1 # Copyright (C) The Arvados Authors. All rights reserved.
3 # SPDX-License-Identifier: Apache-2.0
14 import ruamel.yaml as yaml
16 from cwltool.errors import WorkflowException
17 from cwltool.process import get_feature, UnsupportedRequirement, shortname
18 from cwltool.pathmapper import adjustFileObjs, adjustDirObjs, visit_class
19 from cwltool.utils import aslist
21 import arvados.collection
23 from .arvdocker import arv_docker_get_image
25 from .runner import Runner, arvados_jobs_image, packed_workflow, trim_anonymous_location, remove_redundant_fields
26 from .fsaccess import CollectionFetcher
27 from .pathmapper import NoFollowPathMapper, trim_listing
28 from .perf import Perf
30 logger = logging.getLogger('arvados.cwl-runner')
31 metrics = logging.getLogger('arvados.cwl-runner.metrics')
33 class ArvadosContainer(object):
34 """Submit and manage a Crunch container request for executing a CWL CommandLineTool."""
36 def __init__(self, runner):
37 self.arvrunner = runner
41 def update_pipeline_component(self, r):
44 def run(self, dry_run=False, pull_image=True, **kwargs):
45 # ArvadosCommandTool subclasses from cwltool.CommandLineTool,
46 # which calls makeJobRunner() to get a new ArvadosContainer
47 # object. The fields that define execution such as
48 # command_line, environment, etc are set on the
49 # ArvadosContainer object by CommandLineTool.job() before
53 "command": self.command_line,
54 "owner_uuid": self.arvrunner.project_uuid,
56 "output_path": self.outdir,
58 "priority": kwargs.get("priority"),
62 runtime_constraints = {}
64 if self.arvrunner.secret_store.has_secret(self.command_line):
65 raise WorkflowException("Secret material leaked on command line, only file literals may contain secrets")
67 if self.arvrunner.secret_store.has_secret(self.environment):
68 raise WorkflowException("Secret material leaked in environment, only file literals may contain secrets")
70 resources = self.builder.resources
71 if resources is not None:
72 runtime_constraints["vcpus"] = resources.get("cores", 1)
73 runtime_constraints["ram"] = resources.get("ram") * 2**20
78 "capacity": resources.get("outdirSize", 0) * 2**20
82 "capacity": resources.get("tmpdirSize", 0) * 2**20
86 scheduling_parameters = {}
88 rf = [self.pathmapper.mapper(f) for f in self.pathmapper.referenced_files]
89 rf.sort(key=lambda k: k.resolved)
91 for resolved, target, tp, stg in rf:
94 if prevdir and target.startswith(prevdir):
99 targetdir = os.path.dirname(target)
100 sp = resolved.split("/", 1)
101 pdh = sp[0][5:] # remove "keep:"
102 mounts[targetdir] = {
103 "kind": "collection",
104 "portable_data_hash": pdh
107 if tp == "Directory":
110 path = os.path.dirname(sp[1])
111 if path and path != "/":
112 mounts[targetdir]["path"] = path
113 prevdir = targetdir + "/"
115 with Perf(metrics, "generatefiles %s" % self.name):
116 if self.generatefiles["listing"]:
117 vwd = arvados.collection.Collection(api_client=self.arvrunner.api,
118 keep_client=self.arvrunner.keep_client,
119 num_retries=self.arvrunner.num_retries)
120 generatemapper = NoFollowPathMapper([self.generatefiles], "", "",
123 logger.debug("generatemapper is %s", generatemapper._pathmap)
125 with Perf(metrics, "createfiles %s" % self.name):
126 for f, p in generatemapper.items():
129 elif p.type in ("File", "Directory", "WritableFile", "WritableDirectory"):
130 if p.resolved.startswith("_:"):
133 source, path = self.arvrunner.fs_access.get_collection(p.resolved)
134 vwd.copy(path, p.target, source_collection=source)
135 elif p.type == "CreateFile":
136 if self.arvrunner.secret_store.has_secret(p.resolved):
137 secret_mounts["%s/%s" % (self.outdir, p.target)] = {
139 "content": self.arvrunner.secret_store.retrieve(p.resolved)
142 with vwd.open(p.target, "w") as n:
143 n.write(p.resolved.encode("utf-8"))
145 def keepemptydirs(p):
146 if isinstance(p, arvados.collection.RichCollectionBase):
148 p.open(".keep", "w").close()
155 with Perf(metrics, "generatefiles.save_new %s" % self.name):
158 for f, p in generatemapper.items():
159 if not p.target or self.arvrunner.secret_store.has_secret(p.resolved):
161 mountpoint = "%s/%s" % (self.outdir, p.target)
162 mounts[mountpoint] = {"kind": "collection",
163 "portable_data_hash": vwd.portable_data_hash(),
165 if p.type.startswith("Writable"):
166 mounts[mountpoint]["writable"] = True
168 container_request["environment"] = {"TMPDIR": self.tmpdir, "HOME": self.outdir}
170 container_request["environment"].update(self.environment)
173 sp = self.stdin[6:].split("/", 1)
174 mounts["stdin"] = {"kind": "collection",
175 "portable_data_hash": sp[0],
179 mounts["stderr"] = {"kind": "file",
180 "path": "%s/%s" % (self.outdir, self.stderr)}
183 mounts["stdout"] = {"kind": "file",
184 "path": "%s/%s" % (self.outdir, self.stdout)}
186 (docker_req, docker_is_req) = get_feature(self, "DockerRequirement")
188 docker_req = {"dockerImageId": "arvados/jobs"}
190 container_request["container_image"] = arv_docker_get_image(self.arvrunner.api,
193 self.arvrunner.project_uuid)
195 api_req, _ = get_feature(self, "http://arvados.org/cwl#APIRequirement")
197 runtime_constraints["API"] = True
199 runtime_req, _ = get_feature(self, "http://arvados.org/cwl#RuntimeConstraints")
201 if "keep_cache" in runtime_req:
202 runtime_constraints["keep_cache_ram"] = runtime_req["keep_cache"] * 2**20
203 if "outputDirType" in runtime_req:
204 if runtime_req["outputDirType"] == "local_output_dir":
205 # Currently the default behavior.
207 elif runtime_req["outputDirType"] == "keep_output_dir":
208 mounts[self.outdir]= {
209 "kind": "collection",
213 partition_req, _ = get_feature(self, "http://arvados.org/cwl#PartitionRequirement")
215 scheduling_parameters["partitions"] = aslist(partition_req["partition"])
217 intermediate_output_req, _ = get_feature(self, "http://arvados.org/cwl#IntermediateOutput")
218 if intermediate_output_req:
219 self.output_ttl = intermediate_output_req["outputTTL"]
221 self.output_ttl = self.arvrunner.intermediate_output_ttl
223 if self.output_ttl < 0:
224 raise WorkflowException("Invalid value %d for output_ttl, cannot be less than zero" % container_request["output_ttl"])
226 container_request["output_ttl"] = self.output_ttl
227 container_request["mounts"] = mounts
228 container_request["secret_mounts"] = secret_mounts
229 container_request["runtime_constraints"] = runtime_constraints
230 container_request["scheduling_parameters"] = scheduling_parameters
232 enable_reuse = kwargs.get("enable_reuse", True)
234 reuse_req, _ = get_feature(self, "http://arvados.org/cwl#ReuseRequirement")
236 enable_reuse = reuse_req["enableReuse"]
237 container_request["use_existing"] = enable_reuse
239 if kwargs.get("runnerjob", "").startswith("arvwf:"):
240 wfuuid = kwargs["runnerjob"][6:kwargs["runnerjob"].index("#")]
241 wfrecord = self.arvrunner.api.workflows().get(uuid=wfuuid).execute(num_retries=self.arvrunner.num_retries)
242 if container_request["name"] == "main":
243 container_request["name"] = wfrecord["name"]
244 container_request["properties"]["template_uuid"] = wfuuid
247 response = self.arvrunner.api.container_requests().create(
248 body=container_request
249 ).execute(num_retries=self.arvrunner.num_retries)
251 self.uuid = response["uuid"]
252 self.arvrunner.processes[self.uuid] = self
254 if response["state"] == "Final":
255 logger.info("%s reused container %s", self.arvrunner.label(self), response["container_uuid"])
258 logger.info("%s %s state is %s", self.arvrunner.label(self), response["uuid"], response["state"])
259 except Exception as e:
260 logger.error("%s got error %s" % (self.arvrunner.label(self), str(e)))
261 self.output_callback({}, "permanentFail")
263 def done(self, record):
266 container = self.arvrunner.api.containers().get(
267 uuid=record["container_uuid"]
268 ).execute(num_retries=self.arvrunner.num_retries)
269 if container["state"] == "Complete":
270 rcode = container["exit_code"]
271 if self.successCodes and rcode in self.successCodes:
272 processStatus = "success"
273 elif self.temporaryFailCodes and rcode in self.temporaryFailCodes:
274 processStatus = "temporaryFail"
275 elif self.permanentFailCodes and rcode in self.permanentFailCodes:
276 processStatus = "permanentFail"
278 processStatus = "success"
280 processStatus = "permanentFail"
282 processStatus = "permanentFail"
284 if processStatus == "permanentFail":
285 logc = arvados.collection.CollectionReader(container["log"],
286 api_client=self.arvrunner.api,
287 keep_client=self.arvrunner.keep_client,
288 num_retries=self.arvrunner.num_retries)
289 done.logtail(logc, logger, "%s error log:" % self.arvrunner.label(self))
291 if record["output_uuid"]:
292 if self.arvrunner.trash_intermediate or self.arvrunner.intermediate_output_ttl:
293 # Compute the trash time to avoid requesting the collection record.
294 trash_at = ciso8601.parse_datetime_unaware(record["modified_at"]) + datetime.timedelta(0, self.arvrunner.intermediate_output_ttl)
295 aftertime = " at %s" % trash_at.strftime("%Y-%m-%d %H:%M:%S UTC") if self.arvrunner.intermediate_output_ttl else ""
296 orpart = ", or" if self.arvrunner.trash_intermediate and self.arvrunner.intermediate_output_ttl else ""
297 oncomplete = " upon successful completion of the workflow" if self.arvrunner.trash_intermediate else ""
298 logger.info("%s Intermediate output %s (%s) will be trashed%s%s%s." % (
299 self.arvrunner.label(self), record["output_uuid"], container["output"], aftertime, orpart, oncomplete))
300 self.arvrunner.add_intermediate_output(record["output_uuid"])
302 if container["output"]:
303 outputs = done.done_outputs(self, container, "/tmp", self.outdir, "/keep")
304 except WorkflowException as e:
305 logger.error("%s unable to collect output from %s:\n%s",
306 self.arvrunner.label(self), container["output"], e, exc_info=(e if self.arvrunner.debug else False))
307 processStatus = "permanentFail"
308 except Exception as e:
309 logger.exception("%s while getting output object: %s", self.arvrunner.label(self), e)
310 processStatus = "permanentFail"
312 self.output_callback(outputs, processStatus)
313 if record["uuid"] in self.arvrunner.processes:
314 del self.arvrunner.processes[record["uuid"]]
317 class RunnerContainer(Runner):
318 """Submit and manage a container that runs arvados-cwl-runner."""
320 def arvados_job_spec(self, dry_run=False, pull_image=True, **kwargs):
321 """Create an Arvados container request for this workflow.
323 The returned dict can be used to create a container passed as
324 the +body+ argument to container_requests().create().
327 adjustDirObjs(self.job_order, trim_listing)
328 visit_class(self.job_order, ("File", "Directory"), trim_anonymous_location)
329 visit_class(self.job_order, ("File", "Directory"), remove_redundant_fields)
332 for param in sorted(self.job_order.keys()):
333 if self.secret_store.has_secret(self.job_order[param]):
334 mnt = "/secrets/s%d" % len(secret_mounts)
335 secret_mounts[mnt] = {
337 "content": self.secret_store.retrieve(self.job_order[param])
339 self.job_order[param] = {"$include": mnt}
342 "owner_uuid": self.arvrunner.project_uuid,
344 "output_path": "/var/spool/cwl",
345 "cwd": "/var/spool/cwl",
346 "priority": self.priority,
347 "state": "Committed",
348 "container_image": arvados_jobs_image(self.arvrunner, self.jobs_image),
350 "/var/lib/cwl/cwl.input.json": {
352 "content": self.job_order
356 "path": "/var/spool/cwl/cwl.output.json"
359 "kind": "collection",
363 "secret_mounts": secret_mounts,
364 "runtime_constraints": {
366 "ram": 1024*1024 * self.submit_runner_ram,
369 "use_existing": self.enable_reuse,
373 if self.tool.tool.get("id", "").startswith("keep:"):
374 sp = self.tool.tool["id"].split('/')
375 workflowcollection = sp[0][5:]
376 workflowname = "/".join(sp[1:])
377 workflowpath = "/var/lib/cwl/workflow/%s" % workflowname
378 container_req["mounts"]["/var/lib/cwl/workflow"] = {
379 "kind": "collection",
380 "portable_data_hash": "%s" % workflowcollection
383 packed = packed_workflow(self.arvrunner, self.tool, self.merged_map)
384 workflowpath = "/var/lib/cwl/workflow.json#main"
385 container_req["mounts"]["/var/lib/cwl/workflow.json"] = {
389 if self.tool.tool.get("id", "").startswith("arvwf:"):
390 container_req["properties"]["template_uuid"] = self.tool.tool["id"][6:33]
393 command = ["arvados-cwl-runner", "--local", "--api=containers", "--no-log-timestamps"]
395 command.append("--output-name=" + self.output_name)
396 container_req["output_name"] = self.output_name
399 command.append("--output-tags=" + self.output_tags)
401 if kwargs.get("debug"):
402 command.append("--debug")
404 if self.enable_reuse:
405 command.append("--enable-reuse")
407 command.append("--disable-reuse")
410 command.append("--on-error=" + self.on_error)
412 if self.intermediate_output_ttl:
413 command.append("--intermediate-output-ttl=%d" % self.intermediate_output_ttl)
415 if self.arvrunner.trash_intermediate:
416 command.append("--trash-intermediate")
418 if self.arvrunner.project_uuid:
419 command.append("--project-uuid="+self.arvrunner.project_uuid)
421 command.append("--eval-timeout=%s" % self.arvrunner.eval_timeout)
423 command.extend([workflowpath, "/var/lib/cwl/cwl.input.json"])
425 container_req["command"] = command
430 def run(self, *args, **kwargs):
431 kwargs["keepprefix"] = "keep:"
432 job_spec = self.arvados_job_spec(*args, **kwargs)
433 job_spec.setdefault("owner_uuid", self.arvrunner.project_uuid)
435 response = self.arvrunner.api.container_requests().create(
437 ).execute(num_retries=self.arvrunner.num_retries)
439 self.uuid = response["uuid"]
440 self.arvrunner.processes[self.uuid] = self
442 logger.info("%s submitted container %s", self.arvrunner.label(self), response["uuid"])
444 if response["state"] == "Final":
447 def done(self, record):
449 container = self.arvrunner.api.containers().get(
450 uuid=record["container_uuid"]
451 ).execute(num_retries=self.arvrunner.num_retries)
452 except Exception as e:
453 logger.exception("%s while getting runner container: %s", self.arvrunner.label(self), e)
454 self.arvrunner.output_callback({}, "permanentFail")
456 super(RunnerContainer, self).done(container)
458 if record["uuid"] in self.arvrunner.processes:
459 del self.arvrunner.processes[record["uuid"]]