sdk/cwl/arvados_cwl/arvcontainer.py

   1 # Copyright (C) The Arvados Authors. All rights reserved.
   2 #
   3 # SPDX-License-Identifier: Apache-2.0
   4
   5 from future import standard_library
   6 standard_library.install_aliases()
   7 from builtins import str
   8
   9 import logging
  10 import json
  11 import os
  12 import urllib.request, urllib.parse, urllib.error
  13 import time
  14 import datetime
  15 import ciso8601
  16 import uuid
  17 import math
  18 import re
  19
  20 import arvados_cwl.util
  21 import ruamel.yaml
  22
  23 from cwltool.errors import WorkflowException
  24 from cwltool.process import UnsupportedRequirement, shortname
  25 from cwltool.utils import aslist, adjustFileObjs, adjustDirObjs, visit_class
  26 from cwltool.job import JobBase
  27
  28 import arvados.collection
  29
  30 from .arvdocker import arv_docker_get_image
  31 from . import done
  32 from .runner import Runner, arvados_jobs_image, packed_workflow, trim_anonymous_location, remove_redundant_fields, make_builder
  33 from .fsaccess import CollectionFetcher
  34 from .pathmapper import NoFollowPathMapper, trim_listing
  35 from .perf import Perf
  36 from ._version import __version__
  37
  38 logger = logging.getLogger('arvados.cwl-runner')
  39 metrics = logging.getLogger('arvados.cwl-runner.metrics')
  40
  41 def cleanup_name_for_collection(name):
  42     return name.replace("/", " ")
  43
  44 class ArvadosContainer(JobBase):
  45     """Submit and manage a Crunch container request for executing a CWL CommandLineTool."""
  46
  47     def __init__(self, runner, job_runtime,
  48                  builder,   # type: Builder
  49                  joborder,  # type: Dict[Text, Union[Dict[Text, Any], List, Text]]
  50                  make_path_mapper,  # type: Callable[..., PathMapper]
  51                  requirements,      # type: List[Dict[Text, Text]]
  52                  hints,     # type: List[Dict[Text, Text]]
  53                  name       # type: Text
  54     ):
  55         super(ArvadosContainer, self).__init__(builder, joborder, make_path_mapper, requirements, hints, name)
  56         self.arvrunner = runner
  57         self.job_runtime = job_runtime
  58         self.running = False
  59         self.uuid = None
  60         self.attempt_count = 0
  61
  62     def update_pipeline_component(self, r):
  63         pass
  64
  65     def _required_env(self):
  66         env = {}
  67         env["HOME"] = self.outdir
  68         env["TMPDIR"] = self.tmpdir
  69         return env
  70
  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
  77         # run() is called.
  78
  79         runtimeContext = self.job_runtime
  80
  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)
  85         else:
  86             container_request = {}
  87
  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", {})
  95
  96         container_request["properties"]["cwl_input"] = self.joborder
  97
  98         runtime_constraints = {}
  99
 100         if runtimeContext.project_uuid:
 101             container_request["owner_uuid"] = runtimeContext.project_uuid
 102
 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")
 105
 106         if self.arvrunner.secret_store.has_secret(self.environment):
 107             raise WorkflowException("Secret material leaked in environment, only file literals may contain secrets")
 108
 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)
 113
 114         mounts = {
 115             self.outdir: {
 116                 "kind": "tmp",
 117                 "capacity": math.ceil(resources.get("outdirSize", 0) * 2**20)
 118             },
 119             self.tmpdir: {
 120                 "kind": "tmp",
 121                 "capacity": math.ceil(resources.get("tmpdirSize", 0) * 2**20)
 122             }
 123         }
 124         secret_mounts = {}
 125         scheduling_parameters = {}
 126
 127         rf = [self.pathmapper.mapper(f) for f in self.pathmapper.referenced_files]
 128         rf.sort(key=lambda k: k.resolved)
 129         prevdir = None
 130         for resolved, target, tp, stg in rf:
 131             if not stg:
 132                 continue
 133             if prevdir and target.startswith(prevdir):
 134                 continue
 135             if tp == "Directory":
 136                 targetdir = target
 137             else:
 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
 144             }
 145             if pdh in self.pathmapper.pdh_to_uuid:
 146                 mounts[targetdir]["uuid"] = self.pathmapper.pdh_to_uuid[pdh]
 147             if len(sp) == 2:
 148                 if tp == "Directory":
 149                     path = sp[1]
 150                 else:
 151                     path = os.path.dirname(sp[1])
 152                 if path and path != "/":
 153                     mounts[targetdir]["path"] = path
 154             prevdir = targetdir + "/"
 155
 156         intermediate_collection_info = arvados_cwl.util.get_intermediate_collection_info(self.name, runtimeContext.current_container, runtimeContext.intermediate_output_ttl)
 157
 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"], "", "",
 164                                                     separateDirs=False)
 165
 166                 sorteditems = sorted(generatemapper.items(), key=lambda n: n[1].target)
 167
 168                 logger.debug("generatemapper is %s", sorteditems)
 169
 170                 with Perf(metrics, "createfiles %s" % self.name):
 171                     for f, p in sorteditems:
 172                         if not p.target:
 173                             continue
 174
 175                         if p.target.startswith("/"):
 176                             dst = p.target[len(self.outdir)+1:] if p.target.startswith(self.outdir+"/") else p.target[1:]
 177                         else:
 178                             dst = p.target
 179
 180                         if p.type in ("File", "Directory", "WritableFile", "WritableDirectory"):
 181                             if p.resolved.startswith("_:"):
 182                                 vwd.mkdirs(dst)
 183                             else:
 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] = {
 190                                     "kind": "text",
 191                                     "content": self.arvrunner.secret_store.retrieve(p.resolved)
 192                                 }
 193                             else:
 194                                 with vwd.open(dst, "w") as n:
 195                                     n.write(p.resolved)
 196
 197                 def keepemptydirs(p):
 198                     if isinstance(p, arvados.collection.RichCollectionBase):
 199                         if len(p) == 0:
 200                             p.open(".keep", "w").close()
 201                         else:
 202                             for c in p:
 203                                 keepemptydirs(p[c])
 204
 205                 keepemptydirs(vwd)
 206
 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"])
 214
 215                 prev = None
 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))):
 219                         continue
 220                     if p.target.startswith("/"):
 221                         dst = p.target[len(self.outdir)+1:] if p.target.startswith(self.outdir+"/") else p.target[1:]
 222                     else:
 223                         dst = p.target
 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(),
 227                                           "path": dst}
 228                     if p.type.startswith("Writable"):
 229                         mounts[mountpoint]["writable"] = True
 230                     prev = p.target + "/"
 231
 232         container_request["environment"] = {"TMPDIR": self.tmpdir, "HOME": self.outdir}
 233         if self.environment:
 234             container_request["environment"].update(self.environment)
 235
 236         if self.stdin:
 237             sp = self.stdin[6:].split("/", 1)
 238             mounts["stdin"] = {"kind": "collection",
 239                                 "portable_data_hash": sp[0],
 240                                 "path": sp[1]}
 241
 242         if self.stderr:
 243             mounts["stderr"] = {"kind": "file",
 244                                 "path": "%s/%s" % (self.outdir, self.stderr)}
 245
 246         if self.stdout:
 247             mounts["stdout"] = {"kind": "file",
 248                                 "path": "%s/%s" % (self.outdir, self.stdout)}
 249
 250         (docker_req, docker_is_req) = self.get_requirement("DockerRequirement")
 251
 252         container_request["container_image"] = arv_docker_get_image(self.arvrunner.api,
 253                                                                     docker_req,
 254                                                                     runtimeContext.pull_image,
 255                                                                     runtimeContext)
 256
 257         network_req, _ = self.get_requirement("NetworkAccess")
 258         if network_req:
 259             runtime_constraints["API"] = network_req["networkAccess"]
 260
 261         api_req, _ = self.get_requirement("http://arvados.org/cwl#APIRequirement")
 262         if api_req:
 263             runtime_constraints["API"] = True
 264
 265         use_disk_cache = (self.arvrunner.api.config()["Containers"].get("DefaultKeepCacheRAM", 0) == 0)
 266
 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
 272
 273         runtime_req, _ = self.get_requirement("http://arvados.org/cwl#RuntimeConstraints")
 274         if runtime_req:
 275             if "keep_cache" in runtime_req:
 276                 if use_disk_cache:
 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)
 279                 else:
 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.
 284                     pass
 285                 elif runtime_req["outputDirType"] == "keep_output_dir":
 286                     mounts[self.outdir]= {
 287                         "kind": "collection",
 288                         "writable": True
 289                     }
 290
 291         partition_req, _ = self.get_requirement("http://arvados.org/cwl#PartitionRequirement")
 292         if partition_req:
 293             scheduling_parameters["partitions"] = aslist(partition_req["partition"])
 294
 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"]
 298         else:
 299             self.output_ttl = self.arvrunner.intermediate_output_ttl
 300
 301         if self.output_ttl < 0:
 302             raise WorkflowException("Invalid value %d for output_ttl, cannot be less than zero" % container_request["output_ttl"])
 303
 304
 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"])
 309             else:
 310                 container_request["output_storage_classes"] = runtimeContext.intermediate_storage_classes.strip().split(",")
 311
 312         cuda_req, _ = self.get_requirement("http://commonwl.org/cwltool#CUDARequirement")
 313         if cuda_req:
 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]
 318             }
 319
 320         if runtimeContext.enable_preemptible is False:
 321             scheduling_parameters["preemptible"] = False
 322         else:
 323             preemptible_req, _ = self.get_requirement("http://arvados.org/cwl#UsePreemptible")
 324             if preemptible_req:
 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:
 329                 pass
 330
 331         if self.timelimit is not None and self.timelimit > 0:
 332             scheduling_parameters["max_run_time"] = self.timelimit
 333
 334         extra_submit_params = {}
 335         if runtimeContext.submit_runner_cluster:
 336             extra_submit_params["cluster_id"] = runtimeContext.submit_runner_cluster
 337
 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
 344
 345         enable_reuse = runtimeContext.enable_reuse
 346         if enable_reuse:
 347             reuse_req, _ = self.get_requirement("WorkReuse")
 348             if reuse_req:
 349                 enable_reuse = reuse_req["enableReuse"]
 350             reuse_req, _ = self.get_requirement("http://arvados.org/cwl#ReuseRequirement")
 351             if reuse_req:
 352                 enable_reuse = reuse_req["enableReuse"]
 353         container_request["use_existing"] = enable_reuse
 354
 355         properties_req, _ = self.get_requirement("http://arvados.org/cwl#ProcessProperties")
 356         if properties_req:
 357             for pr in properties_req["processProperties"]:
 358                 container_request["properties"][pr["propertyName"]] = self.builder.do_eval(pr["propertyValue"])
 359
 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"])
 366             else:
 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")
 369
 370         ram_multiplier = [1]
 371
 372         oom_retry_req, _ = self.get_requirement("http://arvados.org/cwl#OutOfMemoryRetry")
 373         if oom_retry_req and oom_retry_req.get('memoryRetryMultipler'):
 374             ram_multiplier.append(oom_retry_req.get('memoryRetryMultipler'))
 375
 376         if runtimeContext.runnerjob.startswith("arvwf:"):
 377             wfuuid = runtimeContext.runnerjob[6:runtimeContext.runnerjob.index("#")]
 378             wfrecord = self.arvrunner.api.workflows().get(uuid=wfuuid).execute(num_retries=self.arvrunner.num_retries)
 379             if container_request["name"] == "main":
 380                 container_request["name"] = wfrecord["name"]
 381             container_request["properties"]["template_uuid"] = wfuuid
 382
 383         if self.attempt_count == 0:
 384             self.output_callback = self.arvrunner.get_wrapped_callback(self.output_callback)
 385
 386         try:
 387             ram = runtime_constraints["ram"]
 388
 389             self.uuid = runtimeContext.submit_request_uuid
 390
 391             for i in ram_multiplier:
 392                 runtime_constraints["ram"] = ram * i
 393
 394                 if self.uuid:
 395                     response = self.arvrunner.api.container_requests().update(
 396                         uuid=self.uuid,
 397                         body=container_request,
 398                         **extra_submit_params
 399                     ).execute(num_retries=self.arvrunner.num_retries)
 400                 else:
 401                     response = self.arvrunner.api.container_requests().create(
 402                         body=container_request,
 403                         **extra_submit_params
 404                     ).execute(num_retries=self.arvrunner.num_retries)
 405                     self.uuid = response["uuid"]
 406
 407                 if response["container_uuid"] is not None:
 408                     break
 409
 410             if response["container_uuid"] is None:
 411                 runtime_constraints["ram"] = ram * ram_multiplier[self.attempt_count]
 412
 413             container_request["state"] = "Committed"
 414             response = self.arvrunner.api.container_requests().update(
 415                 uuid=self.uuid,
 416                 body=container_request,
 417                 **extra_submit_params
 418             ).execute(num_retries=self.arvrunner.num_retries)
 419
 420             self.arvrunner.process_submitted(self)
 421             self.attempt_count += 1
 422
 423             if response["state"] == "Final":
 424                 logger.info("%s reused container %s", self.arvrunner.label(self), response["container_uuid"])
 425             else:
 426                 logger.info("%s %s state is %s", self.arvrunner.label(self), response["uuid"], response["state"])
 427         except Exception as e:
 428             logger.exception("%s error submitting container\n%s", self.arvrunner.label(self), e)
 429             logger.debug("Container request was %s", container_request)
 430             self.output_callback({}, "permanentFail")
 431
 432     def out_of_memory_retry(self, record, container):
 433         oom_retry_req, _ = self.get_requirement("http://arvados.org/cwl#OutOfMemoryRetry")
 434         if oom_retry_req is None:
 435             return False
 436
 437         # Sometimes it gets killed with no warning
 438         if container["exit_code"] == 137:
 439             return True
 440
 441         logc = arvados.collection.CollectionReader(record["log_uuid"],
 442                                                    api_client=self.arvrunner.api,
 443                                                    keep_client=self.arvrunner.keep_client,
 444                                                    num_retries=self.arvrunner.num_retries)
 445
 446         loglines = [""]
 447         def callback(v1, v2, v3):
 448             loglines[0] = v3
 449
 450         done.logtail(logc, callback, "", maxlen=1000)
 451
 452         # Check allocation failure
 453         oom_matches = oom_retry_req.get('memoryErrorRegex') or r'(bad_alloc|out ?of ?memory|memory ?error|container using over 9.% of memory)'
 454         if re.search(oom_matches, loglines[0], re.IGNORECASE | re.MULTILINE):
 455             return True
 456
 457         return False
 458
 459     def done(self, record):
 460         outputs = {}
 461         retried = False
 462         rcode = None
 463         try:
 464             container = self.arvrunner.api.containers().get(
 465                 uuid=record["container_uuid"]
 466             ).execute(num_retries=self.arvrunner.num_retries)
 467             if container["state"] == "Complete":
 468                 rcode = container["exit_code"]
 469                 if self.successCodes and rcode in self.successCodes:
 470                     processStatus = "success"
 471                 elif self.temporaryFailCodes and rcode in self.temporaryFailCodes:
 472                     processStatus = "temporaryFail"
 473                 elif self.permanentFailCodes and rcode in self.permanentFailCodes:
 474                     processStatus = "permanentFail"
 475                 elif rcode == 0:
 476                     processStatus = "success"
 477                 else:
 478                     processStatus = "permanentFail"
 479
 480                 if processStatus == "permanentFail" and self.attempt_count == 1 and self.out_of_memory_retry(record, container):
 481                     logger.warning("%s Container failed with out of memory error, retrying with more RAM.",
 482                                  self.arvrunner.label(self))
 483                     self.job_runtime.submit_request_uuid = None
 484                     self.uuid = None
 485                     self.run(None)
 486                     retried = True
 487                     return
 488
 489                 if rcode == 137:
 490                     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.",
 491                                  self.arvrunner.label(self))
 492             else:
 493                 processStatus = "permanentFail"
 494
 495             if processStatus == "permanentFail" and record["log_uuid"]:
 496                 logc = arvados.collection.CollectionReader(record["log_uuid"],
 497                                                            api_client=self.arvrunner.api,
 498                                                            keep_client=self.arvrunner.keep_client,
 499                                                            num_retries=self.arvrunner.num_retries)
 500                 label = self.arvrunner.label(self)
 501                 done.logtail(
 502                     logc, logger.error,
 503                     "%s (%s) error log:" % (label, record["uuid"]), maxlen=40, include_crunchrun=(rcode is None or rcode > 127))
 504
 505             if record["output_uuid"]:
 506                 if self.arvrunner.trash_intermediate or self.arvrunner.intermediate_output_ttl:
 507                     # Compute the trash time to avoid requesting the collection record.
 508                     trash_at = ciso8601.parse_datetime_as_naive(record["modified_at"]) + datetime.timedelta(0, self.arvrunner.intermediate_output_ttl)
 509                     aftertime = " at %s" % trash_at.strftime("%Y-%m-%d %H:%M:%S UTC") if self.arvrunner.intermediate_output_ttl else ""
 510                     orpart = ", or" if self.arvrunner.trash_intermediate and self.arvrunner.intermediate_output_ttl else ""
 511                     oncomplete = " upon successful completion of the workflow" if self.arvrunner.trash_intermediate else ""
 512                     logger.info("%s Intermediate output %s (%s) will be trashed%s%s%s." % (
 513                         self.arvrunner.label(self), record["output_uuid"], container["output"], aftertime, orpart, oncomplete))
 514                 self.arvrunner.add_intermediate_output(record["output_uuid"])
 515
 516             if container["output"]:
 517                 outputs = done.done_outputs(self, container, "/tmp", self.outdir, "/keep")
 518
 519             properties = record["properties"].copy()
 520             properties["cwl_output"] = outputs
 521             self.arvrunner.api.container_requests().update(
 522                 uuid=self.uuid,
 523                 body={"container_request": {"properties": properties}}
 524             ).execute(num_retries=self.arvrunner.num_retries)
 525         except WorkflowException as e:
 526             # Only include a stack trace if in debug mode.
 527             # A stack trace may obfuscate more useful output about the workflow.
 528             logger.error("%s unable to collect output from %s:\n%s",
 529                          self.arvrunner.label(self), container["output"], e, exc_info=(e if self.arvrunner.debug else False))
 530             processStatus = "permanentFail"
 531         except Exception:
 532             logger.exception("%s while getting output object:", self.arvrunner.label(self))
 533             processStatus = "permanentFail"
 534         finally:
 535             if not retried:
 536                 self.output_callback(outputs, processStatus)
 537
 538
 539 class RunnerContainer(Runner):
 540     """Submit and manage a container that runs arvados-cwl-runner."""
 541
 542     def arvados_job_spec(self, runtimeContext, git_info):
 543         """Create an Arvados container request for this workflow.
 544
 545         The returned dict can be used to create a container passed as
 546         the +body+ argument to container_requests().create().
 547         """
 548
 549         adjustDirObjs(self.job_order, trim_listing)
 550         visit_class(self.job_order, ("File", "Directory"), trim_anonymous_location)
 551         visit_class(self.job_order, ("File", "Directory"), remove_redundant_fields)
 552
 553         secret_mounts = {}
 554         for param in sorted(self.job_order.keys()):
 555             if self.secret_store.has_secret(self.job_order[param]):
 556                 mnt = "/secrets/s%d" % len(secret_mounts)
 557                 secret_mounts[mnt] = {
 558                     "kind": "text",
 559                     "content": self.secret_store.retrieve(self.job_order[param])
 560                 }
 561                 self.job_order[param] = {"$include": mnt}
 562
 563         container_req = {
 564             "name": self.name,
 565             "output_path": "/var/spool/cwl",
 566             "cwd": "/var/spool/cwl",
 567             "priority": self.priority,
 568             "state": "Committed",
 569             "container_image": arvados_jobs_image(self.arvrunner, self.jobs_image, runtimeContext),
 570             "mounts": {
 571                 "/var/lib/cwl/cwl.input.json": {
 572                     "kind": "json",
 573                     "content": self.job_order
 574                 },
 575                 "stdout": {
 576                     "kind": "file",
 577                     "path": "/var/spool/cwl/cwl.output.json"
 578                 },
 579                 "/var/spool/cwl": {
 580                     "kind": "collection",
 581                     "writable": True
 582                 }
 583             },
 584             "secret_mounts": secret_mounts,
 585             "runtime_constraints": {
 586                 "vcpus": math.ceil(self.submit_runner_cores),
 587                 "ram": 1024*1024 * (math.ceil(self.submit_runner_ram) + math.ceil(self.collection_cache_size)),
 588                 "API": True
 589             },
 590             "use_existing": False, # Never reuse the runner container - see #15497.
 591             "properties": {}
 592         }
 593
 594         if self.embedded_tool.tool.get("id", "").startswith("keep:"):
 595             sp = self.embedded_tool.tool["id"].split('/')
 596             workflowcollection = sp[0][5:]
 597             workflowname = "/".join(sp[1:])
 598             workflowpath = "/var/lib/cwl/workflow/%s" % workflowname
 599             container_req["mounts"]["/var/lib/cwl/workflow"] = {
 600                 "kind": "collection",
 601                 "portable_data_hash": "%s" % workflowcollection
 602             }
 603         elif self.embedded_tool.tool.get("id", "").startswith("arvwf:"):
 604             uuid, frg = urllib.parse.urldefrag(self.embedded_tool.tool["id"])
 605             workflowpath = "/var/lib/cwl/workflow.json#" + frg
 606             packedtxt = self.loadingContext.loader.fetch_text(uuid)
 607             yaml = ruamel.yaml.YAML(typ='safe', pure=True)
 608             packed = yaml.load(packedtxt)
 609             container_req["mounts"]["/var/lib/cwl/workflow.json"] = {
 610                 "kind": "json",
 611                 "content": packed
 612             }
 613             container_req["properties"]["template_uuid"] = self.embedded_tool.tool["id"][6:33]
 614         else:
 615             main = self.loadingContext.loader.idx["_:main"]
 616             if main.get("id") == "_:main":
 617                 del main["id"]
 618             workflowpath = "/var/lib/cwl/workflow.json#main"
 619             container_req["mounts"]["/var/lib/cwl/workflow.json"] = {
 620                 "kind": "json",
 621                 "content": main
 622             }
 623
 624         container_req["properties"].update({k.replace("http://arvados.org/cwl#", "arv:"): v for k, v in git_info.items()})
 625
 626         properties_req, _ = self.embedded_tool.get_requirement("http://arvados.org/cwl#ProcessProperties")
 627         if properties_req:
 628             builder = make_builder(self.job_order, self.embedded_tool.hints, self.embedded_tool.requirements, runtimeContext, self.embedded_tool.metadata)
 629             for pr in properties_req["processProperties"]:
 630                 container_req["properties"][pr["propertyName"]] = builder.do_eval(pr["propertyValue"])
 631
 632         # --local means execute the workflow instead of submitting a container request
 633         # --api=containers means use the containers API
 634         # --no-log-timestamps means don't add timestamps (the logging infrastructure does this)
 635         # --disable-validate because we already validated so don't need to do it again
 636         # --eval-timeout is the timeout for javascript invocation
 637         # --parallel-task-count is the number of threads to use for job submission
 638         # --enable/disable-reuse sets desired job reuse
 639         # --collection-cache-size sets aside memory to store collections
 640         command = ["arvados-cwl-runner",
 641                    "--local",
 642                    "--api=containers",
 643                    "--no-log-timestamps",
 644                    "--disable-validate",
 645                    "--disable-color",
 646                    "--eval-timeout=%s" % self.arvrunner.eval_timeout,
 647                    "--thread-count=%s" % self.arvrunner.thread_count,
 648                    "--enable-reuse" if self.enable_reuse else "--disable-reuse",
 649                    "--collection-cache-size=%s" % self.collection_cache_size]
 650
 651         if self.output_name:
 652             command.append("--output-name=" + self.output_name)
 653             container_req["output_name"] = self.output_name
 654
 655         if self.output_tags:
 656             command.append("--output-tags=" + self.output_tags)
 657
 658         if runtimeContext.debug:
 659             command.append("--debug")
 660
 661         if runtimeContext.storage_classes != "default" and runtimeContext.storage_classes:
 662             command.append("--storage-classes=" + runtimeContext.storage_classes)
 663
 664         if runtimeContext.intermediate_storage_classes != "default" and runtimeContext.intermediate_storage_classes:
 665             command.append("--intermediate-storage-classes=" + runtimeContext.intermediate_storage_classes)
 666
 667         if runtimeContext.on_error:
 668             command.append("--on-error=" + self.on_error)
 669
 670         if runtimeContext.intermediate_output_ttl:
 671             command.append("--intermediate-output-ttl=%d" % runtimeContext.intermediate_output_ttl)
 672
 673         if runtimeContext.trash_intermediate:
 674             command.append("--trash-intermediate")
 675
 676         if runtimeContext.project_uuid:
 677             command.append("--project-uuid="+runtimeContext.project_uuid)
 678
 679         if self.enable_dev:
 680             command.append("--enable-dev")
 681
 682         if runtimeContext.enable_preemptible is True:
 683             command.append("--enable-preemptible")
 684
 685         if runtimeContext.enable_preemptible is False:
 686             command.append("--disable-preemptible")
 687
 688         if runtimeContext.varying_url_params:
 689             command.append("--varying-url-params="+runtimeContext.varying_url_params)
 690
 691         if runtimeContext.prefer_cached_downloads:
 692             command.append("--prefer-cached-downloads")
 693
 694         if self.fast_parser:
 695             command.append("--fast-parser")
 696
 697         command.extend([workflowpath, "/var/lib/cwl/cwl.input.json"])
 698
 699         container_req["command"] = command
 700
 701         return container_req
 702
 703
 704     def run(self, runtimeContext):
 705         runtimeContext.keepprefix = "keep:"
 706         job_spec = self.arvados_job_spec(runtimeContext, self.git_info)
 707         if runtimeContext.project_uuid:
 708             job_spec["owner_uuid"] = runtimeContext.project_uuid
 709
 710         extra_submit_params = {}
 711         if runtimeContext.submit_runner_cluster:
 712             extra_submit_params["cluster_id"] = runtimeContext.submit_runner_cluster
 713
 714         if runtimeContext.submit_request_uuid:
 715             if "cluster_id" in extra_submit_params:
 716                 # Doesn't make sense for "update" and actually fails
 717                 del extra_submit_params["cluster_id"]
 718             response = self.arvrunner.api.container_requests().update(
 719                 uuid=runtimeContext.submit_request_uuid,
 720                 body=job_spec,
 721                 **extra_submit_params
 722             ).execute(num_retries=self.arvrunner.num_retries)
 723         else:
 724             response = self.arvrunner.api.container_requests().create(
 725                 body=job_spec,
 726                 **extra_submit_params
 727             ).execute(num_retries=self.arvrunner.num_retries)
 728
 729         self.uuid = response["uuid"]
 730         self.arvrunner.process_submitted(self)
 731
 732         logger.info("%s submitted container_request %s", self.arvrunner.label(self), response["uuid"])
 733
 734         workbench1 = self.arvrunner.api.config()["Services"]["Workbench1"]["ExternalURL"]
 735         workbench2 = self.arvrunner.api.config()["Services"]["Workbench2"]["ExternalURL"]
 736         url = ""
 737         if workbench2:
 738             url = "{}processes/{}".format(workbench2, response["uuid"])
 739         elif workbench1:
 740             url = "{}container_requests/{}".format(workbench1, response["uuid"])
 741         if url:
 742             logger.info("Monitor workflow progress at %s", url)
 743
 744
 745     def done(self, record):
 746         try:
 747             container = self.arvrunner.api.containers().get(
 748                 uuid=record["container_uuid"]
 749             ).execute(num_retries=self.arvrunner.num_retries)
 750             container["log"] = record["log_uuid"]
 751         except Exception:
 752             logger.exception("%s while getting runner container", self.arvrunner.label(self))
 753             self.arvrunner.output_callback({}, "permanentFail")
 754         else:
 755             super(RunnerContainer, self).done(container)