sdk/cwl/arvados_cwl/executor.py

   1 # Copyright (C) The Arvados Authors. All rights reserved.
   2 #
   3 # SPDX-License-Identifier: Apache-2.0
   4
   5 from __future__ import division
   6 from builtins import next
   7 from builtins import object
   8 from builtins import str
   9 from future.utils import viewvalues, viewitems
  10
  11 import argparse
  12 import logging
  13 import os
  14 import sys
  15 import threading
  16 import copy
  17 import json
  18 import re
  19 from functools import partial
  20 import time
  21
  22 from cwltool.errors import WorkflowException
  23 import cwltool.workflow
  24 from schema_salad.sourceline import SourceLine
  25 import schema_salad.validate as validate
  26
  27 import arvados
  28 import arvados.config
  29 from arvados.keep import KeepClient
  30 from arvados.errors import ApiError
  31
  32 import arvados_cwl.util
  33 from .arvcontainer import RunnerContainer
  34 from .runner import Runner, upload_docker, upload_job_order, upload_workflow_deps
  35 from .arvtool import ArvadosCommandTool, validate_cluster_target, ArvadosExpressionTool
  36 from .arvworkflow import ArvadosWorkflow, upload_workflow
  37 from .fsaccess import CollectionFsAccess, CollectionFetcher, collectionResolver, CollectionCache, pdh_size
  38 from .perf import Perf
  39 from .pathmapper import NoFollowPathMapper
  40 from .task_queue import TaskQueue
  41 from .context import ArvLoadingContext, ArvRuntimeContext
  42 from ._version import __version__
  43
  44 from cwltool.process import shortname, UnsupportedRequirement, use_custom_schema
  45 from cwltool.utils import adjustFileObjs, adjustDirObjs, get_listing, visit_class
  46 from cwltool.command_line_tool import compute_checksums
  47 from cwltool.load_tool import load_tool
  48
  49 logger = logging.getLogger('arvados.cwl-runner')
  50 metrics = logging.getLogger('arvados.cwl-runner.metrics')
  51
  52 DEFAULT_PRIORITY = 500
  53
  54 class RuntimeStatusLoggingHandler(logging.Handler):
  55     """
  56     Intercepts logging calls and report them as runtime statuses on runner
  57     containers.
  58     """
  59     def __init__(self, runtime_status_update_func):
  60         super(RuntimeStatusLoggingHandler, self).__init__()
  61         self.runtime_status_update = runtime_status_update_func
  62         self.updatingRuntimeStatus = False
  63
  64     def emit(self, record):
  65         kind = None
  66         if record.levelno >= logging.ERROR:
  67             kind = 'error'
  68         elif record.levelno >= logging.WARNING:
  69             kind = 'warning'
  70         if kind is not None and self.updatingRuntimeStatus is not True:
  71             self.updatingRuntimeStatus = True
  72             try:
  73                 log_msg = record.getMessage()
  74                 if '\n' in log_msg:
  75                     # If the logged message is multi-line, use its first line as status
  76                     # and the rest as detail.
  77                     status, detail = log_msg.split('\n', 1)
  78                     self.runtime_status_update(
  79                         kind,
  80                         "%s: %s" % (record.name, status),
  81                         detail
  82                     )
  83                 else:
  84                     self.runtime_status_update(
  85                         kind,
  86                         "%s: %s" % (record.name, record.getMessage())
  87                     )
  88             finally:
  89                 self.updatingRuntimeStatus = False
  90
  91
  92 class ArvCwlExecutor(object):
  93     """Execute a CWL tool or workflow, submit work (using containers API),
  94     wait for them to complete, and report output.
  95
  96     """
  97
  98     def __init__(self, api_client,
  99                  arvargs=None,
 100                  keep_client=None,
 101                  num_retries=4,
 102                  thread_count=4):
 103
 104         if arvargs is None:
 105             arvargs = argparse.Namespace()
 106             arvargs.work_api = None
 107             arvargs.output_name = None
 108             arvargs.output_tags = None
 109             arvargs.thread_count = 1
 110             arvargs.collection_cache_size = None
 111
 112         self.api = api_client
 113         self.processes = {}
 114         self.workflow_eval_lock = threading.Condition(threading.RLock())
 115         self.final_output = None
 116         self.final_status = None
 117         self.num_retries = num_retries
 118         self.uuid = None
 119         self.stop_polling = threading.Event()
 120         self.poll_api = None
 121         self.pipeline = None
 122         self.final_output_collection = None
 123         self.output_name = arvargs.output_name
 124         self.output_tags = arvargs.output_tags
 125         self.project_uuid = None
 126         self.intermediate_output_ttl = 0
 127         self.intermediate_output_collections = []
 128         self.trash_intermediate = False
 129         self.thread_count = arvargs.thread_count
 130         self.poll_interval = 12
 131         self.loadingContext = None
 132         self.should_estimate_cache_size = True
 133         self.fs_access = None
 134         self.secret_store = None
 135
 136         if keep_client is not None:
 137             self.keep_client = keep_client
 138         else:
 139             self.keep_client = arvados.keep.KeepClient(api_client=self.api, num_retries=self.num_retries)
 140
 141         if arvargs.collection_cache_size:
 142             collection_cache_size = arvargs.collection_cache_size*1024*1024
 143             self.should_estimate_cache_size = False
 144         else:
 145             collection_cache_size = 256*1024*1024
 146
 147         self.collection_cache = CollectionCache(self.api, self.keep_client, self.num_retries,
 148                                                 cap=collection_cache_size)
 149
 150         self.fetcher_constructor = partial(CollectionFetcher,
 151                                            api_client=self.api,
 152                                            fs_access=CollectionFsAccess("", collection_cache=self.collection_cache),
 153                                            num_retries=self.num_retries)
 154
 155         self.work_api = None
 156         expected_api = ["containers"]
 157         for api in expected_api:
 158             try:
 159                 methods = self.api._rootDesc.get('resources')[api]['methods']
 160                 if ('httpMethod' in methods['create'] and
 161                     (arvargs.work_api == api or arvargs.work_api is None)):
 162                     self.work_api = api
 163                     break
 164             except KeyError:
 165                 pass
 166
 167         if not self.work_api:
 168             if arvargs.work_api is None:
 169                 raise Exception("No supported APIs")
 170             else:
 171                 raise Exception("Unsupported API '%s', expected one of %s" % (arvargs.work_api, expected_api))
 172
 173         if self.work_api == "jobs":
 174             logger.error("""
 175 *******************************
 176 The 'jobs' API is no longer supported.
 177 *******************************""")
 178             exit(1)
 179
 180         self.loadingContext = ArvLoadingContext(vars(arvargs))
 181         self.loadingContext.fetcher_constructor = self.fetcher_constructor
 182         self.loadingContext.resolver = partial(collectionResolver, self.api, num_retries=self.num_retries)
 183         self.loadingContext.construct_tool_object = self.arv_make_tool
 184
 185         # Add a custom logging handler to the root logger for runtime status reporting
 186         # if running inside a container
 187         if arvados_cwl.util.get_current_container(self.api, self.num_retries, logger):
 188             root_logger = logging.getLogger('')
 189
 190             # Remove existing RuntimeStatusLoggingHandlers if they exist
 191             handlers = [h for h in root_logger.handlers if not isinstance(h, RuntimeStatusLoggingHandler)]
 192             root_logger.handlers = handlers
 193
 194             handler = RuntimeStatusLoggingHandler(self.runtime_status_update)
 195             root_logger.addHandler(handler)
 196
 197         self.runtimeContext = ArvRuntimeContext(vars(arvargs))
 198         self.runtimeContext.make_fs_access = partial(CollectionFsAccess,
 199                                                      collection_cache=self.collection_cache)
 200
 201         validate_cluster_target(self, self.runtimeContext)
 202
 203
 204     def arv_make_tool(self, toolpath_object, loadingContext):
 205         if "class" in toolpath_object and toolpath_object["class"] == "CommandLineTool":
 206             return ArvadosCommandTool(self, toolpath_object, loadingContext)
 207         elif "class" in toolpath_object and toolpath_object["class"] == "Workflow":
 208             return ArvadosWorkflow(self, toolpath_object, loadingContext)
 209         elif "class" in toolpath_object and toolpath_object["class"] == "ExpressionTool":
 210             return ArvadosExpressionTool(self, toolpath_object, loadingContext)
 211         else:
 212             raise Exception("Unknown tool %s" % toolpath_object.get("class"))
 213
 214     def output_callback(self, out, processStatus):
 215         with self.workflow_eval_lock:
 216             if processStatus == "success":
 217                 logger.info("Overall process status is %s", processStatus)
 218                 state = "Complete"
 219             else:
 220                 logger.error("Overall process status is %s", processStatus)
 221                 state = "Failed"
 222             if self.pipeline:
 223                 self.api.pipeline_instances().update(uuid=self.pipeline["uuid"],
 224                                                         body={"state": state}).execute(num_retries=self.num_retries)
 225             self.final_status = processStatus
 226             self.final_output = out
 227             self.workflow_eval_lock.notifyAll()
 228
 229
 230     def start_run(self, runnable, runtimeContext):
 231         self.task_queue.add(partial(runnable.run, runtimeContext),
 232                             self.workflow_eval_lock, self.stop_polling)
 233
 234     def process_submitted(self, container):
 235         with self.workflow_eval_lock:
 236             self.processes[container.uuid] = container
 237
 238     def process_done(self, uuid, record):
 239         with self.workflow_eval_lock:
 240             j = self.processes[uuid]
 241             logger.info("%s %s is %s", self.label(j), uuid, record["state"])
 242             self.task_queue.add(partial(j.done, record),
 243                                 self.workflow_eval_lock, self.stop_polling)
 244             del self.processes[uuid]
 245
 246     def runtime_status_update(self, kind, message, detail=None):
 247         """
 248         Updates the runtime_status field on the runner container.
 249         Called when there's a need to report errors, warnings or just
 250         activity statuses, for example in the RuntimeStatusLoggingHandler.
 251         """
 252         with self.workflow_eval_lock:
 253             current = None
 254             try:
 255                 current = arvados_cwl.util.get_current_container(self.api, self.num_retries, logger)
 256             except Exception as e:
 257                 logger.info("Couldn't get current container: %s", e)
 258             if current is None:
 259                 return
 260             runtime_status = current.get('runtime_status', {})
 261             # In case of status being an error, only report the first one.
 262             if kind == 'error':
 263                 if not runtime_status.get('error'):
 264                     runtime_status.update({
 265                         'error': message
 266                     })
 267                     if detail is not None:
 268                         runtime_status.update({
 269                             'errorDetail': detail
 270                         })
 271                 # Further errors are only mentioned as a count.
 272                 else:
 273                     # Get anything before an optional 'and N more' string.
 274                     try:
 275                         error_msg = re.match(
 276                             r'^(.*?)(?=\s*\(and \d+ more\)|$)', runtime_status.get('error')).groups()[0]
 277                         more_failures = re.match(
 278                             r'.*\(and (\d+) more\)', runtime_status.get('error'))
 279                     except TypeError:
 280                         # Ignore tests stubbing errors
 281                         return
 282                     if more_failures:
 283                         failure_qty = int(more_failures.groups()[0])
 284                         runtime_status.update({
 285                             'error': "%s (and %d more)" % (error_msg, failure_qty+1)
 286                         })
 287                     else:
 288                         runtime_status.update({
 289                             'error': "%s (and 1 more)" % error_msg
 290                         })
 291             elif kind in ['warning', 'activity']:
 292                 # Record the last warning/activity status without regard of
 293                 # previous occurences.
 294                 runtime_status.update({
 295                     kind: message
 296                 })
 297                 if detail is not None:
 298                     runtime_status.update({
 299                         kind+"Detail": detail
 300                     })
 301             else:
 302                 # Ignore any other status kind
 303                 return
 304             try:
 305                 self.api.containers().update(uuid=current['uuid'],
 306                                             body={
 307                                                 'runtime_status': runtime_status,
 308                                             }).execute(num_retries=self.num_retries)
 309             except Exception as e:
 310                 logger.info("Couldn't update runtime_status: %s", e)
 311
 312     def wrapped_callback(self, cb, obj, st):
 313         with self.workflow_eval_lock:
 314             cb(obj, st)
 315             self.workflow_eval_lock.notifyAll()
 316
 317     def get_wrapped_callback(self, cb):
 318         return partial(self.wrapped_callback, cb)
 319
 320     def on_message(self, event):
 321         if event.get("object_uuid") in self.processes and event["event_type"] == "update":
 322             uuid = event["object_uuid"]
 323             if event["properties"]["new_attributes"]["state"] == "Running":
 324                 with self.workflow_eval_lock:
 325                     j = self.processes[uuid]
 326                     if j.running is False:
 327                         j.running = True
 328                         j.update_pipeline_component(event["properties"]["new_attributes"])
 329                         logger.info("%s %s is Running", self.label(j), uuid)
 330             elif event["properties"]["new_attributes"]["state"] in ("Complete", "Failed", "Cancelled", "Final"):
 331                 self.process_done(uuid, event["properties"]["new_attributes"])
 332
 333     def label(self, obj):
 334         return "[%s %s]" % (self.work_api[0:-1], obj.name)
 335
 336     def poll_states(self):
 337         """Poll status of containers listed in the processes dict.
 338
 339         Runs in a separate thread.
 340         """
 341
 342         try:
 343             remain_wait = self.poll_interval
 344             while True:
 345                 if remain_wait > 0:
 346                     self.stop_polling.wait(remain_wait)
 347                 if self.stop_polling.is_set():
 348                     break
 349                 with self.workflow_eval_lock:
 350                     keys = list(self.processes)
 351                 if not keys:
 352                     remain_wait = self.poll_interval
 353                     continue
 354
 355                 begin_poll = time.time()
 356                 if self.work_api == "containers":
 357                     table = self.poll_api.container_requests()
 358
 359                 pageSize = self.poll_api._rootDesc.get('maxItemsPerResponse', 1000)
 360
 361                 while keys:
 362                     page = keys[:pageSize]
 363                     try:
 364                         proc_states = table.list(filters=[["uuid", "in", page]]).execute(num_retries=self.num_retries)
 365                     except Exception:
 366                         logger.exception("Error checking states on API server: %s")
 367                         remain_wait = self.poll_interval
 368                         continue
 369
 370                     for p in proc_states["items"]:
 371                         self.on_message({
 372                             "object_uuid": p["uuid"],
 373                             "event_type": "update",
 374                             "properties": {
 375                                 "new_attributes": p
 376                             }
 377                         })
 378                     keys = keys[pageSize:]
 379
 380                 finish_poll = time.time()
 381                 remain_wait = self.poll_interval - (finish_poll - begin_poll)
 382         except:
 383             logger.exception("Fatal error in state polling thread.")
 384             with self.workflow_eval_lock:
 385                 self.processes.clear()
 386                 self.workflow_eval_lock.notifyAll()
 387         finally:
 388             self.stop_polling.set()
 389
 390     def add_intermediate_output(self, uuid):
 391         if uuid:
 392             self.intermediate_output_collections.append(uuid)
 393
 394     def trash_intermediate_output(self):
 395         logger.info("Cleaning up intermediate output collections")
 396         for i in self.intermediate_output_collections:
 397             try:
 398                 self.api.collections().delete(uuid=i).execute(num_retries=self.num_retries)
 399             except Exception:
 400                 logger.warning("Failed to delete intermediate output: %s", sys.exc_info()[1], exc_info=(sys.exc_info()[1] if self.debug else False))
 401             except (KeyboardInterrupt, SystemExit):
 402                 break
 403
 404     def check_features(self, obj, parentfield=""):
 405         if isinstance(obj, dict):
 406             if obj.get("class") == "DockerRequirement":
 407                 if obj.get("dockerOutputDirectory"):
 408                     if not obj.get("dockerOutputDirectory").startswith('/'):
 409                         raise SourceLine(obj, "dockerOutputDirectory", validate.ValidationException).makeError(
 410                             "Option 'dockerOutputDirectory' must be an absolute path.")
 411             if obj.get("class") == "InplaceUpdateRequirement":
 412                 if obj["inplaceUpdate"] and parentfield == "requirements":
 413                     raise SourceLine(obj, "class", UnsupportedRequirement).makeError("InplaceUpdateRequirement not supported for keep collections.")
 414             for k,v in viewitems(obj):
 415                 self.check_features(v, parentfield=k)
 416         elif isinstance(obj, list):
 417             for i,v in enumerate(obj):
 418                 with SourceLine(obj, i, UnsupportedRequirement, logger.isEnabledFor(logging.DEBUG)):
 419                     self.check_features(v, parentfield=parentfield)
 420
 421     def make_output_collection(self, name, storage_classes, tagsString, outputObj):
 422         outputObj = copy.deepcopy(outputObj)
 423
 424         files = []
 425         def capture(fileobj):
 426             files.append(fileobj)
 427
 428         adjustDirObjs(outputObj, capture)
 429         adjustFileObjs(outputObj, capture)
 430
 431         generatemapper = NoFollowPathMapper(files, "", "", separateDirs=False)
 432
 433         final = arvados.collection.Collection(api_client=self.api,
 434                                               keep_client=self.keep_client,
 435                                               num_retries=self.num_retries)
 436
 437         for k,v in generatemapper.items():
 438             if v.type == "Directory" and v.resolved.startswith("_:"):
 439                     continue
 440             if v.type == "CreateFile" and (k.startswith("_:") or v.resolved.startswith("_:")):
 441                 with final.open(v.target, "wb") as f:
 442                     f.write(v.resolved.encode("utf-8"))
 443                     continue
 444
 445             if not v.resolved.startswith("keep:"):
 446                 raise Exception("Output source is not in keep or a literal")
 447             sp = v.resolved.split("/")
 448             srccollection = sp[0][5:]
 449             try:
 450                 reader = self.collection_cache.get(srccollection)
 451                 srcpath = "/".join(sp[1:]) if len(sp) > 1 else "."
 452                 final.copy(srcpath, v.target, source_collection=reader, overwrite=False)
 453             except arvados.errors.ArgumentError as e:
 454                 logger.error("Creating CollectionReader for '%s' '%s': %s", k, v, e)
 455                 raise
 456             except IOError as e:
 457                 logger.error("While preparing output collection: %s", e)
 458                 raise
 459
 460         def rewrite(fileobj):
 461             fileobj["location"] = generatemapper.mapper(fileobj["location"]).target
 462             for k in ("listing", "contents", "nameext", "nameroot", "dirname"):
 463                 if k in fileobj:
 464                     del fileobj[k]
 465
 466         adjustDirObjs(outputObj, rewrite)
 467         adjustFileObjs(outputObj, rewrite)
 468
 469         with final.open("cwl.output.json", "w") as f:
 470             res = str(json.dumps(outputObj, sort_keys=True, indent=4, separators=(',',': '), ensure_ascii=False))
 471             f.write(res)
 472
 473         final.save_new(name=name, owner_uuid=self.project_uuid, storage_classes=storage_classes, ensure_unique_name=True)
 474
 475         logger.info("Final output collection %s \"%s\" (%s)", final.portable_data_hash(),
 476                     final.api_response()["name"],
 477                     final.manifest_locator())
 478
 479         final_uuid = final.manifest_locator()
 480         tags = tagsString.split(',')
 481         for tag in tags:
 482              self.api.links().create(body={
 483                 "head_uuid": final_uuid, "link_class": "tag", "name": tag
 484                 }).execute(num_retries=self.num_retries)
 485
 486         def finalcollection(fileobj):
 487             fileobj["location"] = "keep:%s/%s" % (final.portable_data_hash(), fileobj["location"])
 488
 489         adjustDirObjs(outputObj, finalcollection)
 490         adjustFileObjs(outputObj, finalcollection)
 491
 492         return (outputObj, final)
 493
 494     def set_crunch_output(self):
 495         if self.work_api == "containers":
 496             current = arvados_cwl.util.get_current_container(self.api, self.num_retries, logger)
 497             if current is None:
 498                 return
 499             try:
 500                 self.api.containers().update(uuid=current['uuid'],
 501                                              body={
 502                                                  'output': self.final_output_collection.portable_data_hash(),
 503                                              }).execute(num_retries=self.num_retries)
 504                 self.api.collections().update(uuid=self.final_output_collection.manifest_locator(),
 505                                               body={
 506                                                   'is_trashed': True
 507                                               }).execute(num_retries=self.num_retries)
 508             except Exception:
 509                 logger.exception("Setting container output")
 510                 raise
 511
 512     def apply_reqs(self, job_order_object, tool):
 513         if "https://w3id.org/cwl/cwl#requirements" in job_order_object:
 514             if tool.metadata.get("http://commonwl.org/cwltool#original_cwlVersion") == 'v1.0':
 515                 raise WorkflowException(
 516                     "`cwl:requirements` in the input object is not part of CWL "
 517                     "v1.0. You can adjust to use `cwltool:overrides` instead; or you "
 518                     "can set the cwlVersion to v1.1 or greater and re-run with "
 519                     "--enable-dev.")
 520             job_reqs = job_order_object["https://w3id.org/cwl/cwl#requirements"]
 521             for req in job_reqs:
 522                 tool.requirements.append(req)
 523
 524     def arv_executor(self, updated_tool, job_order, runtimeContext, logger=None):
 525         self.debug = runtimeContext.debug
 526
 527         logger.info("Using cluster %s (%s)", self.api.config()["ClusterID"], self.api.config()["Services"]["Controller"]["ExternalURL"])
 528
 529         updated_tool.visit(self.check_features)
 530
 531         self.project_uuid = runtimeContext.project_uuid
 532         self.pipeline = None
 533         self.fs_access = runtimeContext.make_fs_access(runtimeContext.basedir)
 534         self.secret_store = runtimeContext.secret_store
 535
 536         self.trash_intermediate = runtimeContext.trash_intermediate
 537         if self.trash_intermediate and self.work_api != "containers":
 538             raise Exception("--trash-intermediate is only supported with --api=containers.")
 539
 540         self.intermediate_output_ttl = runtimeContext.intermediate_output_ttl
 541         if self.intermediate_output_ttl and self.work_api != "containers":
 542             raise Exception("--intermediate-output-ttl is only supported with --api=containers.")
 543         if self.intermediate_output_ttl < 0:
 544             raise Exception("Invalid value %d for --intermediate-output-ttl, cannot be less than zero" % self.intermediate_output_ttl)
 545
 546         if runtimeContext.submit_request_uuid and self.work_api != "containers":
 547             raise Exception("--submit-request-uuid requires containers API, but using '{}' api".format(self.work_api))
 548
 549         if not runtimeContext.name:
 550             runtimeContext.name = self.name = updated_tool.tool.get("label") or updated_tool.metadata.get("label") or os.path.basename(updated_tool.tool["id"])
 551
 552         # Upload local file references in the job order.
 553         job_order = upload_job_order(self, "%s input" % runtimeContext.name,
 554                                      updated_tool, job_order)
 555
 556         # the last clause means: if it is a command line tool, and we
 557         # are going to wait for the result, and always_submit_runner
 558         # is false, then we don't submit a runner process.
 559
 560         submitting = (runtimeContext.update_workflow or
 561                       runtimeContext.create_workflow or
 562                       (runtimeContext.submit and not
 563                        (updated_tool.tool["class"] == "CommandLineTool" and
 564                         runtimeContext.wait and
 565                         not runtimeContext.always_submit_runner)))
 566
 567         loadingContext = self.loadingContext.copy()
 568         loadingContext.do_validate = False
 569         loadingContext.do_update = False
 570         if submitting:
 571             # Document may have been auto-updated. Reload the original
 572             # document with updating disabled because we want to
 573             # submit the document with its original CWL version, not
 574             # the auto-updated one.
 575             tool = load_tool(updated_tool.tool["id"], loadingContext)
 576         else:
 577             tool = updated_tool
 578
 579         # Upload direct dependencies of workflow steps, get back mapping of files to keep references.
 580         # Also uploads docker images.
 581         merged_map = upload_workflow_deps(self, tool)
 582
 583         # Recreate process object (ArvadosWorkflow or
 584         # ArvadosCommandTool) because tool document may have been
 585         # updated by upload_workflow_deps in ways that modify
 586         # inheritance of hints or requirements.
 587         loadingContext.loader = tool.doc_loader
 588         loadingContext.avsc_names = tool.doc_schema
 589         loadingContext.metadata = tool.metadata
 590         tool = load_tool(tool.tool, loadingContext)
 591
 592         existing_uuid = runtimeContext.update_workflow
 593         if existing_uuid or runtimeContext.create_workflow:
 594             # Create a pipeline template or workflow record and exit.
 595             if self.work_api == "containers":
 596                 return (upload_workflow(self, tool, job_order,
 597                                         self.project_uuid,
 598                                         uuid=existing_uuid,
 599                                         submit_runner_ram=runtimeContext.submit_runner_ram,
 600                                         name=runtimeContext.name,
 601                                         merged_map=merged_map,
 602                                         submit_runner_image=runtimeContext.submit_runner_image),
 603                         "success")
 604
 605         self.apply_reqs(job_order, tool)
 606
 607         self.ignore_docker_for_reuse = runtimeContext.ignore_docker_for_reuse
 608         self.eval_timeout = runtimeContext.eval_timeout
 609
 610         runtimeContext = runtimeContext.copy()
 611         runtimeContext.use_container = True
 612         runtimeContext.tmpdir_prefix = "tmp"
 613         runtimeContext.work_api = self.work_api
 614
 615         if self.work_api == "containers":
 616             if self.ignore_docker_for_reuse:
 617                 raise Exception("--ignore-docker-for-reuse not supported with containers API.")
 618             runtimeContext.outdir = "/var/spool/cwl"
 619             runtimeContext.docker_outdir = "/var/spool/cwl"
 620             runtimeContext.tmpdir = "/tmp"
 621             runtimeContext.docker_tmpdir = "/tmp"
 622
 623         if runtimeContext.priority < 1 or runtimeContext.priority > 1000:
 624             raise Exception("--priority must be in the range 1..1000.")
 625
 626         if self.should_estimate_cache_size:
 627             visited = set()
 628             estimated_size = [0]
 629             def estimate_collection_cache(obj):
 630                 if obj.get("location", "").startswith("keep:"):
 631                     m = pdh_size.match(obj["location"][5:])
 632                     if m and m.group(1) not in visited:
 633                         visited.add(m.group(1))
 634                         estimated_size[0] += int(m.group(2))
 635             visit_class(job_order, ("File", "Directory"), estimate_collection_cache)
 636             runtimeContext.collection_cache_size = max(((estimated_size[0]*192) // (1024*1024))+1, 256)
 637             self.collection_cache.set_cap(runtimeContext.collection_cache_size*1024*1024)
 638
 639         logger.info("Using collection cache size %s MiB", runtimeContext.collection_cache_size)
 640
 641         runnerjob = None
 642         if runtimeContext.submit:
 643             # Submit a runner job to run the workflow for us.
 644             if self.work_api == "containers":
 645                 if submitting:
 646                     tool = RunnerContainer(self, updated_tool,
 647                                            tool, loadingContext, runtimeContext.enable_reuse,
 648                                            self.output_name,
 649                                            self.output_tags,
 650                                            submit_runner_ram=runtimeContext.submit_runner_ram,
 651                                            name=runtimeContext.name,
 652                                            on_error=runtimeContext.on_error,
 653                                            submit_runner_image=runtimeContext.submit_runner_image,
 654                                            intermediate_output_ttl=runtimeContext.intermediate_output_ttl,
 655                                            merged_map=merged_map,
 656                                            priority=runtimeContext.priority,
 657                                            secret_store=self.secret_store,
 658                                            collection_cache_size=runtimeContext.collection_cache_size,
 659                                            collection_cache_is_default=self.should_estimate_cache_size)
 660                 else:
 661                     runtimeContext.runnerjob = tool.tool["id"]
 662
 663         if runtimeContext.cwl_runner_job is not None:
 664             self.uuid = runtimeContext.cwl_runner_job.get('uuid')
 665
 666         jobiter = tool.job(job_order,
 667                            self.output_callback,
 668                            runtimeContext)
 669
 670         if runtimeContext.submit and not runtimeContext.wait:
 671             runnerjob = next(jobiter)
 672             runnerjob.run(runtimeContext)
 673             return (runnerjob.uuid, "success")
 674
 675         current_container = arvados_cwl.util.get_current_container(self.api, self.num_retries, logger)
 676         if current_container:
 677             logger.info("Running inside container %s", current_container.get("uuid"))
 678
 679         self.poll_api = arvados.api('v1', timeout=runtimeContext.http_timeout)
 680         self.polling_thread = threading.Thread(target=self.poll_states)
 681         self.polling_thread.start()
 682
 683         self.task_queue = TaskQueue(self.workflow_eval_lock, self.thread_count)
 684
 685         try:
 686             self.workflow_eval_lock.acquire()
 687
 688             # Holds the lock while this code runs and releases it when
 689             # it is safe to do so in self.workflow_eval_lock.wait(),
 690             # at which point on_message can update job state and
 691             # process output callbacks.
 692
 693             loopperf = Perf(metrics, "jobiter")
 694             loopperf.__enter__()
 695             for runnable in jobiter:
 696                 loopperf.__exit__()
 697
 698                 if self.stop_polling.is_set():
 699                     break
 700
 701                 if self.task_queue.error is not None:
 702                     raise self.task_queue.error
 703
 704                 if runnable:
 705                     with Perf(metrics, "run"):
 706                         self.start_run(runnable, runtimeContext)
 707                 else:
 708                     if (self.task_queue.in_flight + len(self.processes)) > 0:
 709                         self.workflow_eval_lock.wait(3)
 710                     else:
 711                         logger.error("Workflow is deadlocked, no runnable processes and not waiting on any pending processes.")
 712                         break
 713
 714                 if self.stop_polling.is_set():
 715                     break
 716
 717                 loopperf.__enter__()
 718             loopperf.__exit__()
 719
 720             while (self.task_queue.in_flight + len(self.processes)) > 0:
 721                 if self.task_queue.error is not None:
 722                     raise self.task_queue.error
 723                 self.workflow_eval_lock.wait(3)
 724
 725         except UnsupportedRequirement:
 726             raise
 727         except:
 728             if sys.exc_info()[0] is KeyboardInterrupt or sys.exc_info()[0] is SystemExit:
 729                 logger.error("Interrupted, workflow will be cancelled")
 730             elif isinstance(sys.exc_info()[1], WorkflowException):
 731                 logger.error("Workflow execution failed:\n%s", sys.exc_info()[1], exc_info=(sys.exc_info()[1] if self.debug else False))
 732             else:
 733                 logger.exception("Workflow execution failed")
 734
 735             if self.pipeline:
 736                 self.api.pipeline_instances().update(uuid=self.pipeline["uuid"],
 737                                                      body={"state": "Failed"}).execute(num_retries=self.num_retries)
 738
 739             if self.work_api == "containers" and not current_container:
 740                 # Not running in a crunch container, so cancel any outstanding processes.
 741                 for p in self.processes:
 742                     try:
 743                         self.api.container_requests().update(uuid=p,
 744                                                              body={"priority": "0"}
 745                         ).execute(num_retries=self.num_retries)
 746                     except Exception:
 747                         pass
 748         finally:
 749             self.workflow_eval_lock.release()
 750             self.task_queue.drain()
 751             self.stop_polling.set()
 752             self.polling_thread.join()
 753             self.task_queue.join()
 754
 755         if self.final_status == "UnsupportedRequirement":
 756             raise UnsupportedRequirement("Check log for details.")
 757
 758         if self.final_output is None:
 759             raise WorkflowException("Workflow did not return a result.")
 760
 761         if runtimeContext.submit and isinstance(tool, Runner):
 762             logger.info("Final output collection %s", tool.final_output)
 763         else:
 764             if self.output_name is None:
 765                 self.output_name = "Output of %s" % (shortname(tool.tool["id"]))
 766             if self.output_tags is None:
 767                 self.output_tags = ""
 768
 769             storage_classes = runtimeContext.storage_classes.strip().split(",")
 770             self.final_output, self.final_output_collection = self.make_output_collection(self.output_name, storage_classes, self.output_tags, self.final_output)
 771             self.set_crunch_output()
 772
 773         if runtimeContext.compute_checksum:
 774             adjustDirObjs(self.final_output, partial(get_listing, self.fs_access))
 775             adjustFileObjs(self.final_output, partial(compute_checksums, self.fs_access))
 776
 777         if self.trash_intermediate and self.final_status == "success":
 778             self.trash_intermediate_output()
 779
 780         return (self.final_output, self.final_status)