X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/6c43be47cb3756a0e6ffc924572259d1a1c8f2c3..07e339237850f463d18ce6c06446441c6684fc58:/sdk/cwl/arvados_cwl/arvcontainer.py diff --git a/sdk/cwl/arvados_cwl/arvcontainer.py b/sdk/cwl/arvados_cwl/arvcontainer.py index be1140469a..413435db23 100644 --- a/sdk/cwl/arvados_cwl/arvcontainer.py +++ b/sdk/cwl/arvados_cwl/arvcontainer.py @@ -3,7 +3,8 @@ import json import os from cwltool.errors import WorkflowException -from cwltool.process import get_feature, adjustFiles, UnsupportedRequirement, shortname +from cwltool.process import get_feature, UnsupportedRequirement, shortname +from cwltool.pathmapper import adjustFiles import arvados.collection @@ -14,11 +15,12 @@ from .runner import Runner logger = logging.getLogger('arvados.cwl-runner') class ArvadosContainer(object): - """Submit and manage a Crunch job for executing a CWL CommandLineTool.""" + """Submit and manage a Crunch container request for executing a CWL CommandLineTool.""" def __init__(self, runner): self.arvrunner = runner self.running = False + self.uuid = None def update_pipeline_component(self, r): pass @@ -28,52 +30,51 @@ class ArvadosContainer(object): "command": self.command_line, "owner_uuid": self.arvrunner.project_uuid, "name": self.name, - "output_path": "/var/spool/cwl", - "cwd": "/var/spool/cwl", + "output_path": self.outdir, + "cwd": self.outdir, "priority": 1, "state": "Committed" } runtime_constraints = {} mounts = { - "/var/spool/cwl": { + self.outdir: { "kind": "tmp" } } + dirs = set() for f in self.pathmapper.files(): - _, p = self.pathmapper.mapper(f) - mounts[p] = { - "kind": "collection", - "portable_data_hash": p[6:] - } + _, p, tp = self.pathmapper.mapper(f) + if tp == "Directory" and '/' not in p[6:]: + mounts[p] = { + "kind": "collection", + "portable_data_hash": p[6:] + } + dirs.add(p[6:]) + for f in self.pathmapper.files(): + _, p, tp = self.pathmapper.mapper(f) + if p[6:].split("/")[0] not in dirs: + mounts[p] = { + "kind": "collection", + "portable_data_hash": p[6:] + } - if self.generatefiles: + if self.generatefiles["listing"]: raise UnsupportedRequirement("Generate files not supported") - vwd = arvados.collection.Collection(api_client=self.arvrunner.api_client) - container_request["task.vwd"] = {} - for t in self.generatefiles: - if isinstance(self.generatefiles[t], dict): - src, rest = self.arvrunner.fs_access.get_collection(self.generatefiles[t]["path"].replace("$(task.keep)/", "keep:")) - vwd.copy(rest, t, source_collection=src) - else: - with vwd.open(t, "w") as f: - f.write(self.generatefiles[t]) - vwd.save_new() - # TODO - # for t in self.generatefiles: - # container_request["task.vwd"][t] = "$(task.keep)/%s/%s" % (vwd.portable_data_hash(), t) - - container_request["environment"] = {"TMPDIR": "/tmp"} + 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") + if self.stderr: + raise UnsupportedRequirement("Stderr redirection currently not suppported") + if self.stdout: mounts["stdout"] = {"kind": "file", - "path": "/var/spool/cwl/%s" % (self.stdout)} + "path": "%s/%s" % (self.outdir, self.stdout)} (docker_req, docker_is_req) = get_feature(self, "DockerRequirement") if not docker_req: @@ -88,7 +89,6 @@ class ArvadosContainer(object): if resources is not None: runtime_constraints["vcpus"] = resources.get("cores", 1) runtime_constraints["ram"] = resources.get("ram") * 2**20 - #runtime_constraints["min_scratch_mb_per_node"] = resources.get("tmpdirSize", 0) + resources.get("outdirSize", 0) container_request["mounts"] = mounts container_request["runtime_constraints"] = runtime_constraints @@ -98,9 +98,9 @@ class ArvadosContainer(object): body=container_request ).execute(num_retries=self.arvrunner.num_retries) - self.arvrunner.jobs[response["container_uuid"]] = self + self.arvrunner.processes[response["container_uuid"]] = self - logger.info("Container %s (%s) request state is %s", self.name, response["container_uuid"], response["state"]) + logger.info("Container %s (%s) request state is %s", self.name, response["uuid"], response["state"]) if response["state"] == "Final": self.done(response) @@ -111,14 +111,24 @@ class ArvadosContainer(object): def done(self, record): try: if record["state"] == "Complete": - processStatus = "success" + rcode = record["exit_code"] + if self.successCodes and rcode in self.successCodes: + processStatus = "success" + elif self.temporaryFailCodes and rcode in self.temporaryFailCodes: + processStatus = "temporaryFail" + elif self.permanentFailCodes and rcode in self.permanentFailCodes: + processStatus = "permanentFail" + elif rcode == 0: + processStatus = "success" + else: + processStatus = "permanentFail" else: processStatus = "permanentFail" try: outputs = {} if record["output"]: - outputs = done.done(self, record, "/tmp", "/var/spool/cwl", "/keep") + 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" @@ -128,18 +138,17 @@ class ArvadosContainer(object): self.output_callback(outputs, processStatus) finally: - del self.arvrunner.jobs[record["uuid"]] + del self.arvrunner.processes[record["uuid"]] class RunnerContainer(Runner): """Submit and manage a container that runs arvados-cwl-runner.""" def arvados_job_spec(self, dry_run=False, pull_image=True, **kwargs): - """Create an Arvados job specification for this workflow. + """Create an Arvados container request for this workflow. - The returned dict can be used to create a job (i.e., passed as - the +body+ argument to jobs().create()), or as a component in - a pipeline template or pipeline instance. + The returned dict can be used to create a container passed as + the +body+ argument to container_requests().create(). """ workflowmapper = super(RunnerContainer, self).arvados_job_spec(dry_run=dry_run, pull_image=pull_image, **kwargs) @@ -161,7 +170,7 @@ class RunnerContainer(Runner): self.arvrunner.project_uuid) return { - "command": ["arvados-cwl-runner", "--local", "--crunch2", workflowpath, jobpath], + "command": ["arvados-cwl-runner", "--local", "--api=containers", workflowpath, jobpath], "owner_uuid": self.arvrunner.project_uuid, "name": self.name, "output_path": "/var/spool/cwl", @@ -170,7 +179,7 @@ class RunnerContainer(Runner): "state": "Committed", "container_image": container_image, "mounts": { - workflowpath: { + "/var/lib/cwl/workflow": { "kind": "collection", "portable_data_hash": "%s" % workflowcollection }, @@ -181,15 +190,21 @@ class RunnerContainer(Runner): "stdout": { "kind": "file", "path": "/var/spool/cwl/cwl.output.json" + }, + "/var/spool/cwl": { + "kind": "collection", + "writable": True } }, "runtime_constraints": { "vcpus": 1, - "ram": 1024*1024*256 + "ram": 1024*1024*256, + "API": True } } def run(self, *args, **kwargs): + kwargs["keepprefix"] = "keep:" job_spec = self.arvados_job_spec(*args, **kwargs) job_spec.setdefault("owner_uuid", self.arvrunner.project_uuid) @@ -198,7 +213,7 @@ class RunnerContainer(Runner): ).execute(num_retries=self.arvrunner.num_retries) self.uuid = response["uuid"] - self.arvrunner.jobs[response["container_uuid"]] = self + self.arvrunner.processes[response["container_uuid"]] = self logger.info("Submitted container %s", response["uuid"])