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 import logging
   6 import json
   7 import os
   8 import urllib.request, urllib.parse, urllib.error
   9 import time
  10 import datetime
  11 import ciso8601
  12 import uuid
  13 import math
  14 import re
  15
  16 import arvados_cwl.util
  17 import ruamel.yaml
  18
  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
  23
  24 import arvados.collection
  25
  26 import crunchstat_summary.summarizer
  27 import crunchstat_summary.reader
  28
  29 from .arvdocker import arv_docker_get_image
  30 from . import done
  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__
  36
  37 logger = logging.getLogger('arvados.cwl-runner')
  38 metrics = logging.getLogger('arvados.cwl-runner.metrics')
  39
  40 def cleanup_name_for_collection(name):
  41     return name.replace("/", " ")
  42
  43 class ArvadosContainer(JobBase):
  44     """Submit and manage a Crunch container request for executing a CWL CommandLineTool."""
  45
  46     def __init__(self, runner, job_runtime,
  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]]
  52                  name       # type: Text
  53     ):
  54         super(ArvadosContainer, self).__init__(builder, joborder, make_path_mapper, requirements, hints, name)
  55         self.arvrunner = runner
  56         self.job_runtime = job_runtime
  57         self.running = False
  58         self.uuid = None
  59         self.attempt_count = 0
  60
  61     def update_pipeline_component(self, r):
  62         pass
  63
  64     def _required_env(self):
  65         env = {}
  66         env["HOME"] = self.outdir
  67         env["TMPDIR"] = self.tmpdir
  68         return env
  69
  70     def run(self, toplevelRuntimeContext):
  71         # ArvadosCommandTool subclasses from cwltool.CommandLineTool,
  72         # which calls makeJobRunner() to get a new ArvadosContainer
  73         # object.  The fields that define execution such as
  74         # command_line, environment, etc are set on the
  75         # ArvadosContainer object by CommandLineTool.job() before
  76         # run() is called.
  77
  78         runtimeContext = self.job_runtime
  79
  80         if runtimeContext.submit_request_uuid:
  81             container_request = self.arvrunner.api.container_requests().get(
  82                 uuid=runtimeContext.submit_request_uuid
  83             ).execute(num_retries=self.arvrunner.num_retries)
  84         else:
  85             container_request = {}
  86
  87         container_request["command"] = self.command_line
  88         container_request["name"] = self.name
  89         container_request["output_path"] = self.outdir
  90         container_request["cwd"] = self.outdir
  91         container_request["priority"] = runtimeContext.priority
  92         container_request["state"] = "Uncommitted"
  93         container_request.setdefault("properties", {})
  94
  95         container_request["properties"]["cwl_input"] = self.joborder
  96
  97         runtime_constraints = {}
  98
  99         if runtimeContext.project_uuid:
 100             container_request["owner_uuid"] = runtimeContext.project_uuid
 101
 102         if self.arvrunner.secret_store.has_secret(self.command_line):
 103             raise WorkflowException("Secret material leaked on command line, only file literals may contain secrets")
 104
 105         if self.arvrunner.secret_store.has_secret(self.environment):
 106             raise WorkflowException("Secret material leaked in environment, only file literals may contain secrets")
 107
 108         resources = self.builder.resources
 109         if resources is not None:
 110             runtime_constraints["vcpus"] = math.ceil(resources.get("cores", 1))
 111             runtime_constraints["ram"] = math.ceil(resources.get("ram") * 2**20)
 112
 113         mounts = {
 114             self.outdir: {
 115                 "kind": "tmp",
 116                 "capacity": math.ceil(resources.get("outdirSize", 0) * 2**20)
 117             },
 118             self.tmpdir: {
 119                 "kind": "tmp",
 120                 "capacity": math.ceil(resources.get("tmpdirSize", 0) * 2**20)
 121             }
 122         }
 123         secret_mounts = {}
 124         scheduling_parameters = {}
 125
 126         rf = [self.pathmapper.mapper(f) for f in self.pathmapper.referenced_files]
 127         rf.sort(key=lambda k: k.resolved)
 128         prevdir = None
 129         for resolved, target, tp, stg in rf:
 130             if not stg:
 131                 continue
 132             if prevdir and target.startswith(prevdir):
 133                 continue
 134             if tp == "Directory":
 135                 targetdir = target
 136             else:
 137                 targetdir = os.path.dirname(target)
 138             sp = resolved.split("/", 1)
 139             pdh = sp[0][5:]   # remove "keep:"
 140             mounts[targetdir] = {
 141                 "kind": "collection",
 142                 "portable_data_hash": pdh
 143             }
 144             if pdh in self.pathmapper.pdh_to_uuid:
 145                 mounts[targetdir]["uuid"] = self.pathmapper.pdh_to_uuid[pdh]
 146             if len(sp) == 2:
 147                 if tp == "Directory":
 148                     path = sp[1]
 149                 else:
 150                     path = os.path.dirname(sp[1])
 151                 if path and path != "/":
 152                     mounts[targetdir]["path"] = path
 153             prevdir = targetdir + "/"
 154
 155         intermediate_collection_info = arvados_cwl.util.get_intermediate_collection_info(self.name, runtimeContext.current_container, runtimeContext.intermediate_output_ttl)
 156
 157         with Perf(metrics, "generatefiles %s" % self.name):
 158             if self.generatefiles["listing"]:
 159                 vwd = arvados.collection.Collection(api_client=self.arvrunner.api,
 160                                                     keep_client=self.arvrunner.keep_client,
 161                                                     num_retries=self.arvrunner.num_retries)
 162                 generatemapper = NoFollowPathMapper(self.generatefiles["listing"], "", "",
 163                                                     separateDirs=False)
 164
 165                 sorteditems = sorted(generatemapper.items(), key=lambda n: n[1].target)
 166
 167                 logger.debug("generatemapper is %s", sorteditems)
 168
 169                 with Perf(metrics, "createfiles %s" % self.name):
 170                     for f, p in sorteditems:
 171                         if not p.target:
 172                             continue
 173
 174                         if p.target.startswith("/"):
 175                             dst = p.target[len(self.outdir)+1:] if p.target.startswith(self.outdir+"/") else p.target[1:]
 176                         else:
 177                             dst = p.target
 178
 179                         if p.type in ("File", "Directory", "WritableFile", "WritableDirectory"):
 180                             if p.resolved.startswith("_:"):
 181                                 vwd.mkdirs(dst)
 182                             else:
 183                                 source, path = self.arvrunner.fs_access.get_collection(p.resolved)
 184                                 vwd.copy(path or ".", dst, source_collection=source)
 185                         elif p.type == "CreateFile":
 186                             if self.arvrunner.secret_store.has_secret(p.resolved):
 187                                 mountpoint = p.target if p.target.startswith("/") else os.path.join(self.outdir, p.target)
 188                                 secret_mounts[mountpoint] = {
 189                                     "kind": "text",
 190                                     "content": self.arvrunner.secret_store.retrieve(p.resolved)
 191                                 }
 192                             else:
 193                                 with vwd.open(dst, "w") as n:
 194                                     n.write(p.resolved)
 195
 196                 def keepemptydirs(p):
 197                     if isinstance(p, arvados.collection.RichCollectionBase):
 198                         if len(p) == 0:
 199                             p.open(".keep", "w").close()
 200                         else:
 201                             for c in p:
 202                                 keepemptydirs(p[c])
 203
 204                 keepemptydirs(vwd)
 205
 206                 if not runtimeContext.current_container:
 207                     runtimeContext.current_container = arvados_cwl.util.get_current_container(self.arvrunner.api, self.arvrunner.num_retries, logger)
 208                 vwd.save_new(name=intermediate_collection_info["name"],
 209                              owner_uuid=runtimeContext.project_uuid,
 210                              ensure_unique_name=True,
 211                              trash_at=intermediate_collection_info["trash_at"],
 212                              properties=intermediate_collection_info["properties"])
 213
 214                 prev = None
 215                 for f, p in sorteditems:
 216                     if (not p.target or self.arvrunner.secret_store.has_secret(p.resolved) or
 217                         (prev is not None and p.target.startswith(prev))):
 218                         continue
 219                     if p.target.startswith("/"):
 220                         dst = p.target[len(self.outdir)+1:] if p.target.startswith(self.outdir+"/") else p.target[1:]
 221                     else:
 222                         dst = p.target
 223                     mountpoint = p.target if p.target.startswith("/") else os.path.join(self.outdir, p.target)
 224                     mounts[mountpoint] = {"kind": "collection",
 225                                           "portable_data_hash": vwd.portable_data_hash(),
 226                                           "path": dst}
 227                     if p.type.startswith("Writable"):
 228                         mounts[mountpoint]["writable"] = True
 229                     prev = p.target + "/"
 230
 231         container_request["environment"] = {"TMPDIR": self.tmpdir, "HOME": self.outdir}
 232         if self.environment:
 233             container_request["environment"].update(self.environment)
 234
 235         if self.stdin:
 236             sp = self.stdin[6:].split("/", 1)
 237             mounts["stdin"] = {"kind": "collection",
 238                                 "portable_data_hash": sp[0],
 239                                 "path": sp[1]}
 240
 241         if self.stderr:
 242             mounts["stderr"] = {"kind": "file",
 243                                 "path": "%s/%s" % (self.outdir, self.stderr)}
 244
 245         if self.stdout:
 246             mounts["stdout"] = {"kind": "file",
 247                                 "path": "%s/%s" % (self.outdir, self.stdout)}
 248
 249         (docker_req, docker_is_req) = self.get_requirement("DockerRequirement")
 250
 251         container_request["container_image"] = arv_docker_get_image(self.arvrunner.api,
 252                                                                     docker_req,
 253                                                                     runtimeContext.pull_image,
 254                                                                     runtimeContext)
 255
 256         network_req, _ = self.get_requirement("NetworkAccess")
 257         if network_req:
 258             runtime_constraints["API"] = network_req["networkAccess"]
 259
 260         api_req, _ = self.get_requirement("http://arvados.org/cwl#APIRequirement")
 261         if api_req:
 262             runtime_constraints["API"] = True
 263
 264         use_disk_cache = (self.arvrunner.api.config()["Containers"].get("DefaultKeepCacheRAM", 0) == 0)
 265
 266         keep_cache_type_req, _ = self.get_requirement("http://arvados.org/cwl#KeepCacheTypeRequirement")
 267         if keep_cache_type_req:
 268             if "keepCacheType" in keep_cache_type_req:
 269                 if keep_cache_type_req["keepCacheType"] == "ram_cache":
 270                     use_disk_cache = False
 271
 272         runtime_req, _ = self.get_requirement("http://arvados.org/cwl#RuntimeConstraints")
 273         if runtime_req:
 274             if "keep_cache" in runtime_req:
 275                 if use_disk_cache:
 276                     # If DefaultKeepCacheRAM is zero it means we should use disk cache.
 277                     runtime_constraints["keep_cache_disk"] = math.ceil(runtime_req["keep_cache"] * 2**20)
 278                 else:
 279                     runtime_constraints["keep_cache_ram"] = math.ceil(runtime_req["keep_cache"] * 2**20)
 280             if "outputDirType" in runtime_req:
 281                 if runtime_req["outputDirType"] == "local_output_dir":
 282                     # Currently the default behavior.
 283                     pass
 284                 elif runtime_req["outputDirType"] == "keep_output_dir":
 285                     mounts[self.outdir]= {
 286                         "kind": "collection",
 287                         "writable": True
 288                     }
 289
 290         partition_req, _ = self.get_requirement("http://arvados.org/cwl#PartitionRequirement")
 291         if partition_req:
 292             scheduling_parameters["partitions"] = aslist(partition_req["partition"])
 293
 294         intermediate_output_req, _ = self.get_requirement("http://arvados.org/cwl#IntermediateOutput")
 295         if intermediate_output_req:
 296             self.output_ttl = intermediate_output_req["outputTTL"]
 297         else:
 298             self.output_ttl = self.arvrunner.intermediate_output_ttl
 299
 300         if self.output_ttl < 0:
 301             raise WorkflowException("Invalid value %d for output_ttl, cannot be less than zero" % container_request["output_ttl"])
 302
 303
 304         if self.arvrunner.api._rootDesc["revision"] >= "20210628":
 305             storage_class_req, _ = self.get_requirement("http://arvados.org/cwl#OutputStorageClass")
 306             if storage_class_req and storage_class_req.get("intermediateStorageClass"):
 307                 container_request["output_storage_classes"] = aslist(storage_class_req["intermediateStorageClass"])
 308             else:
 309                 container_request["output_storage_classes"] = runtimeContext.intermediate_storage_classes.strip().split(",")
 310
 311         cuda_req, _ = self.get_requirement("http://commonwl.org/cwltool#CUDARequirement")
 312         if cuda_req:
 313             runtime_constraints["cuda"] = {
 314                 "device_count": resources.get("cudaDeviceCount", 1),
 315                 "driver_version": cuda_req["cudaVersionMin"],
 316                 "hardware_capability": aslist(cuda_req["cudaComputeCapability"])[0]
 317             }
 318
 319         if runtimeContext.enable_preemptible is False:
 320             scheduling_parameters["preemptible"] = False
 321         else:
 322             preemptible_req, _ = self.get_requirement("http://arvados.org/cwl#UsePreemptible")
 323             if preemptible_req:
 324                 scheduling_parameters["preemptible"] = preemptible_req["usePreemptible"]
 325             elif runtimeContext.enable_preemptible is True:
 326                 scheduling_parameters["preemptible"] = True
 327             elif runtimeContext.enable_preemptible is None:
 328                 pass
 329
 330         if self.timelimit is not None and self.timelimit > 0:
 331             scheduling_parameters["max_run_time"] = self.timelimit
 332
 333         extra_submit_params = {}
 334         if runtimeContext.submit_runner_cluster:
 335             extra_submit_params["cluster_id"] = runtimeContext.submit_runner_cluster
 336
 337         container_request["output_name"] = cleanup_name_for_collection("Output from step %s" % (self.name))
 338         container_request["output_ttl"] = self.output_ttl
 339         container_request["mounts"] = mounts
 340         container_request["secret_mounts"] = secret_mounts
 341         container_request["runtime_constraints"] = runtime_constraints
 342         container_request["scheduling_parameters"] = scheduling_parameters
 343
 344         enable_reuse = runtimeContext.enable_reuse
 345         if enable_reuse:
 346             reuse_req, _ = self.get_requirement("WorkReuse")
 347             if reuse_req:
 348                 enable_reuse = reuse_req["enableReuse"]
 349             reuse_req, _ = self.get_requirement("http://arvados.org/cwl#ReuseRequirement")
 350             if reuse_req:
 351                 enable_reuse = reuse_req["enableReuse"]
 352         container_request["use_existing"] = enable_reuse
 353
 354         properties_req, _ = self.get_requirement("http://arvados.org/cwl#ProcessProperties")
 355         if properties_req:
 356             for pr in properties_req["processProperties"]:
 357                 container_request["properties"][pr["propertyName"]] = self.builder.do_eval(pr["propertyValue"])
 358
 359         output_properties_req, _ = self.get_requirement("http://arvados.org/cwl#OutputCollectionProperties")
 360         if output_properties_req:
 361             if self.arvrunner.api._rootDesc["revision"] >= "20220510":
 362                 container_request["output_properties"] = {}
 363                 for pr in output_properties_req["outputProperties"]:
 364                     container_request["output_properties"][pr["propertyName"]] = self.builder.do_eval(pr["propertyValue"])
 365             else:
 366                 logger.warning("%s API revision is %s, revision %s is required to support setting properties on output collections.",
 367                                self.arvrunner.label(self), self.arvrunner.api._rootDesc["revision"], "20220510")
 368
 369         ram_multiplier = [1]
 370
 371         oom_retry_req, _ = self.get_requirement("http://arvados.org/cwl#OutOfMemoryRetry")
 372         if oom_retry_req:
 373             if oom_retry_req.get('memoryRetryMultiplier'):
 374                 ram_multiplier.append(oom_retry_req.get('memoryRetryMultiplier'))
 375             elif oom_retry_req.get('memoryRetryMultipler'):
 376                 ram_multiplier.append(oom_retry_req.get('memoryRetryMultipler'))
 377             else:
 378                 ram_multiplier.append(2)
 379
 380         if runtimeContext.runnerjob.startswith("arvwf:"):
 381             wfuuid = runtimeContext.runnerjob[6:runtimeContext.runnerjob.index("#")]
 382             wfrecord = self.arvrunner.api.workflows().get(uuid=wfuuid).execute(num_retries=self.arvrunner.num_retries)
 383             if container_request["name"] == "main":
 384                 container_request["name"] = wfrecord["name"]
 385             container_request["properties"]["template_uuid"] = wfuuid
 386
 387         if self.attempt_count == 0:
 388             self.output_callback = self.arvrunner.get_wrapped_callback(self.output_callback)
 389
 390         try:
 391             ram = runtime_constraints["ram"]
 392
 393             self.uuid = runtimeContext.submit_request_uuid
 394
 395             for i in ram_multiplier:
 396                 runtime_constraints["ram"] = ram * i
 397
 398                 if self.uuid:
 399                     response = self.arvrunner.api.container_requests().update(
 400                         uuid=self.uuid,
 401                         body=container_request,
 402                         **extra_submit_params
 403                     ).execute(num_retries=self.arvrunner.num_retries)
 404                 else:
 405                     response = self.arvrunner.api.container_requests().create(
 406                         body=container_request,
 407                         **extra_submit_params
 408                     ).execute(num_retries=self.arvrunner.num_retries)
 409                     self.uuid = response["uuid"]
 410
 411                 if response["container_uuid"] is not None:
 412                     break
 413
 414             if response["container_uuid"] is None:
 415                 runtime_constraints["ram"] = ram * ram_multiplier[self.attempt_count]
 416
 417             container_request["state"] = "Committed"
 418             response = self.arvrunner.api.container_requests().update(
 419                 uuid=self.uuid,
 420                 body=container_request,
 421                 **extra_submit_params
 422             ).execute(num_retries=self.arvrunner.num_retries)
 423
 424             self.arvrunner.process_submitted(self)
 425             self.attempt_count += 1
 426
 427             if response["state"] == "Final":
 428                 logger.info("%s reused container %s", self.arvrunner.label(self), response["container_uuid"])
 429             else:
 430                 logger.info("%s %s state is %s", self.arvrunner.label(self), response["uuid"], response["state"])
 431         except Exception as e:
 432             logger.exception("%s error submitting container\n%s", self.arvrunner.label(self), e)
 433             logger.debug("Container request was %s", container_request)
 434             self.output_callback({}, "permanentFail")
 435
 436     def out_of_memory_retry(self, record, container):
 437         oom_retry_req, _ = self.get_requirement("http://arvados.org/cwl#OutOfMemoryRetry")
 438         if oom_retry_req is None:
 439             return False
 440
 441         # Sometimes it gets killed with no warning
 442         if container["exit_code"] == 137:
 443             return True
 444
 445         logc = arvados.collection.CollectionReader(record["log_uuid"],
 446                                                    api_client=self.arvrunner.api,
 447                                                    keep_client=self.arvrunner.keep_client,
 448                                                    num_retries=self.arvrunner.num_retries)
 449
 450         loglines = [""]
 451         def callback(v1, v2, v3):
 452             loglines[0] = v3
 453
 454         done.logtail(logc, callback, "", maxlen=1000)
 455
 456         # Check allocation failure
 457         oom_matches = oom_retry_req.get('memoryErrorRegex') or r'(bad_alloc|out ?of ?memory|memory ?error|container using over 9.% of memory)'
 458         if re.search(oom_matches, loglines[0], re.IGNORECASE | re.MULTILINE):
 459             return True
 460
 461         return False
 462
 463     def done(self, record):
 464         outputs = {}
 465         retried = False
 466         rcode = None
 467         try:
 468             container = self.arvrunner.api.containers().get(
 469                 uuid=record["container_uuid"]
 470             ).execute(num_retries=self.arvrunner.num_retries)
 471             if container["state"] == "Complete":
 472                 rcode = container["exit_code"]
 473                 if self.successCodes and rcode in self.successCodes:
 474                     processStatus = "success"
 475                 elif self.temporaryFailCodes and rcode in self.temporaryFailCodes:
 476                     processStatus = "temporaryFail"
 477                 elif self.permanentFailCodes and rcode in self.permanentFailCodes:
 478                     processStatus = "permanentFail"
 479                 elif rcode == 0:
 480                     processStatus = "success"
 481                 else:
 482                     processStatus = "permanentFail"
 483
 484                 if processStatus == "permanentFail" and self.attempt_count == 1 and self.out_of_memory_retry(record, container):
 485                     logger.warning("%s Container failed with out of memory error, retrying with more RAM.",
 486                                  self.arvrunner.label(self))
 487                     self.job_runtime.submit_request_uuid = None
 488                     self.uuid = None
 489                     self.run(None)
 490                     retried = True
 491                     return
 492
 493                 if rcode == 137:
 494                     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.",
 495                                  self.arvrunner.label(self))
 496             else:
 497                 processStatus = "permanentFail"
 498
 499             logc = None
 500             if record["log_uuid"]:
 501                 logc = arvados.collection.Collection(record["log_uuid"],
 502                                                      api_client=self.arvrunner.api,
 503                                                      keep_client=self.arvrunner.keep_client,
 504                                                      num_retries=self.arvrunner.num_retries)
 505
 506             if processStatus == "permanentFail" and logc is not None:
 507                 label = self.arvrunner.label(self)
 508                 done.logtail(
 509                     logc, logger.error,
 510                     "%s (%s) error log:" % (label, record["uuid"]), maxlen=40, include_crunchrun=(rcode is None or rcode > 127))
 511
 512             if record["output_uuid"]:
 513                 if self.arvrunner.trash_intermediate or self.arvrunner.intermediate_output_ttl:
 514                     # Compute the trash time to avoid requesting the collection record.
 515                     trash_at = ciso8601.parse_datetime_as_naive(record["modified_at"]) + datetime.timedelta(0, self.arvrunner.intermediate_output_ttl)
 516                     aftertime = " at %s" % trash_at.strftime("%Y-%m-%d %H:%M:%S UTC") if self.arvrunner.intermediate_output_ttl else ""
 517                     orpart = ", or" if self.arvrunner.trash_intermediate and self.arvrunner.intermediate_output_ttl else ""
 518                     oncomplete = " upon successful completion of the workflow" if self.arvrunner.trash_intermediate else ""
 519                     logger.info("%s Intermediate output %s (%s) will be trashed%s%s%s." % (
 520                         self.arvrunner.label(self), record["output_uuid"], container["output"], aftertime, orpart, oncomplete))
 521                 self.arvrunner.add_intermediate_output(record["output_uuid"])
 522
 523             if container["output"]:
 524                 outputs = done.done_outputs(self, container, "/tmp", self.outdir, "/keep")
 525
 526             properties = record["properties"].copy()
 527             properties["cwl_output"] = outputs
 528             self.arvrunner.api.container_requests().update(
 529                 uuid=self.uuid,
 530                 body={"container_request": {"properties": properties}}
 531             ).execute(num_retries=self.arvrunner.num_retries)
 532
 533             if logc is not None and self.job_runtime.enable_usage_report is not False:
 534                 try:
 535                     summarizer = crunchstat_summary.summarizer.ContainerRequestSummarizer(
 536                         record,
 537                         collection_object=logc,
 538                         label=self.name,
 539                         arv=self.arvrunner.api)
 540                     summarizer.run()
 541                     with logc.open("usage_report.html", "wt") as mr:
 542                         mr.write(summarizer.html_report())
 543                     logc.save()
 544
 545                     # Post warnings about nodes that are under-utilized.
 546                     for rc in summarizer._recommend_gen(lambda x: x):
 547                         self.job_runtime.usage_report_notes.append(rc)
 548
 549                 except Exception as e:
 550                     logger.warning("%s unable to generate resource usage report",
 551                                  self.arvrunner.label(self),
 552                                  exc_info=(e if self.arvrunner.debug else False))
 553
 554         except WorkflowException as e:
 555             # Only include a stack trace if in debug mode.
 556             # A stack trace may obfuscate more useful output about the workflow.
 557             logger.error("%s unable to collect output from %s:\n%s",
 558                          self.arvrunner.label(self), container["output"], e, exc_info=(e if self.arvrunner.debug else False))
 559             processStatus = "permanentFail"
 560         except Exception:
 561             logger.exception("%s while getting output object:", self.arvrunner.label(self))
 562             processStatus = "permanentFail"
 563         finally:
 564             if not retried:
 565                 self.output_callback(outputs, processStatus)
 566
 567
 568 class RunnerContainer(Runner):
 569     """Submit and manage a container that runs arvados-cwl-runner."""
 570
 571     def arvados_job_spec(self, runtimeContext, git_info):
 572         """Create an Arvados container request for this workflow.
 573
 574         The returned dict can be used to create a container passed as
 575         the +body+ argument to container_requests().create().
 576         """
 577
 578         adjustDirObjs(self.job_order, trim_listing)
 579         visit_class(self.job_order, ("File", "Directory"), trim_anonymous_location)
 580         visit_class(self.job_order, ("File", "Directory"), remove_redundant_fields)
 581
 582         secret_mounts = {}
 583         for param in sorted(self.job_order.keys()):
 584             if self.secret_store.has_secret(self.job_order[param]):
 585                 mnt = "/secrets/s%d" % len(secret_mounts)
 586                 secret_mounts[mnt] = {
 587                     "kind": "text",
 588                     "content": self.secret_store.retrieve(self.job_order[param])
 589                 }
 590                 self.job_order[param] = {"$include": mnt}
 591
 592         container_image = arvados_jobs_image(self.arvrunner, self.jobs_image, runtimeContext)
 593
 594         workflow_runner_req, _ = self.embedded_tool.get_requirement("http://arvados.org/cwl#WorkflowRunnerResources")
 595         if workflow_runner_req and workflow_runner_req.get("acrContainerImage"):
 596             container_image = workflow_runner_req.get("acrContainerImage")
 597
 598         container_req = {
 599             "name": self.name,
 600             "output_path": "/var/spool/cwl",
 601             "cwd": "/var/spool/cwl",
 602             "priority": self.priority,
 603             "state": "Committed",
 604             "container_image": container_image,
 605             "mounts": {
 606                 "/var/lib/cwl/cwl.input.json": {
 607                     "kind": "json",
 608                     "content": self.job_order
 609                 },
 610                 "stdout": {
 611                     "kind": "file",
 612                     "path": "/var/spool/cwl/cwl.output.json"
 613                 },
 614                 "/var/spool/cwl": {
 615                     "kind": "collection",
 616                     "writable": True
 617                 }
 618             },
 619             "secret_mounts": secret_mounts,
 620             "runtime_constraints": {
 621                 "vcpus": math.ceil(self.submit_runner_cores),
 622                 "ram": 1024*1024 * (math.ceil(self.submit_runner_ram) + math.ceil(self.collection_cache_size)),
 623                 "API": True
 624             },
 625             "use_existing": self.reuse_runner,
 626             "properties": {}
 627         }
 628
 629         if self.embedded_tool.tool.get("id", "").startswith("keep:"):
 630             sp = self.embedded_tool.tool["id"].split('/')
 631             workflowcollection = sp[0][5:]
 632             workflowname = "/".join(sp[1:])
 633             workflowpath = "/var/lib/cwl/workflow/%s" % workflowname
 634             container_req["mounts"]["/var/lib/cwl/workflow"] = {
 635                 "kind": "collection",
 636                 "portable_data_hash": "%s" % workflowcollection
 637             }
 638         elif self.embedded_tool.tool.get("id", "").startswith("arvwf:"):
 639             uuid, frg = urllib.parse.urldefrag(self.embedded_tool.tool["id"])
 640             workflowpath = "/var/lib/cwl/workflow.json#" + frg
 641             packedtxt = self.loadingContext.loader.fetch_text(uuid)
 642             yaml = ruamel.yaml.YAML(typ='safe', pure=True)
 643             packed = yaml.load(packedtxt)
 644             container_req["mounts"]["/var/lib/cwl/workflow.json"] = {
 645                 "kind": "json",
 646                 "content": packed
 647             }
 648             container_req["properties"]["template_uuid"] = self.embedded_tool.tool["id"][6:33]
 649         elif self.embedded_tool.tool.get("id", "").startswith("file:"):
 650             raise WorkflowException("Tool id '%s' is a local file but expected keep: or arvwf:" % self.embedded_tool.tool.get("id"))
 651         else:
 652             main = self.loadingContext.loader.idx["_:main"]
 653             if main.get("id") == "_:main":
 654                 del main["id"]
 655             workflowpath = "/var/lib/cwl/workflow.json#main"
 656             container_req["mounts"]["/var/lib/cwl/workflow.json"] = {
 657                 "kind": "json",
 658                 "content": main
 659             }
 660
 661         container_req["properties"].update({k.replace("http://arvados.org/cwl#", "arv:"): v for k, v in git_info.items()})
 662
 663         properties_req, _ = self.embedded_tool.get_requirement("http://arvados.org/cwl#ProcessProperties")
 664         if properties_req:
 665             builder = make_builder(self.job_order, self.embedded_tool.hints, self.embedded_tool.requirements, runtimeContext, self.embedded_tool.metadata)
 666             for pr in properties_req["processProperties"]:
 667                 container_req["properties"][pr["propertyName"]] = builder.do_eval(pr["propertyValue"])
 668
 669         # --local means execute the workflow instead of submitting a container request
 670         # --api=containers means use the containers API
 671         # --no-log-timestamps means don't add timestamps (the logging infrastructure does this)
 672         # --disable-validate because we already validated so don't need to do it again
 673         # --eval-timeout is the timeout for javascript invocation
 674         # --parallel-task-count is the number of threads to use for job submission
 675         # --enable/disable-reuse sets desired job reuse
 676         # --collection-cache-size sets aside memory to store collections
 677         command = ["arvados-cwl-runner",
 678                    "--local",
 679                    "--api=containers",
 680                    "--no-log-timestamps",
 681                    "--disable-validate",
 682                    "--disable-color",
 683                    "--eval-timeout=%s" % self.arvrunner.eval_timeout,
 684                    "--thread-count=%s" % self.arvrunner.thread_count,
 685                    "--enable-reuse" if self.enable_reuse else "--disable-reuse",
 686                    "--collection-cache-size=%s" % self.collection_cache_size]
 687
 688         if self.output_name:
 689             command.append("--output-name=" + self.output_name)
 690             container_req["output_name"] = self.output_name
 691
 692         if self.output_tags:
 693             command.append("--output-tags=" + self.output_tags)
 694
 695         if runtimeContext.debug:
 696             command.append("--debug")
 697
 698         if runtimeContext.storage_classes != "default" and runtimeContext.storage_classes:
 699             command.append("--storage-classes=" + runtimeContext.storage_classes)
 700
 701         if runtimeContext.intermediate_storage_classes != "default" and runtimeContext.intermediate_storage_classes:
 702             command.append("--intermediate-storage-classes=" + runtimeContext.intermediate_storage_classes)
 703
 704         if runtimeContext.on_error:
 705             command.append("--on-error=" + self.on_error)
 706
 707         if runtimeContext.intermediate_output_ttl:
 708             command.append("--intermediate-output-ttl=%d" % runtimeContext.intermediate_output_ttl)
 709
 710         if runtimeContext.trash_intermediate:
 711             command.append("--trash-intermediate")
 712
 713         if runtimeContext.project_uuid:
 714             command.append("--project-uuid="+runtimeContext.project_uuid)
 715
 716         if self.enable_dev:
 717             command.append("--enable-dev")
 718
 719         if runtimeContext.enable_preemptible is True:
 720             command.append("--enable-preemptible")
 721
 722         if runtimeContext.enable_preemptible is False:
 723             command.append("--disable-preemptible")
 724
 725         if runtimeContext.varying_url_params:
 726             command.append("--varying-url-params="+runtimeContext.varying_url_params)
 727
 728         if runtimeContext.prefer_cached_downloads:
 729             command.append("--prefer-cached-downloads")
 730
 731         if runtimeContext.enable_usage_report is True:
 732             command.append("--enable-usage-report")
 733
 734         if runtimeContext.enable_usage_report is False:
 735             command.append("--disable-usage-report")
 736
 737         if self.fast_parser:
 738             command.append("--fast-parser")
 739
 740         command.extend([workflowpath, "/var/lib/cwl/cwl.input.json"])
 741
 742         container_req["command"] = command
 743
 744         return container_req
 745
 746
 747     def run(self, runtimeContext):
 748         runtimeContext.keepprefix = "keep:"
 749         job_spec = self.arvados_job_spec(runtimeContext, self.git_info)
 750         if runtimeContext.project_uuid:
 751             job_spec["owner_uuid"] = runtimeContext.project_uuid
 752
 753         extra_submit_params = {}
 754         if runtimeContext.submit_runner_cluster:
 755             extra_submit_params["cluster_id"] = runtimeContext.submit_runner_cluster
 756
 757         if runtimeContext.submit_request_uuid:
 758             if "cluster_id" in extra_submit_params:
 759                 # Doesn't make sense for "update" and actually fails
 760                 del extra_submit_params["cluster_id"]
 761             response = self.arvrunner.api.container_requests().update(
 762                 uuid=runtimeContext.submit_request_uuid,
 763                 body=job_spec,
 764                 **extra_submit_params
 765             ).execute(num_retries=self.arvrunner.num_retries)
 766         else:
 767             response = self.arvrunner.api.container_requests().create(
 768                 body=job_spec,
 769                 **extra_submit_params
 770             ).execute(num_retries=self.arvrunner.num_retries)
 771
 772         self.uuid = response["uuid"]
 773         self.arvrunner.process_submitted(self)
 774
 775         logger.info("%s submitted container_request %s", self.arvrunner.label(self), response["uuid"])
 776
 777         workbench2 = self.arvrunner.api.config()["Services"]["Workbench2"]["ExternalURL"]
 778         if workbench2:
 779             url = "{}processes/{}".format(workbench2, response["uuid"])
 780             logger.info("Monitor workflow progress at %s", url)
 781
 782
 783     def done(self, record):
 784         try:
 785             container = self.arvrunner.api.containers().get(
 786                 uuid=record["container_uuid"]
 787             ).execute(num_retries=self.arvrunner.num_retries)
 788             container["log"] = record["log_uuid"]
 789         except Exception:
 790             logger.exception("%s while getting runner container", self.arvrunner.label(self))
 791             self.arvrunner.output_callback({}, "permanentFail")
 792         else:
 793             super(RunnerContainer, self).done(container)