from cwltool.errors import WorkflowException
from cwltool.process import get_feature, UnsupportedRequirement, shortname
from cwltool.pathmapper import adjustFiles
+from cwltool.utils import aslist
import arvados.collection
from .arvdocker import arv_docker_get_image
from . import done
-from .runner import Runner
+from .runner import Runner, arvados_jobs_image
logger = logging.getLogger('arvados.cwl-runner')
"kind": "tmp"
}
}
+ scheduling_parameters = {}
dirs = set()
for f in self.pathmapper.files():
}
if self.generatefiles["listing"]:
- raise UnsupportedRequirement("Generate files not supported")
+ raise UnsupportedRequirement("InitialWorkDirRequirement not supported with --api=containers")
container_request["environment"] = {"TMPDIR": self.tmpdir, "HOME": self.outdir}
if self.environment:
(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,
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:
+ 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
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.arvrunner.processes[response["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)
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:
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"]]
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)
+
+ 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])
return {
"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",
},
"runtime_constraints": {
"vcpus": 1,
- "ram": 1024*1024*256,
+ "ram": 1024*1024 * self.submit_runner_ram,
"API": True
}
}
).execute(num_retries=self.arvrunner.num_retries)
self.uuid = response["uuid"]
- self.arvrunner.processes[response["container_uuid"]] = self
+ self.arvrunner.processes[response["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"]]
projects / arvados.git / blobdiff