X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/eae48c31bb338689ec67fbc6a14a2e0b1fb5e3b6..9f0854e394baf712f1fbcf3ebd21af215276c0fb:/sdk/cwl/arvados_cwl/__init__.py?ds=inline diff --git a/sdk/cwl/arvados_cwl/__init__.py b/sdk/cwl/arvados_cwl/__init__.py index 92be92d6e0..c7c186390b 100644 --- a/sdk/cwl/arvados_cwl/__init__.py +++ b/sdk/cwl/arvados_cwl/__init__.py @@ -21,13 +21,15 @@ import schema_salad import arvados import arvados.config +from arvados.keep import KeepClient +from arvados.errors import ApiError from .arvcontainer import ArvadosContainer, RunnerContainer from .arvjob import ArvadosJob, RunnerJob, RunnerTemplate from. runner import Runner, upload_instance from .arvtool import ArvadosCommandTool from .arvworkflow import ArvadosWorkflow, upload_workflow -from .fsaccess import CollectionFsAccess +from .fsaccess import CollectionFsAccess, CollectionFetcher from .perf import Perf from .pathmapper import FinalOutputPathMapper from ._version import __version__ @@ -49,7 +51,7 @@ class ArvCwlRunner(object): """ - def __init__(self, api_client, work_api=None, keep_client=None, output_name=None): + def __init__(self, api_client, work_api=None, keep_client=None, output_name=None, output_tags=None, num_retries=4): self.api = api_client self.processes = {} self.lock = threading.Lock() @@ -57,13 +59,14 @@ class ArvCwlRunner(object): self.final_output = None self.final_status = None self.uploaded = {} - self.num_retries = 4 + self.num_retries = num_retries self.uuid = None self.stop_polling = threading.Event() self.poll_api = None self.pipeline = None self.final_output_collection = None self.output_name = output_name + self.output_tags = output_tags self.project_uuid = None if keep_client is not None: @@ -71,7 +74,9 @@ class ArvCwlRunner(object): else: self.keep_client = arvados.keep.KeepClient(api_client=self.api, num_retries=self.num_retries) - for api in ["jobs", "containers"]: + self.work_api = None + expected_api = ["jobs", "containers"] + for api in expected_api: try: methods = self.api._rootDesc.get('resources')[api]['methods'] if ('httpMethod' in methods['create'] and @@ -80,11 +85,12 @@ class ArvCwlRunner(object): break except KeyError: pass + if not self.work_api: if work_api is None: raise Exception("No supported APIs") else: - raise Exception("Unsupported API '%s'" % work_api) + raise Exception("Unsupported API '%s', expected one of %s" % (work_api, expected_api)) def arv_make_tool(self, toolpath_object, **kwargs): kwargs["work_api"] = self.work_api @@ -119,7 +125,7 @@ class ArvCwlRunner(object): logger.info("Job %s (%s) is Running", j.name, uuid) j.running = True j.update_pipeline_component(event["properties"]["new_attributes"]) - elif event["properties"]["new_attributes"]["state"] in ("Complete", "Failed", "Cancelled"): + elif event["properties"]["new_attributes"]["state"] in ("Complete", "Failed", "Cancelled", "Final"): uuid = event["object_uuid"] try: self.cond.acquire() @@ -138,34 +144,43 @@ class ArvCwlRunner(object): Runs in a separate thread. """ - while True: - self.stop_polling.wait(15) - if self.stop_polling.is_set(): - break - with self.lock: - keys = self.processes.keys() - if not keys: - continue + try: + while True: + self.stop_polling.wait(15) + if self.stop_polling.is_set(): + break + with self.lock: + keys = self.processes.keys() + if not keys: + continue - if self.work_api == "containers": - table = self.poll_api.containers() - elif self.work_api == "jobs": - table = self.poll_api.jobs() + if self.work_api == "containers": + table = self.poll_api.container_requests() + elif self.work_api == "jobs": + table = self.poll_api.jobs() - try: - proc_states = table.list(filters=[["uuid", "in", keys]]).execute(num_retries=self.num_retries) - except Exception as e: - logger.warn("Error checking states on API server: %s", e) - continue - - for p in proc_states["items"]: - self.on_message({ - "object_uuid": p["uuid"], - "event_type": "update", - "properties": { - "new_attributes": p - } - }) + try: + proc_states = table.list(filters=[["uuid", "in", keys]]).execute(num_retries=self.num_retries) + except Exception as e: + logger.warn("Error checking states on API server: %s", e) + continue + + for p in proc_states["items"]: + self.on_message({ + "object_uuid": p["uuid"], + "event_type": "update", + "properties": { + "new_attributes": p + } + }) + except: + logger.error("Fatal error in state polling thread.", exc_info=(sys.exc_info()[1] if self.debug else False)) + self.cond.acquire() + self.processes.clear() + self.cond.notify() + self.cond.release() + finally: + self.stop_polling.set() def get_uploaded(self): return self.uploaded.copy() @@ -183,7 +198,7 @@ class ArvCwlRunner(object): for v in obj: self.check_writable(v) - def make_output_collection(self, name, outputObj): + def make_output_collection(self, name, tagsString, outputObj): outputObj = copy.deepcopy(outputObj) files = [] @@ -248,6 +263,13 @@ class ArvCwlRunner(object): final.api_response()["name"], final.manifest_locator()) + final_uuid = final.manifest_locator() + tags = tagsString.split(',') + for tag in tags: + self.api.links().create(body={ + "head_uuid": final_uuid, "link_class": "tag", "name": tag + }).execute(num_retries=self.num_retries) + def finalcollection(fileobj): fileobj["location"] = "keep:%s/%s" % (final.portable_data_hash(), fileobj["location"]) @@ -260,6 +282,12 @@ class ArvCwlRunner(object): if self.work_api == "containers": try: current = self.api.containers().current().execute(num_retries=self.num_retries) + except ApiError as e: + # Status code 404 just means we're not running in a container. + if e.resp.status != 404: + logger.info("Getting current container: %s", e) + return + try: self.api.containers().update(uuid=current['uuid'], body={ 'output': self.final_output_collection.portable_data_hash(), @@ -279,22 +307,28 @@ class ArvCwlRunner(object): tool.visit(self.check_writable) - useruuid = self.api.users().current().execute()["uuid"] - self.project_uuid = kwargs.get("project_uuid") if kwargs.get("project_uuid") else useruuid + self.project_uuid = kwargs.get("project_uuid") self.pipeline = None make_fs_access = kwargs.get("make_fs_access") or partial(CollectionFsAccess, api_client=self.api, keep_client=self.keep_client) self.fs_access = make_fs_access(kwargs["basedir"]) - if kwargs.get("create_template"): - tmpl = RunnerTemplate(self, tool, job_order, kwargs.get("enable_reuse")) - tmpl.save() - # cwltool.main will write our return value to stdout. - return tmpl.uuid - - if kwargs.get("create_workflow") or kwargs.get("update_workflow"): - return upload_workflow(self, tool, job_order, self.project_uuid, kwargs.get("update_workflow")) + existing_uuid = kwargs.get("update_workflow") + if existing_uuid or kwargs.get("create_workflow"): + if self.work_api == "jobs": + tmpl = RunnerTemplate(self, tool, job_order, + kwargs.get("enable_reuse"), + uuid=existing_uuid, + submit_runner_ram=kwargs.get("submit_runner_ram")) + tmpl.save() + # cwltool.main will write our return value to stdout. + return tmpl.uuid + else: + return upload_workflow(self, tool, job_order, + self.project_uuid, + uuid=existing_uuid, + submit_runner_ram=kwargs.get("submit_runner_ram")) self.ignore_docker_for_reuse = kwargs.get("ignore_docker_for_reuse") @@ -325,9 +359,11 @@ class ArvCwlRunner(object): self.output_callback, **kwargs).next() else: - runnerjob = RunnerContainer(self, tool, job_order, kwargs.get("enable_reuse"), self.output_name) + runnerjob = RunnerContainer(self, tool, job_order, kwargs.get("enable_reuse"), self.output_name, + self.output_tags, submit_runner_ram=kwargs.get("submit_runner_ram")) else: - runnerjob = RunnerJob(self, tool, job_order, kwargs.get("enable_reuse"), self.output_name) + runnerjob = RunnerJob(self, tool, job_order, kwargs.get("enable_reuse"), self.output_name, + self.output_tags, submit_runner_ram=kwargs.get("submit_runner_ram")) if not kwargs.get("submit") and "cwl_runner_job" not in kwargs and not self.work_api == "containers": # Create pipeline for local run @@ -366,6 +402,10 @@ class ArvCwlRunner(object): loopperf.__enter__() for runnable in jobiter: loopperf.__exit__() + + if self.stop_polling.is_set(): + break + if runnable: with Perf(metrics, "run"): runnable.run(**kwargs) @@ -387,7 +427,7 @@ class ArvCwlRunner(object): if sys.exc_info()[0] is KeyboardInterrupt: logger.error("Interrupted, marking pipeline as failed") else: - logger.error("Caught unhandled exception, marking pipeline as failed. Error was: %s", sys.exc_info()[1], exc_info=(sys.exc_info()[1] if self.debug else False)) + logger.error("Execution failed: %s", sys.exc_info()[1], exc_info=(sys.exc_info()[1] if self.debug else False)) if self.pipeline: self.api.pipeline_instances().update(uuid=self.pipeline["uuid"], body={"state": "Failed"}).execute(num_retries=self.num_retries) @@ -410,7 +450,9 @@ class ArvCwlRunner(object): else: if self.output_name is None: self.output_name = "Output of %s" % (shortname(tool.tool["id"])) - self.final_output, self.final_output_collection = self.make_output_collection(self.output_name, self.final_output) + if self.output_tags is None: + self.output_tags = "" + self.final_output, self.final_output_collection = self.make_output_collection(self.output_name, self.output_tags, self.final_output) self.set_crunch_output() if self.final_status != "success": @@ -468,6 +510,7 @@ def arg_parser(): # type: () -> argparse.ArgumentParser parser.add_argument("--project-uuid", type=str, metavar="UUID", help="Project that will own the workflow jobs, if not provided, will go to home project.") parser.add_argument("--output-name", type=str, help="Name to use for collection that stores the final output.", default=None) + parser.add_argument("--output-tags", type=str, help="Tags for the final output collection separated by commas, e.g., '--output-tags tag0,tag1,tag2'.", default=None) parser.add_argument("--ignore-docker-for-reuse", action="store_true", help="Ignore Docker image version when deciding whether to reuse past jobs.", default=False) @@ -477,9 +520,10 @@ def arg_parser(): # type: () -> argparse.ArgumentParser default=True, dest="submit") exgroup.add_argument("--local", action="store_false", help="Run workflow on local host (submits jobs to Arvados).", default=True, dest="submit") - exgroup.add_argument("--create-template", action="store_true", help="Create an Arvados pipeline template.") - exgroup.add_argument("--create-workflow", action="store_true", help="Create an Arvados workflow.") - exgroup.add_argument("--update-workflow", type=str, metavar="UUID", help="Update existing Arvados workflow with uuid.") + exgroup.add_argument("--create-template", action="store_true", help="(Deprecated) synonym for --create-workflow.", + dest="create_workflow") + exgroup.add_argument("--create-workflow", action="store_true", help="Create an Arvados workflow (if using the 'containers' API) or pipeline template (if using the 'jobs' API). See --api.") + exgroup.add_argument("--update-workflow", type=str, metavar="UUID", help="Update an existing Arvados workflow or pipeline template with the given UUID.") exgroup = parser.add_mutually_exclusive_group() exgroup.add_argument("--wait", action="store_true", help="After submitting workflow runner job, wait for completion.", @@ -495,6 +539,10 @@ def arg_parser(): # type: () -> argparse.ArgumentParser help="Compute checksum of contents while collecting outputs", dest="compute_checksum") + parser.add_argument("--submit-runner-ram", type=int, + help="RAM (in MiB) required for the workflow runner job (default 1024)", + default=1024) + parser.add_argument("workflow", type=str, nargs="?", default=None, help="The workflow to execute") parser.add_argument("job_order", nargs=argparse.REMAINDER, help="The input object to the workflow.") @@ -517,7 +565,21 @@ def main(args, stdout, stderr, api_client=None, keep_client=None): job_order_object = None arvargs = parser.parse_args(args) - if (arvargs.create_template or arvargs.create_workflow or arvargs.update_workflow) and not arvargs.job_order: + + if arvargs.update_workflow: + if arvargs.update_workflow.find('-7fd4e-') == 5: + want_api = 'containers' + elif arvargs.update_workflow.find('-p5p6p-') == 5: + want_api = 'jobs' + else: + want_api = None + if want_api and arvargs.work_api and want_api != arvargs.work_api: + logger.error('--update-workflow arg {!r} uses {!r} API, but --api={!r} specified'.format( + arvargs.update_workflow, want_api, arvargs.work_api)) + return 1 + arvargs.work_api = want_api + + if (arvargs.create_workflow or arvargs.update_workflow) and not arvargs.job_order: job_order_object = ({}, "") add_arv_hints() @@ -525,7 +587,11 @@ def main(args, stdout, stderr, api_client=None, keep_client=None): try: if api_client is None: api_client=arvados.api('v1', model=OrderedJsonModel()) - runner = ArvCwlRunner(api_client, work_api=arvargs.work_api, keep_client=keep_client, output_name=arvargs.output_name) + if keep_client is None: + keep_client = arvados.keep.KeepClient(api_client=api_client, num_retries=4) + runner = ArvCwlRunner(api_client, work_api=arvargs.work_api, keep_client=keep_client, + num_retries=4, output_name=arvargs.output_name, + output_tags=arvargs.output_tags) except Exception as e: logger.error(e) return 1 @@ -543,6 +609,7 @@ def main(args, stdout, stderr, api_client=None, keep_client=None): arvargs.conformance_test = None arvargs.use_container = True + arvargs.relax_path_checks = True return cwltool.main.main(args=arvargs, stdout=stdout, @@ -551,4 +618,9 @@ def main(args, stdout, stderr, api_client=None, keep_client=None): makeTool=runner.arv_make_tool, versionfunc=versionstring, job_order_object=job_order_object, - make_fs_access=partial(CollectionFsAccess, api_client=api_client)) + make_fs_access=partial(CollectionFsAccess, + api_client=api_client, + keep_client=keep_client), + fetcher_constructor=partial(CollectionFetcher, + api_client=api_client, + keep_client=keep_client))