X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/8f39f0b74291009f036359bd651020e159a75d10..97e40209600c094eac15085627e49da52ab0f517:/sdk/python/arvados/commands/put.py diff --git a/sdk/python/arvados/commands/put.py b/sdk/python/arvados/commands/put.py index d13c2cd50f..0e732eafde 100644 --- a/sdk/python/arvados/commands/put.py +++ b/sdk/python/arvados/commands/put.py @@ -77,8 +77,7 @@ Synonym for --stream. _group.add_argument('--stream', action='store_true', help=""" Store the file content and display the resulting manifest on -stdout. Do not write the manifest to Keep or save a Collection object -in Arvados. +stdout. Do not save a Collection object in Arvados. """) _group.add_argument('--as-manifest', action='store_true', dest='manifest', @@ -174,7 +173,8 @@ Follow file and directory symlinks (default). """) _group.add_argument('--no-follow-links', action='store_false', dest='follow_links', help=""" -Do not follow file and directory symlinks. +Ignore file and directory symlinks. Even paths given explicitly on the +command line will be skipped if they are symlinks. """) @@ -215,6 +215,12 @@ Do not print any debug messages to console. (Any error messages will still be displayed.) """) +run_opts.add_argument('--batch', action='store_true', default=False, + help=""" +Retries with '--no-resume --no-cache' if cached state contains invalid/expired +block signatures. +""") + _group = run_opts.add_mutually_exclusive_group() _group.add_argument('--resume', action='store_true', default=True, help=""" @@ -236,7 +242,7 @@ Do not save upload state in a cache file for resuming. """) _group = upload_opts.add_mutually_exclusive_group() -_group.add_argument('--trash-at', metavar='YYYY-MM-DD HH:MM', default=None, +_group.add_argument('--trash-at', metavar='YYYY-MM-DDTHH:MM', default=None, help=""" Set the trash date of the resulting collection to an absolute date in the future. The accepted format is defined by the ISO 8601 standard. Examples: 20090103, 2009-01-03, 20090103T181505, 2009-01-03T18:15:05.\n @@ -260,9 +266,8 @@ def parse_arguments(arguments): args.paths = ["-" if x == "/dev/stdin" else x for x in args.paths] - if len(args.paths) != 1 or os.path.isdir(args.paths[0]): - if args.filename: - arg_parser.error(""" + if args.filename and (len(args.paths) != 1 or os.path.isdir(args.paths[0])): + arg_parser.error(""" --filename argument cannot be used when storing a directory or multiple files. """) @@ -439,7 +444,7 @@ class ArvPutUploadJob(object): } def __init__(self, paths, resume=True, use_cache=True, reporter=None, - name=None, owner_uuid=None, api_client=None, + name=None, owner_uuid=None, api_client=None, batch_mode=False, ensure_unique_name=False, num_retries=None, put_threads=None, replication_desired=None, filename=None, update_time=60.0, update_collection=None, storage_classes=None, @@ -449,6 +454,7 @@ class ArvPutUploadJob(object): self.paths = paths self.resume = resume self.use_cache = use_cache + self.batch_mode = batch_mode self.update = False self.reporter = reporter # This will set to 0 before start counting, if no special files are going @@ -526,6 +532,9 @@ class ArvPutUploadJob(object): self._write_stdin(self.filename or 'stdin') elif not os.path.exists(path): raise PathDoesNotExistError(u"file or directory '{}' does not exist.".format(path)) + elif (not self.follow_links) and os.path.islink(path): + self.logger.warning("Skipping symlink '{}'".format(path)) + continue elif os.path.isdir(path): # Use absolute paths on cache index so CWD doesn't interfere # with the caching logic. @@ -567,6 +576,9 @@ class ArvPutUploadJob(object): files.sort() for f in files: filepath = os.path.join(root, f) + if not os.path.isfile(filepath): + self.logger.warning("Skipping non-regular file '{}'".format(filepath)) + continue # Add its size to the total bytes count (if applicable) if self.follow_links or (not os.path.islink(filepath)): if self.bytes_expected is not None: @@ -657,15 +669,14 @@ class ArvPutUploadJob(object): else: # The file already exist on remote collection, skip it. pass - self._remote_collection.save(storage_classes=self.storage_classes, - num_retries=self.num_retries, + self._remote_collection.save(num_retries=self.num_retries, trash_at=self._collection_trash_at()) else: - if self.storage_classes is None: - self.storage_classes = ['default'] + if len(self._local_collection) == 0: + self.logger.warning("No files were uploaded, skipping collection creation.") + return self._local_collection.save_new( name=self.name, owner_uuid=self.owner_uuid, - storage_classes=self.storage_classes, ensure_unique_name=self.ensure_unique_name, num_retries=self.num_retries, trash_at=self._collection_trash_at()) @@ -723,6 +734,15 @@ class ArvPutUploadJob(object): self._save_state() except Exception as e: self.logger.error("Unexpected error trying to save cache file: {}".format(e)) + # Keep remote collection's trash_at attribute synced when using relative expire dates + if self._remote_collection is not None and type(self._trash_at) == datetime.timedelta: + try: + self._api_client.collections().update( + uuid=self._remote_collection.manifest_locator(), + body={'trash_at': self._collection_trash_at().strftime("%Y-%m-%dT%H:%M:%S.%fZ")} + ).execute(num_retries=self.num_retries) + except Exception as e: + self.logger.error("Unexpected error trying to update remote collection's expire date: {}".format(e)) else: self.bytes_written = self.bytes_skipped # Call the reporter, if any @@ -734,7 +754,7 @@ class ArvPutUploadJob(object): def _write_stdin(self, filename): output = self._local_collection.open(filename, 'wb') - self._write(sys.stdin, output) + self._write(sys.stdin.buffer, output) output.close() def _check_file(self, source, filename): @@ -858,7 +878,10 @@ class ArvPutUploadJob(object): update_collection): try: self._remote_collection = arvados.collection.Collection( - update_collection, api_client=self._api_client) + update_collection, + api_client=self._api_client, + storage_classes_desired=self.storage_classes, + num_retries=self.num_retries) except arvados.errors.ApiError as error: raise CollectionUpdateError("Cannot read collection {} ({})".format(update_collection, error)) else: @@ -895,13 +918,20 @@ class ArvPutUploadJob(object): # No cache file, set empty state self._state = copy.deepcopy(self.EMPTY_STATE) if not self._cached_manifest_valid(): - raise ResumeCacheInvalidError() + if not self.batch_mode: + raise ResumeCacheInvalidError() + else: + self.logger.info("Invalid signatures on cache file '{}' while being run in 'batch mode' -- continuing anyways.".format(self._cache_file.name)) + self.use_cache = False # Don't overwrite preexisting cache file. + self._state = copy.deepcopy(self.EMPTY_STATE) # Load the previous manifest so we can check if files were modified remotely. self._local_collection = arvados.collection.Collection( self._state['manifest'], replication_desired=self.replication_desired, + storage_classes_desired=self.storage_classes, put_threads=self.put_threads, - api_client=self._api_client) + api_client=self._api_client, + num_retries=self.num_retries) def _cached_manifest_valid(self): """ @@ -993,6 +1023,9 @@ class ArvPutUploadJob(object): def collection_name(self): return self._my_collection().api_response()['name'] if self._my_collection().api_response() else None + def collection_trash_at(self): + return self._my_collection().get_trash_at() + def manifest_locator(self): return self._my_collection().manifest_locator() @@ -1103,7 +1136,7 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr, logging.getLogger('arvados').handlers[0].setFormatter(formatter) if api_client is None: - api_client = arvados.api('v1', request_id=request_id) + api_client = arvados.api('v1', request_id=request_id, num_retries=args.retries) if install_sig_handlers: arv_cmd.install_signal_handlers() @@ -1111,6 +1144,14 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr, # Trash arguments validation trash_at = None if args.trash_at is not None: + # ciso8601 considers YYYYMM as invalid but YYYY-MM as valid, so here we + # make sure the user provides a complete YYYY-MM-DD date. + if not re.match(r'^\d{4}(?P-?)\d{2}?(?P=dash)\d{2}', args.trash_at): + logger.error("--trash-at argument format invalid, use --help to see examples.") + sys.exit(1) + # Check if no time information was provided. In that case, assume end-of-day. + if re.match(r'^\d{4}(?P-?)\d{2}?(?P=dash)\d{2}$', args.trash_at): + args.trash_at += 'T23:59:59' try: trash_at = ciso8601.parse_datetime(args.trash_at) except: @@ -1119,19 +1160,22 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr, else: if trash_at.tzinfo is not None: # Timezone aware datetime provided. - delta = trash_at.utcoffset() + utcoffset = -trash_at.utcoffset() else: # Timezone naive datetime provided. Assume is local. - delta = datetime.datetime.now() - datetime.datetime.utcnow() + if time.daylight: + utcoffset = datetime.timedelta(seconds=time.altzone) + else: + utcoffset = datetime.timedelta(seconds=time.timezone) # Convert to UTC timezone naive datetime. - trash_at = trash_at.replace(tzinfo=None) - delta + trash_at = trash_at.replace(tzinfo=None) + utcoffset if trash_at <= datetime.datetime.utcnow(): - logger.error("--trash-at argument should be set in the future") + logger.error("--trash-at argument must be set in the future") sys.exit(1) if args.trash_after is not None: if args.trash_after < 1: - logger.error("--trash-after argument should be >= 1") + logger.error("--trash-after argument must be >= 1") sys.exit(1) trash_at = datetime.timedelta(seconds=(args.trash_after * 24 * 60 * 60)) @@ -1172,11 +1216,7 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr, # Split storage-classes argument storage_classes = None if args.storage_classes: - storage_classes = args.storage_classes.strip().split(',') - if len(storage_classes) > 1: - logger.error("Multiple storage classes are not supported currently.") - sys.exit(1) - + storage_classes = args.storage_classes.strip().replace(' ', '').split(',') # Setup exclude regex from all the --exclude arguments provided name_patterns = [] @@ -1225,6 +1265,7 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr, writer = ArvPutUploadJob(paths = args.paths, resume = args.resume, use_cache = args.use_cache, + batch_mode= args.batch, filename = args.filename, reporter = reporter, api_client = api_client, @@ -1253,7 +1294,8 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr, " or been created with another Arvados user's credentials.", " Switch user or use one of the following options to restart upload:", " --no-resume to start a new resume cache.", - " --no-cache to disable resume cache."])) + " --no-cache to disable resume cache.", + " --batch to ignore the resume cache if invalid."])) sys.exit(1) except (CollectionUpdateError, PathDoesNotExistError) as error: logger.error("\n".join([ @@ -1276,7 +1318,7 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr, output = None try: writer.start(save_collection=not(args.stream or args.raw)) - except arvados.errors.ApiError as error: + except (arvados.errors.ApiError, arvados.errors.KeepWriteError) as error: logger.error("\n".join([ "arv-put: %s" % str(error)])) sys.exit(1) @@ -1291,12 +1333,23 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr, output = writer.manifest_text() elif args.raw: output = ','.join(writer.data_locators()) - else: + elif writer.manifest_locator() is not None: try: + expiration_notice = "" + if writer.collection_trash_at() is not None: + # Get the local timezone-naive version, and log it with timezone information. + if time.daylight: + local_trash_at = writer.collection_trash_at().replace(tzinfo=None) - datetime.timedelta(seconds=time.altzone) + else: + local_trash_at = writer.collection_trash_at().replace(tzinfo=None) - datetime.timedelta(seconds=time.timezone) + expiration_notice = ". It will expire on {} {}.".format( + local_trash_at.strftime("%Y-%m-%d %H:%M:%S"), time.strftime("%z")) if args.update_collection: - logger.info(u"Collection updated: '{}'".format(writer.collection_name())) + logger.info(u"Collection updated: '{}'{}".format( + writer.collection_name(), expiration_notice)) else: - logger.info(u"Collection saved as '{}'".format(writer.collection_name())) + logger.info(u"Collection saved as '{}'{}".format( + writer.collection_name(), expiration_notice)) if args.portable_data_hash: output = writer.portable_data_hash() else: @@ -1306,6 +1359,8 @@ def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr, "arv-put: Error creating Collection on project: {}.".format( error)) status = 1 + else: + status = 1 # Print the locator (uuid) of the new collection. if output is None: