X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/18ddae75beb91b0792ad34a9b4996316715e67b7..4554374c672ee56608c9ddbd6a48486fe20c90d1:/sdk/python/arvados/commands/arv_copy.py diff --git a/sdk/python/arvados/commands/arv_copy.py b/sdk/python/arvados/commands/arv_copy.py index 308bd78028..0ba3f0a483 100755 --- a/sdk/python/arvados/commands/arv_copy.py +++ b/sdk/python/arvados/commands/arv_copy.py @@ -1,4 +1,6 @@ -#! /usr/bin/env python +# Copyright (C) The Arvados Authors. All rights reserved. +# +# SPDX-License-Identifier: Apache-2.0 # arv-copy [--recursive] [--no-recursive] object-uuid src dst # @@ -16,7 +18,14 @@ # instances src and dst. If either of these files is not found, # arv-copy will issue an error. +from __future__ import division +from future import standard_library +from future.utils import listvalues +standard_library.install_aliases() +from past.builtins import basestring +from builtins import object import argparse +import contextlib import getpass import os import re @@ -24,6 +33,7 @@ import shutil import sys import logging import tempfile +import urllib.parse import arvados import arvados.config @@ -31,6 +41,12 @@ import arvados.keep import arvados.util import arvados.commands._util as arv_cmd import arvados.commands.keepdocker +import ruamel.yaml as yaml + +from arvados.api import OrderedJsonModel +from arvados._version import __version__ + +COMMIT_HASH_RE = re.compile(r'^[0-9a-f]{1,40}$') logger = logging.getLogger('arvados.arv-copy') @@ -46,9 +62,18 @@ local_repo_dir = {} # destination collection UUIDs. collections_copied = {} +# Set of (repository, script_version) two-tuples of commits copied in git. +scripts_copied = set() + +# The owner_uuid of the object being copied +src_owner_uuid = None + def main(): copy_opts = argparse.ArgumentParser(add_help=False) + copy_opts.add_argument( + '--version', action='version', version="%s %s" % (sys.argv[0], __version__), + help='Print version and exit.') copy_opts.add_argument( '-v', '--verbose', dest='verbose', action='store_true', help='Verbose output.') @@ -61,12 +86,15 @@ def main(): copy_opts.add_argument( '-f', '--force', dest='force', action='store_true', help='Perform copy even if the object appears to exist at the remote destination.') + copy_opts.add_argument( + '--force-filters', action='store_true', default=False, + help="Copy pipeline template filters verbatim, even if they act differently on the destination cluster.") copy_opts.add_argument( '--src', dest='source_arvados', required=True, - help='The name of the source Arvados instance (required). May be either a pathname to a config file, or the basename of a file in $HOME/.config/arvados/instance_name.conf.') + help='The name of the source Arvados instance (required) - points at an Arvados config file. May be either a pathname to a config file, or (for example) "foo" as shorthand for $HOME/.config/arvados/foo.conf.') copy_opts.add_argument( '--dst', dest='destination_arvados', required=True, - help='The name of the destination Arvados instance (required). May be either a pathname to a config file, or the basename of a file in $HOME/.config/arvados/instance_name.conf.') + help='The name of the destination Arvados instance (required) - points at an Arvados config file. May be either a pathname to a config file, or (for example) "foo" as shorthand for $HOME/.config/arvados/foo.conf.') copy_opts.add_argument( '--recursive', dest='recursive', action='store_true', help='Recursively copy any dependencies for this object. (default)') @@ -79,6 +107,13 @@ def main(): copy_opts.add_argument( '--project-uuid', dest='project_uuid', help='The UUID of the project at the destination to which the pipeline should be copied.') + copy_opts.add_argument( + '--allow-git-http-src', action="store_true", + help='Allow cloning git repositories over insecure http') + copy_opts.add_argument( + '--allow-git-http-dst', action="store_true", + help='Allow pushing git repositories over insecure http') + copy_opts.add_argument( 'object_uuid', help='The UUID of the object to be copied.') @@ -86,7 +121,7 @@ def main(): copy_opts.set_defaults(recursive=True) parser = argparse.ArgumentParser( - description='Copy a pipeline instance, template or collection from one Arvados instance to another.', + description='Copy a pipeline instance, template, workflow, or collection from one Arvados instance to another.', parents=[copy_opts, arv_cmd.retry_opt]) args = parser.parse_args() @@ -105,21 +140,27 @@ def main(): # Identify the kind of object we have been given, and begin copying. t = uuid_type(src_arv, args.object_uuid) if t == 'Collection': + set_src_owner_uuid(src_arv.collections(), args.object_uuid, args) result = copy_collection(args.object_uuid, src_arv, dst_arv, args) elif t == 'PipelineInstance': + set_src_owner_uuid(src_arv.pipeline_instances(), args.object_uuid, args) result = copy_pipeline_instance(args.object_uuid, src_arv, dst_arv, args) elif t == 'PipelineTemplate': + set_src_owner_uuid(src_arv.pipeline_templates(), args.object_uuid, args) result = copy_pipeline_template(args.object_uuid, src_arv, dst_arv, args) + elif t == 'Workflow': + set_src_owner_uuid(src_arv.workflows(), args.object_uuid, args) + result = copy_workflow(args.object_uuid, src_arv, dst_arv, args) else: abort("cannot copy object {} of type {}".format(args.object_uuid, t)) # Clean up any outstanding temp git repositories. - for d in local_repo_dir.values(): + for d in listvalues(local_repo_dir): shutil.rmtree(d, ignore_errors=True) # If no exception was thrown and the response does not have an @@ -132,6 +173,11 @@ def main(): logger.info("Success: created copy with uuid {}".format(result['uuid'])) exit(0) +def set_src_owner_uuid(resource, uuid, args): + global src_owner_uuid + c = resource.get(uuid=uuid).execute(num_retries=args.retries) + src_owner_uuid = c.get("owner_uuid") + # api_for_instance(instance_name) # # Creates an API client for the Arvados instance identified by @@ -166,11 +212,19 @@ def api_for_instance(instance_name): client = arvados.api('v1', host=cfg['ARVADOS_API_HOST'], token=cfg['ARVADOS_API_TOKEN'], - insecure=api_is_insecure) + insecure=api_is_insecure, + model=OrderedJsonModel()) else: abort('need ARVADOS_API_HOST and ARVADOS_API_TOKEN for {}'.format(instance_name)) return client +# Check if git is available +def check_git_availability(): + try: + arvados.util.run_command(['git', '--help']) + except Exception: + abort('git command is not available. Please ensure git is installed.') + # copy_pipeline_instance(pi_uuid, src, dst, args) # # Copies a pipeline instance identified by pi_uuid from src to dst. @@ -192,9 +246,11 @@ def api_for_instance(instance_name): # def copy_pipeline_instance(pi_uuid, src, dst, args): # Fetch the pipeline instance record. - pi = src.pipeline_instances().get(uuid=pi_uuid).execute() + pi = src.pipeline_instances().get(uuid=pi_uuid).execute(num_retries=args.retries) if args.recursive: + check_git_availability() + if not args.dst_git_repo: abort('--dst-git-repo is required when copying a pipeline recursively.') # Copy the pipeline template and save the copied template. @@ -204,7 +260,7 @@ def copy_pipeline_instance(pi_uuid, src, dst, args): # Copy input collections, docker images and git repos. pi = copy_collections(pi, src, dst, args) - copy_git_repos(pi, src, dst, args.dst_git_repo) + copy_git_repos(pi, src, dst, args.dst_git_repo, args) copy_docker_images(pi, src, dst, args) # Update the fields of the pipeline instance with the copied @@ -228,9 +284,97 @@ def copy_pipeline_instance(pi_uuid, src, dst, args): del pi['uuid'] - new_pi = dst.pipeline_instances().create(body=pi, ensure_unique_name=True).execute() + new_pi = dst.pipeline_instances().create(body=pi, ensure_unique_name=True).execute(num_retries=args.retries) return new_pi +def filter_iter(arg): + """Iterate a filter string-or-list. + + Pass in a filter field that can either be a string or list. + This will iterate elements as if the field had been written as a list. + """ + if isinstance(arg, basestring): + return iter((arg,)) + else: + return iter(arg) + +def migrate_repository_filter(repo_filter, src_repository, dst_repository): + """Update a single repository filter in-place for the destination. + + If the filter checks that the repository is src_repository, it is + updated to check that the repository is dst_repository. If it does + anything else, this function raises ValueError. + """ + if src_repository is None: + raise ValueError("component does not specify a source repository") + elif dst_repository is None: + raise ValueError("no destination repository specified to update repository filter") + elif repo_filter[1:] == ['=', src_repository]: + repo_filter[2] = dst_repository + elif repo_filter[1:] == ['in', [src_repository]]: + repo_filter[2] = [dst_repository] + else: + raise ValueError("repository filter is not a simple source match") + +def migrate_script_version_filter(version_filter): + """Update a single script_version filter in-place for the destination. + + Currently this function checks that all the filter operands are Git + commit hashes. If they're not, it raises ValueError to indicate that + the filter is not portable. It could be extended to make other + transformations in the future. + """ + if not all(COMMIT_HASH_RE.match(v) for v in filter_iter(version_filter[2])): + raise ValueError("script_version filter is not limited to commit hashes") + +def attr_filtered(filter_, *attr_names): + """Return True if filter_ applies to any of attr_names, else False.""" + return any((name == 'any') or (name in attr_names) + for name in filter_iter(filter_[0])) + +@contextlib.contextmanager +def exception_handler(handler, *exc_types): + """If any exc_types are raised in the block, call handler on the exception.""" + try: + yield + except exc_types as error: + handler(error) + +def migrate_components_filters(template_components, dst_git_repo): + """Update template component filters in-place for the destination. + + template_components is a dictionary of components in a pipeline template. + This method walks over each component's filters, and updates them to have + identical semantics on the destination cluster. It returns a list of + error strings that describe what filters could not be updated safely. + + dst_git_repo is the name of the destination Git repository, which can + be None if that is not known. + """ + errors = [] + for cname, cspec in template_components.items(): + def add_error(errmsg): + errors.append("{}: {}".format(cname, errmsg)) + if not isinstance(cspec, dict): + add_error("value is not a component definition") + continue + src_repository = cspec.get('repository') + filters = cspec.get('filters', []) + if not isinstance(filters, list): + add_error("filters are not a list") + continue + for cfilter in filters: + if not (isinstance(cfilter, list) and (len(cfilter) == 3)): + add_error("malformed filter {!r}".format(cfilter)) + continue + if attr_filtered(cfilter, 'repository'): + with exception_handler(add_error, ValueError): + migrate_repository_filter(cfilter, src_repository, dst_git_repo) + if attr_filtered(cfilter, 'script_version'): + with exception_handler(add_error, ValueError): + migrate_script_version_filter(cfilter) + return errors + # copy_pipeline_template(pt_uuid, src, dst, args) # # Copies a pipeline template identified by pt_uuid from src to dst. @@ -245,14 +389,22 @@ def copy_pipeline_instance(pi_uuid, src, dst, args): # def copy_pipeline_template(pt_uuid, src, dst, args): # fetch the pipeline template from the source instance - pt = src.pipeline_templates().get(uuid=pt_uuid).execute() + pt = src.pipeline_templates().get(uuid=pt_uuid).execute(num_retries=args.retries) + + if not args.force_filters: + filter_errors = migrate_components_filters(pt['components'], args.dst_git_repo) + if filter_errors: + abort("Template filters cannot be copied safely. Use --force-filters to copy anyway.\n" + + "\n".join(filter_errors)) if args.recursive: + check_git_availability() + if not args.dst_git_repo: abort('--dst-git-repo is required when copying a pipeline recursively.') # Copy input collections, docker images and git repos. pt = copy_collections(pt, src, dst, args) - copy_git_repos(pt, src, dst, args.dst_git_repo) + copy_git_repos(pt, src, dst, args.dst_git_repo, args) copy_docker_images(pt, src, dst, args) pt['description'] = "Pipeline template copied from {}\n\n{}".format( @@ -263,7 +415,65 @@ def copy_pipeline_template(pt_uuid, src, dst, args): pt['owner_uuid'] = args.project_uuid - return dst.pipeline_templates().create(body=pt, ensure_unique_name=True).execute() + return dst.pipeline_templates().create(body=pt, ensure_unique_name=True).execute(num_retries=args.retries) + +# copy_workflow(wf_uuid, src, dst, args) +# +# Copies a workflow identified by wf_uuid from src to dst. +# +# If args.recursive is True, also copy any collections +# referenced in the workflow definition yaml. +# +# The owner_uuid of the new workflow is set to any given +# project_uuid or the user who copied the template. +# +# Returns the copied workflow object. +# +def copy_workflow(wf_uuid, src, dst, args): + # fetch the workflow from the source instance + wf = src.workflows().get(uuid=wf_uuid).execute(num_retries=args.retries) + + # copy collections and docker images + if args.recursive: + wf_def = yaml.safe_load(wf["definition"]) + if wf_def is not None: + locations = [] + docker_images = {} + graph = wf_def.get('$graph', None) + if graph is not None: + workflow_collections(graph, locations, docker_images) + else: + workflow_collections(wf_def, locations, docker_images) + + if locations: + copy_collections(locations, src, dst, args) + + for image in docker_images: + copy_docker_image(image, docker_images[image], src, dst, args) + + # copy the workflow itself + del wf['uuid'] + wf['owner_uuid'] = args.project_uuid + return dst.workflows().create(body=wf).execute(num_retries=args.retries) + +def workflow_collections(obj, locations, docker_images): + if isinstance(obj, dict): + loc = obj.get('location', None) + if loc is not None: + if loc.startswith("keep:"): + locations.append(loc[5:]) + + docker_image = obj.get('dockerImageId', None) or obj.get('dockerPull', None) + if docker_image is not None: + ds = docker_image.split(":", 1) + tag = ds[1] if len(ds)==2 else 'latest' + docker_images[ds[0]] = tag + + for x in obj: + workflow_collections(obj[x], locations, docker_images) + elif isinstance(obj, list): + for x in obj: + workflow_collections(x, locations, docker_images) # copy_collections(obj, src, dst, args) # @@ -301,13 +511,39 @@ def copy_collections(obj, src, dst, args): obj = arvados.util.portable_data_hash_pattern.sub(copy_collection_fn, obj) obj = arvados.util.collection_uuid_pattern.sub(copy_collection_fn, obj) return obj - elif type(obj) == dict: - return {v: copy_collections(obj[v], src, dst, args) for v in obj} - elif type(obj) == list: - return [copy_collections(v, src, dst, args) for v in obj] + elif isinstance(obj, dict): + return type(obj)((v, copy_collections(obj[v], src, dst, args)) + for v in obj) + elif isinstance(obj, list): + return type(obj)(copy_collections(v, src, dst, args) for v in obj) return obj -# copy_git_repos(p, src, dst, dst_repo) +def migrate_jobspec(jobspec, src, dst, dst_repo, args): + """Copy a job's script to the destination repository, and update its record. + + Given a jobspec dictionary, this function finds the referenced script from + src and copies it to dst and dst_repo. It also updates jobspec in place to + refer to names on the destination. + """ + repo = jobspec.get('repository') + if repo is None: + return + # script_version is the "script_version" parameter from the source + # component or job. If no script_version was supplied in the + # component or job, it is a mistake in the pipeline, but for the + # purposes of copying the repository, default to "master". + script_version = jobspec.get('script_version') or 'master' + script_key = (repo, script_version) + if script_key not in scripts_copied: + copy_git_repo(repo, src, dst, dst_repo, script_version, args) + scripts_copied.add(script_key) + jobspec['repository'] = dst_repo + repo_dir = local_repo_dir[repo] + for version_key in ['script_version', 'supplied_script_version']: + if version_key in jobspec: + jobspec[version_key] = git_rev_parse(jobspec[version_key], repo_dir) + +# copy_git_repos(p, src, dst, dst_repo, args) # # Copies all git repositories referenced by pipeline instance or # template 'p' from src to dst. @@ -324,34 +560,11 @@ def copy_collections(obj, src, dst, args): # The pipeline object is updated in place with the new repository # names. The return value is undefined. # -def copy_git_repos(p, src, dst, dst_repo): - copied = set() - for c in p['components']: - component = p['components'][c] - if 'repository' in component: - repo = component['repository'] - script_version = component.get('script_version', None) - if repo not in copied: - copy_git_repo(repo, src, dst, dst_repo, script_version) - copied.add(repo) - component['repository'] = dst_repo - if script_version: - repo_dir = local_repo_dir[repo] - component['script_version'] = git_rev_parse(script_version, repo_dir) +def copy_git_repos(p, src, dst, dst_repo, args): + for component in p['components'].values(): + migrate_jobspec(component, src, dst, dst_repo, args) if 'job' in component: - j = component['job'] - if 'repository' in j: - repo = j['repository'] - script_version = j.get('script_version', None) - if repo not in copied: - copy_git_repo(repo, src, dst, dst_repo, script_version) - copied.add(repo) - j['repository'] = dst_repo - repo_dir = local_repo_dir[repo] - if script_version: - j['script_version'] = git_rev_parse(script_version, repo_dir) - if 'supplied_script_version' in j: - j['supplied_script_version'] = git_rev_parse(j['supplied_script_version'], repo_dir) + migrate_jobspec(component['job'], src, dst, dst_repo, args) def total_collection_size(manifest_text): """Return the total number of bytes in this collection (excluding @@ -377,9 +590,11 @@ def create_collection_from(c, src, dst, args): available.""" collection_uuid = c['uuid'] - del c['uuid'] + if not c["name"]: + c['name'] = "copied from " + collection_uuid + if 'properties' in c: del c['properties'] @@ -392,12 +607,10 @@ def create_collection_from(c, src, dst, args): for link_class in ("docker_image_repo+tag", "docker_image_hash"): docker_links = src.links().list(filters=[["head_uuid", "=", collection_uuid], ["link_class", "=", link_class]]).execute(num_retries=args.retries)['items'] - for d in docker_links: - body={ - 'head_uuid': dst_collection['uuid'], - 'link_class': link_class, - 'name': d['name'], - } + for src_link in docker_links: + body = {key: src_link[key] + for key in ['link_class', 'name', 'properties']} + body['head_uuid'] = dst_collection['uuid'] body['owner_uuid'] = args.project_uuid lk = dst.links().create(body=body).execute(num_retries=args.retries) @@ -428,7 +641,55 @@ def create_collection_from(c, src, dst, args): # hash will not change. # def copy_collection(obj_uuid, src, dst, args): - c = src.collections().get(uuid=obj_uuid).execute() + if arvados.util.keep_locator_pattern.match(obj_uuid): + # If the obj_uuid is a portable data hash, it might not be + # uniquely identified with a particular collection. As a + # result, it is ambiguous as to what name to use for the copy. + # Apply some heuristics to pick which collection to get the + # name from. + srccol = src.collections().list( + filters=[['portable_data_hash', '=', obj_uuid]], + order="created_at asc" + ).execute(num_retries=args.retries) + + items = srccol.get("items") + + if not items: + logger.warning("Could not find collection with portable data hash %s", obj_uuid) + return + + c = None + + if len(items) == 1: + # There's only one collection with the PDH, so use that. + c = items[0] + if not c: + # See if there is a collection that's in the same project + # as the root item (usually a pipeline) being copied. + for i in items: + if i.get("owner_uuid") == src_owner_uuid and i.get("name"): + c = i + break + if not c: + # Didn't find any collections located in the same project, so + # pick the oldest collection that has a name assigned to it. + for i in items: + if i.get("name"): + c = i + break + if not c: + # None of the collections have names (?!), so just pick the + # first one. + c = items[0] + + # list() doesn't return manifest text (and we don't want it to, + # because we don't need the same maninfest text sent to us 50 + # times) so go and retrieve the collection object directly + # which will include the manifest text. + c = src.collections().get(uuid=c["uuid"]).execute(num_retries=args.retries) + else: + # Assume this is an actual collection uuid, so fetch it directly. + c = src.collections().get(uuid=obj_uuid).execute(num_retries=args.retries) # If a collection with this hash already exists at the # destination, and 'force' is not true, just return that @@ -440,14 +701,16 @@ def copy_collection(obj_uuid, src, dst, args): colhash = c['uuid'] dstcol = dst.collections().list( filters=[['portable_data_hash', '=', colhash]] - ).execute() + ).execute(num_retries=args.retries) if dstcol['items_available'] > 0: for d in dstcol['items']: if ((args.project_uuid == d['owner_uuid']) and - (c['name'] == d['name']) and + (c.get('name') == d['name']) and (c['portable_data_hash'] == d['portable_data_hash'])): return d - + c['manifest_text'] = dst.collections().get( + uuid=dstcol['items'][0]['uuid'] + ).execute(num_retries=args.retries)['manifest_text'] return create_collection_from(c, src, dst, args) # Fetch the collection's manifest. @@ -468,31 +731,30 @@ def copy_collection(obj_uuid, src, dst, args): else: progress_writer = None - for line in manifest.splitlines(True): + for line in manifest.splitlines(): words = line.split() - dst_manifest_line = words[0] + dst_manifest += words[0] for word in words[1:]: try: loc = arvados.KeepLocator(word) - blockhash = loc.md5sum - # copy this block if we haven't seen it before - # (otherwise, just reuse the existing dst_locator) - if blockhash not in dst_locators: - logger.debug("Copying block %s (%s bytes)", blockhash, loc.size) - if progress_writer: - progress_writer.report(obj_uuid, bytes_written, bytes_expected) - data = src_keep.get(word) - dst_locator = dst_keep.put(data) - dst_locators[blockhash] = dst_locator - bytes_written += loc.size - dst_manifest_line += ' ' + dst_locators[blockhash] except ValueError: # If 'word' can't be parsed as a locator, # presume it's a filename. - dst_manifest_line += ' ' + word - dst_manifest += dst_manifest_line - if line.endswith("\n"): - dst_manifest += "\n" + dst_manifest += ' ' + word + continue + blockhash = loc.md5sum + # copy this block if we haven't seen it before + # (otherwise, just reuse the existing dst_locator) + if blockhash not in dst_locators: + logger.debug("Copying block %s (%s bytes)", blockhash, loc.size) + if progress_writer: + progress_writer.report(obj_uuid, bytes_written, bytes_expected) + data = src_keep.get(word) + dst_locator = dst_keep.put(data) + dst_locators[blockhash] = dst_locator + bytes_written += loc.size + dst_manifest += ' ' + dst_locators[blockhash] + dst_manifest += "\n" if progress_writer: progress_writer.report(obj_uuid, bytes_written, bytes_expected) @@ -501,11 +763,59 @@ def copy_collection(obj_uuid, src, dst, args): # Copy the manifest and save the collection. logger.debug('saving %s with manifest: <%s>', obj_uuid, dst_manifest) - dst_keep.put(dst_manifest.encode('utf-8')) c['manifest_text'] = dst_manifest return create_collection_from(c, src, dst, args) -# copy_git_repo(src_git_repo, src, dst, dst_git_repo, script_version) +def select_git_url(api, repo_name, retries, allow_insecure_http, allow_insecure_http_opt): + r = api.repositories().list( + filters=[['name', '=', repo_name]]).execute(num_retries=retries) + if r['items_available'] != 1: + raise Exception('cannot identify repo {}; {} repos found' + .format(repo_name, r['items_available'])) + + https_url = [c for c in r['items'][0]["clone_urls"] if c.startswith("https:")] + http_url = [c for c in r['items'][0]["clone_urls"] if c.startswith("http:")] + other_url = [c for c in r['items'][0]["clone_urls"] if not c.startswith("http")] + + priority = https_url + other_url + http_url + + git_config = [] + git_url = None + for url in priority: + if url.startswith("http"): + u = urllib.parse.urlsplit(url) + baseurl = urllib.parse.urlunsplit((u.scheme, u.netloc, "", "", "")) + git_config = ["-c", "credential.%s/.username=none" % baseurl, + "-c", "credential.%s/.helper=!cred(){ cat >/dev/null; if [ \"$1\" = get ]; then echo password=$ARVADOS_API_TOKEN; fi; };cred" % baseurl] + else: + git_config = [] + + try: + logger.debug("trying %s", url) + arvados.util.run_command(["git"] + git_config + ["ls-remote", url], + env={"HOME": os.environ["HOME"], + "ARVADOS_API_TOKEN": api.api_token, + "GIT_ASKPASS": "/bin/false"}) + except arvados.errors.CommandFailedError: + pass + else: + git_url = url + break + + if not git_url: + raise Exception('Cannot access git repository, tried {}' + .format(priority)) + + if git_url.startswith("http:"): + if allow_insecure_http: + logger.warning("Using insecure git url %s but will allow this because %s", git_url, allow_insecure_http_opt) + else: + raise Exception("Refusing to use insecure git url %s, use %s if you really want this." % (git_url, allow_insecure_http_opt)) + + return (git_url, git_config) + + +# copy_git_repo(src_git_repo, src, dst, dst_git_repo, script_version, args) # # Copies commits from git repository 'src_git_repo' on Arvados # instance 'src' to 'dst_git_repo' on 'dst'. Both src_git_repo @@ -515,65 +825,50 @@ def copy_collection(obj_uuid, src, dst, args): # All commits will be copied to a destination branch named for the # source repository URL. # -# Because users cannot create their own repositories, the -# destination repository must already exist. +# The destination repository must already exist. # # The user running this command must be authenticated # to both repositories. # -def copy_git_repo(src_git_repo, src, dst, dst_git_repo, script_version): +def copy_git_repo(src_git_repo, src, dst, dst_git_repo, script_version, args): # Identify the fetch and push URLs for the git repositories. - r = src.repositories().list( - filters=[['name', '=', src_git_repo]]).execute() - if r['items_available'] != 1: - raise Exception('cannot identify source repo {}; {} repos found' - .format(src_git_repo, r['items_available'])) - src_git_url = r['items'][0]['fetch_url'] - logger.debug('src_git_url: {}'.format(src_git_url)) - r = dst.repositories().list( - filters=[['name', '=', dst_git_repo]]).execute() - if r['items_available'] != 1: - raise Exception('cannot identify destination repo {}; {} repos found' - .format(dst_git_repo, r['items_available'])) - dst_git_push_url = r['items'][0]['push_url'] - logger.debug('dst_git_push_url: {}'.format(dst_git_push_url)) + (src_git_url, src_git_config) = select_git_url(src, src_git_repo, args.retries, args.allow_git_http_src, "--allow-git-http-src") + (dst_git_url, dst_git_config) = select_git_url(dst, dst_git_repo, args.retries, args.allow_git_http_dst, "--allow-git-http-dst") - # script_version is the "script_version" parameter from the source - # component or job. It is used here to tie the destination branch - # to the commit that was used on the source. If no script_version - # was supplied in the component or job, it is a mistake in the pipeline, - # but for the purposes of copying the repository, default to "master". - # - if not script_version: - script_version = "master" + logger.debug('src_git_url: {}'.format(src_git_url)) + logger.debug('dst_git_url: {}'.format(dst_git_url)) dst_branch = re.sub(r'\W+', '_', "{}_{}".format(src_git_url, script_version)) - # Copy git commits from src repo to dst repo (but only if - # we have not already copied this repo in this session). - # - if src_git_repo in local_repo_dir: - logger.debug('already copied src repo %s, skipping', src_git_repo) - else: - tmprepo = tempfile.mkdtemp() - local_repo_dir[src_git_repo] = tmprepo + # Copy git commits from src repo to dst repo. + if src_git_repo not in local_repo_dir: + local_repo_dir[src_git_repo] = tempfile.mkdtemp() arvados.util.run_command( - ["git", "clone", "--bare", src_git_url, tmprepo], - cwd=os.path.dirname(tmprepo)) + ["git"] + src_git_config + ["clone", "--bare", src_git_url, + local_repo_dir[src_git_repo]], + cwd=os.path.dirname(local_repo_dir[src_git_repo]), + env={"HOME": os.environ["HOME"], + "ARVADOS_API_TOKEN": src.api_token, + "GIT_ASKPASS": "/bin/false"}) arvados.util.run_command( - ["git", "branch", dst_branch, script_version], - cwd=tmprepo) - arvados.util.run_command(["git", "remote", "add", "dst", dst_git_push_url], cwd=tmprepo) - arvados.util.run_command(["git", "push", "dst", dst_branch], cwd=tmprepo) - + ["git", "remote", "add", "dst", dst_git_url], + cwd=local_repo_dir[src_git_repo]) + arvados.util.run_command( + ["git", "branch", dst_branch, script_version], + cwd=local_repo_dir[src_git_repo]) + arvados.util.run_command(["git"] + dst_git_config + ["push", "dst", dst_branch], + cwd=local_repo_dir[src_git_repo], + env={"HOME": os.environ["HOME"], + "ARVADOS_API_TOKEN": dst.api_token, + "GIT_ASKPASS": "/bin/false"}) def copy_docker_images(pipeline, src, dst, args): """Copy any docker images named in the pipeline components' runtime_constraints field from src to dst.""" logger.debug('copy_docker_images: {}'.format(pipeline['uuid'])) - for c_name, c_info in pipeline['components'].iteritems(): + for c_name, c_info in pipeline['components'].items(): if ('runtime_constraints' in c_info and 'docker_image' in c_info['runtime_constraints']): copy_docker_image( @@ -594,12 +889,16 @@ def copy_docker_image(docker_image, docker_image_tag, src, dst, args): # Find the link identifying this docker image. docker_image_list = arvados.commands.keepdocker.list_images_in_arv( src, args.retries, docker_image, docker_image_tag) - image_uuid, image_info = docker_image_list[0] - logger.debug('copying collection {} {}'.format(image_uuid, image_info)) - - # Copy the collection it refers to. - dst_image_col = copy_collection(image_uuid, src, dst, args) - + if docker_image_list: + image_uuid, image_info = docker_image_list[0] + logger.debug('copying collection {} {}'.format(image_uuid, image_info)) + + # Copy the collection it refers to. + dst_image_col = copy_collection(image_uuid, src, dst, args) + elif arvados.util.keep_locator_pattern.match(docker_image): + dst_image_col = copy_collection(docker_image, src, dst, args) + else: + logger.warning('Could not find docker image {}:{}'.format(docker_image, docker_image_tag)) # git_rev_parse(rev, repo) #