X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/a07c8bfac95524f3074d11cda0d6689b5e7bf9cc..10b3fe2ae3a37ee473684177aa6e4e9f090a230e:/sdk/cwl/arvados_cwl/arvcontainer.py diff --git a/sdk/cwl/arvados_cwl/arvcontainer.py b/sdk/cwl/arvados_cwl/arvcontainer.py index 1fda412217..20601243b1 100644 --- a/sdk/cwl/arvados_cwl/arvcontainer.py +++ b/sdk/cwl/arvados_cwl/arvcontainer.py @@ -2,18 +2,24 @@ 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 +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.""" @@ -34,7 +40,8 @@ class ArvadosContainer(object): "output_path": self.outdir, "cwd": self.outdir, "priority": 1, - "state": "Committed" + "state": "Committed", + "properties": {} } runtime_constraints = {} mounts = { @@ -46,23 +53,54 @@ 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) + pdh, path = res[5:].split("/", 1) + if pdh not in dirs: mounts[p] = { "kind": "collection", - "portable_data_hash": p[6:] + "portable_data_hash": pdh } - - if self.generatefiles["listing"]: - raise UnsupportedRequirement("Generate files not supported") + if path: + mounts[p]["path"] = path + + 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: @@ -80,7 +118,7 @@ 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, @@ -99,7 +137,7 @@ 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: @@ -110,29 +148,37 @@ 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 ).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) - - logger.info("Container request %s (%s) state is %s with container %s %s", self.name, response["uuid"], response["state"], container["uuid"], container["state"]) + self.uuid = response["uuid"] + self.arvrunner.processes[self.uuid] = self - if container["state"] in ("Complete", "Cancelled"): - self.done(container) + 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): 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: @@ -146,27 +192,27 @@ 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" - - self.output_callback(outputs, processStatus) + 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"]: + 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): @@ -179,52 +225,20 @@ 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) - - 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" - - 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]) + adjustDirObjs(self.job_order, trim_listing) - 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": arvados_jobs_image(self.arvrunner), + "container_image": arvados_jobs_image(self.arvrunner, self.jobs_image), "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", @@ -237,11 +251,57 @@ class RunnerContainer(Runner): }, "runtime_constraints": { "vcpus": 1, - "ram": 1024*1024*256, + "ram": 1024*1024 * self.submit_runner_ram, "API": True - } + }, + "properties": {} } + 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 + } + else: + packed = packed_workflow(self.arvrunner, self.tool) + workflowpath = "/var/lib/cwl/workflow.json#main" + container_req["mounts"]["/var/lib/cwl/workflow.json"] = { + "kind": "json", + "content": packed + } + 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", "--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 + + return container_req + + def run(self, *args, **kwargs): kwargs["keepprefix"] = "keep:" job_spec = self.arvados_job_spec(*args, **kwargs) @@ -252,9 +312,23 @@ 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"]) + logger.info("%s submitted container %s", self.arvrunner.label(self), 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("%s while getting runner container: %s", self.arvrunner.label(self), e) + self.arvrunner.output_callback({}, "permanentFail") + else: + super(RunnerContainer, self).done(container) + finally: + if record["uuid"] in self.arvrunner.processes: + del self.arvrunner.processes[record["uuid"]]