X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/2126d3a9115620040260847a9e29fd5c73b466ab..HEAD:/sdk/python/arvados/commands/put.py diff --git a/sdk/python/arvados/commands/put.py b/sdk/python/arvados/commands/put.py index 5773cb4f98..ce2e5375b7 100644 --- a/sdk/python/arvados/commands/put.py +++ b/sdk/python/arvados/commands/put.py @@ -2,10 +2,6 @@ # # SPDX-License-Identifier: Apache-2.0 -from __future__ import division -from future.utils import listitems, listvalues -from builtins import str -from builtins import object import argparse import arvados import arvados.collection @@ -30,10 +26,12 @@ import threading import time import traceback +from pathlib import Path + from apiclient import errors as apiclient_errors from arvados._version import __version__ -from arvados.util import keep_locator_pattern +import arvados.util import arvados.commands._util as arv_cmd api_client = None @@ -77,8 +75,7 @@ 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 +171,8 @@ 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 +213,12 @@ 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=""" @@ -260,9 +264,8 @@ 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. """) @@ -350,7 +353,7 @@ class ArvPutLogFormatter(logging.Formatter): class ResumeCache(object): - CACHE_DIR = '.cache/arvados/arv-put' + CACHE_DIR = 'arv-put' def __init__(self, file_spec): self.cache_file = open(file_spec, 'a+') @@ -367,9 +370,14 @@ class ResumeCache(object): md5.update(b'-1') elif args.filename: md5.update(args.filename.encode()) - return os.path.join( - arv_cmd.make_home_conf_dir(cls.CACHE_DIR, 0o700, 'raise'), - md5.hexdigest()) + cache_path = Path(cls.CACHE_DIR) + if len(cache_path.parts) == 1: + cache_path = arvados.util._BaseDirectories('CACHE').storage_path(cache_path) + else: + # Note this is a noop if cache_path is absolute, which is what we want. + cache_path = Path.home() / cache_path + cache_path.mkdir(parents=True, exist_ok=True, mode=0o700) + return str(cache_path / md5.hexdigest()) def _lock_file(self, fileobj): try: @@ -432,14 +440,14 @@ class ResumeCache(object): class ArvPutUploadJob(object): - CACHE_DIR = '.cache/arvados/arv-put' + CACHE_DIR = 'arv-put' EMPTY_STATE = { 'manifest' : None, # Last saved manifest checkpoint 'files' : {} # Previous run file list: {path : {size, mtime}} } 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 +457,7 @@ 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 +535,9 @@ 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 +579,9 @@ 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 +672,14 @@ 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()) @@ -685,7 +699,7 @@ class ArvPutUploadJob(object): Recursively get the total size of the collection """ size = 0 - for item in listvalues(collection): + for item in collection.values(): if isinstance(item, arvados.collection.Collection) or isinstance(item, arvados.collection.Subcollection): size += self._collection_size(item) else: @@ -743,7 +757,7 @@ 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): @@ -852,11 +866,14 @@ class ArvPutUploadJob(object): md5.update(b'\0'.join([p.encode() for p in realpaths])) if self.filename: md5.update(self.filename.encode()) - cache_filename = md5.hexdigest() - cache_filepath = os.path.join( - arv_cmd.make_home_conf_dir(self.CACHE_DIR, 0o700, 'raise'), - cache_filename) - return cache_filepath + cache_path = Path(self.CACHE_DIR) + if len(cache_path.parts) == 1: + cache_path = arvados.util._BaseDirectories('CACHE').storage_path(cache_path) + else: + # Note this is a noop if cache_path is absolute, which is what we want. + cache_path = Path.home() / cache_path + cache_path.mkdir(parents=True, exist_ok=True, mode=0o700) + return str(cache_path / md5.hexdigest()) def _setup_state(self, update_collection): """ @@ -867,7 +884,10 @@ 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: @@ -904,13 +924,20 @@ 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): """ @@ -925,7 +952,7 @@ class ArvPutUploadJob(object): oldest_exp = None oldest_loc = None block_found = False - for m in keep_locator_pattern.finditer(self._state['manifest']): + for m in arvados.util.keep_locator_pattern.finditer(self._state['manifest']): loc = m.group(0) try: exp = datetime.datetime.utcfromtimestamp(int(loc.split('@')[1], 16)) @@ -957,7 +984,7 @@ class ArvPutUploadJob(object): def collection_file_paths(self, col, path_prefix='.'): """Return a list of file paths by recursively go through the entire collection `col`""" file_paths = [] - for name, item in listitems(col): + for name, item in col.items(): if isinstance(item, arvados.arvfile.ArvadosFile): file_paths.append(os.path.join(path_prefix, name)) elif isinstance(item, arvados.collection.Subcollection): @@ -1037,7 +1064,7 @@ class ArvPutUploadJob(object): locators.append(loc) return locators elif isinstance(item, arvados.collection.Collection): - l = [self._datablocks_on_item(x) for x in listvalues(item)] + l = [self._datablocks_on_item(x) for x in item.values()] # Fast list flattener method taken from: # http://stackoverflow.com/questions/952914/making-a-flat-list-out-of-list-of-lists-in-python return [loc for sublist in l for loc in sublist] @@ -1115,7 +1142,7 @@ 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() @@ -1195,11 +1222,7 @@ 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 = [] @@ -1248,6 +1271,7 @@ 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, @@ -1276,7 +1300,8 @@ 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([ @@ -1299,7 +1324,7 @@ 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) @@ -1314,7 +1339,7 @@ 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: @@ -1340,6 +1365,8 @@ 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: