Merge branch '8784-dir-listings'
[arvados.git] / sdk / python / arvados / commands / keepdocker.py
index e48a6d15472cc2c90cd4af7e35251b653aa87cce..2c1ec738d1b797415532a6d7f88e9974a758294d 100644 (file)
@@ -1,15 +1,20 @@
-#!/usr/bin/env python
+# Copyright (C) The Arvados Authors. All rights reserved.
+#
+# SPDX-License-Identifier: Apache-2.0
 
+from builtins import next
 import argparse
 import collections
 import datetime
 import errno
 import json
 import os
+import re
 import subprocess
 import sys
 import tarfile
 import tempfile
+import shutil
 import _strptime
 
 from operator import itemgetter
@@ -19,7 +24,16 @@ import arvados
 import arvados.util
 import arvados.commands._util as arv_cmd
 import arvados.commands.put as arv_put
+from arvados.collection import CollectionReader
 import ciso8601
+import logging
+import arvados.config
+
+from arvados._version import __version__
+
+logger = logging.getLogger('arvados.keepdocker')
+logger.setLevel(logging.DEBUG if arvados.config.get('ARVADOS_DEBUG')
+                else logging.INFO)
 
 EARLIEST_DATETIME = datetime.datetime(datetime.MINYEAR, 1, 1, 0, 0, 0)
 STAT_CACHE_ERRORS = (IOError, OSError, ValueError)
@@ -28,9 +42,15 @@ DockerImage = collections.namedtuple(
     'DockerImage', ['repo', 'tag', 'hash', 'created', 'vsize'])
 
 keepdocker_parser = argparse.ArgumentParser(add_help=False)
+keepdocker_parser.add_argument(
+    '--version', action='version', version="%s %s" % (sys.argv[0], __version__),
+    help='Print version and exit.')
 keepdocker_parser.add_argument(
     '-f', '--force', action='store_true', default=False,
     help="Re-upload the image even if it already exists on the server")
+keepdocker_parser.add_argument(
+    '--force-image-format', action='store_true', default=False,
+    help="Proceed even if the image format is not supported by the server")
 
 _group = keepdocker_parser.add_mutually_exclusive_group()
 _group.add_argument(
@@ -76,6 +96,35 @@ def check_docker(proc, description):
         raise DockerError("docker {} returned status code {}".
                           format(description, proc.returncode))
 
+def docker_image_format(image_hash):
+    """Return the registry format ('v1' or 'v2') of the given image."""
+    cmd = popen_docker(['inspect', '--format={{.Id}}', image_hash],
+                        stdout=subprocess.PIPE)
+    try:
+        image_id = next(cmd.stdout).decode().strip()
+        if image_id.startswith('sha256:'):
+            return 'v2'
+        elif ':' not in image_id:
+            return 'v1'
+        else:
+            return 'unknown'
+    finally:
+        check_docker(cmd, "inspect")
+
+def docker_image_compatible(api, image_hash):
+    supported = api._rootDesc.get('dockerImageFormats', [])
+    if not supported:
+        logger.warning("server does not specify supported image formats (see docker_image_formats in server config).")
+        return False
+
+    fmt = docker_image_format(image_hash)
+    if fmt in supported:
+        return True
+    else:
+        logger.error("image format is {!r} " \
+            "but server supports only {!r}".format(fmt, supported))
+        return False
+
 def docker_images():
     # Yield a DockerImage tuple for each installed image.
     list_proc = popen_docker(['images', '--no-trunc'], stdout=subprocess.PIPE)
@@ -269,7 +318,7 @@ def list_images_in_arv(api_client, num_retries, image_name=None, image_tag=None)
     # and add image listings for them, retaining the API server preference
     # sorting.
     images_start_size = len(images)
-    for collection_uuid, link in hash_link_map.iteritems():
+    for collection_uuid, link in hash_link_map.items():
         if not seen_image_names[collection_uuid]:
             images.append(_new_image_listing(link, link['name']))
     if len(images) > images_start_size:
@@ -283,15 +332,30 @@ def list_images_in_arv(api_client, num_retries, image_name=None, image_tag=None)
     return [(image['collection'], image) for image in images
             if image['collection'] in existing_coll_uuids]
 
-def main(arguments=None):
+def items_owned_by(owner_uuid, arv_items):
+    return (item for item in arv_items if item['owner_uuid'] == owner_uuid)
+
+def _uuid2pdh(api, uuid):
+    return api.collections().list(
+        filters=[['uuid', '=', uuid]],
+        select=['portable_data_hash'],
+    ).execute()['items'][0]['portable_data_hash']
+
+def main(arguments=None, stdout=sys.stdout):
     args = arg_parser.parse_args(arguments)
     api = arvados.api('v1')
 
     if args.image is None or args.image == 'images':
-        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")))
+        fmt = "{:30}  {:10}  {:12}  {:29}  {:20}\n"
+        stdout.write(fmt.format("REPOSITORY", "TAG", "IMAGE ID", "COLLECTION", "CREATED"))
+        try:
+            for i, j in list_images_in_arv(api, args.retries):
+                stdout.write(fmt.format(j["repo"], j["tag"], j["dockerhash"][0:12], i, j["timestamp"].strftime("%c")))
+        except IOError as e:
+            if e.errno == errno.EPIPE:
+                pass
+            else:
+                raise
         sys.exit(0)
 
     # Pull the image if requested, unless the image is specified as a hash
@@ -302,9 +366,17 @@ def main(arguments=None):
     try:
         image_hash = find_one_image_hash(args.image, args.tag)
     except DockerError as error:
-        print >>sys.stderr, "arv-keepdocker:", error.message
+        logger.error(error.message)
         sys.exit(1)
 
+    if not docker_image_compatible(api, image_hash):
+        if args.force_image_format:
+            logger.warning("forcing incompatible image")
+        else:
+            logger.error("refusing to store " \
+                "incompatible format (use --force-image-format to override)")
+            sys.exit(1)
+
     image_repo_tag = '{}:{}'.format(args.image, args.tag) if not image_hash.startswith(args.image.lower()) else None
 
     if args.name is None:
@@ -326,10 +398,10 @@ def main(arguments=None):
                 num_retries=args.retries)['uuid']
 
         # Find image hash tags
-        existing_links = api.links().list(
+        existing_links = _get_docker_links(
+            api, args.retries,
             filters=[['link_class', '=', 'docker_image_hash'],
-                     ['name', '=', image_hash]]
-            ).execute(num_retries=args.retries)['items']
+                     ['name', '=', image_hash]])
         if existing_links:
             # get readable collections
             collections = api.collections().list(
@@ -339,21 +411,18 @@ def main(arguments=None):
 
             if collections:
                 # check for repo+tag links on these collections
-                existing_repo_tag = (api.links().list(
-                    filters=[['link_class', '=', 'docker_image_repo+tag'],
-                             ['name', '=', image_repo_tag],
-                             ['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]
-
-                if owned_col:
-                    # already have a collection owned by this project
-                    coll_uuid = owned_col[0]['uuid']
+                if image_repo_tag:
+                    existing_repo_tag = _get_docker_links(
+                        api, args.retries,
+                        filters=[['link_class', '=', 'docker_image_repo+tag'],
+                                 ['name', '=', image_repo_tag],
+                                 ['head_uuid', 'in', [c["uuid"] for c in collections]]])
                 else:
+                    existing_repo_tag = []
+
+                try:
+                    coll_uuid = next(items_owned_by(parent_project_uuid, collections))['uuid']
+                except StopIteration:
                     # create new collection owned by the project
                     coll_uuid = api.collections().create(
                         body={"manifest_text": collections[0]['manifest_text'],
@@ -363,19 +432,20 @@ def main(arguments=None):
                         ).execute(num_retries=args.retries)['uuid']
 
                 link_base = {'owner_uuid': parent_project_uuid,
-                             'head_uuid':  coll_uuid }
+                             'head_uuid':  coll_uuid,
+                             'properties': existing_links[0]['properties']}
 
-                if not owned_img:
+                if not any(items_owned_by(parent_project_uuid, existing_links)):
                     # create image link owned by the project
                     make_link(api, args.retries,
                               'docker_image_hash', image_hash, **link_base)
 
-                if not owned_rep and image_repo_tag:
+                if image_repo_tag and not any(items_owned_by(parent_project_uuid, existing_repo_tag)):
                     # create repo+tag link owned by the project
                     make_link(api, args.retries, 'docker_image_repo+tag',
                               image_repo_tag, **link_base)
 
-                print(coll_uuid)
+                stdout.write(coll_uuid + "\n")
 
                 sys.exit(0)
 
@@ -393,12 +463,17 @@ def main(arguments=None):
         put_args += ['--name', collection_name]
 
     coll_uuid = arv_put.main(
-        put_args + ['--filename', outfile_name, image_file.name]).strip()
+        put_args + ['--filename', outfile_name, image_file.name], stdout=stdout).strip()
 
     # Read the image metadata and make Arvados links from it.
     image_file.seek(0)
     image_tar = tarfile.open(fileobj=image_file)
-    json_file = image_tar.extractfile(image_tar.getmember(image_hash + '/json'))
+    image_hash_type, _, raw_image_hash = image_hash.rpartition(':')
+    if image_hash_type:
+        json_filename = raw_image_hash + '.json'
+    else:
+        json_filename = raw_image_hash + '/json'
+    json_file = image_tar.extractfile(image_tar.getmember(json_filename))
     image_metadata = json.load(json_file)
     json_file.close()
     image_tar.close()