1 # Copyright (C) The Arvados Authors. All rights reserved.
3 # SPDX-License-Identifier: Apache-2.0
8 import urllib.request, urllib.parse, urllib.error
16 import arvados_cwl.util
19 from cwltool.errors import WorkflowException
20 from cwltool.process import UnsupportedRequirement, shortname
21 from cwltool.utils import aslist, adjustFileObjs, adjustDirObjs, visit_class
22 from cwltool.job import JobBase
24 import arvados.collection
26 import crunchstat_summary.summarizer
27 import crunchstat_summary.reader
29 from .arvdocker import arv_docker_get_image
31 from .runner import Runner, arvados_jobs_image, packed_workflow, trim_anonymous_location, remove_redundant_fields, make_builder
32 from .fsaccess import CollectionFetcher
33 from .pathmapper import NoFollowPathMapper, trim_listing
34 from .perf import Perf
35 from ._version import __version__
37 logger = logging.getLogger('arvados.cwl-runner')
38 metrics = logging.getLogger('arvados.cwl-runner.metrics')
40 def cleanup_name_for_collection(name):
41 return name.replace("/", " ")
43 class ArvadosContainer(JobBase):
44 """Submit and manage a Crunch container request for executing a CWL CommandLineTool."""
46 def __init__(self, runner, job_runtime, globpatterns,
47 builder, # type: Builder
48 joborder, # type: Dict[Text, Union[Dict[Text, Any], List, Text]]
49 make_path_mapper, # type: Callable[..., PathMapper]
50 requirements, # type: List[Dict[Text, Text]]
51 hints, # type: List[Dict[Text, Text]]
54 super(ArvadosContainer, self).__init__(builder, joborder, make_path_mapper, requirements, hints, name)
55 self.arvrunner = runner
56 self.job_runtime = job_runtime
59 self.attempt_count = 0
60 self.globpatterns = globpatterns
62 def update_pipeline_component(self, r):
65 def _required_env(self):
67 env["HOME"] = self.outdir
68 env["TMPDIR"] = self.tmpdir
71 def run(self, toplevelRuntimeContext):
72 # ArvadosCommandTool subclasses from cwltool.CommandLineTool,
73 # which calls makeJobRunner() to get a new ArvadosContainer
74 # object. The fields that define execution such as
75 # command_line, environment, etc are set on the
76 # ArvadosContainer object by CommandLineTool.job() before
79 runtimeContext = self.job_runtime
81 if runtimeContext.submit_request_uuid:
82 container_request = self.arvrunner.api.container_requests().get(
83 uuid=runtimeContext.submit_request_uuid
84 ).execute(num_retries=self.arvrunner.num_retries)
86 container_request = {}
88 container_request["command"] = self.command_line
89 container_request["name"] = self.name
90 container_request["output_path"] = self.outdir
91 container_request["cwd"] = self.outdir
92 container_request["priority"] = runtimeContext.priority
93 container_request["state"] = "Uncommitted"
94 container_request.setdefault("properties", {})
96 container_request["properties"]["cwl_input"] = self.joborder
98 runtime_constraints = {}
100 if runtimeContext.project_uuid:
101 container_request["owner_uuid"] = runtimeContext.project_uuid
103 if self.arvrunner.secret_store.has_secret(self.command_line):
104 raise WorkflowException("Secret material leaked on command line, only file literals may contain secrets")
106 if self.arvrunner.secret_store.has_secret(self.environment):
107 raise WorkflowException("Secret material leaked in environment, only file literals may contain secrets")
109 resources = self.builder.resources
110 if resources is not None:
111 runtime_constraints["vcpus"] = math.ceil(resources.get("cores", 1))
112 runtime_constraints["ram"] = math.ceil(resources.get("ram") * 2**20)
117 "capacity": math.ceil(resources.get("outdirSize", 0) * 2**20)
121 "capacity": math.ceil(resources.get("tmpdirSize", 0) * 2**20)
125 scheduling_parameters = {}
127 rf = [self.pathmapper.mapper(f) for f in self.pathmapper.referenced_files]
128 rf.sort(key=lambda k: k.resolved)
130 for resolved, target, tp, stg in rf:
133 if prevdir and target.startswith(prevdir):
135 if tp == "Directory":
138 targetdir = os.path.dirname(target)
139 sp = resolved.split("/", 1)
140 pdh = sp[0][5:] # remove "keep:"
141 mounts[targetdir] = {
142 "kind": "collection",
143 "portable_data_hash": pdh
145 if pdh in self.pathmapper.pdh_to_uuid:
146 mounts[targetdir]["uuid"] = self.pathmapper.pdh_to_uuid[pdh]
148 if tp == "Directory":
151 path = os.path.dirname(sp[1])
152 if path and path != "/":
153 mounts[targetdir]["path"] = path
154 prevdir = targetdir + "/"
156 intermediate_collection_info = arvados_cwl.util.get_intermediate_collection_info(self.name, runtimeContext.current_container, runtimeContext.intermediate_output_ttl)
158 with Perf(metrics, "generatefiles %s" % self.name):
159 if self.generatefiles["listing"]:
160 vwd = arvados.collection.Collection(api_client=self.arvrunner.api,
161 keep_client=self.arvrunner.keep_client,
162 num_retries=self.arvrunner.num_retries)
163 generatemapper = NoFollowPathMapper(self.generatefiles["listing"], "", "",
166 sorteditems = sorted(generatemapper.items(), key=lambda n: n[1].target)
168 logger.debug("generatemapper is %s", sorteditems)
170 with Perf(metrics, "createfiles %s" % self.name):
171 for f, p in sorteditems:
175 if p.target.startswith("/"):
176 dst = p.target[len(self.outdir)+1:] if p.target.startswith(self.outdir+"/") else p.target[1:]
180 if p.type in ("File", "Directory", "WritableFile", "WritableDirectory"):
181 if p.resolved.startswith("_:"):
184 source, path = self.arvrunner.fs_access.get_collection(p.resolved)
185 vwd.copy(path or ".", dst, source_collection=source)
186 elif p.type == "CreateFile":
187 if self.arvrunner.secret_store.has_secret(p.resolved):
188 mountpoint = p.target if p.target.startswith("/") else os.path.join(self.outdir, p.target)
189 secret_mounts[mountpoint] = {
191 "content": self.arvrunner.secret_store.retrieve(p.resolved)
194 with vwd.open(dst, "w") as n:
197 def keepemptydirs(p):
198 if isinstance(p, arvados.collection.RichCollectionBase):
200 p.open(".keep", "w").close()
207 if not runtimeContext.current_container:
208 runtimeContext.current_container = arvados_cwl.util.get_current_container(self.arvrunner.api, self.arvrunner.num_retries, logger)
209 vwd.save_new(name=intermediate_collection_info["name"],
210 owner_uuid=runtimeContext.project_uuid,
211 ensure_unique_name=True,
212 trash_at=intermediate_collection_info["trash_at"],
213 properties=intermediate_collection_info["properties"])
216 for f, p in sorteditems:
217 if (not p.target or self.arvrunner.secret_store.has_secret(p.resolved) or
218 (prev is not None and p.target.startswith(prev))):
220 if p.target.startswith("/"):
221 dst = p.target[len(self.outdir)+1:] if p.target.startswith(self.outdir+"/") else p.target[1:]
224 mountpoint = p.target if p.target.startswith("/") else os.path.join(self.outdir, p.target)
225 mounts[mountpoint] = {"kind": "collection",
226 "portable_data_hash": vwd.portable_data_hash(),
228 if p.type.startswith("Writable"):
229 mounts[mountpoint]["writable"] = True
230 prev = p.target + "/"
232 container_request["environment"] = {"TMPDIR": self.tmpdir, "HOME": self.outdir}
234 container_request["environment"].update(self.environment)
237 sp = self.stdin[6:].split("/", 1)
238 mounts["stdin"] = {"kind": "collection",
239 "portable_data_hash": sp[0],
243 mounts["stderr"] = {"kind": "file",
244 "path": "%s/%s" % (self.outdir, self.stderr)}
247 mounts["stdout"] = {"kind": "file",
248 "path": "%s/%s" % (self.outdir, self.stdout)}
250 (docker_req, docker_is_req) = self.get_requirement("DockerRequirement")
252 container_request["container_image"] = arv_docker_get_image(self.arvrunner.api,
254 runtimeContext.pull_image,
257 network_req, _ = self.get_requirement("NetworkAccess")
259 runtime_constraints["API"] = network_req["networkAccess"]
261 api_req, _ = self.get_requirement("http://arvados.org/cwl#APIRequirement")
263 runtime_constraints["API"] = True
265 use_disk_cache = (self.arvrunner.api.config()["Containers"].get("DefaultKeepCacheRAM", 0) == 0)
267 keep_cache_type_req, _ = self.get_requirement("http://arvados.org/cwl#KeepCacheTypeRequirement")
268 if keep_cache_type_req:
269 if "keepCacheType" in keep_cache_type_req:
270 if keep_cache_type_req["keepCacheType"] == "ram_cache":
271 use_disk_cache = False
273 runtime_req, _ = self.get_requirement("http://arvados.org/cwl#RuntimeConstraints")
275 if "keep_cache" in runtime_req:
277 # If DefaultKeepCacheRAM is zero it means we should use disk cache.
278 runtime_constraints["keep_cache_disk"] = math.ceil(runtime_req["keep_cache"] * 2**20)
280 runtime_constraints["keep_cache_ram"] = math.ceil(runtime_req["keep_cache"] * 2**20)
281 if "outputDirType" in runtime_req:
282 if runtime_req["outputDirType"] == "local_output_dir":
283 # Currently the default behavior.
285 elif runtime_req["outputDirType"] == "keep_output_dir":
286 mounts[self.outdir]= {
287 "kind": "collection",
291 partition_req, _ = self.get_requirement("http://arvados.org/cwl#PartitionRequirement")
293 scheduling_parameters["partitions"] = aslist(partition_req["partition"])
295 intermediate_output_req, _ = self.get_requirement("http://arvados.org/cwl#IntermediateOutput")
296 if intermediate_output_req:
297 self.output_ttl = intermediate_output_req["outputTTL"]
299 self.output_ttl = self.arvrunner.intermediate_output_ttl
301 if self.output_ttl < 0:
302 raise WorkflowException("Invalid value %d for output_ttl, cannot be less than zero" % container_request["output_ttl"])
305 if self.arvrunner.api._rootDesc["revision"] >= "20210628":
306 storage_class_req, _ = self.get_requirement("http://arvados.org/cwl#OutputStorageClass")
307 if storage_class_req and storage_class_req.get("intermediateStorageClass"):
308 container_request["output_storage_classes"] = aslist(storage_class_req["intermediateStorageClass"])
310 container_request["output_storage_classes"] = runtimeContext.intermediate_storage_classes.strip().split(",")
312 cuda_req, _ = self.get_requirement("http://commonwl.org/cwltool#CUDARequirement")
314 runtime_constraints["cuda"] = {
315 "device_count": resources.get("cudaDeviceCount", 1),
316 "driver_version": cuda_req["cudaVersionMin"],
317 "hardware_capability": aslist(cuda_req["cudaComputeCapability"])[0]
320 if runtimeContext.enable_preemptible is False:
321 scheduling_parameters["preemptible"] = False
323 preemptible_req, _ = self.get_requirement("http://arvados.org/cwl#UsePreemptible")
325 scheduling_parameters["preemptible"] = preemptible_req["usePreemptible"]
326 elif runtimeContext.enable_preemptible is True:
327 scheduling_parameters["preemptible"] = True
328 elif runtimeContext.enable_preemptible is None:
331 if self.timelimit is not None and self.timelimit > 0:
332 scheduling_parameters["max_run_time"] = self.timelimit
334 extra_submit_params = {}
335 if runtimeContext.submit_runner_cluster:
336 extra_submit_params["cluster_id"] = runtimeContext.submit_runner_cluster
338 container_request["output_name"] = cleanup_name_for_collection("Output from step %s" % (self.name))
339 container_request["output_ttl"] = self.output_ttl
340 container_request["mounts"] = mounts
341 container_request["secret_mounts"] = secret_mounts
342 container_request["runtime_constraints"] = runtime_constraints
343 container_request["scheduling_parameters"] = scheduling_parameters
345 enable_reuse = runtimeContext.enable_reuse
347 reuse_req, _ = self.get_requirement("WorkReuse")
349 enable_reuse = reuse_req["enableReuse"]
350 reuse_req, _ = self.get_requirement("http://arvados.org/cwl#ReuseRequirement")
352 enable_reuse = reuse_req["enableReuse"]
353 container_request["use_existing"] = enable_reuse
355 properties_req, _ = self.get_requirement("http://arvados.org/cwl#ProcessProperties")
357 for pr in properties_req["processProperties"]:
358 container_request["properties"][pr["propertyName"]] = self.builder.do_eval(pr["propertyValue"])
360 output_properties_req, _ = self.get_requirement("http://arvados.org/cwl#OutputCollectionProperties")
361 if output_properties_req:
362 if self.arvrunner.api._rootDesc["revision"] >= "20220510":
363 container_request["output_properties"] = {}
364 for pr in output_properties_req["outputProperties"]:
365 container_request["output_properties"][pr["propertyName"]] = self.builder.do_eval(pr["propertyValue"])
367 logger.warning("%s API revision is %s, revision %s is required to support setting properties on output collections.",
368 self.arvrunner.label(self), self.arvrunner.api._rootDesc["revision"], "20220510")
370 if self.arvrunner.api._rootDesc["revision"] >= "20240502" and self.globpatterns:
372 for gb in self.globpatterns:
373 gb = self.builder.do_eval(gb)
376 for gbeval in aslist(gb):
377 if gbeval.startswith(self.outdir+"/"):
378 gbeval = gbeval[len(self.outdir)+1:]
379 if gbeval in (self.outdir, "", "."):
380 output_glob.append("**")
382 output_glob.append(gbeval)
383 output_glob.append(gbeval + "/**")
385 if "**" in output_glob:
386 # if it's going to match all, prefer not to provide it
391 # Tools should either use cwl.output.json or output
392 # binding glob. Unfortunately one CWL conformance test
393 # has both, but that test uses cwl.output.json return
394 # a string, not a File.
396 # It could happen that a tool uses cwl.output.json,
397 # references a file, but also uses a glob, and that
398 # would break, because we don't know about it ahead of
399 # time. However I don't think any conformance tests
400 # do this and I'd even be inclined to rule that out of
401 # order in the CWL spec if it came up.
402 output_glob.append("cwl.output.json")
403 container_request["output_glob"] = output_glob
407 oom_retry_req, _ = self.get_requirement("http://arvados.org/cwl#OutOfMemoryRetry")
409 if oom_retry_req.get('memoryRetryMultiplier'):
410 ram_multiplier.append(oom_retry_req.get('memoryRetryMultiplier'))
411 elif oom_retry_req.get('memoryRetryMultipler'):
412 ram_multiplier.append(oom_retry_req.get('memoryRetryMultipler'))
414 ram_multiplier.append(2)
416 if runtimeContext.runnerjob.startswith("arvwf:"):
417 wfuuid = runtimeContext.runnerjob[6:runtimeContext.runnerjob.index("#")]
418 wfrecord = self.arvrunner.api.workflows().get(uuid=wfuuid).execute(num_retries=self.arvrunner.num_retries)
419 if container_request["name"] == "main":
420 container_request["name"] = wfrecord["name"]
421 container_request["properties"]["template_uuid"] = wfuuid
423 if self.attempt_count == 0:
424 self.output_callback = self.arvrunner.get_wrapped_callback(self.output_callback)
427 ram = runtime_constraints["ram"]
429 self.uuid = runtimeContext.submit_request_uuid
431 for i in ram_multiplier:
432 runtime_constraints["ram"] = ram * i
435 response = self.arvrunner.api.container_requests().update(
437 body=container_request,
438 **extra_submit_params
439 ).execute(num_retries=self.arvrunner.num_retries)
441 response = self.arvrunner.api.container_requests().create(
442 body=container_request,
443 **extra_submit_params
444 ).execute(num_retries=self.arvrunner.num_retries)
445 self.uuid = response["uuid"]
447 if response["container_uuid"] is not None:
450 if response["container_uuid"] is None:
451 runtime_constraints["ram"] = ram * ram_multiplier[self.attempt_count]
453 container_request["state"] = "Committed"
454 response = self.arvrunner.api.container_requests().update(
456 body=container_request,
457 **extra_submit_params
458 ).execute(num_retries=self.arvrunner.num_retries)
460 self.arvrunner.process_submitted(self)
461 self.attempt_count += 1
463 if response["state"] == "Final":
464 logger.info("%s reused container %s", self.arvrunner.label(self), response["container_uuid"])
466 logger.info("%s %s state is %s", self.arvrunner.label(self), response["uuid"], response["state"])
467 except Exception as e:
468 logger.exception("%s error submitting container\n%s", self.arvrunner.label(self), e)
469 logger.debug("Container request was %s", container_request)
470 self.output_callback({}, "permanentFail")
472 def out_of_memory_retry(self, record, container):
473 oom_retry_req, _ = self.get_requirement("http://arvados.org/cwl#OutOfMemoryRetry")
474 if oom_retry_req is None:
477 # Sometimes it gets killed with no warning
478 if container["exit_code"] == 137:
481 logc = arvados.collection.CollectionReader(record["log_uuid"],
482 api_client=self.arvrunner.api,
483 keep_client=self.arvrunner.keep_client,
484 num_retries=self.arvrunner.num_retries)
487 def callback(v1, v2, v3):
490 done.logtail(logc, callback, "", maxlen=1000)
492 # Check allocation failure
493 oom_matches = oom_retry_req.get('memoryErrorRegex') or r'(bad_alloc|out ?of ?memory|memory ?error|container using over 9.% of memory)'
494 if re.search(oom_matches, loglines[0], re.IGNORECASE | re.MULTILINE):
499 def done(self, record):
504 container = self.arvrunner.api.containers().get(
505 uuid=record["container_uuid"]
506 ).execute(num_retries=self.arvrunner.num_retries)
507 if container["state"] == "Complete":
508 rcode = container["exit_code"]
509 if self.successCodes and rcode in self.successCodes:
510 processStatus = "success"
511 elif self.temporaryFailCodes and rcode in self.temporaryFailCodes:
512 processStatus = "temporaryFail"
513 elif self.permanentFailCodes and rcode in self.permanentFailCodes:
514 processStatus = "permanentFail"
516 processStatus = "success"
518 processStatus = "permanentFail"
520 if processStatus == "permanentFail" and self.attempt_count == 1 and self.out_of_memory_retry(record, container):
521 logger.warning("%s Container failed with out of memory error, retrying with more RAM.",
522 self.arvrunner.label(self))
523 self.job_runtime.submit_request_uuid = None
530 logger.warning("%s Container may have been killed for using too much RAM. Try resubmitting with a higher 'ramMin' or use the arv:OutOfMemoryRetry feature.",
531 self.arvrunner.label(self))
533 processStatus = "permanentFail"
536 if record["log_uuid"]:
537 logc = arvados.collection.Collection(record["log_uuid"],
538 api_client=self.arvrunner.api,
539 keep_client=self.arvrunner.keep_client,
540 num_retries=self.arvrunner.num_retries)
542 if processStatus == "permanentFail" and logc is not None:
543 label = self.arvrunner.label(self)
546 "%s (%s) error log:" % (label, record["uuid"]), maxlen=40, include_crunchrun=(rcode is None or rcode > 127))
548 if record["output_uuid"]:
549 if self.arvrunner.trash_intermediate or self.arvrunner.intermediate_output_ttl:
550 # Compute the trash time to avoid requesting the collection record.
551 trash_at = ciso8601.parse_datetime_as_naive(record["modified_at"]) + datetime.timedelta(0, self.arvrunner.intermediate_output_ttl)
552 aftertime = " at %s" % trash_at.strftime("%Y-%m-%d %H:%M:%S UTC") if self.arvrunner.intermediate_output_ttl else ""
553 orpart = ", or" if self.arvrunner.trash_intermediate and self.arvrunner.intermediate_output_ttl else ""
554 oncomplete = " upon successful completion of the workflow" if self.arvrunner.trash_intermediate else ""
555 logger.info("%s Intermediate output %s (%s) will be trashed%s%s%s." % (
556 self.arvrunner.label(self), record["output_uuid"], container["output"], aftertime, orpart, oncomplete))
557 self.arvrunner.add_intermediate_output(record["output_uuid"])
559 if container["output"]:
560 outputs = done.done_outputs(self, container, "/tmp", self.outdir, "/keep")
562 properties = record["properties"].copy()
563 properties["cwl_output"] = outputs
564 self.arvrunner.api.container_requests().update(
566 body={"container_request": {"properties": properties}}
567 ).execute(num_retries=self.arvrunner.num_retries)
569 if logc is not None and self.job_runtime.enable_usage_report is not False:
571 summarizer = crunchstat_summary.summarizer.ContainerRequestSummarizer(
573 collection_object=logc,
575 arv=self.arvrunner.api)
577 with logc.open("usage_report.html", "wt") as mr:
578 mr.write(summarizer.html_report())
581 # Post warnings about nodes that are under-utilized.
582 for rc in summarizer._recommend_gen(lambda x: x):
583 self.job_runtime.usage_report_notes.append(rc)
585 except Exception as e:
586 logger.warning("%s unable to generate resource usage report",
587 self.arvrunner.label(self),
588 exc_info=(e if self.arvrunner.debug else False))
590 except WorkflowException as e:
591 # Only include a stack trace if in debug mode.
592 # A stack trace may obfuscate more useful output about the workflow.
593 logger.error("%s unable to collect output from %s:\n%s",
594 self.arvrunner.label(self), container["output"], e, exc_info=(e if self.arvrunner.debug else False))
595 processStatus = "permanentFail"
597 logger.exception("%s while getting output object:", self.arvrunner.label(self))
598 processStatus = "permanentFail"
601 self.output_callback(outputs, processStatus)
604 class RunnerContainer(Runner):
605 """Submit and manage a container that runs arvados-cwl-runner."""
607 def arvados_job_spec(self, runtimeContext, git_info):
608 """Create an Arvados container request for this workflow.
610 The returned dict can be used to create a container passed as
611 the +body+ argument to container_requests().create().
614 adjustDirObjs(self.job_order, trim_listing)
615 visit_class(self.job_order, ("File", "Directory"), trim_anonymous_location)
616 visit_class(self.job_order, ("File", "Directory"), remove_redundant_fields)
619 for param in sorted(self.job_order.keys()):
620 if self.secret_store.has_secret(self.job_order[param]):
621 mnt = "/secrets/s%d" % len(secret_mounts)
622 secret_mounts[mnt] = {
624 "content": self.secret_store.retrieve(self.job_order[param])
626 self.job_order[param] = {"$include": mnt}
628 container_image = arvados_jobs_image(self.arvrunner, self.jobs_image, runtimeContext)
630 workflow_runner_req, _ = self.embedded_tool.get_requirement("http://arvados.org/cwl#WorkflowRunnerResources")
631 if workflow_runner_req and workflow_runner_req.get("acrContainerImage"):
632 container_image = workflow_runner_req.get("acrContainerImage")
636 "output_path": "/var/spool/cwl",
637 "cwd": "/var/spool/cwl",
638 "priority": self.priority,
639 "state": "Committed",
640 "container_image": container_image,
642 "/var/lib/cwl/cwl.input.json": {
644 "content": self.job_order
648 "path": "/var/spool/cwl/cwl.output.json"
651 "kind": "collection",
655 "secret_mounts": secret_mounts,
656 "runtime_constraints": {
657 "vcpus": math.ceil(self.submit_runner_cores),
658 "ram": 1024*1024 * (math.ceil(self.submit_runner_ram) + math.ceil(self.collection_cache_size)),
661 "use_existing": self.reuse_runner,
665 if self.embedded_tool.tool.get("id", "").startswith("keep:"):
666 sp = self.embedded_tool.tool["id"].split('/')
667 workflowcollection = sp[0][5:]
668 workflowname = "/".join(sp[1:])
669 workflowpath = "/var/lib/cwl/workflow/%s" % workflowname
670 container_req["mounts"]["/var/lib/cwl/workflow"] = {
671 "kind": "collection",
672 "portable_data_hash": "%s" % workflowcollection
674 elif self.embedded_tool.tool.get("id", "").startswith("arvwf:"):
675 uuid, frg = urllib.parse.urldefrag(self.embedded_tool.tool["id"])
676 workflowpath = "/var/lib/cwl/workflow.json#" + frg
677 packedtxt = self.loadingContext.loader.fetch_text(uuid)
678 yaml = ruamel.yaml.YAML(typ='safe', pure=True)
679 packed = yaml.load(packedtxt)
680 container_req["mounts"]["/var/lib/cwl/workflow.json"] = {
684 container_req["properties"]["template_uuid"] = self.embedded_tool.tool["id"][6:33]
685 elif self.embedded_tool.tool.get("id", "").startswith("file:"):
686 raise WorkflowException("Tool id '%s' is a local file but expected keep: or arvwf:" % self.embedded_tool.tool.get("id"))
688 main = self.loadingContext.loader.idx["_:main"]
689 if main.get("id") == "_:main":
691 workflowpath = "/var/lib/cwl/workflow.json#main"
692 container_req["mounts"]["/var/lib/cwl/workflow.json"] = {
697 container_req["properties"].update({k.replace("http://arvados.org/cwl#", "arv:"): v for k, v in git_info.items()})
699 properties_req, _ = self.embedded_tool.get_requirement("http://arvados.org/cwl#ProcessProperties")
701 builder = make_builder(self.job_order, self.embedded_tool.hints, self.embedded_tool.requirements, runtimeContext, self.embedded_tool.metadata)
702 for pr in properties_req["processProperties"]:
703 container_req["properties"][pr["propertyName"]] = builder.do_eval(pr["propertyValue"])
705 # --local means execute the workflow instead of submitting a container request
706 # --api=containers means use the containers API
707 # --no-log-timestamps means don't add timestamps (the logging infrastructure does this)
708 # --disable-validate because we already validated so don't need to do it again
709 # --eval-timeout is the timeout for javascript invocation
710 # --parallel-task-count is the number of threads to use for job submission
711 # --enable/disable-reuse sets desired job reuse
712 # --collection-cache-size sets aside memory to store collections
713 command = ["arvados-cwl-runner",
716 "--no-log-timestamps",
717 "--disable-validate",
719 "--eval-timeout=%s" % self.arvrunner.eval_timeout,
720 "--thread-count=%s" % self.arvrunner.thread_count,
721 "--enable-reuse" if self.enable_reuse else "--disable-reuse",
722 "--collection-cache-size=%s" % self.collection_cache_size]
725 command.append("--output-name=" + self.output_name)
726 container_req["output_name"] = self.output_name
729 command.append("--output-tags=" + self.output_tags)
731 if runtimeContext.debug:
732 command.append("--debug")
734 if runtimeContext.storage_classes != "default" and runtimeContext.storage_classes:
735 command.append("--storage-classes=" + runtimeContext.storage_classes)
737 if runtimeContext.intermediate_storage_classes != "default" and runtimeContext.intermediate_storage_classes:
738 command.append("--intermediate-storage-classes=" + runtimeContext.intermediate_storage_classes)
740 if runtimeContext.on_error:
741 command.append("--on-error=" + self.on_error)
743 if runtimeContext.intermediate_output_ttl:
744 command.append("--intermediate-output-ttl=%d" % runtimeContext.intermediate_output_ttl)
746 if runtimeContext.trash_intermediate:
747 command.append("--trash-intermediate")
749 if runtimeContext.project_uuid:
750 command.append("--project-uuid="+runtimeContext.project_uuid)
753 command.append("--enable-dev")
755 if runtimeContext.enable_preemptible is True:
756 command.append("--enable-preemptible")
758 if runtimeContext.enable_preemptible is False:
759 command.append("--disable-preemptible")
761 if runtimeContext.varying_url_params:
762 command.append("--varying-url-params="+runtimeContext.varying_url_params)
764 if runtimeContext.prefer_cached_downloads:
765 command.append("--prefer-cached-downloads")
767 if runtimeContext.enable_usage_report is True:
768 command.append("--enable-usage-report")
770 if runtimeContext.enable_usage_report is False:
771 command.append("--disable-usage-report")
774 command.append("--fast-parser")
776 command.extend([workflowpath, "/var/lib/cwl/cwl.input.json"])
778 container_req["command"] = command
783 def run(self, runtimeContext):
784 runtimeContext.keepprefix = "keep:"
785 job_spec = self.arvados_job_spec(runtimeContext, self.git_info)
786 if runtimeContext.project_uuid:
787 job_spec["owner_uuid"] = runtimeContext.project_uuid
789 extra_submit_params = {}
790 if runtimeContext.submit_runner_cluster:
791 extra_submit_params["cluster_id"] = runtimeContext.submit_runner_cluster
793 if runtimeContext.submit_request_uuid:
794 if "cluster_id" in extra_submit_params:
795 # Doesn't make sense for "update" and actually fails
796 del extra_submit_params["cluster_id"]
797 response = self.arvrunner.api.container_requests().update(
798 uuid=runtimeContext.submit_request_uuid,
800 **extra_submit_params
801 ).execute(num_retries=self.arvrunner.num_retries)
803 response = self.arvrunner.api.container_requests().create(
805 **extra_submit_params
806 ).execute(num_retries=self.arvrunner.num_retries)
808 self.uuid = response["uuid"]
809 self.arvrunner.process_submitted(self)
811 logger.info("%s submitted container_request %s", self.arvrunner.label(self), response["uuid"])
813 workbench2 = self.arvrunner.api.config()["Services"]["Workbench2"]["ExternalURL"]
815 url = "{}processes/{}".format(workbench2, response["uuid"])
816 logger.info("Monitor workflow progress at %s", url)
819 def done(self, record):
821 container = self.arvrunner.api.containers().get(
822 uuid=record["container_uuid"]
823 ).execute(num_retries=self.arvrunner.num_retries)
824 container["log"] = record["log_uuid"]
826 logger.exception("%s while getting runner container", self.arvrunner.label(self))
827 self.arvrunner.output_callback({}, "permanentFail")
829 super(RunnerContainer, self).done(container)