1 # Copyright (C) The Arvados Authors. All rights reserved.
3 # SPDX-License-Identifier: Apache-2.0
5 from __future__ import division
6 from future.utils import listitems, listvalues
7 from builtins import str
8 from builtins import object
11 import arvados.collection
32 from apiclient import errors as apiclient_errors
33 from arvados._version import __version__
35 import arvados.commands._util as arv_cmd
37 CAUGHT_SIGNALS = [signal.SIGINT, signal.SIGQUIT, signal.SIGTERM]
40 upload_opts = argparse.ArgumentParser(add_help=False)
42 upload_opts.add_argument('--version', action='version',
43 version="%s %s" % (sys.argv[0], __version__),
44 help='Print version and exit.')
45 upload_opts.add_argument('paths', metavar='path', type=str, nargs='*',
47 Local file or directory. If path is a directory reference with a trailing
48 slash, then just upload the directory's contents; otherwise upload the
49 directory itself. Default: read from standard input.
52 _group = upload_opts.add_mutually_exclusive_group()
54 _group.add_argument('--max-manifest-depth', type=int, metavar='N',
55 default=-1, help=argparse.SUPPRESS)
57 _group.add_argument('--normalize', action='store_true',
59 Normalize the manifest by re-ordering files and streams after writing
63 _group.add_argument('--dry-run', action='store_true', default=False,
65 Don't actually upload files, but only check if any file should be
66 uploaded. Exit with code=2 when files are pending for upload.
69 _group = upload_opts.add_mutually_exclusive_group()
71 _group.add_argument('--as-stream', action='store_true', dest='stream',
76 _group.add_argument('--stream', action='store_true',
78 Store the file content and display the resulting manifest on
79 stdout. Do not write the manifest to Keep or save a Collection object
83 _group.add_argument('--as-manifest', action='store_true', dest='manifest',
85 Synonym for --manifest.
88 _group.add_argument('--in-manifest', action='store_true', dest='manifest',
90 Synonym for --manifest.
93 _group.add_argument('--manifest', action='store_true',
95 Store the file data and resulting manifest in Keep, save a Collection
96 object in Arvados, and display the manifest locator (Collection uuid)
97 on stdout. This is the default behavior.
100 _group.add_argument('--as-raw', action='store_true', dest='raw',
105 _group.add_argument('--raw', action='store_true',
107 Store the file content and display the data block locators on stdout,
108 separated by commas, with a trailing newline. Do not store a
112 upload_opts.add_argument('--update-collection', type=str, default=None,
113 dest='update_collection', metavar="UUID", help="""
114 Update an existing collection identified by the given Arvados collection
115 UUID. All new local files will be uploaded.
118 upload_opts.add_argument('--use-filename', type=str, default=None,
119 dest='filename', help="""
120 Synonym for --filename.
123 upload_opts.add_argument('--filename', type=str, default=None,
125 Use the given filename in the manifest, instead of the name of the
126 local file. This is useful when "-" or "/dev/stdin" is given as an
127 input file. It can be used only if there is exactly one path given and
128 it is not a directory. Implies --manifest.
131 upload_opts.add_argument('--portable-data-hash', action='store_true',
133 Print the portable data hash instead of the Arvados UUID for the collection
134 created by the upload.
137 upload_opts.add_argument('--replication', type=int, metavar='N', default=None,
139 Set the replication level for the new collection: how many different
140 physical storage devices (e.g., disks) should have a copy of each data
141 block. Default is to use the server-provided default (if any) or 2.
144 upload_opts.add_argument('--threads', type=int, metavar='N', default=None,
146 Set the number of upload threads to be used. Take into account that
147 using lots of threads will increase the RAM requirements. Default is
149 On high latency installations, using a greater number will improve
153 run_opts = argparse.ArgumentParser(add_help=False)
155 run_opts.add_argument('--project-uuid', metavar='UUID', help="""
156 Store the collection in the specified project, instead of your Home
160 run_opts.add_argument('--name', help="""
161 Save the collection with the specified name.
164 run_opts.add_argument('--exclude', metavar='PATTERN', default=[],
165 action='append', help="""
166 Exclude files and directories whose names match the given glob pattern. When
167 using a path-like pattern like 'subdir/*.txt', all text files inside 'subdir'
168 directory, relative to the provided input dirs will be excluded.
169 When using a filename pattern like '*.txt', any text file will be excluded
170 no matter where is placed.
171 You can specify multiple patterns by using this argument more than once.
174 _group = run_opts.add_mutually_exclusive_group()
175 _group.add_argument('--progress', action='store_true',
177 Display human-readable progress on stderr (bytes and, if possible,
178 percentage of total data size). This is the default behavior when
182 _group.add_argument('--no-progress', action='store_true',
184 Do not display human-readable progress on stderr, even if stderr is a
188 _group.add_argument('--batch-progress', action='store_true',
190 Display machine-readable progress on stderr (bytes and, if known,
194 _group = run_opts.add_mutually_exclusive_group()
195 _group.add_argument('--resume', action='store_true', default=True,
197 Continue interrupted uploads from cached state (default).
199 _group.add_argument('--no-resume', action='store_false', dest='resume',
201 Do not continue interrupted uploads from cached state.
204 _group = run_opts.add_mutually_exclusive_group()
205 _group.add_argument('--follow-links', action='store_true', default=True,
206 dest='follow_links', help="""
207 Follow file and directory symlinks (default).
209 _group.add_argument('--no-follow-links', action='store_false', dest='follow_links',
211 Do not follow file and directory symlinks.
214 _group = run_opts.add_mutually_exclusive_group()
215 _group.add_argument('--cache', action='store_true', dest='use_cache', default=True,
217 Save upload state in a cache file for resuming (default).
219 _group.add_argument('--no-cache', action='store_false', dest='use_cache',
221 Do not save upload state in a cache file for resuming.
224 arg_parser = argparse.ArgumentParser(
225 description='Copy data from the local filesystem to Keep.',
226 parents=[upload_opts, run_opts, arv_cmd.retry_opt])
228 def parse_arguments(arguments):
229 args = arg_parser.parse_args(arguments)
231 if len(args.paths) == 0:
234 args.paths = ["-" if x == "/dev/stdin" else x for x in args.paths]
236 if len(args.paths) != 1 or os.path.isdir(args.paths[0]):
239 --filename argument cannot be used when storing a directory or
243 # Turn on --progress by default if stderr is a tty.
244 if (not (args.batch_progress or args.no_progress)
245 and os.isatty(sys.stderr.fileno())):
248 # Turn off --resume (default) if --no-cache is used.
249 if not args.use_cache:
252 if args.paths == ['-']:
253 if args.update_collection:
255 --update-collection cannot be used when reading from stdin.
258 args.use_cache = False
259 if not args.filename:
260 args.filename = 'stdin'
262 # Remove possible duplicated patterns
263 if len(args.exclude) > 0:
264 args.exclude = list(set(args.exclude))
269 class PathDoesNotExistError(Exception):
273 class CollectionUpdateError(Exception):
277 class ResumeCacheConflict(Exception):
281 class ArvPutArgumentConflict(Exception):
285 class ArvPutUploadIsPending(Exception):
289 class ArvPutUploadNotPending(Exception):
293 class FileUploadList(list):
294 def __init__(self, dry_run=False):
296 self.dry_run = dry_run
298 def append(self, other):
300 raise ArvPutUploadIsPending()
301 super(FileUploadList, self).append(other)
304 class ResumeCache(object):
305 CACHE_DIR = '.cache/arvados/arv-put'
307 def __init__(self, file_spec):
308 self.cache_file = open(file_spec, 'a+')
309 self._lock_file(self.cache_file)
310 self.filename = self.cache_file.name
313 def make_path(cls, args):
315 md5.update(arvados.config.get('ARVADOS_API_HOST', '!nohost').encode())
316 realpaths = sorted(os.path.realpath(path) for path in args.paths)
317 md5.update(b'\0'.join([p.encode() for p in realpaths]))
318 if any(os.path.isdir(path) for path in realpaths):
321 md5.update(args.filename.encode())
323 arv_cmd.make_home_conf_dir(cls.CACHE_DIR, 0o700, 'raise'),
326 def _lock_file(self, fileobj):
328 fcntl.flock(fileobj, fcntl.LOCK_EX | fcntl.LOCK_NB)
330 raise ResumeCacheConflict("{} locked".format(fileobj.name))
333 self.cache_file.seek(0)
334 return json.load(self.cache_file)
336 def check_cache(self, api_client=None, num_retries=0):
341 if "_finished_streams" in state and len(state["_finished_streams"]) > 0:
342 locator = state["_finished_streams"][0][1][0]
343 elif "_current_stream_locators" in state and len(state["_current_stream_locators"]) > 0:
344 locator = state["_current_stream_locators"][0]
345 if locator is not None:
346 kc = arvados.keep.KeepClient(api_client=api_client)
347 kc.head(locator, num_retries=num_retries)
348 except Exception as e:
353 def save(self, data):
355 new_cache_fd, new_cache_name = tempfile.mkstemp(
356 dir=os.path.dirname(self.filename))
357 self._lock_file(new_cache_fd)
358 new_cache = os.fdopen(new_cache_fd, 'r+')
359 json.dump(data, new_cache)
360 os.rename(new_cache_name, self.filename)
361 except (IOError, OSError, ResumeCacheConflict) as error:
363 os.unlink(new_cache_name)
364 except NameError: # mkstemp failed.
367 self.cache_file.close()
368 self.cache_file = new_cache
371 self.cache_file.close()
375 os.unlink(self.filename)
376 except OSError as error:
377 if error.errno != errno.ENOENT: # That's what we wanted anyway.
383 self.__init__(self.filename)
386 class ArvPutUploadJob(object):
387 CACHE_DIR = '.cache/arvados/arv-put'
389 'manifest' : None, # Last saved manifest checkpoint
390 'files' : {} # Previous run file list: {path : {size, mtime}}
393 def __init__(self, paths, resume=True, use_cache=True, reporter=None,
394 name=None, owner_uuid=None,
395 ensure_unique_name=False, num_retries=None,
396 put_threads=None, replication_desired=None,
397 filename=None, update_time=60.0, update_collection=None,
398 logger=logging.getLogger('arvados.arv_put'), dry_run=False,
399 follow_links=True, exclude_paths=[], exclude_names=None):
402 self.use_cache = use_cache
404 self.reporter = reporter
405 # This will set to 0 before start counting, if no special files are going
407 self.bytes_expected = None
408 self.bytes_written = 0
409 self.bytes_skipped = 0
411 self.owner_uuid = owner_uuid
412 self.ensure_unique_name = ensure_unique_name
413 self.num_retries = num_retries
414 self.replication_desired = replication_desired
415 self.put_threads = put_threads
416 self.filename = filename
417 self._state_lock = threading.Lock()
418 self._state = None # Previous run state (file list & manifest)
419 self._current_files = [] # Current run file list
420 self._cache_file = None
421 self._collection_lock = threading.Lock()
422 self._remote_collection = None # Collection being updated (if asked)
423 self._local_collection = None # Collection from previous run manifest
424 self._file_paths = set() # Files to be updated in remote collection
425 self._stop_checkpointer = threading.Event()
426 self._checkpointer = threading.Thread(target=self._update_task)
427 self._checkpointer.daemon = True
428 self._update_task_time = update_time # How many seconds wait between update runs
429 self._files_to_upload = FileUploadList(dry_run=dry_run)
430 self._upload_started = False
432 self.dry_run = dry_run
433 self._checkpoint_before_quit = True
434 self.follow_links = follow_links
435 self.exclude_paths = exclude_paths
436 self.exclude_names = exclude_names
438 if not self.use_cache and self.resume:
439 raise ArvPutArgumentConflict('resume cannot be True when use_cache is False')
441 # Check for obvious dry-run responses
442 if self.dry_run and (not self.use_cache or not self.resume):
443 raise ArvPutUploadIsPending()
445 # Load cached data if any and if needed
446 self._setup_state(update_collection)
448 # Build the upload file list, excluding requested files and counting the
449 # bytes expected to be uploaded.
450 self._build_upload_list()
452 def _build_upload_list(self):
454 Scan the requested paths to count file sizes, excluding files & dirs if requested
455 and building the upload file list.
457 # If there aren't special files to be read, reset total bytes count to zero
459 if not any(filter(lambda p: not (os.path.isfile(p) or os.path.isdir(p)),
461 self.bytes_expected = 0
463 for path in self.paths:
464 # Test for stdin first, in case some file named '-' exist
467 raise ArvPutUploadIsPending()
468 self._write_stdin(self.filename or 'stdin')
469 elif not os.path.exists(path):
470 raise PathDoesNotExistError("file or directory '{}' does not exist.".format(path))
471 elif os.path.isdir(path):
472 # Use absolute paths on cache index so CWD doesn't interfere
473 # with the caching logic.
475 path = os.path.abspath(path)
476 if orig_path[-1:] == os.sep:
477 # When passing a directory reference with a trailing slash,
478 # its contents should be uploaded directly to the
482 # When passing a directory reference with no trailing slash,
483 # upload the directory to the collection's root.
484 prefixdir = os.path.dirname(path)
486 for root, dirs, files in os.walk(path,
487 followlinks=self.follow_links):
488 root_relpath = os.path.relpath(root, path)
489 if root_relpath == '.':
491 # Exclude files/dirs by full path matching pattern
492 if self.exclude_paths:
493 dirs[:] = list(filter(
495 [pathname_match(os.path.join(root_relpath, d),
497 for pat in self.exclude_paths]),
501 [pathname_match(os.path.join(root_relpath, f),
503 for pat in self.exclude_paths]),
505 # Exclude files/dirs by name matching pattern
506 if self.exclude_names is not None:
507 dirs[:] = list(filter(lambda d: not self.exclude_names.match(d), dirs))
508 files = list(filter(lambda f: not self.exclude_names.match(f), files))
509 # Make os.walk()'s dir traversing order deterministic
513 filepath = os.path.join(root, f)
514 # Add its size to the total bytes count (if applicable)
515 if self.follow_links or (not os.path.islink(filepath)):
516 if self.bytes_expected is not None:
517 self.bytes_expected += os.path.getsize(filepath)
518 self._check_file(filepath,
519 os.path.join(root[len(prefixdir):], f))
521 filepath = os.path.abspath(path)
522 # Add its size to the total bytes count (if applicable)
523 if self.follow_links or (not os.path.islink(filepath)):
524 if self.bytes_expected is not None:
525 self.bytes_expected += os.path.getsize(filepath)
526 self._check_file(filepath,
527 self.filename or os.path.basename(path))
528 # If dry-mode is on, and got up to this point, then we should notify that
529 # there aren't any file to upload.
531 raise ArvPutUploadNotPending()
532 # Remove local_collection's files that don't exist locally anymore, so the
533 # bytes_written count is correct.
534 for f in self.collection_file_paths(self._local_collection,
536 if f != 'stdin' and f != self.filename and not f in self._file_paths:
537 self._local_collection.remove(f)
539 def start(self, save_collection):
541 Start supporting thread & file uploading
543 self._checkpointer.start()
545 # Update bytes_written from current local collection and
546 # report initial progress.
549 self._upload_started = True # Used by the update thread to start checkpointing
551 except (SystemExit, Exception) as e:
552 self._checkpoint_before_quit = False
553 # Log stack trace only when Ctrl-C isn't pressed (SIGINT)
554 # Note: We're expecting SystemExit instead of
555 # KeyboardInterrupt because we have a custom signal
556 # handler in place that raises SystemExit with the catched
558 if isinstance(e, PathDoesNotExistError):
559 # We aren't interested in the traceback for this case
561 elif not isinstance(e, SystemExit) or e.code != -2:
562 self.logger.warning("Abnormal termination:\n{}".format(
563 traceback.format_exc()))
567 # Stop the thread before doing anything else
568 self._stop_checkpointer.set()
569 self._checkpointer.join()
570 if self._checkpoint_before_quit:
571 # Commit all pending blocks & one last _update()
572 self._local_collection.manifest_text()
573 self._update(final=True)
575 self.save_collection()
577 self._cache_file.close()
579 def save_collection(self):
581 # Check if files should be updated on the remote collection.
582 for fp in self._file_paths:
583 remote_file = self._remote_collection.find(fp)
585 # File don't exist on remote collection, copy it.
586 self._remote_collection.copy(fp, fp, self._local_collection)
587 elif remote_file != self._local_collection.find(fp):
588 # A different file exist on remote collection, overwrite it.
589 self._remote_collection.copy(fp, fp, self._local_collection, overwrite=True)
591 # The file already exist on remote collection, skip it.
593 self._remote_collection.save(num_retries=self.num_retries)
595 self._local_collection.save_new(
596 name=self.name, owner_uuid=self.owner_uuid,
597 ensure_unique_name=self.ensure_unique_name,
598 num_retries=self.num_retries)
600 def destroy_cache(self):
603 os.unlink(self._cache_filename)
604 except OSError as error:
605 # That's what we wanted anyway.
606 if error.errno != errno.ENOENT:
608 self._cache_file.close()
610 def _collection_size(self, collection):
612 Recursively get the total size of the collection
615 for item in listvalues(collection):
616 if isinstance(item, arvados.collection.Collection) or isinstance(item, arvados.collection.Subcollection):
617 size += self._collection_size(item)
622 def _update_task(self):
624 Periodically called support task. File uploading is
625 asynchronous so we poll status from the collection.
627 while not self._stop_checkpointer.wait(1 if not self._upload_started else self._update_task_time):
630 def _update(self, final=False):
632 Update cached manifest text and report progress.
634 if self._upload_started:
635 with self._collection_lock:
636 self.bytes_written = self._collection_size(self._local_collection)
639 manifest = self._local_collection.manifest_text()
641 # Get the manifest text without comitting pending blocks
642 manifest = self._local_collection.manifest_text(strip=False,
646 with self._state_lock:
647 self._state['manifest'] = manifest
651 except Exception as e:
652 self.logger.error("Unexpected error trying to save cache file: {}".format(e))
654 self.bytes_written = self.bytes_skipped
655 # Call the reporter, if any
656 self.report_progress()
658 def report_progress(self):
659 if self.reporter is not None:
660 self.reporter(self.bytes_written, self.bytes_expected)
662 def _write_stdin(self, filename):
663 output = self._local_collection.open(filename, 'wb')
664 self._write(sys.stdin, output)
667 def _check_file(self, source, filename):
669 Check if this file needs to be uploaded
671 # Ignore symlinks when requested
672 if (not self.follow_links) and os.path.islink(source):
675 should_upload = False
676 new_file_in_cache = False
677 # Record file path for updating the remote collection before exiting
678 self._file_paths.add(filename)
680 with self._state_lock:
681 # If no previous cached data on this file, store it for an eventual
683 if source not in self._state['files']:
684 self._state['files'][source] = {
685 'mtime': os.path.getmtime(source),
686 'size' : os.path.getsize(source)
688 new_file_in_cache = True
689 cached_file_data = self._state['files'][source]
691 # Check if file was already uploaded (at least partially)
692 file_in_local_collection = self._local_collection.find(filename)
694 # If not resuming, upload the full file.
697 # New file detected from last run, upload it.
698 elif new_file_in_cache:
700 # Local file didn't change from last run.
701 elif cached_file_data['mtime'] == os.path.getmtime(source) and cached_file_data['size'] == os.path.getsize(source):
702 if not file_in_local_collection:
703 # File not uploaded yet, upload it completely
705 elif file_in_local_collection.permission_expired():
706 # Permission token expired, re-upload file. This will change whenever
707 # we have a API for refreshing tokens.
709 self._local_collection.remove(filename)
710 elif cached_file_data['size'] == file_in_local_collection.size():
711 # File already there, skip it.
712 self.bytes_skipped += cached_file_data['size']
713 elif cached_file_data['size'] > file_in_local_collection.size():
714 # File partially uploaded, resume!
715 resume_offset = file_in_local_collection.size()
716 self.bytes_skipped += resume_offset
719 # Inconsistent cache, re-upload the file
721 self._local_collection.remove(filename)
722 self.logger.warning("Uploaded version of file '{}' is bigger than local version, will re-upload it from scratch.".format(source))
723 # Local file differs from cached data, re-upload it.
725 if file_in_local_collection:
726 self._local_collection.remove(filename)
731 self._files_to_upload.append((source, resume_offset, filename))
732 except ArvPutUploadIsPending:
733 # This could happen when running on dry-mode, close cache file to
734 # avoid locking issues.
735 self._cache_file.close()
738 def _upload_files(self):
739 for source, resume_offset, filename in self._files_to_upload:
740 with open(source, 'rb') as source_fd:
741 with self._state_lock:
742 self._state['files'][source]['mtime'] = os.path.getmtime(source)
743 self._state['files'][source]['size'] = os.path.getsize(source)
744 if resume_offset > 0:
745 # Start upload where we left off
746 output = self._local_collection.open(filename, 'ab')
747 source_fd.seek(resume_offset)
750 output = self._local_collection.open(filename, 'wb')
751 self._write(source_fd, output)
752 output.close(flush=False)
754 def _write(self, source_fd, output):
756 data = source_fd.read(arvados.config.KEEP_BLOCK_SIZE)
761 def _my_collection(self):
762 return self._remote_collection if self.update else self._local_collection
764 def _setup_state(self, update_collection):
766 Create a new cache file or load a previously existing one.
768 # Load an already existing collection for update
769 if update_collection and re.match(arvados.util.collection_uuid_pattern,
772 self._remote_collection = arvados.collection.Collection(update_collection)
773 except arvados.errors.ApiError as error:
774 raise CollectionUpdateError("Cannot read collection {} ({})".format(update_collection, error))
777 elif update_collection:
778 # Collection locator provided, but unknown format
779 raise CollectionUpdateError("Collection locator unknown: '{}'".format(update_collection))
782 # Set up cache file name from input paths.
784 md5.update(arvados.config.get('ARVADOS_API_HOST', '!nohost').encode())
785 realpaths = sorted(os.path.realpath(path) for path in self.paths)
786 md5.update(b'\0'.join([p.encode() for p in realpaths]))
788 md5.update(self.filename.encode())
789 cache_filename = md5.hexdigest()
790 cache_filepath = os.path.join(
791 arv_cmd.make_home_conf_dir(self.CACHE_DIR, 0o700, 'raise'),
793 if self.resume and os.path.exists(cache_filepath):
794 self.logger.info("Resuming upload from cache file {}".format(cache_filepath))
795 self._cache_file = open(cache_filepath, 'a+')
797 # --no-resume means start with a empty cache file.
798 self.logger.info("Creating new cache file at {}".format(cache_filepath))
799 self._cache_file = open(cache_filepath, 'w+')
800 self._cache_filename = self._cache_file.name
801 self._lock_file(self._cache_file)
802 self._cache_file.seek(0)
804 with self._state_lock:
807 self._state = json.load(self._cache_file)
808 if not set(['manifest', 'files']).issubset(set(self._state.keys())):
809 # Cache at least partially incomplete, set up new cache
810 self._state = copy.deepcopy(self.EMPTY_STATE)
812 # Cache file empty, set up new cache
813 self._state = copy.deepcopy(self.EMPTY_STATE)
815 self.logger.info("No cache usage requested for this run.")
816 # No cache file, set empty state
817 self._state = copy.deepcopy(self.EMPTY_STATE)
818 # Load the previous manifest so we can check if files were modified remotely.
819 self._local_collection = arvados.collection.Collection(self._state['manifest'], replication_desired=self.replication_desired, put_threads=self.put_threads)
821 def collection_file_paths(self, col, path_prefix='.'):
822 """Return a list of file paths by recursively go through the entire collection `col`"""
824 for name, item in listitems(col):
825 if isinstance(item, arvados.arvfile.ArvadosFile):
826 file_paths.append(os.path.join(path_prefix, name))
827 elif isinstance(item, arvados.collection.Subcollection):
828 new_prefix = os.path.join(path_prefix, name)
829 file_paths += self.collection_file_paths(item, path_prefix=new_prefix)
832 def _lock_file(self, fileobj):
834 fcntl.flock(fileobj, fcntl.LOCK_EX | fcntl.LOCK_NB)
836 raise ResumeCacheConflict("{} locked".format(fileobj.name))
838 def _save_state(self):
840 Atomically save current state into cache.
842 with self._state_lock:
843 # We're not using copy.deepcopy() here because it's a lot slower
844 # than json.dumps(), and we're already needing JSON format to be
846 state = json.dumps(self._state)
848 new_cache = tempfile.NamedTemporaryFile(
850 dir=os.path.dirname(self._cache_filename), delete=False)
851 self._lock_file(new_cache)
852 new_cache.write(state)
855 os.rename(new_cache.name, self._cache_filename)
856 except (IOError, OSError, ResumeCacheConflict) as error:
857 self.logger.error("There was a problem while saving the cache file: {}".format(error))
859 os.unlink(new_cache_name)
860 except NameError: # mkstemp failed.
863 self._cache_file.close()
864 self._cache_file = new_cache
866 def collection_name(self):
867 return self._my_collection().api_response()['name'] if self._my_collection().api_response() else None
869 def manifest_locator(self):
870 return self._my_collection().manifest_locator()
872 def portable_data_hash(self):
873 pdh = self._my_collection().portable_data_hash()
874 m = self._my_collection().stripped_manifest().encode()
875 local_pdh = '{}+{}'.format(hashlib.md5(m).hexdigest(), len(m))
877 logger.warning("\n".join([
878 "arv-put: API server provided PDH differs from local manifest.",
879 " This should not happen; showing API server version."]))
882 def manifest_text(self, stream_name=".", strip=False, normalize=False):
883 return self._my_collection().manifest_text(stream_name, strip, normalize)
885 def _datablocks_on_item(self, item):
887 Return a list of datablock locators, recursively navigating
888 through subcollections
890 if isinstance(item, arvados.arvfile.ArvadosFile):
893 return ["d41d8cd98f00b204e9800998ecf8427e+0"]
896 for segment in item.segments():
897 loc = segment.locator
900 elif isinstance(item, arvados.collection.Collection):
901 l = [self._datablocks_on_item(x) for x in listvalues(item)]
902 # Fast list flattener method taken from:
903 # http://stackoverflow.com/questions/952914/making-a-flat-list-out-of-list-of-lists-in-python
904 return [loc for sublist in l for loc in sublist]
908 def data_locators(self):
909 with self._collection_lock:
910 # Make sure all datablocks are flushed before getting the locators
911 self._my_collection().manifest_text()
912 datablocks = self._datablocks_on_item(self._my_collection())
915 _machine_format = "{} {}: {{}} written {{}} total\n".format(sys.argv[0],
918 # Simulate glob.glob() matching behavior without the need to scan the filesystem
919 # Note: fnmatch() doesn't work correctly when used with pathnames. For example the
920 # pattern 'tests/*.py' will match 'tests/run_test.py' and also 'tests/subdir/run_test.py',
921 # so instead we're using it on every path component.
922 def pathname_match(pathname, pattern):
923 name = pathname.split(os.sep)
924 # Fix patterns like 'some/subdir/' or 'some//subdir'
925 pat = [x for x in pattern.split(os.sep) if x != '']
926 if len(name) != len(pat):
928 for i in range(len(name)):
929 if not fnmatch.fnmatch(name[i], pat[i]):
933 def machine_progress(bytes_written, bytes_expected):
934 return _machine_format.format(
935 bytes_written, -1 if (bytes_expected is None) else bytes_expected)
937 def human_progress(bytes_written, bytes_expected):
939 return "\r{}M / {}M {:.1%} ".format(
940 bytes_written >> 20, bytes_expected >> 20,
941 float(bytes_written) / bytes_expected)
943 return "\r{} ".format(bytes_written)
945 def progress_writer(progress_func, outfile=sys.stderr):
946 def write_progress(bytes_written, bytes_expected):
947 outfile.write(progress_func(bytes_written, bytes_expected))
948 return write_progress
950 def exit_signal_handler(sigcode, frame):
953 def desired_project_uuid(api_client, project_uuid, num_retries):
955 query = api_client.users().current()
956 elif arvados.util.user_uuid_pattern.match(project_uuid):
957 query = api_client.users().get(uuid=project_uuid)
958 elif arvados.util.group_uuid_pattern.match(project_uuid):
959 query = api_client.groups().get(uuid=project_uuid)
961 raise ValueError("Not a valid project UUID: {}".format(project_uuid))
962 return query.execute(num_retries=num_retries)['uuid']
964 def main(arguments=None, stdout=sys.stdout, stderr=sys.stderr):
967 logger = logging.getLogger('arvados.arv_put')
968 logger.setLevel(logging.INFO)
969 args = parse_arguments(arguments)
971 if api_client is None:
972 api_client = arvados.api('v1')
974 # Determine the name to use
976 if args.stream or args.raw:
977 logger.error("Cannot use --name with --stream or --raw")
979 elif args.update_collection:
980 logger.error("Cannot use --name with --update-collection")
982 collection_name = args.name
984 collection_name = "Saved at {} by {}@{}".format(
985 datetime.datetime.utcnow().strftime("%Y-%m-%d %H:%M:%S UTC"),
986 pwd.getpwuid(os.getuid()).pw_name,
987 socket.gethostname())
989 if args.project_uuid and (args.stream or args.raw):
990 logger.error("Cannot use --project-uuid with --stream or --raw")
993 # Determine the parent project
995 project_uuid = desired_project_uuid(api_client, args.project_uuid,
997 except (apiclient_errors.Error, ValueError) as error:
1002 reporter = progress_writer(human_progress)
1003 elif args.batch_progress:
1004 reporter = progress_writer(machine_progress)
1008 # Setup exclude regex from all the --exclude arguments provided
1011 exclude_names = None
1012 if len(args.exclude) > 0:
1013 # We're supporting 2 kinds of exclusion patterns:
1014 # 1) --exclude '*.jpg' (file/dir name patterns, will only match
1016 # 2) --exclude 'foo/bar' (file/dir path patterns, will match the
1017 # entire path, and should be relative to
1018 # any input dir argument)
1019 for p in args.exclude:
1020 # Only relative paths patterns allowed
1021 if p.startswith(os.sep):
1022 logger.error("Cannot use absolute paths with --exclude")
1024 if os.path.dirname(p):
1025 # We don't support of path patterns with '.' or '..'
1026 p_parts = p.split(os.sep)
1027 if '.' in p_parts or '..' in p_parts:
1029 "Cannot use path patterns that include '.' or '..'")
1031 # Path search pattern
1032 exclude_paths.append(p)
1034 # Name-only search pattern
1035 name_patterns.append(p)
1036 # For name only matching, we can combine all patterns into a single regexp,
1037 # for better performance.
1038 exclude_names = re.compile('|'.join(
1039 [fnmatch.translate(p) for p in name_patterns]
1040 )) if len(name_patterns) > 0 else None
1041 # Show the user the patterns to be used, just in case they weren't specified inside
1042 # quotes and got changed by the shell expansion.
1043 logger.info("Exclude patterns: {}".format(args.exclude))
1045 # If this is used by a human, and there's at least one directory to be
1046 # uploaded, the expected bytes calculation can take a moment.
1047 if args.progress and any([os.path.isdir(f) for f in args.paths]):
1048 logger.info("Calculating upload size, this could take some time...")
1050 writer = ArvPutUploadJob(paths = args.paths,
1051 resume = args.resume,
1052 use_cache = args.use_cache,
1053 filename = args.filename,
1054 reporter = reporter,
1055 num_retries = args.retries,
1056 replication_desired = args.replication,
1057 put_threads = args.threads,
1058 name = collection_name,
1059 owner_uuid = project_uuid,
1060 ensure_unique_name = True,
1061 update_collection = args.update_collection,
1063 dry_run=args.dry_run,
1064 follow_links=args.follow_links,
1065 exclude_paths=exclude_paths,
1066 exclude_names=exclude_names)
1067 except ResumeCacheConflict:
1068 logger.error("\n".join([
1069 "arv-put: Another process is already uploading this data.",
1070 " Use --no-cache if this is really what you want."]))
1072 except CollectionUpdateError as error:
1073 logger.error("\n".join([
1074 "arv-put: %s" % str(error)]))
1076 except ArvPutUploadIsPending:
1077 # Dry run check successful, return proper exit code.
1079 except ArvPutUploadNotPending:
1080 # No files pending for upload
1082 except PathDoesNotExistError as error:
1083 logger.error("\n".join([
1084 "arv-put: %s" % str(error)]))
1087 # Install our signal handler for each code in CAUGHT_SIGNALS, and save
1089 orig_signal_handlers = {sigcode: signal.signal(sigcode, exit_signal_handler)
1090 for sigcode in CAUGHT_SIGNALS}
1092 if not args.dry_run and not args.update_collection and args.resume and writer.bytes_written > 0:
1093 logger.warning("\n".join([
1094 "arv-put: Resuming previous upload from last checkpoint.",
1095 " Use the --no-resume option to start over."]))
1097 if not args.dry_run:
1098 writer.report_progress()
1101 writer.start(save_collection=not(args.stream or args.raw))
1102 except arvados.errors.ApiError as error:
1103 logger.error("\n".join([
1104 "arv-put: %s" % str(error)]))
1107 if args.progress: # Print newline to split stderr from stdout for humans.
1112 output = writer.manifest_text(normalize=True)
1114 output = writer.manifest_text()
1116 output = ','.join(writer.data_locators())
1119 if args.update_collection:
1120 logger.info("Collection updated: '{}'".format(writer.collection_name()))
1122 logger.info("Collection saved as '{}'".format(writer.collection_name()))
1123 if args.portable_data_hash:
1124 output = writer.portable_data_hash()
1126 output = writer.manifest_locator()
1127 except apiclient_errors.Error as error:
1129 "arv-put: Error creating Collection on project: {}.".format(
1133 # Print the locator (uuid) of the new collection.
1135 status = status or 1
1137 stdout.write(output)
1138 if not output.endswith('\n'):
1141 for sigcode, orig_handler in listitems(orig_signal_handlers):
1142 signal.signal(sigcode, orig_handler)
1151 if __name__ == '__main__':