X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/9e9506f9acce0d11a82f7abcea31680761d6427b..6ac59725fedc0a958d60d85740b47deb25addb5b:/sdk/python/arvados/commands/keepdocker.py diff --git a/sdk/python/arvados/commands/keepdocker.py b/sdk/python/arvados/commands/keepdocker.py index 5be4ef4e5e..933fd77dd7 100644 --- a/sdk/python/arvados/commands/keepdocker.py +++ b/sdk/python/arvados/commands/keepdocker.py @@ -9,7 +9,7 @@ import subprocess import sys import tarfile import tempfile -import textwrap +import _strptime from collections import namedtuple from stat import * @@ -48,7 +48,7 @@ keepdocker_parser.add_argument( # --progress/--no-progress/--batch-progress and --resume/--no-resume. arg_parser = argparse.ArgumentParser( description="Upload or list Docker images in Arvados", - parents=[keepdocker_parser, arv_put.run_opts]) + parents=[keepdocker_parser, arv_put.run_opts, arv_cmd.retry_opt]) class DockerError(Exception): pass @@ -117,7 +117,8 @@ def stat_cache_name(image_file): return getattr(image_file, 'name', image_file) + '.stat' def pull_image(image_name, image_tag): - check_docker(popen_docker(['pull', '-t', image_tag, image_name]), "pull") + check_docker(popen_docker(['pull', '{}:{}'.format(image_name, image_tag)]), + "pull") def save_image(image_hash, image_file): # Save the specified Docker image to image_file, then try to save its @@ -153,9 +154,10 @@ def prep_image_file(filename): image_file = open(file_path, 'w+b' if need_save else 'rb') return image_file, need_save -def make_link(link_class, link_name, **link_attrs): +def make_link(api_client, num_retries, link_class, link_name, **link_attrs): link_attrs.update({'link_class': link_class, 'name': link_name}) - return arvados.api('v1').links().create(body=link_attrs).execute() + return api_client.links().create(body=link_attrs).execute( + num_retries=num_retries) def ptimestamp(t): s = t.split(".") @@ -163,8 +165,26 @@ def ptimestamp(t): t = s[0] + s[1][-1:] return datetime.datetime.strptime(t, "%Y-%m-%dT%H:%M:%SZ") -def list_images_in_arv(): - existing_links = arvados.api('v1').links().list(filters=[['link_class', 'in', ['docker_image_hash', 'docker_image_repo+tag']]]).execute()['items'] +def list_images_in_arv(api_client, num_retries, image_name=None, image_tag=None): + """List all Docker images known to the api_client with image_name and + image_tag. If no image_name is given, defaults to listing all + Docker images. + + Returns a list of tuples representing matching Docker images, + sorted in preference order (i.e. the first collection in the list + is the one that the API server would use). Each tuple is a + (collection_uuid, collection_info) pair, where collection_info is + a dict with fields "dockerhash", "repo", "tag", and "timestamp". + + """ + docker_image_filters = [['link_class', 'in', ['docker_image_hash', 'docker_image_repo+tag']]] + if image_name: + image_link_name = "{}:{}".format(image_name, image_tag or 'latest') + docker_image_filters.append(['name', '=', image_link_name]) + + existing_links = api_client.links().list( + filters=docker_image_filters + ).execute(num_retries=num_retries)['items'] images = {} for link in existing_links: collection_uuid = link["head_uuid"] @@ -188,18 +208,18 @@ def list_images_in_arv(): else: images[collection_uuid]["timestamp"] = ptimestamp(link["created_at"]) - st = sorted(images.items(), lambda a, b: cmp(b[1]["timestamp"], a[1]["timestamp"])) + return sorted(images.items(), lambda a, b: cmp(b[1]["timestamp"], a[1]["timestamp"])) - fmt = "{:30} {:10} {:12} {:29} {:20}" - print fmt.format("REPOSITORY", "TAG", "IMAGE ID", "COLLECTION", "CREATED") - for i, j in st: - print(fmt.format(j["repo"], j["tag"], j["dockerhash"][0:11], i, j["timestamp"].strftime("%c"))) def main(arguments=None): args = arg_parser.parse_args(arguments) + api = arvados.api('v1') if args.image is None or args.image == 'images': - list_images_in_arv() + fmt = "{:30} {:10} {:12} {:29} {:20}" + print fmt.format("REPOSITORY", "TAG", "IMAGE ID", "COLLECTION", "CREATED") + for i, j in list_images_in_arv(api, args.retries): + print(fmt.format(j["repo"], j["tag"], j["dockerhash"][0:12], i, j["timestamp"].strftime("%c"))) sys.exit(0) # Pull the image if requested, unless the image is specified as a hash @@ -213,66 +233,79 @@ def main(arguments=None): print >>sys.stderr, "arv-keepdocker:", error.message sys.exit(1) - image_repo_tag = '{}:{}'.format(args.image, args.tag) + image_repo_tag = '{}:{}'.format(args.image, args.tag) if not image_hash.startswith(args.image.lower()) else None if args.name is None: - collection_name = 'Docker image {} {}'.format(image_repo_tag, image_hash[0:11]) + if image_repo_tag: + collection_name = 'Docker image {} {}'.format(image_repo_tag, image_hash[0:12]) + else: + collection_name = 'Docker image {}'.format(image_hash[0:12]) else: collection_name = args.name - api = arvados.api('v1') - if not args.force: # Check if this image is already in Arvados. # Project where everything should be owned - parent_project_uuid = args.project_uuid if args.project_uuid else api.users().current().execute()['uuid'] + if args.project_uuid: + parent_project_uuid = args.project_uuid + else: + parent_project_uuid = api.users().current().execute( + num_retries=args.retries)['uuid'] # Find image hash tags existing_links = api.links().list( filters=[['link_class', '=', 'docker_image_hash'], - ['name', '=', image_hash]]).execute()['items'] + ['name', '=', image_hash]] + ).execute(num_retries=args.retries)['items'] if existing_links: # get readable collections collections = api.collections().list( - filters=[['uuid', 'in', [link['head_uuid'] for link in existing_links]]], - select=["uuid", "owner_uuid", "name", "manifest_text"]).execute()['items'] + filters=[['uuid', 'in', [link['head_uuid'] for link in existing_links]]], + select=["uuid", "owner_uuid", "name", "manifest_text"] + ).execute(num_retries=args.retries)['items'] if collections: # check for repo+tag links on these collections - existing_repo_tag = api.links().list( + existing_repo_tag = (api.links().list( filters=[['link_class', '=', 'docker_image_repo+tag'], ['name', '=', image_repo_tag], - ['head_uuid', 'in', collections]]).execute()['items'] + ['head_uuid', 'in', collections]] + ).execute(num_retries=args.retries)['items']) if image_repo_tag else [] # Filter on elements owned by the parent project owned_col = [c for c in collections if c['owner_uuid'] == parent_project_uuid] - owned_img = [c for c in existing_links if c['owner_uuid'] == parent_project_uuid] - owned_rep = [c for c in existing_repo_tag if c['owner_uuid'] == parent_project_uuid] + owned_img = [c for c in existing_links if c['owner_uuid'] == parent_project_uuid] + owned_rep = [c for c in existing_repo_tag if c['owner_uuid'] == parent_project_uuid] if owned_col: # already have a collection owned by this project coll_uuid = owned_col[0]['uuid'] else: # create new collection owned by the project - coll_uuid = api.collections().create(body={"manifest_text": collections[0]['manifest_text'], - "name": collection_name, - "owner_uuid": parent_project_uuid}).execute()['uuid'] - - link_base = {'owner_uuid': parent_project_uuid, + coll_uuid = api.collections().create( + body={"manifest_text": collections[0]['manifest_text'], + "name": collection_name, + "owner_uuid": parent_project_uuid}, + ensure_unique_name=True + ).execute(num_retries=args.retries)['uuid'] + + link_base = {'owner_uuid': parent_project_uuid, 'head_uuid': coll_uuid } if not owned_img: # create image link owned by the project - make_link('docker_image_hash', image_hash, **link_base) + make_link(api, args.retries, + 'docker_image_hash', image_hash, **link_base) - if not owned_rep: + if not owned_rep and image_repo_tag: # create repo+tag link owned by the project - make_link('docker_image_repo+tag', image_repo_tag, **link_base) + make_link(api, args.retries, 'docker_image_repo+tag', + image_repo_tag, **link_base) print(coll_uuid) - sys.exit(0) + sys.exit(0) # Open a file for the saved image, and write it if needed. outfile_name = '{}.tar'.format(image_hash) @@ -303,10 +336,10 @@ def main(arguments=None): if args.project_uuid is not None: link_base['owner_uuid'] = args.project_uuid - make_link('docker_image_hash', image_hash, **link_base) - if not image_hash.startswith(args.image.lower()): - make_link('docker_image_repo+tag', image_repo_tag, - **link_base) + make_link(api, args.retries, 'docker_image_hash', image_hash, **link_base) + if image_repo_tag: + make_link(api, args.retries, + 'docker_image_repo+tag', image_repo_tag, **link_base) # Clean up. image_file.close()