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
+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')
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
"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"
}
}
+ scheduling_parameters = {}
+ 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:
- 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,
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)
+
+ api_req, _ = get_feature(self, "http://arvados.org/cwl#APIRequirement")
+ if api_req:
+ runtime_constraints["API"] = True
+
+ 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"]
+
+ 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(
self.arvrunner.processes[response["container_uuid"]] = self
- logger.info("Container %s (%s) request state is %s", self.name, response["container_uuid"], response["state"])
+ 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"])
- if response["state"] == "Final":
- self.done(response)
+ if container["state"] in ("Complete", "Cancelled"):
+ self.done(container)
except Exception as e:
logger.error("Got error %s" % str(e))
self.output_callback({}, "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"
"""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)
- with arvados.collection.Collection(api_client=self.arvrunner.api) as jobobj:
+ 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)
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.enable_reuse:
+ command.append("--enable-reuse")
+ else:
+ command.append("--disable-reuse")
+
+ command.extend([workflowpath, jobpath])
return {
- "command": ["arvados-cwl-runner", "--local", "--api=containers", workflowpath, jobpath],
+ "command": command,
"owner_uuid": self.arvrunner.project_uuid,
"name": self.name,
"output_path": "/var/spool/cwl",
"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",
projects / arvados.git / blobdiff