X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/9c3cc0f61751720cfdea62717934746d1aa32b72..aaa964ac44f923985d6a6eb40c179f62c13ca8ce:/sdk/cwl/arvados_cwl/arvcontainer.py diff --git a/sdk/cwl/arvados_cwl/arvcontainer.py b/sdk/cwl/arvados_cwl/arvcontainer.py index 1fda412217..cea2b1091d 100644 --- a/sdk/cwl/arvados_cwl/arvcontainer.py +++ b/sdk/cwl/arvados_cwl/arvcontainer.py @@ -2,6 +2,8 @@ import logging import json import os +import ruamel.yaml as yaml + from cwltool.errors import WorkflowException from cwltool.process import get_feature, UnsupportedRequirement, shortname from cwltool.pathmapper import adjustFiles @@ -12,6 +14,7 @@ import arvados.collection from .arvdocker import arv_docker_get_image from . import done from .runner import Runner, arvados_jobs_image +from .fsaccess import CollectionFetcher logger = logging.getLogger('arvados.cwl-runner') @@ -62,7 +65,7 @@ class ArvadosContainer(object): } if self.generatefiles["listing"]: - raise UnsupportedRequirement("Generate files not supported") + raise UnsupportedRequirement("InitialWorkDirRequirement not supported with --api=containers") container_request["environment"] = {"TMPDIR": self.tmpdir, "HOME": self.outdir} if self.environment: @@ -115,24 +118,23 @@ class ArvadosContainer(object): body=container_request ).execute(num_retries=self.arvrunner.num_retries) - self.arvrunner.processes[response["container_uuid"]] = self - - container = self.arvrunner.api.containers().get( - uuid=response["container_uuid"] - ).execute(num_retries=self.arvrunner.num_retries) + self.arvrunner.processes[response["uuid"]] = self - logger.info("Container request %s (%s) state is %s with container %s %s", self.name, response["uuid"], response["state"], container["uuid"], container["state"]) + logger.info("Container request %s (%s) state is %s", self.name, response["uuid"], response["state"]) - if container["state"] in ("Complete", "Cancelled"): - self.done(container) + if response["state"] == "Final": + self.done(response) except Exception as e: logger.error("Got error %s" % str(e)) self.output_callback({}, "permanentFail") def done(self, record): try: - if record["state"] == "Complete": - rcode = record["exit_code"] + container = self.arvrunner.api.containers().get( + uuid=record["container_uuid"] + ).execute(num_retries=self.arvrunner.num_retries) + if container["state"] == "Complete": + rcode = container["exit_code"] if self.successCodes and rcode in self.successCodes: processStatus = "success" elif self.temporaryFailCodes and rcode in self.temporaryFailCodes: @@ -146,24 +148,14 @@ class ArvadosContainer(object): else: processStatus = "permanentFail" - try: - outputs = {} - if record["output"]: - outputs = done.done(self, record, "/tmp", self.outdir, "/keep") - except WorkflowException as e: - logger.error("Error while collecting output for container %s:\n%s", self.name, e, exc_info=(e if self.arvrunner.debug else False)) - processStatus = "permanentFail" - except Exception as e: - logger.exception("Got unknown exception while collecting output for container %s:", self.name) - processStatus = "permanentFail" - - # Note: Currently, on error output_callback is expecting an empty dict, - # anything else will fail. - if not isinstance(outputs, dict): - logger.error("Unexpected output type %s '%s'", type(outputs), outputs) - outputs = {} - processStatus = "permanentFail" + outputs = {} + if container["output"]: + try: + outputs = done.done_outputs(self, container, "/tmp", self.outdir, "/keep") + except Exception as e: + logger.error("Got error %s" % str(e)) + self.output_callback({}, "permanentFail") self.output_callback(outputs, processStatus) finally: del self.arvrunner.processes[record["uuid"]] @@ -188,28 +180,9 @@ class RunnerContainer(Runner): json.dump(self.job_order, f, sort_keys=True, indent=4) jobobj.save_new(owner_uuid=self.arvrunner.project_uuid) - workflowname = os.path.basename(self.tool.tool["id"]) - workflowpath = "/var/lib/cwl/workflow/%s" % workflowname - workflowcollection = workflowmapper.mapper(self.tool.tool["id"])[1] - workflowcollection = workflowcollection[5:workflowcollection.index('/')] jobpath = "/var/lib/cwl/job/cwl.input.json" - command = ["arvados-cwl-runner", "--local", "--api=containers"] - if self.output_name: - command.append("--output-name=" + self.output_name) - - if self.output_tags: - command.append("--output-tags=" + self.output_tags) - - if self.enable_reuse: - command.append("--enable-reuse") - else: - command.append("--disable-reuse") - - command.extend([workflowpath, jobpath]) - - return { - "command": command, + container_req = { "owner_uuid": self.arvrunner.project_uuid, "name": self.name, "output_path": "/var/spool/cwl", @@ -218,10 +191,6 @@ class RunnerContainer(Runner): "state": "Committed", "container_image": arvados_jobs_image(self.arvrunner), "mounts": { - "/var/lib/cwl/workflow": { - "kind": "collection", - "portable_data_hash": "%s" % workflowcollection - }, jobpath: { "kind": "collection", "portable_data_hash": "%s/cwl.input.json" % jobobj.portable_data_hash() @@ -237,11 +206,50 @@ class RunnerContainer(Runner): }, "runtime_constraints": { "vcpus": 1, - "ram": 1024*1024*256, + "ram": 1024*1024 * self.submit_runner_ram, "API": True } } + workflowcollection = workflowmapper.mapper(self.tool.tool["id"])[1] + if workflowcollection.startswith("keep:"): + workflowcollection = workflowcollection[5:workflowcollection.index('/')] + workflowname = os.path.basename(self.tool.tool["id"]) + workflowpath = "/var/lib/cwl/workflow/%s" % workflowname + container_req["mounts"]["/var/lib/cwl/workflow"] = { + "kind": "collection", + "portable_data_hash": "%s" % workflowcollection + } + elif workflowcollection.startswith("arvwf:"): + workflowpath = "/var/lib/cwl/workflow.json#main" + fetcher = CollectionFetcher({}, None, + api_client=self.arvrunner.api, + keep_client=self.arvrunner.keep_client) + wfobj = yaml.safe_load(fetcher.fetch_text(workflowcollection)) + container_req["mounts"]["/var/lib/cwl/workflow.json"] = { + "kind": "json", + "json": wfobj + } + + command = ["arvados-cwl-runner", "--local", "--api=containers"] + if self.output_name: + command.append("--output-name=" + self.output_name) + + if self.output_tags: + command.append("--output-tags=" + self.output_tags) + + if self.enable_reuse: + command.append("--enable-reuse") + else: + command.append("--disable-reuse") + + command.extend([workflowpath, jobpath]) + + container_req["command"] = command + + return container_req + + def run(self, *args, **kwargs): kwargs["keepprefix"] = "keep:" job_spec = self.arvados_job_spec(*args, **kwargs) @@ -252,9 +260,23 @@ class RunnerContainer(Runner): ).execute(num_retries=self.arvrunner.num_retries) self.uuid = response["uuid"] - self.arvrunner.processes[response["container_uuid"]] = self + self.arvrunner.processes[response["uuid"]] = self logger.info("Submitted container %s", response["uuid"]) - if response["state"] in ("Complete", "Failed", "Cancelled"): + if response["state"] == "Final": self.done(response) + + def done(self, record): + try: + container = self.arvrunner.api.containers().get( + uuid=record["container_uuid"] + ).execute(num_retries=self.arvrunner.num_retries) + except Exception as e: + logger.exception("While getting runner container: %s", e) + self.arvrunner.output_callback({}, "permanentFail") + del self.arvrunner.processes[record["uuid"]] + else: + super(RunnerContainer, self).done(container) + finally: + del self.arvrunner.processes[record["uuid"]]