X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/91dc5f1d7f5ad9eb2640f6089e2d0476cbf87c8e..8d4ec10fc26d93d282845c789cd61da79e4b2836:/sdk/cwl/arvados_cwl/arvcontainer.py diff --git a/sdk/cwl/arvados_cwl/arvcontainer.py b/sdk/cwl/arvados_cwl/arvcontainer.py index a3220f97b7..827e92d679 100644 --- a/sdk/cwl/arvados_cwl/arvcontainer.py +++ b/sdk/cwl/arvados_cwl/arvcontainer.py @@ -2,6 +2,10 @@ import logging import json import os +import ruamel.yaml as yaml + +from schema_salad.sourceline import SourceLine + from cwltool.errors import WorkflowException from cwltool.process import get_feature, UnsupportedRequirement, shortname from cwltool.pathmapper import adjustFiles @@ -11,7 +15,8 @@ import arvados.collection from .arvdocker import arv_docker_get_image from . import done -from .runner import Runner +from .runner import Runner, arvados_jobs_image +from .fsaccess import CollectionFetcher logger = logging.getLogger('arvados.cwl-runner') @@ -34,7 +39,8 @@ class ArvadosContainer(object): "output_path": self.outdir, "cwd": self.outdir, "priority": 1, - "state": "Committed" + "state": "Committed", + "properties": {} } runtime_constraints = {} mounts = { @@ -42,6 +48,7 @@ class ArvadosContainer(object): "kind": "tmp" } } + scheduling_parameters = {} dirs = set() for f in self.pathmapper.files(): @@ -61,17 +68,17 @@ class ArvadosContainer(object): } if self.generatefiles["listing"]: - raise UnsupportedRequirement("Generate files not supported") + raise SourceLine(self.tool.get_requirement("InitialWorkDirRequirement")[0], None, UnsupportedRequirement).makeError("InitialWorkDirRequirement not supported with --api=containers") container_request["environment"] = {"TMPDIR": self.tmpdir, "HOME": self.outdir} if self.environment: container_request["environment"].update(self.environment) if self.stdin: - raise UnsupportedRequirement("Stdin redirection currently not suppported") + raise SourceLine(self.tool.tool, "stdin", UnsupportedRequirement).makeError("Stdin redirection currently not suppported") if self.stderr: - raise UnsupportedRequirement("Stderr redirection currently not suppported") + raise SourceLine(self.tool.tool, "stderr", UnsupportedRequirement).makeError("Stderr redirection currently not suppported") if self.stdout: mounts["stdout"] = {"kind": "file", @@ -79,7 +86,7 @@ class ArvadosContainer(object): (docker_req, docker_is_req) = get_feature(self, "DockerRequirement") if not docker_req: - docker_req = {"dockerImageId": "arvados/jobs"} + docker_req = {"dockerImageId": arvados_jobs_image(self.arvrunner)} container_request["container_image"] = arv_docker_get_image(self.arvrunner.api, docker_req, @@ -97,23 +104,34 @@ class ArvadosContainer(object): runtime_req, _ = get_feature(self, "http://arvados.org/cwl#RuntimeConstraints") if runtime_req: - logger.warn("RuntimeConstraints not yet supported by container API") + if "keep_cache" in runtime_req: + runtime_constraints["keep_cache_ram"] = runtime_req["keep_cache"] partition_req, _ = get_feature(self, "http://arvados.org/cwl#PartitionRequirement") if partition_req: - runtime_constraints["partition"] = aslist(partition_req["partition"]) + scheduling_parameters["partitions"] = aslist(partition_req["partition"]) container_request["mounts"] = mounts container_request["runtime_constraints"] = runtime_constraints + container_request["use_existing"] = kwargs.get("enable_reuse", True) + container_request["scheduling_parameters"] = scheduling_parameters + + if kwargs.get("runnerjob", "").startswith("arvwf:"): + wfuuid = kwargs["runnerjob"][6:kwargs["runnerjob"].index("#")] + wfrecord = self.arvrunner.api.workflows().get(uuid=wfuuid).execute(num_retries=self.arvrunner.num_retries) + if container_request["name"] == "main": + container_request["name"] = wfrecord["name"] + container_request["properties"]["template_uuid"] = wfuuid try: response = self.arvrunner.api.container_requests().create( body=container_request ).execute(num_retries=self.arvrunner.num_retries) - self.arvrunner.processes[response["container_uuid"]] = self + self.uuid = response["uuid"] + self.arvrunner.processes[self.uuid] = self - logger.info("Container %s (%s) request state is %s", self.name, response["uuid"], response["state"]) + logger.info("Container request %s (%s) state is %s", self.name, response["uuid"], response["state"]) if response["state"] == "Final": self.done(response) @@ -123,8 +141,11 @@ class ArvadosContainer(object): 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: @@ -138,17 +159,14 @@ 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 container outputs:\n%s", e, exc_info=(e if self.arvrunner.debug else False)) - processStatus = "permanentFail" - except Exception as e: - logger.exception("Got unknown exception while collecting job 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"]] @@ -166,46 +184,18 @@ class RunnerContainer(Runner): workflowmapper = super(RunnerContainer, self).arvados_job_spec(dry_run=dry_run, pull_image=pull_image, **kwargs) - with arvados.collection.Collection(api_client=self.arvrunner.api, - keep_client=self.arvrunner.keep_client, - num_retries=self.arvrunner.num_retries) as jobobj: - with jobobj.open("cwl.input.json", "w") as f: - 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" - - container_image = arv_docker_get_image(self.arvrunner.api, - {"dockerImageId": "arvados/jobs"}, - pull_image, - self.arvrunner.project_uuid) - - command = ["arvados-cwl-runner", "--local", "--api=containers"] - if self.output_name: - command.append("--output-name=" + self.output_name) - command.extend([workflowpath, jobpath]) - - return { - "command": command, + container_req = { "owner_uuid": self.arvrunner.project_uuid, "name": self.name, "output_path": "/var/spool/cwl", "cwd": "/var/spool/cwl", "priority": 1, "state": "Committed", - "container_image": container_image, + "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() + "/var/lib/cwl/cwl.input.json": { + "kind": "json", + "content": self.job_order }, "stdout": { "kind": "file", @@ -218,11 +208,53 @@ class RunnerContainer(Runner): }, "runtime_constraints": { "vcpus": 1, - "ram": 1024*1024*256, + "ram": 1024*1024 * self.submit_runner_ram, "API": True - } + }, + "properties": {} } + 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" + wfuuid = workflowcollection[6:workflowcollection.index("#")] + wfrecord = self.arvrunner.api.workflows().get(uuid=wfuuid).execute(num_retries=self.arvrunner.num_retries) + wfobj = yaml.safe_load(wfrecord["definition"]) + if container_req["name"].startswith("arvwf:"): + container_req["name"] = wfrecord["name"] + container_req["mounts"]["/var/lib/cwl/workflow.json"] = { + "kind": "json", + "json": wfobj + } + container_req["properties"]["template_uuid"] = wfuuid + + 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, "/var/lib/cwl/cwl.input.json"]) + + container_req["command"] = command + + return container_req + + def run(self, *args, **kwargs): kwargs["keepprefix"] = "keep:" job_spec = self.arvados_job_spec(*args, **kwargs) @@ -233,9 +265,23 @@ class RunnerContainer(Runner): ).execute(num_retries=self.arvrunner.num_retries) self.uuid = response["uuid"] - self.arvrunner.processes[response["container_uuid"]] = self + self.arvrunner.processes[self.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"]]