X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/83139cd2df6b7d86f352bee46d1ad86146bd093b..fecb5eb18b9cf15459de8eba44b6e545962d8cd4:/sdk/cwl/arvados_cwl/__init__.py diff --git a/sdk/cwl/arvados_cwl/__init__.py b/sdk/cwl/arvados_cwl/__init__.py index ef4f785529..3144592fc9 100644 --- a/sdk/cwl/arvados_cwl/__init__.py +++ b/sdk/cwl/arvados_cwl/__init__.py @@ -24,14 +24,16 @@ import arvados.config 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 .perf import Perf from .pathmapper import FinalOutputPathMapper +from ._version import __version__ from cwltool.pack import pack -from cwltool.process import shortname, UnsupportedRequirement +from cwltool.process import shortname, UnsupportedRequirement, getListing from cwltool.pathmapper import adjustFileObjs, adjustDirObjs from cwltool.draft2tool import compute_checksums from arvados.api import OrderedJsonModel @@ -47,7 +49,7 @@ class ArvCwlRunner(object): """ - def __init__(self, api_client, work_api=None): + def __init__(self, api_client, work_api=None, keep_client=None, output_name=None): self.api = api_client self.processes = {} self.lock = threading.Lock() @@ -57,18 +59,32 @@ class ArvCwlRunner(object): self.uploaded = {} self.num_retries = 4 self.uuid = None - self.work_api = work_api self.stop_polling = threading.Event() self.poll_api = None self.pipeline = None self.final_output_collection = None + self.output_name = output_name + self.project_uuid = None - if self.work_api is None: - # todo: autodetect API to use. - self.work_api = "jobs" + if keep_client is not None: + self.keep_client = keep_client + else: + self.keep_client = arvados.keep.KeepClient(api_client=self.api, num_retries=self.num_retries) - if self.work_api not in ("containers", "jobs"): - raise Exception("Unsupported API '%s'" % self.work_api) + for api in ["jobs", "containers"]: + try: + methods = self.api._rootDesc.get('resources')[api]['methods'] + if ('httpMethod' in methods['create'] and + (work_api == api or work_api is None)): + self.work_api = api + 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) def arv_make_tool(self, toolpath_object, **kwargs): kwargs["work_api"] = self.work_api @@ -108,7 +124,8 @@ class ArvCwlRunner(object): try: self.cond.acquire() j = self.processes[uuid] - logger.info("Job %s (%s) is %s", j.name, uuid, event["properties"]["new_attributes"]["state"]) + txt = self.work_api[0].upper() + self.work_api[1:-1] + logger.info("%s %s (%s) is %s", txt, j.name, uuid, event["properties"]["new_attributes"]["state"]) with Perf(metrics, "done %s" % j.name): j.done(event["properties"]["new_attributes"]) self.cond.notify() @@ -178,17 +195,24 @@ class ArvCwlRunner(object): generatemapper = FinalOutputPathMapper(files, "", "", separateDirs=False) - final = arvados.collection.Collection() + final = arvados.collection.Collection(api_client=self.api, + keep_client=self.keep_client, + num_retries=self.num_retries) srccollections = {} for k,v in generatemapper.items(): sp = k.split("/") srccollection = sp[0][5:] if srccollection not in srccollections: - srccollections[srccollection] = arvados.collection.CollectionReader(srccollection) + srccollections[srccollection] = arvados.collection.CollectionReader( + srccollection, + api_client=self.api, + keep_client=self.keep_client, + num_retries=self.num_retries) reader = srccollections[srccollection] try: - final.copy("/".join(sp[1:]), v.target, source_collection=reader, overwrite=False) + srcpath = "/".join(sp[1:]) if len(sp) > 1 else "." + final.copy(srcpath, v.target, source_collection=reader, overwrite=False) except IOError as e: logger.warn("While preparing output collection: %s", e) @@ -202,14 +226,34 @@ class ArvCwlRunner(object): adjustFileObjs(outputObj, rewrite) with final.open("cwl.output.json", "w") as f: - json.dump(outputObj, f, sort_keys=True, indent=4) + json.dump(outputObj, f, sort_keys=True, indent=4, separators=(',',': ')) final.save_new(name=name, owner_uuid=self.project_uuid, ensure_unique_name=True) - logger.info("Final output collection %s (%s)", final.portable_data_hash(), final.manifest_locator()) + logger.info("Final output collection %s \"%s\" (%s)", final.portable_data_hash(), + final.api_response()["name"], + final.manifest_locator()) self.final_output_collection = final + def set_crunch_output(self): + if self.work_api == "containers": + try: + current = self.api.containers().current().execute(num_retries=self.num_retries) + self.api.containers().update(uuid=current['uuid'], + body={ + 'output': self.final_output_collection.portable_data_hash(), + }).execute(num_retries=self.num_retries) + except Exception as e: + logger.info("Setting container output: %s", e) + elif self.work_api == "jobs" and "TASK_UUID" in os.environ: + self.api.job_tasks().update(uuid=os.environ["TASK_UUID"], + body={ + 'output': self.final_output_collection.portable_data_hash(), + 'success': self.final_status == "success", + 'progress':1.0 + }).execute(num_retries=self.num_retries) + def arv_executor(self, tool, job_order, **kwargs): self.debug = kwargs.get("debug") @@ -218,7 +262,9 @@ class ArvCwlRunner(object): useruuid = self.api.users().current().execute()["uuid"] self.project_uuid = kwargs.get("project_uuid") if kwargs.get("project_uuid") else useruuid self.pipeline = None - make_fs_access = kwargs.get("make_fs_access") or partial(CollectionFsAccess, api_client=self.api) + 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"): @@ -249,6 +295,8 @@ class ArvCwlRunner(object): kwargs["docker_outdir"] = "$(task.outdir)" kwargs["tmpdir"] = "$(task.tmpdir)" + upload_instance(self, shortname(tool.tool["id"]), tool, job_order) + runnerjob = None if kwargs.get("submit"): if self.work_api == "containers": @@ -257,9 +305,9 @@ class ArvCwlRunner(object): self.output_callback, **kwargs).next() else: - runnerjob = RunnerContainer(self, tool, job_order, kwargs.get("enable_reuse")) + runnerjob = RunnerContainer(self, tool, job_order, kwargs.get("enable_reuse"), self.output_name) else: - runnerjob = RunnerJob(self, tool, job_order, kwargs.get("enable_reuse")) + runnerjob = RunnerJob(self, tool, job_order, kwargs.get("enable_reuse"), self.output_name) if not kwargs.get("submit") and "cwl_runner_job" not in kwargs and not self.work_api == "containers": # Create pipeline for local run @@ -272,7 +320,7 @@ class ArvCwlRunner(object): logger.info("Pipeline instance %s", self.pipeline["uuid"]) if runnerjob and not kwargs.get("wait"): - runnerjob.run() + runnerjob.run(wait=kwargs.get("wait")) return runnerjob.uuid self.poll_api = arvados.api('v1') @@ -334,19 +382,22 @@ class ArvCwlRunner(object): if self.final_status == "UnsupportedRequirement": raise UnsupportedRequirement("Check log for details.") - if self.final_status != "success": - raise WorkflowException("Workflow failed.") - if self.final_output is None: raise WorkflowException("Workflow did not return a result.") - if kwargs.get("submit"): + if kwargs.get("submit") and isinstance(runnerjob, Runner): logger.info("Final output collection %s", runnerjob.final_output) else: - self.make_output_collection("Output of %s" % (shortname(tool.tool["id"])), - self.final_output) + if self.output_name is None: + self.output_name = "Output of %s" % (shortname(tool.tool["id"])) + self.make_output_collection(self.output_name, self.final_output) + self.set_crunch_output() + + if self.final_status != "success": + raise WorkflowException("Workflow failed.") if kwargs.get("compute_checksum"): + adjustDirObjs(self.final_output, partial(getListing, self.fs_access)) adjustFileObjs(self.final_output, partial(compute_checksums, self.fs_access)) return self.final_output @@ -359,7 +410,7 @@ def versionstring(): arvpkg = pkg_resources.require("arvados-python-client") cwlpkg = pkg_resources.require("cwltool") - return "%s %s, %s %s, %s %s" % (sys.argv[0], arvcwlpkg[0].version, + return "%s %s %s, %s %s, %s %s" % (sys.argv[0], __version__, arvcwlpkg[0].version, "arvados-python-client", arvpkg[0].version, "cwltool", cwlpkg[0].version) @@ -396,6 +447,7 @@ def arg_parser(): # type: () -> argparse.ArgumentParser help="") 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("--ignore-docker-for-reuse", action="store_true", help="Ignore Docker image version when deciding whether to reuse past jobs.", default=False) @@ -417,7 +469,7 @@ def arg_parser(): # type: () -> argparse.ArgumentParser parser.add_argument("--api", type=str, default=None, dest="work_api", - help="Select work submission API, one of 'jobs' or 'containers'.") + help="Select work submission API, one of 'jobs' or 'containers'. Default is 'jobs' if that API is available, otherwise 'containers'.") parser.add_argument("--compute-checksum", action="store_true", default=False, help="Compute checksum of contents while collecting outputs", @@ -433,13 +485,14 @@ def add_arv_hints(): res = pkg_resources.resource_stream(__name__, 'arv-cwl-schema.yml') cache["http://arvados.org/cwl"] = res.read() res.close() - _, cwlnames, _, _ = cwltool.process.get_schema("v1.0") + document_loader, cwlnames, _, _ = cwltool.process.get_schema("v1.0") _, extnames, _, _ = schema_salad.schema.load_schema("http://arvados.org/cwl", cache=cache) for n in extnames.names: if not cwlnames.has_name("http://arvados.org/cwl#"+n, ""): cwlnames.add_name("http://arvados.org/cwl#"+n, "", extnames.get_name(n, "")) + document_loader.idx["http://arvados.org/cwl#"+n] = {} -def main(args, stdout, stderr, api_client=None): +def main(args, stdout, stderr, api_client=None, keep_client=None): parser = arg_parser() job_order_object = None @@ -452,7 +505,7 @@ def main(args, stdout, stderr, api_client=None): try: if api_client is None: api_client=arvados.api('v1', model=OrderedJsonModel()) - runner = ArvCwlRunner(api_client, work_api=arvargs.work_api) + runner = ArvCwlRunner(api_client, work_api=arvargs.work_api, keep_client=keep_client, output_name=arvargs.output_name) except Exception as e: logger.error(e) return 1