X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/3bbb988777079718338e3e6cb9c6c9b5399be800..0ecea550fde014578e71004360b700cdfeae4909:/sdk/python/arvados/commands/run.py diff --git a/sdk/python/arvados/commands/run.py b/sdk/python/arvados/commands/run.py index f2bf0f353b..46a56ca558 100644 --- a/sdk/python/arvados/commands/run.py +++ b/sdk/python/arvados/commands/run.py @@ -11,21 +11,40 @@ import put import time import subprocess import logging +import sys +import errno import arvados.commands._util as arv_cmd +import arvados.collection + +from arvados._version import __version__ logger = logging.getLogger('arvados.arv-run') +logger.setLevel(logging.INFO) arvrun_parser = argparse.ArgumentParser(parents=[arv_cmd.retry_opt]) -arvrun_parser.add_argument('--dry-run', action="store_true", help="Print out the pipeline that would be submitted and exit") -arvrun_parser.add_argument('--local', action="store_true", help="Run locally using arv-run-pipeline-instance") -arvrun_parser.add_argument('--docker-image', type=str, default="arvados/jobs", help="Docker image to use, default arvados/jobs") -arvrun_parser.add_argument('--ignore-rcode', action="store_true", help="Commands that return non-zero return codes should not be considered failed.") -arvrun_parser.add_argument('--no-reuse', action="store_true", help="Do not reuse past jobs.") -arvrun_parser.add_argument('--no-wait', action="store_true", help="Do not wait and display logs after submitting command, just exit.") -arvrun_parser.add_argument('--project-uuid', type=str, help="Parent project of the pipeline") -arvrun_parser.add_argument('--git-dir', type=str, default="", help="Git repository passed to arv-crunch-job when using --local") -arvrun_parser.add_argument('--repository', type=str, default="arvados", help="repository field of component, default 'arvados'") -arvrun_parser.add_argument('--script-version', type=str, default="master", help="script_version field of component, default 'master'") +arvrun_parser.add_argument('--dry-run', action="store_true", + help="Print out the pipeline that would be submitted and exit") +arvrun_parser.add_argument('--local', action="store_true", + help="Run locally using arv-run-pipeline-instance") +arvrun_parser.add_argument('--docker-image', type=str, + help="Docker image to use, otherwise use instance default.") +arvrun_parser.add_argument('--ignore-rcode', action="store_true", + help="Commands that return non-zero return codes should not be considered failed.") +arvrun_parser.add_argument('--no-reuse', action="store_true", + help="Do not reuse past jobs.") +arvrun_parser.add_argument('--no-wait', action="store_true", + help="Do not wait and display logs after submitting command, just exit.") +arvrun_parser.add_argument('--project-uuid', type=str, + help="Parent project of the pipeline") +arvrun_parser.add_argument('--git-dir', type=str, default="", + help="Git repository passed to arv-crunch-job when using --local") +arvrun_parser.add_argument('--repository', type=str, default="arvados", + help="repository field of component, default 'arvados'") +arvrun_parser.add_argument('--script-version', type=str, default="master", + help="script_version field of component, default 'master'") +arvrun_parser.add_argument('--version', action='version', + version="%s %s" % (sys.argv[0], __version__), + help='Print version and exit.') arvrun_parser.add_argument('args', nargs=argparse.REMAINDER) class ArvFile(object): @@ -33,6 +52,12 @@ class ArvFile(object): self.prefix = prefix self.fn = fn + def __hash__(self): + return (self.prefix+self.fn).__hash__() + + def __eq__(self, other): + return (self.prefix == other.prefix) and (self.fn == other.fn) + class UploadFile(ArvFile): pass @@ -51,7 +76,7 @@ def is_in_collection(root, branch): else: sp = os.path.split(root) return is_in_collection(sp[0], os.path.join(sp[1], branch)) - except IOError, OSError: + except (IOError, OSError): return (None, None) # Determine the project to place the output of this command by searching upward @@ -72,7 +97,7 @@ def determine_project(root, current_user): else: sp = os.path.split(root) return determine_project(sp[0], current_user) - except IOError, OSError: + except (IOError, OSError): return current_user # Determine if string corresponds to a file, and if that file is part of a @@ -80,26 +105,118 @@ def determine_project(root, current_user): # ArvFile() (file already exists in a collection), UploadFile() (file needs to # be uploaded to a collection), or simply returns prefix+fn (which yields the # original parameter string). -def statfile(prefix, fn): +def statfile(prefix, fn, fnPattern="$(file %s/%s)", dirPattern="$(dir %s/%s/)"): absfn = os.path.abspath(fn) - if os.path.exists(absfn): + try: st = os.stat(absfn) - if stat.S_ISREG(st.st_mode): - sp = os.path.split(absfn) - (pdh, branch) = is_in_collection(sp[0], sp[1]) - if pdh: - return ArvFile(prefix, "$(file %s/%s)" % (pdh, branch)) + sp = os.path.split(absfn) + (pdh, branch) = is_in_collection(sp[0], sp[1]) + if pdh: + if stat.S_ISREG(st.st_mode): + return ArvFile(prefix, fnPattern % (pdh, branch)) + elif stat.S_ISDIR(st.st_mode): + return ArvFile(prefix, dirPattern % (pdh, branch)) else: - # trim leading '/' for path prefix test later - return UploadFile(prefix, absfn[1:]) - if stat.S_ISDIR(st.st_mode): - sp = os.path.split(absfn) - (pdh, branch) = is_in_collection(sp[0], sp[1]) - if pdh: - return ArvFile(prefix, "$(dir %s/%s/)" % (pdh, branch)) + raise Exception("%s is not a regular file or directory" % absfn) + else: + # trim leading '/' for path prefix test later + return UploadFile(prefix, absfn[1:]) + except OSError as e: + if e.errno == errno.ENOENT: + pass + else: + raise return prefix+fn +def write_file(collection, pathprefix, fn): + with open(os.path.join(pathprefix, fn)) as src: + dst = collection.open(fn, "w") + r = src.read(1024*128) + while r: + dst.write(r) + r = src.read(1024*128) + dst.close(flush=False) + +def uploadfiles(files, api, dry_run=False, num_retries=0, + project=None, + fnPattern="$(file %s/%s)", + name=None): + # Find the smallest path prefix that includes all the files that need to be uploaded. + # This starts at the root and iteratively removes common parent directory prefixes + # until all file paths no longer have a common parent. + n = True + pathprefix = "/" + while n: + pathstep = None + for c in files: + if pathstep is None: + sp = c.fn.split('/') + if len(sp) < 2: + # no parent directories left + n = False + break + # path step takes next directory + pathstep = sp[0] + "/" + else: + # check if pathstep is common prefix for all files + if not c.fn.startswith(pathstep): + n = False + break + if n: + # pathstep is common parent directory for all files, so remove the prefix + # from each path + pathprefix += pathstep + for c in files: + c.fn = c.fn[len(pathstep):] + + logger.info("Upload local files: \"%s\"", '" "'.join([c.fn for c in files])) + + if dry_run: + logger.info("$(input) is %s", pathprefix.rstrip('/')) + pdh = "$(input)" + else: + files = sorted(files, key=lambda x: x.fn) + collection = arvados.collection.Collection(api_client=api, num_retries=num_retries) + prev = "" + for f in files: + localpath = os.path.join(pathprefix, f.fn) + if prev and localpath.startswith(prev+"/"): + # If this path is inside an already uploaded subdirectory, + # don't redundantly re-upload it. + # e.g. we uploaded /tmp/foo and the next file is /tmp/foo/bar + # skip it because it starts with "/tmp/foo/" + continue + prev = localpath + if os.path.isfile(localpath): + write_file(collection, pathprefix, f.fn) + elif os.path.isdir(localpath): + for root, dirs, iterfiles in os.walk(localpath): + root = root[len(pathprefix):] + for src in iterfiles: + write_file(collection, pathprefix, os.path.join(root, src)) + + filters=[["portable_data_hash", "=", collection.portable_data_hash()], + ["name", "like", name+"%"]] + if project: + filters.append(["owner_uuid", "=", project]) + + exists = api.collections().list(filters=filters, limit=1).execute(num_retries=num_retries) + + if exists["items"]: + item = exists["items"][0] + pdh = item["portable_data_hash"] + logger.info("Using collection %s (%s)", pdh, item["uuid"]) + else: + collection.save_new(name=name, owner_uuid=project, ensure_unique_name=True) + pdh = collection.portable_data_hash() + logger.info("Uploaded to %s (%s)", pdh, collection.manifest_locator()) + + for c in files: + c.keepref = "%s/%s" % (pdh, c.fn) + c.fn = fnPattern % (pdh, c.fn) + + def main(arguments=None): args = arvrun_parser.parse_args(arguments) @@ -178,62 +295,9 @@ def main(arguments=None): command[i] = statfile(m.group(1), m.group(2)) break - n = True - pathprefix = "/" files = [c for command in slots[1:] for c in command if isinstance(c, UploadFile)] - if len(files) > 0: - # Find the smallest path prefix that includes all the files that need to be uploaded. - # This starts at the root and iteratively removes common parent directory prefixes - # until all file pathes no longer have a common parent. - while n: - pathstep = None - for c in files: - if pathstep is None: - sp = c.fn.split('/') - if len(sp) < 2: - # no parent directories left - n = False - break - # path step takes next directory - pathstep = sp[0] + "/" - else: - # check if pathstep is common prefix for all files - if not c.fn.startswith(pathstep): - n = False - break - if n: - # pathstep is common parent directory for all files, so remove the prefix - # from each path - pathprefix += pathstep - for c in files: - c.fn = c.fn[len(pathstep):] - - orgdir = os.getcwd() - os.chdir(pathprefix) - - print("Upload local files: \"%s\"" % '" "'.join([c.fn for c in files])) - - if args.dry_run: - print("$(input) is %s" % pathprefix.rstrip('/')) - pdh = "$(input)" - else: - files = sorted(files, key=lambda x: x.fn) - collection = arvados.CollectionWriter(api, num_retries=args.retries) - stream = None - for f in files: - sp = os.path.split(f.fn) - if sp[0] != stream: - stream = sp[0] - collection.start_new_stream(stream) - collection.write_file(f.fn, sp[1]) - item = api.collections().create(body={"owner_uuid": project, "manifest_text": collection.manifest_text()}).execute() - pdh = item["portable_data_hash"] - print "Uploaded to %s" % item["uuid"] - - for c in files: - c.fn = "$(file %s/%s)" % (pdh, c.fn) - - os.chdir(orgdir) + if files: + uploadfiles(files, api, dry_run=args.dry_run, num_retries=args.retries, project=project) for i in xrange(1, len(slots)): slots[i] = [("%s%s" % (c.prefix, c.fn)) if isinstance(c, ArvFile) else c for c in slots[i]] @@ -244,11 +308,12 @@ def main(arguments=None): "repository": args.repository, "script_parameters": { }, - "runtime_constraints": { - "docker_image": args.docker_image - } + "runtime_constraints": {} } + if args.docker_image: + component["runtime_constraints"]["docker_image"] = args.docker_image + task_foreach = [] group_parser = argparse.ArgumentParser() group_parser.add_argument('-b', '--batch-size', type=int) @@ -298,7 +363,7 @@ def main(arguments=None): else: pipeline["owner_uuid"] = project pi = api.pipeline_instances().create(body=pipeline, ensure_unique_name=True).execute() - print "Running pipeline %s" % pi["uuid"] + logger.info("Running pipeline %s", pi["uuid"]) if args.local: subprocess.call(["arv-run-pipeline-instance", "--instance", pi["uuid"], "--run-jobs-here"] + (["--no-reuse"] if args.no_reuse else [])) @@ -306,11 +371,11 @@ def main(arguments=None): ws.main(["--pipeline", pi["uuid"]]) pi = api.pipeline_instances().get(uuid=pi["uuid"]).execute() - print "Pipeline is %s" % pi["state"] + logger.info("Pipeline is %s", pi["state"]) if "output_uuid" in pi["components"]["command"]: - print "Output is %s" % pi["components"]["command"]["output_uuid"] + logger.info("Output is %s", pi["components"]["command"]["output_uuid"]) else: - print "No output" + logger.info("No output") if __name__ == '__main__': main()