X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/0d62edcb9d25bf4dcdb20d8872ea7b438e12fc59..8ed521f7fd1e48e1e415125745ed8c6627a62c91:/sdk/cwl/arvados_cwl/arvcontainer.py diff --git a/sdk/cwl/arvados_cwl/arvcontainer.py b/sdk/cwl/arvados_cwl/arvcontainer.py index 2c2aa7ff26..3090936121 100644 --- a/sdk/cwl/arvados_cwl/arvcontainer.py +++ b/sdk/cwl/arvados_cwl/arvcontainer.py @@ -6,17 +6,20 @@ import ruamel.yaml as yaml from cwltool.errors import WorkflowException from cwltool.process import get_feature, UnsupportedRequirement, shortname -from cwltool.pathmapper import adjustFiles +from cwltool.pathmapper import adjustFiles, adjustDirObjs from cwltool.utils import aslist import arvados.collection from .arvdocker import arv_docker_get_image from . import done -from .runner import Runner, arvados_jobs_image +from .runner import Runner, arvados_jobs_image, packed_workflow, trim_listing from .fsaccess import CollectionFetcher +from .pathmapper import NoFollowPathMapper +from .perf import Perf logger = logging.getLogger('arvados.cwl-runner') +metrics = logging.getLogger('arvados.cwl-runner.metrics') class ArvadosContainer(object): """Submit and manage a Crunch container request for executing a CWL CommandLineTool.""" @@ -37,7 +40,8 @@ class ArvadosContainer(object): "output_path": self.outdir, "cwd": self.outdir, "priority": 1, - "state": "Committed" + "state": "Committed", + "properties": {} } runtime_constraints = {} mounts = { @@ -49,23 +53,61 @@ class ArvadosContainer(object): dirs = set() for f in self.pathmapper.files(): - _, p, tp = self.pathmapper.mapper(f) - if tp == "Directory" and '/' not in p[6:]: + pdh, p, tp = self.pathmapper.mapper(f) + if tp == "Directory" and '/' not in pdh: mounts[p] = { "kind": "collection", - "portable_data_hash": p[6:] + "portable_data_hash": pdh[5:] } - dirs.add(p[6:]) + dirs.add(pdh) + for f in self.pathmapper.files(): - _, p, tp = self.pathmapper.mapper(f) - if p[6:].split("/")[0] not in dirs: + res, p, tp = self.pathmapper.mapper(f) + if res.startswith("keep:"): + res = res[5:] + elif res.startswith("/keep/"): + res = res[6:] + else: + continue + sp = res.split("/", 1) + pdh = sp[0] + if pdh not in dirs: mounts[p] = { "kind": "collection", - "portable_data_hash": p[6:] + "portable_data_hash": pdh } - - if self.generatefiles["listing"]: - raise UnsupportedRequirement("InitialWorkDirRequirement not supported with --api=containers") + if len(sp) == 2: + mounts[p]["path"] = sp[1] + + with Perf(metrics, "generatefiles %s" % self.name): + if self.generatefiles["listing"]: + vwd = arvados.collection.Collection(api_client=self.arvrunner.api, + keep_client=self.arvrunner.keep_client, + num_retries=self.arvrunner.num_retries) + generatemapper = NoFollowPathMapper([self.generatefiles], "", "", + separateDirs=False) + + with Perf(metrics, "createfiles %s" % self.name): + for f, p in generatemapper.items(): + if not p.target: + pass + elif p.type in ("File", "Directory"): + source, path = self.arvrunner.fs_access.get_collection(p.resolved) + vwd.copy(path, p.target, source_collection=source) + elif p.type == "CreateFile": + with vwd.open(p.target, "w") as n: + n.write(p.resolved.encode("utf-8")) + + with Perf(metrics, "generatefiles.save_new %s" % self.name): + vwd.save_new() + + for f, p in generatemapper.items(): + if not p.target: + continue + mountpoint = "%s/%s" % (self.outdir, p.target) + mounts[mountpoint] = {"kind": "collection", + "portable_data_hash": vwd.portable_data_hash(), + "path": p.target} container_request["environment"] = {"TMPDIR": self.tmpdir, "HOME": self.outdir} if self.environment: @@ -83,7 +125,7 @@ class ArvadosContainer(object): (docker_req, docker_is_req) = get_feature(self, "DockerRequirement") if not docker_req: - docker_req = {"dockerImageId": arvados_jobs_image(self.arvrunner)} + docker_req = {"dockerImageId": "arvados/jobs"} container_request["container_image"] = arv_docker_get_image(self.arvrunner.api, docker_req, @@ -102,7 +144,7 @@ class ArvadosContainer(object): runtime_req, _ = get_feature(self, "http://arvados.org/cwl#RuntimeConstraints") if runtime_req: if "keep_cache" in runtime_req: - runtime_constraints["keep_cache_ram"] = runtime_req["keep_cache"] + runtime_constraints["keep_cache_ram"] = runtime_req["keep_cache"] * 2**20 partition_req, _ = get_feature(self, "http://arvados.org/cwl#PartitionRequirement") if partition_req: @@ -113,6 +155,13 @@ class ArvadosContainer(object): 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 @@ -121,12 +170,13 @@ class ArvadosContainer(object): self.uuid = response["uuid"] self.arvrunner.processes[self.uuid] = self - logger.info("Container request %s (%s) state is %s", self.name, response["uuid"], response["state"]) - if response["state"] == "Final": + logger.info("%s reused container %s", self.arvrunner.label(self), response["container_uuid"]) self.done(response) + else: + logger.info("%s %s state is %s", self.arvrunner.label(self), response["uuid"], response["state"]) except Exception as e: - logger.error("Got error %s" % str(e)) + logger.error("%s got error %s" % (self.arvrunner.label(self), str(e))) self.output_callback({}, "permanentFail") def done(self, record): @@ -149,17 +199,27 @@ class ArvadosContainer(object): else: processStatus = "permanentFail" - outputs = {} + if processStatus == "permanentFail": + logc = arvados.collection.CollectionReader(container["log"], + api_client=self.arvrunner.api, + keep_client=self.arvrunner.keep_client, + num_retries=self.arvrunner.num_retries) + done.logtail(logc, logger, "%s error log:" % self.arvrunner.label(self)) + 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) + outputs = done.done_outputs(self, container, "/tmp", self.outdir, "/keep") + except WorkflowException as e: + logger.error("%s unable to collect output from %s:\n%s", + self.arvrunner.label(self), container["output"], e, exc_info=(e if self.arvrunner.debug else False)) + processStatus = "permanentFail" + except Exception as e: + logger.exception("%s while getting output object: %s", self.arvrunner.label(self), e) + processStatus = "permanentFail" finally: - del self.arvrunner.processes[record["uuid"]] + self.output_callback(outputs, processStatus) + if record["uuid"] in self.arvrunner.processes: + del self.arvrunner.processes[record["uuid"]] class RunnerContainer(Runner): @@ -172,7 +232,7 @@ class RunnerContainer(Runner): the +body+ argument to container_requests().create(). """ - workflowmapper = super(RunnerContainer, self).arvados_job_spec(dry_run=dry_run, pull_image=pull_image, **kwargs) + adjustDirObjs(self.job_order, trim_listing) container_req = { "owner_uuid": self.arvrunner.project_uuid, @@ -181,7 +241,7 @@ class RunnerContainer(Runner): "cwd": "/var/spool/cwl", "priority": 1, "state": "Committed", - "container_image": arvados_jobs_image(self.arvrunner), + "container_image": arvados_jobs_image(self.arvrunner, self.jobs_image), "mounts": { "/var/lib/cwl/cwl.input.json": { "kind": "json", @@ -204,40 +264,44 @@ class RunnerContainer(Runner): "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"]) + if self.tool.tool.get("id", "").startswith("keep:"): + sp = self.tool.tool["id"].split('/') + workflowcollection = sp[0][5:] + workflowname = "/".join(sp[1:]) 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:"): + } + else: + packed = packed_workflow(self.arvrunner, self.tool) workflowpath = "/var/lib/cwl/workflow.json#main" - wfuuid = workflowcollection[6:workflowcollection.index("#")] - wfrecord = self.arvrunner.api.workflows().get(uuid=wfuuid).execute() - 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 + "content": packed } - container_req["properties"]["template_uuid"] = wfuuid + if self.tool.tool.get("id", "").startswith("arvwf:"): + container_req["properties"]["template_uuid"] = self.tool.tool["id"][6:33] - command = ["arvados-cwl-runner", "--local", "--api=containers"] + command = ["arvados-cwl-runner", "--local", "--api=containers", "--no-log-timestamps"] if self.output_name: command.append("--output-name=" + self.output_name) + container_req["output_name"] = self.output_name if self.output_tags: command.append("--output-tags=" + self.output_tags) + if kwargs.get("debug"): + command.append("--debug") + if self.enable_reuse: command.append("--enable-reuse") else: command.append("--disable-reuse") + if self.on_error: + command.append("--on-error=" + self.on_error) + command.extend([workflowpath, "/var/lib/cwl/cwl.input.json"]) container_req["command"] = command @@ -257,7 +321,7 @@ class RunnerContainer(Runner): self.uuid = response["uuid"] self.arvrunner.processes[self.uuid] = self - logger.info("Submitted container %s", response["uuid"]) + logger.info("%s submitted container %s", self.arvrunner.label(self), response["uuid"]) if response["state"] == "Final": self.done(response) @@ -268,10 +332,10 @@ class RunnerContainer(Runner): uuid=record["container_uuid"] ).execute(num_retries=self.arvrunner.num_retries) except Exception as e: - logger.exception("While getting runner container: %s", e) + logger.exception("%s while getting runner container: %s", self.arvrunner.label(self), 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"]] + if record["uuid"] in self.arvrunner.processes: + del self.arvrunner.processes[record["uuid"]]