1 # Copyright (C) The Arvados Authors. All rights reserved.
3 # SPDX-License-Identifier: Apache-2.0
5 from future import standard_library
6 standard_library.install_aliases()
7 from builtins import str
12 import urllib.request, urllib.parse, urllib.error
19 import arvados_cwl.util
20 import ruamel.yaml as yaml
22 from cwltool.errors import WorkflowException
23 from cwltool.process import UnsupportedRequirement, shortname
24 from cwltool.pathmapper import adjustFileObjs, adjustDirObjs, visit_class
25 from cwltool.utils import aslist
26 from cwltool.job import JobBase
28 import arvados.collection
30 from .arvdocker import arv_docker_get_image
32 from .runner import Runner, arvados_jobs_image, packed_workflow, trim_anonymous_location, remove_redundant_fields
33 from .fsaccess import CollectionFetcher
34 from .pathmapper import NoFollowPathMapper, trim_listing
35 from .perf import Perf
37 logger = logging.getLogger('arvados.cwl-runner')
38 metrics = logging.getLogger('arvados.cwl-runner.metrics')
40 class ArvadosContainer(JobBase):
41 """Submit and manage a Crunch container request for executing a CWL CommandLineTool."""
43 def __init__(self, runner, job_runtime,
44 builder, # type: Builder
45 joborder, # type: Dict[Text, Union[Dict[Text, Any], List, Text]]
46 make_path_mapper, # type: Callable[..., PathMapper]
47 requirements, # type: List[Dict[Text, Text]]
48 hints, # type: List[Dict[Text, Text]]
51 super(ArvadosContainer, self).__init__(builder, joborder, make_path_mapper, requirements, hints, name)
52 self.arvrunner = runner
53 self.job_runtime = job_runtime
57 def update_pipeline_component(self, r):
60 def run(self, runtimeContext):
61 # ArvadosCommandTool subclasses from cwltool.CommandLineTool,
62 # which calls makeJobRunner() to get a new ArvadosContainer
63 # object. The fields that define execution such as
64 # command_line, environment, etc are set on the
65 # ArvadosContainer object by CommandLineTool.job() before
68 runtimeContext = self.job_runtime
71 "command": self.command_line,
73 "output_path": self.outdir,
75 "priority": runtimeContext.priority,
79 runtime_constraints = {}
81 if runtimeContext.project_uuid:
82 container_request["owner_uuid"] = runtimeContext.project_uuid
84 if self.arvrunner.secret_store.has_secret(self.command_line):
85 raise WorkflowException("Secret material leaked on command line, only file literals may contain secrets")
87 if self.arvrunner.secret_store.has_secret(self.environment):
88 raise WorkflowException("Secret material leaked in environment, only file literals may contain secrets")
90 resources = self.builder.resources
91 if resources is not None:
92 runtime_constraints["vcpus"] = math.ceil(resources.get("cores", 1))
93 runtime_constraints["ram"] = math.ceil(resources.get("ram") * 2**20)
98 "capacity": math.ceil(resources.get("outdirSize", 0) * 2**20)
102 "capacity": math.ceil(resources.get("tmpdirSize", 0) * 2**20)
106 scheduling_parameters = {}
108 rf = [self.pathmapper.mapper(f) for f in self.pathmapper.referenced_files]
109 rf.sort(key=lambda k: k.resolved)
111 for resolved, target, tp, stg in rf:
114 if prevdir and target.startswith(prevdir):
116 if tp == "Directory":
119 targetdir = os.path.dirname(target)
120 sp = resolved.split("/", 1)
121 pdh = sp[0][5:] # remove "keep:"
122 mounts[targetdir] = {
123 "kind": "collection",
124 "portable_data_hash": pdh
126 if pdh in self.pathmapper.pdh_to_uuid:
127 mounts[targetdir]["uuid"] = self.pathmapper.pdh_to_uuid[pdh]
129 if tp == "Directory":
132 path = os.path.dirname(sp[1])
133 if path and path != "/":
134 mounts[targetdir]["path"] = path
135 prevdir = targetdir + "/"
137 with Perf(metrics, "generatefiles %s" % self.name):
138 if self.generatefiles["listing"]:
139 vwd = arvados.collection.Collection(api_client=self.arvrunner.api,
140 keep_client=self.arvrunner.keep_client,
141 num_retries=self.arvrunner.num_retries)
142 generatemapper = NoFollowPathMapper(self.generatefiles["listing"], "", "",
145 sorteditems = sorted(generatemapper.items(), key=lambda n: n[1].target)
147 logger.debug("generatemapper is %s", sorteditems)
149 with Perf(metrics, "createfiles %s" % self.name):
150 for f, p in sorteditems:
153 elif p.type in ("File", "Directory", "WritableFile", "WritableDirectory"):
154 if p.resolved.startswith("_:"):
157 source, path = self.arvrunner.fs_access.get_collection(p.resolved)
158 vwd.copy(path or ".", p.target, source_collection=source)
159 elif p.type == "CreateFile":
160 if self.arvrunner.secret_store.has_secret(p.resolved):
161 secret_mounts["%s/%s" % (self.outdir, p.target)] = {
163 "content": self.arvrunner.secret_store.retrieve(p.resolved)
166 with vwd.open(p.target, "w") as n:
169 def keepemptydirs(p):
170 if isinstance(p, arvados.collection.RichCollectionBase):
172 p.open(".keep", "w").close()
179 if not runtimeContext.current_container:
180 runtimeContext.current_container = arvados_cwl.util.get_current_container(self.arvrunner.api, self.arvrunner.num_retries, logger)
181 info = arvados_cwl.util.get_intermediate_collection_info(self.name, runtimeContext.current_container, runtimeContext.intermediate_output_ttl)
182 vwd.save_new(name=info["name"],
183 owner_uuid=runtimeContext.project_uuid,
184 ensure_unique_name=True,
185 trash_at=info["trash_at"],
186 properties=info["properties"])
189 for f, p in sorteditems:
190 if (not p.target or self.arvrunner.secret_store.has_secret(p.resolved) or
191 (prev is not None and p.target.startswith(prev))):
193 mountpoint = "%s/%s" % (self.outdir, p.target)
194 mounts[mountpoint] = {"kind": "collection",
195 "portable_data_hash": vwd.portable_data_hash(),
197 if p.type.startswith("Writable"):
198 mounts[mountpoint]["writable"] = True
199 prev = p.target + "/"
201 container_request["environment"] = {"TMPDIR": self.tmpdir, "HOME": self.outdir}
203 container_request["environment"].update(self.environment)
206 sp = self.stdin[6:].split("/", 1)
207 mounts["stdin"] = {"kind": "collection",
208 "portable_data_hash": sp[0],
212 mounts["stderr"] = {"kind": "file",
213 "path": "%s/%s" % (self.outdir, self.stderr)}
216 mounts["stdout"] = {"kind": "file",
217 "path": "%s/%s" % (self.outdir, self.stdout)}
219 (docker_req, docker_is_req) = self.get_requirement("DockerRequirement")
221 docker_req = {"dockerImageId": "arvados/jobs"}
223 container_request["container_image"] = arv_docker_get_image(self.arvrunner.api,
225 runtimeContext.pull_image,
226 runtimeContext.project_uuid)
228 network_req, _ = self.get_requirement("NetworkAccess")
230 runtime_constraints["API"] = network_req["networkAccess"]
232 api_req, _ = self.get_requirement("http://arvados.org/cwl#APIRequirement")
234 runtime_constraints["API"] = True
236 runtime_req, _ = self.get_requirement("http://arvados.org/cwl#RuntimeConstraints")
238 if "keep_cache" in runtime_req:
239 runtime_constraints["keep_cache_ram"] = math.ceil(runtime_req["keep_cache"] * 2**20)
240 if "outputDirType" in runtime_req:
241 if runtime_req["outputDirType"] == "local_output_dir":
242 # Currently the default behavior.
244 elif runtime_req["outputDirType"] == "keep_output_dir":
245 mounts[self.outdir]= {
246 "kind": "collection",
250 partition_req, _ = self.get_requirement("http://arvados.org/cwl#PartitionRequirement")
252 scheduling_parameters["partitions"] = aslist(partition_req["partition"])
254 intermediate_output_req, _ = self.get_requirement("http://arvados.org/cwl#IntermediateOutput")
255 if intermediate_output_req:
256 self.output_ttl = intermediate_output_req["outputTTL"]
258 self.output_ttl = self.arvrunner.intermediate_output_ttl
260 if self.output_ttl < 0:
261 raise WorkflowException("Invalid value %d for output_ttl, cannot be less than zero" % container_request["output_ttl"])
263 if self.timelimit is not None and self.timelimit > 0:
264 scheduling_parameters["max_run_time"] = self.timelimit
266 extra_submit_params = {}
267 if runtimeContext.submit_runner_cluster:
268 extra_submit_params["cluster_id"] = runtimeContext.submit_runner_cluster
270 container_request["output_name"] = "Output for step %s" % (self.name)
271 container_request["output_ttl"] = self.output_ttl
272 container_request["mounts"] = mounts
273 container_request["secret_mounts"] = secret_mounts
274 container_request["runtime_constraints"] = runtime_constraints
275 container_request["scheduling_parameters"] = scheduling_parameters
277 enable_reuse = runtimeContext.enable_reuse
279 reuse_req, _ = self.get_requirement("WorkReuse")
281 enable_reuse = reuse_req["enableReuse"]
282 reuse_req, _ = self.get_requirement("http://arvados.org/cwl#ReuseRequirement")
284 enable_reuse = reuse_req["enableReuse"]
285 container_request["use_existing"] = enable_reuse
287 if runtimeContext.runnerjob.startswith("arvwf:"):
288 wfuuid = runtimeContext.runnerjob[6:runtimeContext.runnerjob.index("#")]
289 wfrecord = self.arvrunner.api.workflows().get(uuid=wfuuid).execute(num_retries=self.arvrunner.num_retries)
290 if container_request["name"] == "main":
291 container_request["name"] = wfrecord["name"]
292 container_request["properties"]["template_uuid"] = wfuuid
294 self.output_callback = self.arvrunner.get_wrapped_callback(self.output_callback)
297 if runtimeContext.submit_request_uuid:
298 response = self.arvrunner.api.container_requests().update(
299 uuid=runtimeContext.submit_request_uuid,
300 body=container_request,
301 **extra_submit_params
302 ).execute(num_retries=self.arvrunner.num_retries)
304 response = self.arvrunner.api.container_requests().create(
305 body=container_request,
306 **extra_submit_params
307 ).execute(num_retries=self.arvrunner.num_retries)
309 self.uuid = response["uuid"]
310 self.arvrunner.process_submitted(self)
312 if response["state"] == "Final":
313 logger.info("%s reused container %s", self.arvrunner.label(self), response["container_uuid"])
315 logger.info("%s %s state is %s", self.arvrunner.label(self), response["uuid"], response["state"])
317 logger.exception("%s got an error", self.arvrunner.label(self))
318 self.output_callback({}, "permanentFail")
320 def done(self, record):
323 container = self.arvrunner.api.containers().get(
324 uuid=record["container_uuid"]
325 ).execute(num_retries=self.arvrunner.num_retries)
326 if container["state"] == "Complete":
327 rcode = container["exit_code"]
328 if self.successCodes and rcode in self.successCodes:
329 processStatus = "success"
330 elif self.temporaryFailCodes and rcode in self.temporaryFailCodes:
331 processStatus = "temporaryFail"
332 elif self.permanentFailCodes and rcode in self.permanentFailCodes:
333 processStatus = "permanentFail"
335 processStatus = "success"
337 processStatus = "permanentFail"
339 processStatus = "permanentFail"
341 if processStatus == "permanentFail" and record["log_uuid"]:
342 logc = arvados.collection.CollectionReader(record["log_uuid"],
343 api_client=self.arvrunner.api,
344 keep_client=self.arvrunner.keep_client,
345 num_retries=self.arvrunner.num_retries)
346 label = self.arvrunner.label(self)
349 "%s (%s) error log:" % (label, record["uuid"]), maxlen=40)
351 if record["output_uuid"]:
352 if self.arvrunner.trash_intermediate or self.arvrunner.intermediate_output_ttl:
353 # Compute the trash time to avoid requesting the collection record.
354 trash_at = ciso8601.parse_datetime_as_naive(record["modified_at"]) + datetime.timedelta(0, self.arvrunner.intermediate_output_ttl)
355 aftertime = " at %s" % trash_at.strftime("%Y-%m-%d %H:%M:%S UTC") if self.arvrunner.intermediate_output_ttl else ""
356 orpart = ", or" if self.arvrunner.trash_intermediate and self.arvrunner.intermediate_output_ttl else ""
357 oncomplete = " upon successful completion of the workflow" if self.arvrunner.trash_intermediate else ""
358 logger.info("%s Intermediate output %s (%s) will be trashed%s%s%s." % (
359 self.arvrunner.label(self), record["output_uuid"], container["output"], aftertime, orpart, oncomplete))
360 self.arvrunner.add_intermediate_output(record["output_uuid"])
362 if container["output"]:
363 outputs = done.done_outputs(self, container, "/tmp", self.outdir, "/keep")
364 except WorkflowException as e:
365 # Only include a stack trace if in debug mode.
366 # A stack trace may obfuscate more useful output about the workflow.
367 logger.error("%s unable to collect output from %s:\n%s",
368 self.arvrunner.label(self), container["output"], e, exc_info=(e if self.arvrunner.debug else False))
369 processStatus = "permanentFail"
371 logger.exception("%s while getting output object:", self.arvrunner.label(self))
372 processStatus = "permanentFail"
374 self.output_callback(outputs, processStatus)
377 class RunnerContainer(Runner):
378 """Submit and manage a container that runs arvados-cwl-runner."""
380 def arvados_job_spec(self, runtimeContext):
381 """Create an Arvados container request for this workflow.
383 The returned dict can be used to create a container passed as
384 the +body+ argument to container_requests().create().
387 adjustDirObjs(self.job_order, trim_listing)
388 visit_class(self.job_order, ("File", "Directory"), trim_anonymous_location)
389 visit_class(self.job_order, ("File", "Directory"), remove_redundant_fields)
392 for param in sorted(self.job_order.keys()):
393 if self.secret_store.has_secret(self.job_order[param]):
394 mnt = "/secrets/s%d" % len(secret_mounts)
395 secret_mounts[mnt] = {
397 "content": self.secret_store.retrieve(self.job_order[param])
399 self.job_order[param] = {"$include": mnt}
403 "output_path": "/var/spool/cwl",
404 "cwd": "/var/spool/cwl",
405 "priority": self.priority,
406 "state": "Committed",
407 "container_image": arvados_jobs_image(self.arvrunner, self.jobs_image),
409 "/var/lib/cwl/cwl.input.json": {
411 "content": self.job_order
415 "path": "/var/spool/cwl/cwl.output.json"
418 "kind": "collection",
422 "secret_mounts": secret_mounts,
423 "runtime_constraints": {
424 "vcpus": math.ceil(self.submit_runner_cores),
425 "ram": 1024*1024 * (math.ceil(self.submit_runner_ram) + math.ceil(self.collection_cache_size)),
428 "use_existing": False, # Never reuse the runner container - see #15497.
432 if self.embedded_tool.tool.get("id", "").startswith("keep:"):
433 sp = self.embedded_tool.tool["id"].split('/')
434 workflowcollection = sp[0][5:]
435 workflowname = "/".join(sp[1:])
436 workflowpath = "/var/lib/cwl/workflow/%s" % workflowname
437 container_req["mounts"]["/var/lib/cwl/workflow"] = {
438 "kind": "collection",
439 "portable_data_hash": "%s" % workflowcollection
442 packed = packed_workflow(self.arvrunner, self.embedded_tool, self.merged_map)
443 workflowpath = "/var/lib/cwl/workflow.json#main"
444 container_req["mounts"]["/var/lib/cwl/workflow.json"] = {
448 if self.embedded_tool.tool.get("id", "").startswith("arvwf:"):
449 container_req["properties"]["template_uuid"] = self.embedded_tool.tool["id"][6:33]
452 # --local means execute the workflow instead of submitting a container request
453 # --api=containers means use the containers API
454 # --no-log-timestamps means don't add timestamps (the logging infrastructure does this)
455 # --disable-validate because we already validated so don't need to do it again
456 # --eval-timeout is the timeout for javascript invocation
457 # --parallel-task-count is the number of threads to use for job submission
458 # --enable/disable-reuse sets desired job reuse
459 # --collection-cache-size sets aside memory to store collections
460 command = ["arvados-cwl-runner",
463 "--no-log-timestamps",
464 "--disable-validate",
465 "--eval-timeout=%s" % self.arvrunner.eval_timeout,
466 "--thread-count=%s" % self.arvrunner.thread_count,
467 "--enable-reuse" if self.enable_reuse else "--disable-reuse",
468 "--collection-cache-size=%s" % self.collection_cache_size]
471 command.append("--output-name=" + self.output_name)
472 container_req["output_name"] = self.output_name
475 command.append("--output-tags=" + self.output_tags)
477 if runtimeContext.debug:
478 command.append("--debug")
480 if runtimeContext.storage_classes != "default":
481 command.append("--storage-classes=" + runtimeContext.storage_classes)
484 command.append("--on-error=" + self.on_error)
486 if self.intermediate_output_ttl:
487 command.append("--intermediate-output-ttl=%d" % self.intermediate_output_ttl)
489 if self.arvrunner.trash_intermediate:
490 command.append("--trash-intermediate")
492 if self.arvrunner.project_uuid:
493 command.append("--project-uuid="+self.arvrunner.project_uuid)
496 command.append("--enable-dev")
498 command.extend([workflowpath, "/var/lib/cwl/cwl.input.json"])
500 container_req["command"] = command
505 def run(self, runtimeContext):
506 runtimeContext.keepprefix = "keep:"
507 job_spec = self.arvados_job_spec(runtimeContext)
508 if self.arvrunner.project_uuid:
509 job_spec["owner_uuid"] = self.arvrunner.project_uuid
511 extra_submit_params = {}
512 if runtimeContext.submit_runner_cluster:
513 extra_submit_params["cluster_id"] = runtimeContext.submit_runner_cluster
515 if runtimeContext.submit_request_uuid:
516 if "cluster_id" in extra_submit_params:
517 # Doesn't make sense for "update" and actually fails
518 del extra_submit_params["cluster_id"]
519 response = self.arvrunner.api.container_requests().update(
520 uuid=runtimeContext.submit_request_uuid,
522 **extra_submit_params
523 ).execute(num_retries=self.arvrunner.num_retries)
525 response = self.arvrunner.api.container_requests().create(
527 **extra_submit_params
528 ).execute(num_retries=self.arvrunner.num_retries)
530 self.uuid = response["uuid"]
531 self.arvrunner.process_submitted(self)
533 logger.info("%s submitted container_request %s", self.arvrunner.label(self), response["uuid"])
535 def done(self, record):
537 container = self.arvrunner.api.containers().get(
538 uuid=record["container_uuid"]
539 ).execute(num_retries=self.arvrunner.num_retries)
540 container["log"] = record["log_uuid"]
542 logger.exception("%s while getting runner container", self.arvrunner.label(self))
543 self.arvrunner.output_callback({}, "permanentFail")
545 super(RunnerContainer, self).done(container)