKeyType.CSEK: self.csek_hash_to_keywrapper[keywrapper.crypto_key_sha256] = keywrapper # Also include the encryption_key, since it should be used to decrypt and # then encrypt if the object's CSEK should remain the same. if self.boto_file_encryption_sha256 is not None: self.csek_hash_to_keywrapper[self.boto_file_encryption_sha256] = ( self.boto_file_encryption_keywrapper) if self.boto_file_encryption_keywrapper is None: msg = '\n'.join( textwrap.wrap( 'NOTE: No encryption_key was specified in the boto configuration ' 'file, so gsutil will not provide an encryption key in its rewrite ' 'API requests. This will decrypt the objects unless they are in ' 'buckets with a default KMS key set, in which case the service ' 'will automatically encrypt the rewritten objects with that key.') ) print('%s\n' % msg, file=sys.stderr) # Perform rewrite requests in parallel (-m) mode, if requested. self.Apply(_RewriteFuncWrapper, name_expansion_iterator, _RewriteExceptionHandler, fail_on_error=(not self.continue_on_error), shared_attrs=['op_failure_count'], seek_ahead_iterator=seek_ahead_iterator) if self.op_failure_count: plural_str = 's' if self.op_failure_count else '' raise CommandException('%d file%s/object%s could not be rewritten.' % (self.op_failure_count, plural_str, plural_str)) return 0
def RunCommand(self): """Command entry point for the rm command.""" # self.recursion_requested is initialized in command.py (so it can be # checked in parent class for all commands). self.continue_on_error = self.parallel_operations self.read_args_from_stdin = False self.all_versions = False if self.sub_opts: for o, unused_a in self.sub_opts: if o == '-a': self.all_versions = True elif o == '-f': self.continue_on_error = True elif o == '-I': self.read_args_from_stdin = True elif o == '-r' or o == '-R': self.recursion_requested = True self.all_versions = True if self.read_args_from_stdin: if self.args: raise CommandException( 'No arguments allowed with the -I flag.') url_strs = StdinIterator() else: if not self.args: raise CommandException( 'The rm command (without -I) expects at ' 'least one URL.') url_strs = self.args # Tracks number of object deletes that failed. self.op_failure_count = 0 # Tracks if any buckets were missing. self.bucket_not_found_count = 0 # Tracks buckets that are slated for recursive deletion. bucket_urls_to_delete = [] self.bucket_strings_to_delete = [] if self.recursion_requested: bucket_fields = ['id'] for url_str in url_strs: url = StorageUrlFromString(url_str) if url.IsBucket() or url.IsProvider(): for blr in self.WildcardIterator(url_str).IterBuckets( bucket_fields=bucket_fields): bucket_urls_to_delete.append(blr.storage_url) self.bucket_strings_to_delete.append(url_str) self.preconditions = PreconditionsFromHeaders(self.headers or {}) try: # Expand wildcards, dirs, buckets, and bucket subdirs in URLs. name_expansion_iterator = NameExpansionIterator( self.command_name, self.debug, self.logger, self.gsutil_api, url_strs, self.recursion_requested, project_id=self.project_id, all_versions=self.all_versions, continue_on_error=self.continue_on_error or self.parallel_operations) seek_ahead_iterator = None # Cannot seek ahead with stdin args, since we can only iterate them # once without buffering in memory. if not self.read_args_from_stdin: seek_ahead_iterator = SeekAheadNameExpansionIterator( self.command_name, self.debug, self.GetSeekAheadGsutilApi(), url_strs, self.recursion_requested, all_versions=self.all_versions, project_id=self.project_id) # Perform remove requests in parallel (-m) mode, if requested, using # configured number of parallel processes and threads. Otherwise, # perform requests with sequential function calls in current process. self.Apply( _RemoveFuncWrapper, name_expansion_iterator, _RemoveExceptionHandler, fail_on_error=(not self.continue_on_error), shared_attrs=['op_failure_count', 'bucket_not_found_count'], seek_ahead_iterator=seek_ahead_iterator) # Assuming the bucket has versioning enabled, url's that don't map to # objects should throw an error even with all_versions, since the prior # round of deletes only sends objects to a history table. # This assumption that rm -a is only called for versioned buckets should be # corrected, but the fix is non-trivial. except CommandException as e: # Don't raise if there are buckets to delete -- it's valid to say: # gsutil rm -r gs://some_bucket # if the bucket is empty. if _ExceptionMatchesBucketToDelete(self.bucket_strings_to_delete, e): DecrementFailureCount() else: raise except ServiceException as e: if not self.continue_on_error: raise if self.bucket_not_found_count: raise CommandException( 'Encountered non-existent bucket during listing') if self.op_failure_count and not self.continue_on_error: raise CommandException('Some files could not be removed.') # If this was a gsutil rm -r command covering any bucket subdirs, # remove any dir_$folder$ objects (which are created by various web UI # tools to simulate folders). if self.recursion_requested: folder_object_wildcards = [] for url_str in url_strs: url = StorageUrlFromString(url_str) if url.IsObject(): folder_object_wildcards.append('%s**_$folder$' % url_str) if folder_object_wildcards: self.continue_on_error = True try: name_expansion_iterator = NameExpansionIterator( self.command_name, self.debug, self.logger, self.gsutil_api, folder_object_wildcards, self.recursion_requested, project_id=self.project_id, all_versions=self.all_versions) # When we're removing folder objects, always continue on error self.Apply(_RemoveFuncWrapper, name_expansion_iterator, _RemoveFoldersExceptionHandler, fail_on_error=False) except CommandException as e: # Ignore exception from name expansion due to an absent folder file. if not e.reason.startswith(NO_URLS_MATCHED_GENERIC): raise # Now that all data has been deleted, delete any bucket URLs. for url in bucket_urls_to_delete: self.logger.info('Removing %s...', url) @Retry(NotEmptyException, tries=3, timeout_secs=1) def BucketDeleteWithRetry(): self.gsutil_api.DeleteBucket(url.bucket_name, provider=url.scheme) BucketDeleteWithRetry() if self.op_failure_count: plural_str = 's' if self.op_failure_count else '' raise CommandException( '%d file%s/object%s could not be removed.' % (self.op_failure_count, plural_str, plural_str)) return 0
def RunCommand(self): """Command entry point for the rm command.""" # self.recursion_requested is initialized in command.py (so it can be # checked in parent class for all commands). self.continue_on_error = self.parallel_operations self.read_args_from_stdin = False self.all_versions = False if self.sub_opts: for o, unused_a in self.sub_opts: if o == '-a': self.all_versions = True elif o == '-f': self.continue_on_error = True elif o == '-I': self.read_args_from_stdin = True elif o == '-r' or o == '-R': self.recursion_requested = True self.all_versions = True if self.read_args_from_stdin: if self.args: raise CommandException( 'No arguments allowed with the -I flag.') url_strs = StdinIterator() else: if not self.args: raise CommandException( 'The rm command (without -I) expects at ' 'least one URL.') url_strs = self.args # Tracks number of object deletes that failed. self.op_failure_count = 0 # Tracks if any buckets were missing. self.bucket_not_found_count = 0 # Tracks buckets that are slated for recursive deletion. bucket_urls_to_delete = [] self.bucket_strings_to_delete = [] if self.recursion_requested: bucket_fields = ['id'] for url_str in url_strs: url = StorageUrlFromString(url_str) if url.IsBucket() or url.IsProvider(): for blr in self.WildcardIterator(url_str).IterBuckets( bucket_fields=bucket_fields): bucket_urls_to_delete.append(blr.storage_url) self.bucket_strings_to_delete.append(url_str) self.preconditions = PreconditionsFromHeaders(self.headers or {}) try: # Expand wildcards, dirs, buckets, and bucket subdirs in URLs. name_expansion_iterator = NameExpansionIterator( self.command_name, self.debug, self.logger, self.gsutil_api, url_strs, self.recursion_requested, project_id=self.project_id, all_versions=self.all_versions, continue_on_error=self.continue_on_error or self.parallel_operations) seek_ahead_iterator = None # Cannot seek ahead with stdin args, since we can only iterate them # once without buffering in memory. if not self.read_args_from_stdin: seek_ahead_iterator = SeekAheadNameExpansionIterator( self.command_name, self.debug, self.GetSeekAheadGsutilApi(), url_strs, self.recursion_requested, all_versions=self.all_versions, project_id=self.project_id) # Perform remove requests in parallel (-m) mode, if requested, using # configured number of parallel processes and threads. Otherwise, # perform requests with sequential function calls in current process. self.Apply( _RemoveFuncWrapper, name_expansion_iterator, _RemoveExceptionHandler, fail_on_error=(not self.continue_on_error), shared_attrs=['op_failure_count', 'bucket_not_found_count'], seek_ahead_iterator=seek_ahead_iterator) # Assuming the bucket has versioning enabled, url's that don't map to # objects should throw an error even with all_versions, since the prior # round of deletes only sends objects to a history table. # This assumption that rm -a is only called for versioned buckets should be # corrected, but the fix is non-trivial. except CommandException as e: # Don't raise if there are buckets to delete -- it's valid to say: # gsutil rm -r gs://some_bucket # if the bucket is empty. if _ExceptionMatchesBucketToDelete(self.bucket_strings_to_delete, e): DecrementFailureCount() else: raise except ServiceException, e: if not self.continue_on_error: raise