from .arvcontainer import ArvadosContainer, RunnerContainer
from .arvjob import ArvadosJob, RunnerJob, RunnerTemplate
-from. runner import Runner, upload_instance
+from. runner import Runner, upload_docker, upload_job_order, upload_workflow_deps
from .arvtool import ArvadosCommandTool
from .arvworkflow import ArvadosWorkflow, upload_workflow
from .fsaccess import CollectionFsAccess, CollectionFetcher, collectionResolver
metrics = logging.getLogger('arvados.cwl-runner.metrics')
logger.setLevel(logging.INFO)
+arvados.log_handler.setFormatter(logging.Formatter(
+ '%(asctime)s %(name)s %(levelname)s: %(message)s',
+ '%Y-%m-%d %H:%M:%S'))
class ArvCwlRunner(object):
"""Execute a CWL tool or workflow, submit work (using either jobs or
uuid = event["object_uuid"]
with self.lock:
j = self.processes[uuid]
- logger.info("Job %s (%s) is Running", j.name, uuid)
+ logger.info("%s %s is Running", self.label(j), uuid)
j.running = True
j.update_pipeline_component(event["properties"]["new_attributes"])
elif event["properties"]["new_attributes"]["state"] in ("Complete", "Failed", "Cancelled", "Final"):
try:
self.cond.acquire()
j = self.processes[uuid]
- txt = self.work_api[0].upper() + self.work_api[1:-1]
- logger.info("%s %s (%s) is %s", txt, j.name, uuid, event["properties"]["new_attributes"]["state"])
+ logger.info("%s %s is %s", self.label(j), uuid, event["properties"]["new_attributes"]["state"])
with Perf(metrics, "done %s" % j.name):
j.done(event["properties"]["new_attributes"])
self.cond.notify()
finally:
self.cond.release()
+ def label(self, obj):
+ return "[%s %s]" % (self.work_api[0:-1], obj.name)
+
def poll_states(self):
"""Poll status of jobs or containers listed in the processes dict.
for p in proc_states["items"]:
self.on_message({
- "object_uuid": p["uuid"],
+ "object_uuid": p["uuid"],
"event_type": "update",
"properties": {
"new_attributes": p
keep_client=self.keep_client)
self.fs_access = make_fs_access(kwargs["basedir"])
+ if not kwargs.get("name"):
+ kwargs["name"] = self.name = tool.tool.get("label") or tool.metadata.get("label") or os.path.basename(tool.tool["id"])
+
+ # Upload direct dependencies of workflow steps, get back mapping of files to keep references.
+ # Also uploads docker images.
+ upload_workflow_deps(self, tool)
+
+ # Reload tool object which may have been updated by
+ # upload_workflow_deps
+ tool = self.arv_make_tool(tool.doc_loader.idx[tool.tool["id"]],
+ makeTool=self.arv_make_tool,
+ loader=tool.doc_loader,
+ avsc_names=tool.doc_schema,
+ metadata=tool.metadata)
+
+ # Upload local file references in the job order.
+ job_order = upload_job_order(self, "%s input" % kwargs["name"],
+ tool, job_order)
+
existing_uuid = kwargs.get("update_workflow")
if existing_uuid or kwargs.get("create_workflow"):
+ # Create a pipeline template or workflow record and exit.
if self.work_api == "jobs":
tmpl = RunnerTemplate(self, tool, job_order,
kwargs.get("enable_reuse"),
uuid=existing_uuid,
submit_runner_ram=kwargs.get("submit_runner_ram"),
- name=kwargs.get("name"))
+ name=kwargs["name"])
tmpl.save()
# cwltool.main will write our return value to stdout.
- return tmpl.uuid
- else:
- return upload_workflow(self, tool, job_order,
- self.project_uuid,
- uuid=existing_uuid,
- submit_runner_ram=kwargs.get("submit_runner_ram"),
- name=kwargs.get("name"))
+ return (tmpl.uuid, "success")
+ elif self.work_api == "containers":
+ return (upload_workflow(self, tool, job_order,
+ self.project_uuid,
+ uuid=existing_uuid,
+ submit_runner_ram=kwargs.get("submit_runner_ram"),
+ name=kwargs["name"]),
+ "success")
self.ignore_docker_for_reuse = kwargs.get("ignore_docker_for_reuse")
kwargs["enable_reuse"] = kwargs.get("enable_reuse")
kwargs["use_container"] = True
kwargs["tmpdir_prefix"] = "tmp"
- kwargs["on_error"] = "continue"
kwargs["compute_checksum"] = kwargs.get("compute_checksum")
- if not kwargs["name"]:
- del kwargs["name"]
-
if self.work_api == "containers":
kwargs["outdir"] = "/var/spool/cwl"
kwargs["docker_outdir"] = "/var/spool/cwl"
kwargs["docker_outdir"] = "$(task.outdir)"
kwargs["tmpdir"] = "$(task.tmpdir)"
- upload_instance(self, shortname(tool.tool["id"]), tool, job_order)
-
runnerjob = None
if kwargs.get("submit"):
+ # Submit a runner job to run the workflow for us.
if self.work_api == "containers":
if tool.tool["class"] == "CommandLineTool":
kwargs["runnerjob"] = tool.tool["id"]
+ upload_dependencies(self,
+ kwargs["name"],
+ tool.doc_loader,
+ tool.tool,
+ tool.tool["id"],
+ False)
runnerjob = tool.job(job_order,
self.output_callback,
**kwargs).next()
else:
- runnerjob = RunnerContainer(self, tool, job_order, kwargs.get("enable_reuse"), self.output_name,
- self.output_tags, submit_runner_ram=kwargs.get("submit_runner_ram"),
- name=kwargs.get("name"))
- else:
- runnerjob = RunnerJob(self, tool, job_order, kwargs.get("enable_reuse"), self.output_name,
- self.output_tags, submit_runner_ram=kwargs.get("submit_runner_ram"),
- name=kwargs.get("name"))
+ runnerjob = RunnerContainer(self, tool, job_order, kwargs.get("enable_reuse"),
+ self.output_name,
+ self.output_tags,
+ submit_runner_ram=kwargs.get("submit_runner_ram"),
+ name=kwargs.get("name"),
+ on_error=kwargs.get("on_error"),
+ submit_runner_image=kwargs.get("submit_runner_image"))
+ elif self.work_api == "jobs":
+ runnerjob = RunnerJob(self, tool, job_order, kwargs.get("enable_reuse"),
+ self.output_name,
+ self.output_tags,
+ submit_runner_ram=kwargs.get("submit_runner_ram"),
+ name=kwargs.get("name"),
+ on_error=kwargs.get("on_error"),
+ submit_runner_image=kwargs.get("submit_runner_image"))
- if not kwargs.get("submit") and "cwl_runner_job" not in kwargs and not self.work_api == "containers":
+ if not kwargs.get("submit") and "cwl_runner_job" not in kwargs and self.work_api == "jobs":
# Create pipeline for local run
self.pipeline = self.api.pipeline_instances().create(
body={
if runnerjob and not kwargs.get("wait"):
runnerjob.run(wait=kwargs.get("wait"))
- return runnerjob.uuid
+ return (runnerjob.uuid, "success")
self.poll_api = arvados.api('v1')
self.polling_thread = threading.Thread(target=self.poll_states)
self.final_output, self.final_output_collection = self.make_output_collection(self.output_name, self.output_tags, self.final_output)
self.set_crunch_output()
- if self.final_status != "success":
- raise WorkflowException("Workflow failed.")
-
if kwargs.get("compute_checksum"):
adjustDirObjs(self.final_output, partial(getListing, self.fs_access))
adjustFileObjs(self.final_output, partial(compute_checksums, self.fs_access))
- return self.final_output
+ return (self.final_output, self.final_status)
def versionstring():
exgroup.add_argument("--no-wait", action="store_false", help="Submit workflow runner job and exit.",
default=True, dest="wait")
+ exgroup = parser.add_mutually_exclusive_group()
+ exgroup.add_argument("--log-timestamps", action="store_true", help="Prefix logging lines with timestamp",
+ default=True, dest="log_timestamps")
+ exgroup.add_argument("--no-log-timestamps", action="store_false", help="No timestamp on logging lines",
+ default=True, dest="log_timestamps")
+
parser.add_argument("--api", type=str,
default=None, dest="work_api",
help="Select work submission API, one of 'jobs' or 'containers'. Default is 'jobs' if that API is available, otherwise 'containers'.")
help="RAM (in MiB) required for the workflow runner job (default 1024)",
default=1024)
+ parser.add_argument("--submit-runner-image", type=str,
+ help="Docker image for workflow runner job, default arvados/jobs:%s" % __version__,
+ default=None)
+
parser.add_argument("--name", type=str,
help="Name to use for workflow execution instance.",
default=None)
+ parser.add_argument("--on-error", type=str,
+ help="Desired workflow behavior when a step fails. One of 'stop' or 'continue'. "
+ "Default is 'continue'.", default="continue", choices=("stop", "continue"))
+
parser.add_argument("workflow", type=str, nargs="?", default=None, help="The workflow to execute")
parser.add_argument("job_order", nargs=argparse.REMAINDER, help="The input object to the workflow.")
metrics.setLevel(logging.DEBUG)
logging.getLogger("cwltool.metrics").setLevel(logging.DEBUG)
+ if arvargs.log_timestamps:
+ arvados.log_handler.setFormatter(logging.Formatter(
+ '%(asctime)s %(name)s %(levelname)s: %(message)s',
+ '%Y-%m-%d %H:%M:%S'))
+ else:
+ arvados.log_handler.setFormatter(logging.Formatter('%(name)s %(levelname)s: %(message)s'))
+
arvargs.conformance_test = None
arvargs.use_container = True
arvargs.relax_path_checks = True
fetcher_constructor=partial(CollectionFetcher,
api_client=api_client,
keep_client=keep_client),
- resolver=partial(collectionResolver, api_client))
+ resolver=partial(collectionResolver, api_client),
+ logger_handler=arvados.log_handler)
projects / arvados.git / blobdiff