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
  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 .arvjob import RunnerJob, RunnerTemplate
  35 from .runner import Runner, upload_docker, upload_job_order, upload_workflow_deps
  36 from .arvtool import ArvadosCommandTool, validate_cluster_target, ArvadosExpressionTool
  37 from .arvworkflow import ArvadosWorkflow, upload_workflow
  38 from .fsaccess import CollectionFsAccess, CollectionFetcher, collectionResolver, CollectionCache, pdh_size
  39 from .perf import Perf
  40 from .pathmapper import NoFollowPathMapper
  41 from .task_queue import TaskQueue
  42 from .context import ArvLoadingContext, ArvRuntimeContext
  43 from ._version import __version__
  44
  45 from cwltool.process import shortname, UnsupportedRequirement, use_custom_schema
  46 from cwltool.pathmapper import adjustFileObjs, adjustDirObjs, get_listing, visit_class
  47 from cwltool.command_line_tool import compute_checksums
  48 from cwltool.load_tool import load_tool
  49
  50 logger = logging.getLogger('arvados.cwl-runner')
  51 metrics = logging.getLogger('arvados.cwl-runner.metrics')
  52
  53 DEFAULT_PRIORITY = 500
  54
  55 class RuntimeStatusLoggingHandler(logging.Handler):
  56     """
  57     Intercepts logging calls and report them as runtime statuses on runner
  58     containers.
  59     """
  60     def __init__(self, runtime_status_update_func):
  61         super(RuntimeStatusLoggingHandler, self).__init__()
  62         self.runtime_status_update = runtime_status_update_func
  63         self.updatingRuntimeStatus = False
  64
  65     def emit(self, record):
  66         kind = None
  67         if record.levelno >= logging.ERROR:
  68             kind = 'error'
  69         elif record.levelno >= logging.WARNING:
  70             kind = 'warning'
  71         if kind is not None and self.updatingRuntimeStatus is not True:
  72             self.updatingRuntimeStatus = True
  73             try:
  74                 log_msg = record.getMessage()
  75                 if '\n' in log_msg:
  76                     # If the logged message is multi-line, use its first line as status
  77                     # and the rest as detail.
  78                     status, detail = log_msg.split('\n', 1)
  79                     self.runtime_status_update(
  80                         kind,
  81                         "%s: %s" % (record.name, status),
  82                         detail
  83                     )
  84                 else:
  85                     self.runtime_status_update(
  86                         kind,
  87                         "%s: %s" % (record.name, record.getMessage())
  88                     )
  89             finally:
  90                 self.updatingRuntimeStatus = False
  91
  92
  93 class ArvCwlExecutor(object):
  94     """Execute a CWL tool or workflow, submit work (using either jobs or
  95     containers API), wait for them to complete, and report output.
  96
  97     """
  98
  99     def __init__(self, api_client,
 100                  arvargs=None,
 101                  keep_client=None,
 102                  num_retries=4,
 103                  thread_count=4):
 104
 105         if arvargs is None:
 106             arvargs = argparse.Namespace()
 107             arvargs.work_api = None
 108             arvargs.output_name = None
 109             arvargs.output_tags = None
 110             arvargs.thread_count = 1
 111             arvargs.collection_cache_size = None
 112
 113         self.api = api_client
 114         self.processes = {}
 115         self.workflow_eval_lock = threading.Condition(threading.RLock())
 116         self.final_output = None
 117         self.final_status = None
 118         self.num_retries = num_retries
 119         self.uuid = None
 120         self.stop_polling = threading.Event()
 121         self.poll_api = None
 122         self.pipeline = None
 123         self.final_output_collection = None
 124         self.output_name = arvargs.output_name
 125         self.output_tags = arvargs.output_tags
 126         self.project_uuid = None
 127         self.intermediate_output_ttl = 0
 128         self.intermediate_output_collections = []
 129         self.trash_intermediate = False
 130         self.thread_count = arvargs.thread_count
 131         self.poll_interval = 12
 132         self.loadingContext = None
 133         self.should_estimate_cache_size = True
 134
 135         if keep_client is not None:
 136             self.keep_client = keep_client
 137         else:
 138             self.keep_client = arvados.keep.KeepClient(api_client=self.api, num_retries=self.num_retries)
 139
 140         if arvargs.collection_cache_size:
 141             collection_cache_size = arvargs.collection_cache_size*1024*1024
 142             self.should_estimate_cache_size = False
 143         else:
 144             collection_cache_size = 256*1024*1024
 145
 146         self.collection_cache = CollectionCache(self.api, self.keep_client, self.num_retries,
 147                                                 cap=collection_cache_size)
 148
 149         self.fetcher_constructor = partial(CollectionFetcher,
 150                                            api_client=self.api,
 151                                            fs_access=CollectionFsAccess("", collection_cache=self.collection_cache),
 152                                            num_retries=self.num_retries)
 153
 154         self.work_api = None
 155         expected_api = ["jobs", "containers"]
 156         for api in expected_api:
 157             try:
 158                 methods = self.api._rootDesc.get('resources')[api]['methods']
 159                 if ('httpMethod' in methods['create'] and
 160                     (arvargs.work_api == api or arvargs.work_api is None)):
 161                     self.work_api = api
 162                     break
 163             except KeyError:
 164                 pass
 165
 166         if not self.work_api:
 167             if arvargs.work_api is None:
 168                 raise Exception("No supported APIs")
 169             else:
 170                 raise Exception("Unsupported API '%s', expected one of %s" % (arvargs.work_api, expected_api))
 171
 172         if self.work_api == "jobs":
 173             logger.warning("""
 174 *******************************
 175 Using the deprecated 'jobs' API.
 176
 177 To get rid of this warning:
 178
 179 Users: read about migrating at
 180 http://doc.arvados.org/user/cwl/cwl-style.html#migrate
 181 and use the option --api=containers
 182
 183 Admins: configure the cluster to disable the 'jobs' API as described at:
 184 http://doc.arvados.org/install/install-api-server.html#disable_api_methods
 185 *******************************""")
 186
 187         self.loadingContext = ArvLoadingContext(vars(arvargs))
 188         self.loadingContext.fetcher_constructor = self.fetcher_constructor
 189         self.loadingContext.resolver = partial(collectionResolver, self.api, num_retries=self.num_retries)
 190         self.loadingContext.construct_tool_object = self.arv_make_tool
 191
 192         # Add a custom logging handler to the root logger for runtime status reporting
 193         # if running inside a container
 194         if arvados_cwl.util.get_current_container(self.api, self.num_retries, logger):
 195             root_logger = logging.getLogger('')
 196
 197             # Remove existing RuntimeStatusLoggingHandlers if they exist
 198             handlers = [h for h in root_logger.handlers if not isinstance(h, RuntimeStatusLoggingHandler)]
 199             root_logger.handlers = handlers
 200
 201             handler = RuntimeStatusLoggingHandler(self.runtime_status_update)
 202             root_logger.addHandler(handler)
 203
 204         self.runtimeContext = ArvRuntimeContext(vars(arvargs))
 205         self.runtimeContext.make_fs_access = partial(CollectionFsAccess,
 206                                                      collection_cache=self.collection_cache)
 207
 208         validate_cluster_target(self, self.runtimeContext)
 209
 210
 211     def arv_make_tool(self, toolpath_object, loadingContext):
 212         if "class" in toolpath_object and toolpath_object["class"] == "CommandLineTool":
 213             return ArvadosCommandTool(self, toolpath_object, loadingContext)
 214         elif "class" in toolpath_object and toolpath_object["class"] == "Workflow":
 215             return ArvadosWorkflow(self, toolpath_object, loadingContext)
 216         elif "class" in toolpath_object and toolpath_object["class"] == "ExpressionTool":
 217             return ArvadosExpressionTool(self, toolpath_object, loadingContext)
 218         else:
 219             raise Exception("Unknown tool %s" % toolpath_object.get("class"))
 220
 221     def output_callback(self, out, processStatus):
 222         with self.workflow_eval_lock:
 223             if processStatus == "success":
 224                 logger.info("Overall process status is %s", processStatus)
 225                 state = "Complete"
 226             else:
 227                 logger.error("Overall process status is %s", processStatus)
 228                 state = "Failed"
 229             if self.pipeline:
 230                 self.api.pipeline_instances().update(uuid=self.pipeline["uuid"],
 231                                                         body={"state": state}).execute(num_retries=self.num_retries)
 232             self.final_status = processStatus
 233             self.final_output = out
 234             self.workflow_eval_lock.notifyAll()
 235
 236
 237     def start_run(self, runnable, runtimeContext):
 238         self.task_queue.add(partial(runnable.run, runtimeContext),
 239                             self.workflow_eval_lock, self.stop_polling)
 240
 241     def process_submitted(self, container):
 242         with self.workflow_eval_lock:
 243             self.processes[container.uuid] = container
 244
 245     def process_done(self, uuid, record):
 246         with self.workflow_eval_lock:
 247             j = self.processes[uuid]
 248             logger.info("%s %s is %s", self.label(j), uuid, record["state"])
 249             self.task_queue.add(partial(j.done, record),
 250                                 self.workflow_eval_lock, self.stop_polling)
 251             del self.processes[uuid]
 252
 253     def runtime_status_update(self, kind, message, detail=None):
 254         """
 255         Updates the runtime_status field on the runner container.
 256         Called when there's a need to report errors, warnings or just
 257         activity statuses, for example in the RuntimeStatusLoggingHandler.
 258         """
 259         with self.workflow_eval_lock:
 260             current = arvados_cwl.util.get_current_container(self.api, self.num_retries, logger)
 261             if current is None:
 262                 return
 263             runtime_status = current.get('runtime_status', {})
 264             # In case of status being an error, only report the first one.
 265             if kind == 'error':
 266                 if not runtime_status.get('error'):
 267                     runtime_status.update({
 268                         'error': message
 269                     })
 270                     if detail is not None:
 271                         runtime_status.update({
 272                             'errorDetail': detail
 273                         })
 274                 # Further errors are only mentioned as a count.
 275                 else:
 276                     # Get anything before an optional 'and N more' string.
 277                     try:
 278                         error_msg = re.match(
 279                             r'^(.*?)(?=\s*\(and \d+ more\)|$)', runtime_status.get('error')).groups()[0]
 280                         more_failures = re.match(
 281                             r'.*\(and (\d+) more\)', runtime_status.get('error'))
 282                     except TypeError:
 283                         # Ignore tests stubbing errors
 284                         return
 285                     if more_failures:
 286                         failure_qty = int(more_failures.groups()[0])
 287                         runtime_status.update({
 288                             'error': "%s (and %d more)" % (error_msg, failure_qty+1)
 289                         })
 290                     else:
 291                         runtime_status.update({
 292                             'error': "%s (and 1 more)" % error_msg
 293                         })
 294             elif kind in ['warning', 'activity']:
 295                 # Record the last warning/activity status without regard of
 296                 # previous occurences.
 297                 runtime_status.update({
 298                     kind: message
 299                 })
 300                 if detail is not None:
 301                     runtime_status.update({
 302                         kind+"Detail": detail
 303                     })
 304             else:
 305                 # Ignore any other status kind
 306                 return
 307             try:
 308                 self.api.containers().update(uuid=current['uuid'],
 309                                             body={
 310                                                 'runtime_status': runtime_status,
 311                                             }).execute(num_retries=self.num_retries)
 312             except Exception as e:
 313                 logger.info("Couldn't update runtime_status: %s", e)
 314
 315     def wrapped_callback(self, cb, obj, st):
 316         with self.workflow_eval_lock:
 317             cb(obj, st)
 318             self.workflow_eval_lock.notifyAll()
 319
 320     def get_wrapped_callback(self, cb):
 321         return partial(self.wrapped_callback, cb)
 322
 323     def on_message(self, event):
 324         if event.get("object_uuid") in self.processes and event["event_type"] == "update":
 325             uuid = event["object_uuid"]
 326             if event["properties"]["new_attributes"]["state"] == "Running":
 327                 with self.workflow_eval_lock:
 328                     j = self.processes[uuid]
 329                     if j.running is False:
 330                         j.running = True
 331                         j.update_pipeline_component(event["properties"]["new_attributes"])
 332                         logger.info("%s %s is Running", self.label(j), uuid)
 333             elif event["properties"]["new_attributes"]["state"] in ("Complete", "Failed", "Cancelled", "Final"):
 334                 self.process_done(uuid, event["properties"]["new_attributes"])
 335
 336     def label(self, obj):
 337         return "[%s %s]" % (self.work_api[0:-1], obj.name)
 338
 339     def poll_states(self):
 340         """Poll status of jobs or containers listed in the processes dict.
 341
 342         Runs in a separate thread.
 343         """
 344
 345         try:
 346             remain_wait = self.poll_interval
 347             while True:
 348                 if remain_wait > 0:
 349                     self.stop_polling.wait(remain_wait)
 350                 if self.stop_polling.is_set():
 351                     break
 352                 with self.workflow_eval_lock:
 353                     keys = list(self.processes)
 354                 if not keys:
 355                     remain_wait = self.poll_interval
 356                     continue
 357
 358                 begin_poll = time.time()
 359                 if self.work_api == "containers":
 360                     table = self.poll_api.container_requests()
 361                 elif self.work_api == "jobs":
 362                     table = self.poll_api.jobs()
 363
 364                 pageSize = self.poll_api._rootDesc.get('maxItemsPerResponse', 1000)
 365
 366                 while keys:
 367                     page = keys[:pageSize]
 368                     keys = keys[pageSize:]
 369                     try:
 370                         proc_states = table.list(filters=[["uuid", "in", page]]).execute(num_retries=self.num_retries)
 371                     except Exception:
 372                         logger.exception("Error checking states on API server: %s")
 373                         remain_wait = self.poll_interval
 374                         continue
 375
 376                     for p in proc_states["items"]:
 377                         self.on_message({
 378                             "object_uuid": p["uuid"],
 379                             "event_type": "update",
 380                             "properties": {
 381                                 "new_attributes": p
 382                             }
 383                         })
 384                 finish_poll = time.time()
 385                 remain_wait = self.poll_interval - (finish_poll - begin_poll)
 386         except:
 387             logger.exception("Fatal error in state polling thread.")
 388             with self.workflow_eval_lock:
 389                 self.processes.clear()
 390                 self.workflow_eval_lock.notifyAll()
 391         finally:
 392             self.stop_polling.set()
 393
 394     def add_intermediate_output(self, uuid):
 395         if uuid:
 396             self.intermediate_output_collections.append(uuid)
 397
 398     def trash_intermediate_output(self):
 399         logger.info("Cleaning up intermediate output collections")
 400         for i in self.intermediate_output_collections:
 401             try:
 402                 self.api.collections().delete(uuid=i).execute(num_retries=self.num_retries)
 403             except Exception:
 404                 logger.warning("Failed to delete intermediate output: %s", sys.exc_info()[1], exc_info=(sys.exc_info()[1] if self.debug else False))
 405             except (KeyboardInterrupt, SystemExit):
 406                 break
 407
 408     def check_features(self, obj):
 409         if isinstance(obj, dict):
 410             if obj.get("writable") and self.work_api != "containers":
 411                 raise SourceLine(obj, "writable", UnsupportedRequirement).makeError("InitialWorkDir feature 'writable: true' not supported with --api=jobs")
 412             if obj.get("class") == "DockerRequirement":
 413                 if obj.get("dockerOutputDirectory"):
 414                     if self.work_api != "containers":
 415                         raise SourceLine(obj, "dockerOutputDirectory", UnsupportedRequirement).makeError(
 416                             "Option 'dockerOutputDirectory' of DockerRequirement not supported with --api=jobs.")
 417                     if not obj.get("dockerOutputDirectory").startswith('/'):
 418                         raise SourceLine(obj, "dockerOutputDirectory", validate.ValidationException).makeError(
 419                             "Option 'dockerOutputDirectory' must be an absolute path.")
 420             if obj.get("class") == "http://commonwl.org/cwltool#Secrets" and self.work_api != "containers":
 421                 raise SourceLine(obj, "class", UnsupportedRequirement).makeError("Secrets not supported with --api=jobs")
 422             for v in viewvalues(obj):
 423                 self.check_features(v)
 424         elif isinstance(obj, list):
 425             for i,v in enumerate(obj):
 426                 with SourceLine(obj, i, UnsupportedRequirement, logger.isEnabledFor(logging.DEBUG)):
 427                     self.check_features(v)
 428
 429     def make_output_collection(self, name, storage_classes, tagsString, outputObj):
 430         outputObj = copy.deepcopy(outputObj)
 431
 432         files = []
 433         def capture(fileobj):
 434             files.append(fileobj)
 435
 436         adjustDirObjs(outputObj, capture)
 437         adjustFileObjs(outputObj, capture)
 438
 439         generatemapper = NoFollowPathMapper(files, "", "", separateDirs=False)
 440
 441         final = arvados.collection.Collection(api_client=self.api,
 442                                               keep_client=self.keep_client,
 443                                               num_retries=self.num_retries)
 444
 445         for k,v in generatemapper.items():
 446             if v.type == "Directory" and v.resolved.startswith("_:"):
 447                     continue
 448             if v.type == "CreateFile" and (k.startswith("_:") or v.resolved.startswith("_:")):
 449                 with final.open(v.target, "wb") as f:
 450                     f.write(v.resolved.encode("utf-8"))
 451                     continue
 452
 453             if not v.resolved.startswith("keep:"):
 454                 raise Exception("Output source is not in keep or a literal")
 455             sp = v.resolved.split("/")
 456             srccollection = sp[0][5:]
 457             try:
 458                 reader = self.collection_cache.get(srccollection)
 459                 srcpath = "/".join(sp[1:]) if len(sp) > 1 else "."
 460                 final.copy(srcpath, v.target, source_collection=reader, overwrite=False)
 461             except arvados.errors.ArgumentError as e:
 462                 logger.error("Creating CollectionReader for '%s' '%s': %s", k, v, e)
 463                 raise
 464             except IOError as e:
 465                 logger.error("While preparing output collection: %s", e)
 466                 raise
 467
 468         def rewrite(fileobj):
 469             fileobj["location"] = generatemapper.mapper(fileobj["location"]).target
 470             for k in ("listing", "contents", "nameext", "nameroot", "dirname"):
 471                 if k in fileobj:
 472                     del fileobj[k]
 473
 474         adjustDirObjs(outputObj, rewrite)
 475         adjustFileObjs(outputObj, rewrite)
 476
 477         with final.open("cwl.output.json", "w") as f:
 478             res = str(json.dumps(outputObj, sort_keys=True, indent=4, separators=(',',': '), ensure_ascii=False))
 479             f.write(res)
 480
 481         final.save_new(name=name, owner_uuid=self.project_uuid, storage_classes=storage_classes, ensure_unique_name=True)
 482
 483         logger.info("Final output collection %s \"%s\" (%s)", final.portable_data_hash(),
 484                     final.api_response()["name"],
 485                     final.manifest_locator())
 486
 487         final_uuid = final.manifest_locator()
 488         tags = tagsString.split(',')
 489         for tag in tags:
 490              self.api.links().create(body={
 491                 "head_uuid": final_uuid, "link_class": "tag", "name": tag
 492                 }).execute(num_retries=self.num_retries)
 493
 494         def finalcollection(fileobj):
 495             fileobj["location"] = "keep:%s/%s" % (final.portable_data_hash(), fileobj["location"])
 496
 497         adjustDirObjs(outputObj, finalcollection)
 498         adjustFileObjs(outputObj, finalcollection)
 499
 500         return (outputObj, final)
 501
 502     def set_crunch_output(self):
 503         if self.work_api == "containers":
 504             current = arvados_cwl.util.get_current_container(self.api, self.num_retries, logger)
 505             if current is None:
 506                 return
 507             try:
 508                 self.api.containers().update(uuid=current['uuid'],
 509                                              body={
 510                                                  'output': self.final_output_collection.portable_data_hash(),
 511                                              }).execute(num_retries=self.num_retries)
 512                 self.api.collections().update(uuid=self.final_output_collection.manifest_locator(),
 513                                               body={
 514                                                   'is_trashed': True
 515                                               }).execute(num_retries=self.num_retries)
 516             except Exception:
 517                 logger.exception("Setting container output")
 518                 return
 519         elif self.work_api == "jobs" and "TASK_UUID" in os.environ:
 520             self.api.job_tasks().update(uuid=os.environ["TASK_UUID"],
 521                                    body={
 522                                        'output': self.final_output_collection.portable_data_hash(),
 523                                        'success': self.final_status == "success",
 524                                        'progress':1.0
 525                                    }).execute(num_retries=self.num_retries)
 526
 527     def arv_executor(self, tool, job_order, runtimeContext, logger=None):
 528         self.debug = runtimeContext.debug
 529
 530         tool.visit(self.check_features)
 531
 532         self.project_uuid = runtimeContext.project_uuid
 533         self.pipeline = None
 534         self.fs_access = runtimeContext.make_fs_access(runtimeContext.basedir)
 535         self.secret_store = runtimeContext.secret_store
 536
 537         self.trash_intermediate = runtimeContext.trash_intermediate
 538         if self.trash_intermediate and self.work_api != "containers":
 539             raise Exception("--trash-intermediate is only supported with --api=containers.")
 540
 541         self.intermediate_output_ttl = runtimeContext.intermediate_output_ttl
 542         if self.intermediate_output_ttl and self.work_api != "containers":
 543             raise Exception("--intermediate-output-ttl is only supported with --api=containers.")
 544         if self.intermediate_output_ttl < 0:
 545             raise Exception("Invalid value %d for --intermediate-output-ttl, cannot be less than zero" % self.intermediate_output_ttl)
 546
 547         if runtimeContext.submit_request_uuid and self.work_api != "containers":
 548             raise Exception("--submit-request-uuid requires containers API, but using '{}' api".format(self.work_api))
 549
 550         if not runtimeContext.name:
 551             runtimeContext.name = self.name = tool.tool.get("label") or tool.metadata.get("label") or os.path.basename(tool.tool["id"])
 552
 553         submitting = (runtimeContext.update_workflow or
 554                       runtimeContext.create_workflow or
 555                       (runtimeContext.submit and not
 556                        (tool.tool["class"] == "CommandLineTool" and
 557                         runtimeContext.wait and
 558                         not runtimeContext.always_submit_runner)))
 559
 560         loadingContext = self.loadingContext.copy()
 561         loadingContext.do_validate = False
 562         loadingContext.do_update = False
 563         if submitting:
 564             # Document may have been auto-updated. Reload the original
 565             # document with updating disabled because we want to
 566             # submit the original document, not the auto-updated one.
 567             tool = load_tool(tool.tool["id"], loadingContext)
 568
 569         # Upload direct dependencies of workflow steps, get back mapping of files to keep references.
 570         # Also uploads docker images.
 571         merged_map = upload_workflow_deps(self, tool)
 572
 573         # Recreate process object (ArvadosWorkflow or
 574         # ArvadosCommandTool) because tool document may have been
 575         # updated by upload_workflow_deps in ways that modify
 576         # inheritance of hints or requirements.
 577         tool = load_tool(tool.tool, loadingContext)
 578
 579         # Upload local file references in the job order.
 580         job_order = upload_job_order(self, "%s input" % runtimeContext.name,
 581                                      tool, job_order)
 582
 583         existing_uuid = runtimeContext.update_workflow
 584         if existing_uuid or runtimeContext.create_workflow:
 585             # Create a pipeline template or workflow record and exit.
 586             if self.work_api == "jobs":
 587                 tmpl = RunnerTemplate(self, tool, job_order,
 588                                       runtimeContext.enable_reuse,
 589                                       uuid=existing_uuid,
 590                                       submit_runner_ram=runtimeContext.submit_runner_ram,
 591                                       name=runtimeContext.name,
 592                                       merged_map=merged_map,
 593                                       loadingContext=loadingContext)
 594                 tmpl.save()
 595                 # cwltool.main will write our return value to stdout.
 596                 return (tmpl.uuid, "success")
 597             elif self.work_api == "containers":
 598                 return (upload_workflow(self, tool, job_order,
 599                                         self.project_uuid,
 600                                         uuid=existing_uuid,
 601                                         submit_runner_ram=runtimeContext.submit_runner_ram,
 602                                         name=runtimeContext.name,
 603                                         merged_map=merged_map),
 604                         "success")
 605
 606         self.ignore_docker_for_reuse = runtimeContext.ignore_docker_for_reuse
 607         self.eval_timeout = runtimeContext.eval_timeout
 608
 609         runtimeContext = runtimeContext.copy()
 610         runtimeContext.use_container = True
 611         runtimeContext.tmpdir_prefix = "tmp"
 612         runtimeContext.work_api = self.work_api
 613
 614         if self.work_api == "containers":
 615             if self.ignore_docker_for_reuse:
 616                 raise Exception("--ignore-docker-for-reuse not supported with containers API.")
 617             runtimeContext.outdir = "/var/spool/cwl"
 618             runtimeContext.docker_outdir = "/var/spool/cwl"
 619             runtimeContext.tmpdir = "/tmp"
 620             runtimeContext.docker_tmpdir = "/tmp"
 621         elif self.work_api == "jobs":
 622             if runtimeContext.priority != DEFAULT_PRIORITY:
 623                 raise Exception("--priority not implemented for jobs API.")
 624             runtimeContext.outdir = "$(task.outdir)"
 625             runtimeContext.docker_outdir = "$(task.outdir)"
 626             runtimeContext.tmpdir = "$(task.tmpdir)"
 627
 628         if runtimeContext.priority < 1 or runtimeContext.priority > 1000:
 629             raise Exception("--priority must be in the range 1..1000.")
 630
 631         if self.should_estimate_cache_size:
 632             visited = set()
 633             estimated_size = [0]
 634             def estimate_collection_cache(obj):
 635                 if obj.get("location", "").startswith("keep:"):
 636                     m = pdh_size.match(obj["location"][5:])
 637                     if m and m.group(1) not in visited:
 638                         visited.add(m.group(1))
 639                         estimated_size[0] += int(m.group(2))
 640             visit_class(job_order, ("File", "Directory"), estimate_collection_cache)
 641             runtimeContext.collection_cache_size = max(((estimated_size[0]*192) // (1024*1024))+1, 256)
 642             self.collection_cache.set_cap(runtimeContext.collection_cache_size*1024*1024)
 643
 644         logger.info("Using collection cache size %s MiB", runtimeContext.collection_cache_size)
 645
 646         runnerjob = None
 647         if runtimeContext.submit:
 648             # Submit a runner job to run the workflow for us.
 649             if self.work_api == "containers":
 650                 loadingContext.loader = tool.doc_loader
 651                 loadingContext.avsc_names = tool.doc_schema
 652                 if tool.tool["class"] == "CommandLineTool" and runtimeContext.wait and (not runtimeContext.always_submit_runner):
 653                     runtimeContext.runnerjob = tool.tool["id"]
 654                 else:
 655                     tool = RunnerContainer(self, tool, loadingContext, runtimeContext.enable_reuse,
 656                                                 self.output_name,
 657                                                 self.output_tags,
 658                                                 submit_runner_ram=runtimeContext.submit_runner_ram,
 659                                                 name=runtimeContext.name,
 660                                                 on_error=runtimeContext.on_error,
 661                                                 submit_runner_image=runtimeContext.submit_runner_image,
 662                                                 intermediate_output_ttl=runtimeContext.intermediate_output_ttl,
 663                                                 merged_map=merged_map,
 664                                                 priority=runtimeContext.priority,
 665                                                 secret_store=self.secret_store,
 666                                                 collection_cache_size=runtimeContext.collection_cache_size,
 667                                                 collection_cache_is_default=self.should_estimate_cache_size)
 668             elif self.work_api == "jobs":
 669                 tool = RunnerJob(self, tool, loadingContext, runtimeContext.enable_reuse,
 670                                       self.output_name,
 671                                       self.output_tags,
 672                                       submit_runner_ram=runtimeContext.submit_runner_ram,
 673                                       name=runtimeContext.name,
 674                                       on_error=runtimeContext.on_error,
 675                                       submit_runner_image=runtimeContext.submit_runner_image,
 676                                       merged_map=merged_map)
 677         elif runtimeContext.cwl_runner_job is None and self.work_api == "jobs":
 678             # Create pipeline for local run
 679             self.pipeline = self.api.pipeline_instances().create(
 680                 body={
 681                     "owner_uuid": self.project_uuid,
 682                     "name": runtimeContext.name if runtimeContext.name else shortname(tool.tool["id"]),
 683                     "components": {},
 684                     "state": "RunningOnClient"}).execute(num_retries=self.num_retries)
 685             logger.info("Pipeline instance %s", self.pipeline["uuid"])
 686
 687         if runtimeContext.cwl_runner_job is not None:
 688             self.uuid = runtimeContext.cwl_runner_job.get('uuid')
 689
 690         jobiter = tool.job(job_order,
 691                            self.output_callback,
 692                            runtimeContext)
 693
 694         if runtimeContext.submit and not runtimeContext.wait:
 695             runnerjob = next(jobiter)
 696             runnerjob.run(runtimeContext)
 697             return (runnerjob.uuid, "success")
 698
 699         current_container = arvados_cwl.util.get_current_container(self.api, self.num_retries, logger)
 700         if current_container:
 701             logger.info("Running inside container %s", current_container.get("uuid"))
 702
 703         self.poll_api = arvados.api('v1', timeout=runtimeContext.http_timeout)
 704         self.polling_thread = threading.Thread(target=self.poll_states)
 705         self.polling_thread.start()
 706
 707         self.task_queue = TaskQueue(self.workflow_eval_lock, self.thread_count)
 708
 709         try:
 710             self.workflow_eval_lock.acquire()
 711
 712             # Holds the lock while this code runs and releases it when
 713             # it is safe to do so in self.workflow_eval_lock.wait(),
 714             # at which point on_message can update job state and
 715             # process output callbacks.
 716
 717             loopperf = Perf(metrics, "jobiter")
 718             loopperf.__enter__()
 719             for runnable in jobiter:
 720                 loopperf.__exit__()
 721
 722                 if self.stop_polling.is_set():
 723                     break
 724
 725                 if self.task_queue.error is not None:
 726                     raise self.task_queue.error
 727
 728                 if runnable:
 729                     with Perf(metrics, "run"):
 730                         self.start_run(runnable, runtimeContext)
 731                 else:
 732                     if (self.task_queue.in_flight + len(self.processes)) > 0:
 733                         self.workflow_eval_lock.wait(3)
 734                     else:
 735                         logger.error("Workflow is deadlocked, no runnable processes and not waiting on any pending processes.")
 736                         break
 737
 738                 if self.stop_polling.is_set():
 739                     break
 740
 741                 loopperf.__enter__()
 742             loopperf.__exit__()
 743
 744             while (self.task_queue.in_flight + len(self.processes)) > 0:
 745                 if self.task_queue.error is not None:
 746                     raise self.task_queue.error
 747                 self.workflow_eval_lock.wait(3)
 748
 749         except UnsupportedRequirement:
 750             raise
 751         except:
 752             if sys.exc_info()[0] is KeyboardInterrupt or sys.exc_info()[0] is SystemExit:
 753                 logger.error("Interrupted, workflow will be cancelled")
 754             elif isinstance(sys.exc_info()[1], WorkflowException):
 755                 logger.error("Workflow execution failed:\n%s", sys.exc_info()[1], exc_info=(sys.exc_info()[1] if self.debug else False))
 756             else:
 757                 logger.exception("Workflow execution failed")
 758
 759             if self.pipeline:
 760                 self.api.pipeline_instances().update(uuid=self.pipeline["uuid"],
 761                                                      body={"state": "Failed"}).execute(num_retries=self.num_retries)
 762             if runtimeContext.submit and isinstance(tool, Runner):
 763                 runnerjob = tool
 764                 if runnerjob.uuid and self.work_api == "containers":
 765                     self.api.container_requests().update(uuid=runnerjob.uuid,
 766                                                      body={"priority": "0"}).execute(num_retries=self.num_retries)
 767         finally:
 768             self.workflow_eval_lock.release()
 769             self.task_queue.drain()
 770             self.stop_polling.set()
 771             self.polling_thread.join()
 772             self.task_queue.join()
 773
 774         if self.final_status == "UnsupportedRequirement":
 775             raise UnsupportedRequirement("Check log for details.")
 776
 777         if self.final_output is None:
 778             raise WorkflowException("Workflow did not return a result.")
 779
 780         if runtimeContext.submit and isinstance(tool, Runner):
 781             logger.info("Final output collection %s", tool.final_output)
 782         else:
 783             if self.output_name is None:
 784                 self.output_name = "Output of %s" % (shortname(tool.tool["id"]))
 785             if self.output_tags is None:
 786                 self.output_tags = ""
 787
 788             storage_classes = runtimeContext.storage_classes.strip().split(",")
 789             self.final_output, self.final_output_collection = self.make_output_collection(self.output_name, storage_classes, self.output_tags, self.final_output)
 790             self.set_crunch_output()
 791
 792         if runtimeContext.compute_checksum:
 793             adjustDirObjs(self.final_output, partial(get_listing, self.fs_access))
 794             adjustFileObjs(self.final_output, partial(compute_checksums, self.fs_access))
 795
 796         if self.trash_intermediate and self.final_status == "success":
 797             self.trash_intermediate_output()
 798
 799         return (self.final_output, self.final_status)