sdk/cwl/arvados_cwl/__init__.py

   1 #!/usr/bin/env python
   2
   3 # Implement cwl-runner interface for submitting and running work on Arvados, using
   4 # either the Crunch jobs API or Crunch containers API.
   5
   6 import argparse
   7 import logging
   8 import os
   9 import sys
  10 import threading
  11 import hashlib
  12 import copy
  13 import json
  14 from functools import partial
  15 import pkg_resources  # part of setuptools
  16
  17 from cwltool.errors import WorkflowException
  18 import cwltool.main
  19 import cwltool.workflow
  20 import cwltool.process
  21 import schema_salad
  22 from schema_salad.sourceline import SourceLine
  23
  24 import arvados
  25 import arvados.config
  26 from arvados.keep import KeepClient
  27 from arvados.errors import ApiError
  28
  29 from .arvcontainer import ArvadosContainer, RunnerContainer
  30 from .arvjob import ArvadosJob, RunnerJob, RunnerTemplate
  31 from. runner import Runner, upload_instance
  32 from .arvtool import ArvadosCommandTool
  33 from .arvworkflow import ArvadosWorkflow, upload_workflow
  34 from .fsaccess import CollectionFsAccess, CollectionFetcher, collectionResolver
  35 from .perf import Perf
  36 from .pathmapper import FinalOutputPathMapper
  37 from ._version import __version__
  38
  39 from cwltool.pack import pack
  40 from cwltool.process import shortname, UnsupportedRequirement, getListing
  41 from cwltool.pathmapper import adjustFileObjs, adjustDirObjs
  42 from cwltool.draft2tool import compute_checksums
  43 from arvados.api import OrderedJsonModel
  44
  45 logger = logging.getLogger('arvados.cwl-runner')
  46 metrics = logging.getLogger('arvados.cwl-runner.metrics')
  47 logger.setLevel(logging.INFO)
  48
  49 arvados.log_handler.setFormatter(logging.Formatter(
  50         '%(asctime)s %(name)s %(levelname)s: %(message)s',
  51         '%Y-%m-%d %H:%M:%S'))
  52
  53 class ArvCwlRunner(object):
  54     """Execute a CWL tool or workflow, submit work (using either jobs or
  55     containers API), wait for them to complete, and report output.
  56
  57     """
  58
  59     def __init__(self, api_client, work_api=None, keep_client=None, output_name=None, output_tags=None, num_retries=4):
  60         self.api = api_client
  61         self.processes = {}
  62         self.lock = threading.Lock()
  63         self.cond = threading.Condition(self.lock)
  64         self.final_output = None
  65         self.final_status = None
  66         self.uploaded = {}
  67         self.num_retries = num_retries
  68         self.uuid = None
  69         self.stop_polling = threading.Event()
  70         self.poll_api = None
  71         self.pipeline = None
  72         self.final_output_collection = None
  73         self.output_name = output_name
  74         self.output_tags = output_tags
  75         self.project_uuid = None
  76
  77         if keep_client is not None:
  78             self.keep_client = keep_client
  79         else:
  80             self.keep_client = arvados.keep.KeepClient(api_client=self.api, num_retries=self.num_retries)
  81
  82         self.work_api = None
  83         expected_api = ["jobs", "containers"]
  84         for api in expected_api:
  85             try:
  86                 methods = self.api._rootDesc.get('resources')[api]['methods']
  87                 if ('httpMethod' in methods['create'] and
  88                     (work_api == api or work_api is None)):
  89                     self.work_api = api
  90                     break
  91             except KeyError:
  92                 pass
  93
  94         if not self.work_api:
  95             if work_api is None:
  96                 raise Exception("No supported APIs")
  97             else:
  98                 raise Exception("Unsupported API '%s', expected one of %s" % (work_api, expected_api))
  99
 100     def arv_make_tool(self, toolpath_object, **kwargs):
 101         kwargs["work_api"] = self.work_api
 102         kwargs["fetcher_constructor"] = partial(CollectionFetcher,
 103                                                 api_client=self.api,
 104                                                 keep_client=self.keep_client)
 105         if "class" in toolpath_object and toolpath_object["class"] == "CommandLineTool":
 106             return ArvadosCommandTool(self, toolpath_object, **kwargs)
 107         elif "class" in toolpath_object and toolpath_object["class"] == "Workflow":
 108             return ArvadosWorkflow(self, toolpath_object, **kwargs)
 109         else:
 110             return cwltool.workflow.defaultMakeTool(toolpath_object, **kwargs)
 111
 112     def output_callback(self, out, processStatus):
 113         if processStatus == "success":
 114             logger.info("Overall process status is %s", processStatus)
 115             if self.pipeline:
 116                 self.api.pipeline_instances().update(uuid=self.pipeline["uuid"],
 117                                                      body={"state": "Complete"}).execute(num_retries=self.num_retries)
 118         else:
 119             logger.warn("Overall process status is %s", processStatus)
 120             if self.pipeline:
 121                 self.api.pipeline_instances().update(uuid=self.pipeline["uuid"],
 122                                                      body={"state": "Failed"}).execute(num_retries=self.num_retries)
 123         self.final_status = processStatus
 124         self.final_output = out
 125
 126     def on_message(self, event):
 127         if "object_uuid" in event:
 128             if event["object_uuid"] in self.processes and event["event_type"] == "update":
 129                 if event["properties"]["new_attributes"]["state"] == "Running" and self.processes[event["object_uuid"]].running is False:
 130                     uuid = event["object_uuid"]
 131                     with self.lock:
 132                         j = self.processes[uuid]
 133                         logger.info("%s %s is Running", self.label(j), uuid)
 134                         j.running = True
 135                         j.update_pipeline_component(event["properties"]["new_attributes"])
 136                 elif event["properties"]["new_attributes"]["state"] in ("Complete", "Failed", "Cancelled", "Final"):
 137                     uuid = event["object_uuid"]
 138                     try:
 139                         self.cond.acquire()
 140                         j = self.processes[uuid]
 141                         logger.info("%s %s is %s", self.label(j), uuid, event["properties"]["new_attributes"]["state"])
 142                         with Perf(metrics, "done %s" % j.name):
 143                             j.done(event["properties"]["new_attributes"])
 144                         self.cond.notify()
 145                     finally:
 146                         self.cond.release()
 147
 148     def label(self, obj):
 149         return "[%s %s]" % (self.work_api[0:-1], obj.name)
 150
 151     def poll_states(self):
 152         """Poll status of jobs or containers listed in the processes dict.
 153
 154         Runs in a separate thread.
 155         """
 156
 157         try:
 158             while True:
 159                 self.stop_polling.wait(15)
 160                 if self.stop_polling.is_set():
 161                     break
 162                 with self.lock:
 163                     keys = self.processes.keys()
 164                 if not keys:
 165                     continue
 166
 167                 if self.work_api == "containers":
 168                     table = self.poll_api.container_requests()
 169                 elif self.work_api == "jobs":
 170                     table = self.poll_api.jobs()
 171
 172                 try:
 173                     proc_states = table.list(filters=[["uuid", "in", keys]]).execute(num_retries=self.num_retries)
 174                 except Exception as e:
 175                     logger.warn("Error checking states on API server: %s", e)
 176                     continue
 177
 178                 for p in proc_states["items"]:
 179                     self.on_message({
 180                         "object_uuid": p["uuid"],
 181                         "event_type": "update",
 182                         "properties": {
 183                             "new_attributes": p
 184                         }
 185                     })
 186         except:
 187             logger.error("Fatal error in state polling thread.", exc_info=(sys.exc_info()[1] if self.debug else False))
 188             self.cond.acquire()
 189             self.processes.clear()
 190             self.cond.notify()
 191             self.cond.release()
 192         finally:
 193             self.stop_polling.set()
 194
 195     def get_uploaded(self):
 196         return self.uploaded.copy()
 197
 198     def add_uploaded(self, src, pair):
 199         self.uploaded[src] = pair
 200
 201     def check_features(self, obj):
 202         if isinstance(obj, dict):
 203             if obj.get("class") == "InitialWorkDirRequirement":
 204                 if self.work_api == "containers":
 205                     raise UnsupportedRequirement("InitialWorkDirRequirement not supported with --api=containers")
 206             if obj.get("writable"):
 207                 raise SourceLine(obj, "writable", UnsupportedRequirement).makeError("InitialWorkDir feature 'writable: true' not supported")
 208             if obj.get("class") == "CommandLineTool":
 209                 if self.work_api == "containers":
 210                     if obj.get("stdin"):
 211                         raise SourceLine(obj, "stdin", UnsupportedRequirement).makeError("Stdin redirection currently not suppported with --api=containers")
 212                     if obj.get("stderr"):
 213                         raise SourceLine(obj, "stderr", UnsupportedRequirement).makeError("Stderr redirection currently not suppported with --api=containers")
 214             for v in obj.itervalues():
 215                 self.check_features(v)
 216         elif isinstance(obj, list):
 217             for i,v in enumerate(obj):
 218                 with SourceLine(obj, i, UnsupportedRequirement):
 219                     self.check_features(v)
 220
 221     def make_output_collection(self, name, tagsString, outputObj):
 222         outputObj = copy.deepcopy(outputObj)
 223
 224         files = []
 225         def capture(fileobj):
 226             files.append(fileobj)
 227
 228         adjustDirObjs(outputObj, capture)
 229         adjustFileObjs(outputObj, capture)
 230
 231         generatemapper = FinalOutputPathMapper(files, "", "", separateDirs=False)
 232
 233         final = arvados.collection.Collection(api_client=self.api,
 234                                               keep_client=self.keep_client,
 235                                               num_retries=self.num_retries)
 236
 237         srccollections = {}
 238         for k,v in generatemapper.items():
 239             if k.startswith("_:"):
 240                 if v.type == "Directory":
 241                     continue
 242                 if v.type == "CreateFile":
 243                     with final.open(v.target, "wb") as f:
 244                         f.write(v.resolved.encode("utf-8"))
 245                     continue
 246
 247             if not k.startswith("keep:"):
 248                 raise Exception("Output source is not in keep or a literal")
 249             sp = k.split("/")
 250             srccollection = sp[0][5:]
 251             if srccollection not in srccollections:
 252                 try:
 253                     srccollections[srccollection] = arvados.collection.CollectionReader(
 254                         srccollection,
 255                         api_client=self.api,
 256                         keep_client=self.keep_client,
 257                         num_retries=self.num_retries)
 258                 except arvados.errors.ArgumentError as e:
 259                     logger.error("Creating CollectionReader for '%s' '%s': %s", k, v, e)
 260                     raise
 261             reader = srccollections[srccollection]
 262             try:
 263                 srcpath = "/".join(sp[1:]) if len(sp) > 1 else "."
 264                 final.copy(srcpath, v.target, source_collection=reader, overwrite=False)
 265             except IOError as e:
 266                 logger.warn("While preparing output collection: %s", e)
 267
 268         def rewrite(fileobj):
 269             fileobj["location"] = generatemapper.mapper(fileobj["location"]).target
 270             for k in ("basename", "listing", "contents"):
 271                 if k in fileobj:
 272                     del fileobj[k]
 273
 274         adjustDirObjs(outputObj, rewrite)
 275         adjustFileObjs(outputObj, rewrite)
 276
 277         with final.open("cwl.output.json", "w") as f:
 278             json.dump(outputObj, f, sort_keys=True, indent=4, separators=(',',': '))
 279
 280         final.save_new(name=name, owner_uuid=self.project_uuid, ensure_unique_name=True)
 281
 282         logger.info("Final output collection %s \"%s\" (%s)", final.portable_data_hash(),
 283                     final.api_response()["name"],
 284                     final.manifest_locator())
 285
 286         final_uuid = final.manifest_locator()
 287         tags = tagsString.split(',')
 288         for tag in tags:
 289              self.api.links().create(body={
 290                 "head_uuid": final_uuid, "link_class": "tag", "name": tag
 291                 }).execute(num_retries=self.num_retries)
 292
 293         def finalcollection(fileobj):
 294             fileobj["location"] = "keep:%s/%s" % (final.portable_data_hash(), fileobj["location"])
 295
 296         adjustDirObjs(outputObj, finalcollection)
 297         adjustFileObjs(outputObj, finalcollection)
 298
 299         return (outputObj, final)
 300
 301     def set_crunch_output(self):
 302         if self.work_api == "containers":
 303             try:
 304                 current = self.api.containers().current().execute(num_retries=self.num_retries)
 305             except ApiError as e:
 306                 # Status code 404 just means we're not running in a container.
 307                 if e.resp.status != 404:
 308                     logger.info("Getting current container: %s", e)
 309                 return
 310             try:
 311                 self.api.containers().update(uuid=current['uuid'],
 312                                              body={
 313                                                  'output': self.final_output_collection.portable_data_hash(),
 314                                              }).execute(num_retries=self.num_retries)
 315             except Exception as e:
 316                 logger.info("Setting container output: %s", e)
 317         elif self.work_api == "jobs" and "TASK_UUID" in os.environ:
 318             self.api.job_tasks().update(uuid=os.environ["TASK_UUID"],
 319                                    body={
 320                                        'output': self.final_output_collection.portable_data_hash(),
 321                                        'success': self.final_status == "success",
 322                                        'progress':1.0
 323                                    }).execute(num_retries=self.num_retries)
 324
 325     def arv_executor(self, tool, job_order, **kwargs):
 326         self.debug = kwargs.get("debug")
 327
 328         tool.visit(self.check_features)
 329
 330         self.project_uuid = kwargs.get("project_uuid")
 331         self.pipeline = None
 332         make_fs_access = kwargs.get("make_fs_access") or partial(CollectionFsAccess,
 333                                                                  api_client=self.api,
 334                                                                  keep_client=self.keep_client)
 335         self.fs_access = make_fs_access(kwargs["basedir"])
 336
 337         existing_uuid = kwargs.get("update_workflow")
 338         if existing_uuid or kwargs.get("create_workflow"):
 339             if self.work_api == "jobs":
 340                 tmpl = RunnerTemplate(self, tool, job_order,
 341                                       kwargs.get("enable_reuse"),
 342                                       uuid=existing_uuid,
 343                                       submit_runner_ram=kwargs.get("submit_runner_ram"),
 344                                       name=kwargs.get("name"))
 345                 tmpl.save()
 346                 # cwltool.main will write our return value to stdout.
 347                 return tmpl.uuid
 348             else:
 349                 return upload_workflow(self, tool, job_order,
 350                                        self.project_uuid,
 351                                        uuid=existing_uuid,
 352                                        submit_runner_ram=kwargs.get("submit_runner_ram"),
 353                                        name=kwargs.get("name"))
 354
 355         self.ignore_docker_for_reuse = kwargs.get("ignore_docker_for_reuse")
 356
 357         kwargs["make_fs_access"] = make_fs_access
 358         kwargs["enable_reuse"] = kwargs.get("enable_reuse")
 359         kwargs["use_container"] = True
 360         kwargs["tmpdir_prefix"] = "tmp"
 361         kwargs["compute_checksum"] = kwargs.get("compute_checksum")
 362
 363         if not kwargs["name"]:
 364             del kwargs["name"]
 365
 366         if self.work_api == "containers":
 367             kwargs["outdir"] = "/var/spool/cwl"
 368             kwargs["docker_outdir"] = "/var/spool/cwl"
 369             kwargs["tmpdir"] = "/tmp"
 370             kwargs["docker_tmpdir"] = "/tmp"
 371         elif self.work_api == "jobs":
 372             kwargs["outdir"] = "$(task.outdir)"
 373             kwargs["docker_outdir"] = "$(task.outdir)"
 374             kwargs["tmpdir"] = "$(task.tmpdir)"
 375
 376         upload_instance(self, shortname(tool.tool["id"]), tool, job_order)
 377
 378         runnerjob = None
 379         if kwargs.get("submit"):
 380             if self.work_api == "containers":
 381                 if tool.tool["class"] == "CommandLineTool":
 382                     kwargs["runnerjob"] = tool.tool["id"]
 383                     runnerjob = tool.job(job_order,
 384                                          self.output_callback,
 385                                          **kwargs).next()
 386                 else:
 387                     runnerjob = RunnerContainer(self, tool, job_order, kwargs.get("enable_reuse"), self.output_name,
 388                                                 self.output_tags, submit_runner_ram=kwargs.get("submit_runner_ram"),
 389                                                 name=kwargs.get("name"), on_error=kwargs.get("on_error"))
 390             else:
 391                 runnerjob = RunnerJob(self, tool, job_order, kwargs.get("enable_reuse"), self.output_name,
 392                                       self.output_tags, submit_runner_ram=kwargs.get("submit_runner_ram"),
 393                                       name=kwargs.get("name"), on_error=kwargs.get("on_error"))
 394
 395         if not kwargs.get("submit") and "cwl_runner_job" not in kwargs and not self.work_api == "containers":
 396             # Create pipeline for local run
 397             self.pipeline = self.api.pipeline_instances().create(
 398                 body={
 399                     "owner_uuid": self.project_uuid,
 400                     "name": kwargs["name"] if kwargs.get("name") else shortname(tool.tool["id"]),
 401                     "components": {},
 402                     "state": "RunningOnClient"}).execute(num_retries=self.num_retries)
 403             logger.info("Pipeline instance %s", self.pipeline["uuid"])
 404
 405         if runnerjob and not kwargs.get("wait"):
 406             runnerjob.run(wait=kwargs.get("wait"))
 407             return runnerjob.uuid
 408
 409         self.poll_api = arvados.api('v1')
 410         self.polling_thread = threading.Thread(target=self.poll_states)
 411         self.polling_thread.start()
 412
 413         if runnerjob:
 414             jobiter = iter((runnerjob,))
 415         else:
 416             if "cwl_runner_job" in kwargs:
 417                 self.uuid = kwargs.get("cwl_runner_job").get('uuid')
 418             jobiter = tool.job(job_order,
 419                                self.output_callback,
 420                                **kwargs)
 421
 422         try:
 423             self.cond.acquire()
 424             # Will continue to hold the lock for the duration of this code
 425             # except when in cond.wait(), at which point on_message can update
 426             # job state and process output callbacks.
 427
 428             loopperf = Perf(metrics, "jobiter")
 429             loopperf.__enter__()
 430             for runnable in jobiter:
 431                 loopperf.__exit__()
 432
 433                 if self.stop_polling.is_set():
 434                     break
 435
 436                 if runnable:
 437                     with Perf(metrics, "run"):
 438                         runnable.run(**kwargs)
 439                 else:
 440                     if self.processes:
 441                         self.cond.wait(1)
 442                     else:
 443                         logger.error("Workflow is deadlocked, no runnable jobs and not waiting on any pending jobs.")
 444                         break
 445                 loopperf.__enter__()
 446             loopperf.__exit__()
 447
 448             while self.processes:
 449                 self.cond.wait(1)
 450
 451         except UnsupportedRequirement:
 452             raise
 453         except:
 454             if sys.exc_info()[0] is KeyboardInterrupt:
 455                 logger.error("Interrupted, marking pipeline as failed")
 456             else:
 457                 logger.error("Execution failed: %s", sys.exc_info()[1], exc_info=(sys.exc_info()[1] if self.debug else False))
 458             if self.pipeline:
 459                 self.api.pipeline_instances().update(uuid=self.pipeline["uuid"],
 460                                                      body={"state": "Failed"}).execute(num_retries=self.num_retries)
 461             if runnerjob and runnerjob.uuid and self.work_api == "containers":
 462                 self.api.container_requests().update(uuid=runnerjob.uuid,
 463                                                      body={"priority": "0"}).execute(num_retries=self.num_retries)
 464         finally:
 465             self.cond.release()
 466             self.stop_polling.set()
 467             self.polling_thread.join()
 468
 469         if self.final_status == "UnsupportedRequirement":
 470             raise UnsupportedRequirement("Check log for details.")
 471
 472         if self.final_output is None:
 473             raise WorkflowException("Workflow did not return a result.")
 474
 475         if kwargs.get("submit") and isinstance(runnerjob, Runner):
 476             logger.info("Final output collection %s", runnerjob.final_output)
 477         else:
 478             if self.output_name is None:
 479                 self.output_name = "Output of %s" % (shortname(tool.tool["id"]))
 480             if self.output_tags is None:
 481                 self.output_tags = ""
 482             self.final_output, self.final_output_collection = self.make_output_collection(self.output_name, self.output_tags, self.final_output)
 483             self.set_crunch_output()
 484
 485         if self.final_status != "success":
 486             raise WorkflowException("Workflow failed.")
 487
 488         if kwargs.get("compute_checksum"):
 489             adjustDirObjs(self.final_output, partial(getListing, self.fs_access))
 490             adjustFileObjs(self.final_output, partial(compute_checksums, self.fs_access))
 491
 492         return self.final_output
 493
 494
 495 def versionstring():
 496     """Print version string of key packages for provenance and debugging."""
 497
 498     arvcwlpkg = pkg_resources.require("arvados-cwl-runner")
 499     arvpkg = pkg_resources.require("arvados-python-client")
 500     cwlpkg = pkg_resources.require("cwltool")
 501
 502     return "%s %s %s, %s %s, %s %s" % (sys.argv[0], __version__, arvcwlpkg[0].version,
 503                                     "arvados-python-client", arvpkg[0].version,
 504                                     "cwltool", cwlpkg[0].version)
 505
 506
 507 def arg_parser():  # type: () -> argparse.ArgumentParser
 508     parser = argparse.ArgumentParser(description='Arvados executor for Common Workflow Language')
 509
 510     parser.add_argument("--basedir", type=str,
 511                         help="Base directory used to resolve relative references in the input, default to directory of input object file or current directory (if inputs piped/provided on command line).")
 512     parser.add_argument("--outdir", type=str, default=os.path.abspath('.'),
 513                         help="Output directory, default current directory")
 514
 515     parser.add_argument("--eval-timeout",
 516                         help="Time to wait for a Javascript expression to evaluate before giving an error, default 20s.",
 517                         type=float,
 518                         default=20)
 519     parser.add_argument("--version", action="store_true", help="Print version and exit")
 520
 521     exgroup = parser.add_mutually_exclusive_group()
 522     exgroup.add_argument("--verbose", action="store_true", help="Default logging")
 523     exgroup.add_argument("--quiet", action="store_true", help="Only print warnings and errors.")
 524     exgroup.add_argument("--debug", action="store_true", help="Print even more logging")
 525
 526     parser.add_argument("--metrics", action="store_true", help="Print timing metrics")
 527
 528     parser.add_argument("--tool-help", action="store_true", help="Print command line help for tool")
 529
 530     exgroup = parser.add_mutually_exclusive_group()
 531     exgroup.add_argument("--enable-reuse", action="store_true",
 532                         default=True, dest="enable_reuse",
 533                         help="")
 534     exgroup.add_argument("--disable-reuse", action="store_false",
 535                         default=True, dest="enable_reuse",
 536                         help="")
 537
 538     parser.add_argument("--project-uuid", type=str, metavar="UUID", help="Project that will own the workflow jobs, if not provided, will go to home project.")
 539     parser.add_argument("--output-name", type=str, help="Name to use for collection that stores the final output.", default=None)
 540     parser.add_argument("--output-tags", type=str, help="Tags for the final output collection separated by commas, e.g., '--output-tags tag0,tag1,tag2'.", default=None)
 541     parser.add_argument("--ignore-docker-for-reuse", action="store_true",
 542                         help="Ignore Docker image version when deciding whether to reuse past jobs.",
 543                         default=False)
 544
 545     exgroup = parser.add_mutually_exclusive_group()
 546     exgroup.add_argument("--submit", action="store_true", help="Submit workflow to run on Arvados.",
 547                         default=True, dest="submit")
 548     exgroup.add_argument("--local", action="store_false", help="Run workflow on local host (submits jobs to Arvados).",
 549                         default=True, dest="submit")
 550     exgroup.add_argument("--create-template", action="store_true", help="(Deprecated) synonym for --create-workflow.",
 551                          dest="create_workflow")
 552     exgroup.add_argument("--create-workflow", action="store_true", help="Create an Arvados workflow (if using the 'containers' API) or pipeline template (if using the 'jobs' API). See --api.")
 553     exgroup.add_argument("--update-workflow", type=str, metavar="UUID", help="Update an existing Arvados workflow or pipeline template with the given UUID.")
 554
 555     exgroup = parser.add_mutually_exclusive_group()
 556     exgroup.add_argument("--wait", action="store_true", help="After submitting workflow runner job, wait for completion.",
 557                         default=True, dest="wait")
 558     exgroup.add_argument("--no-wait", action="store_false", help="Submit workflow runner job and exit.",
 559                         default=True, dest="wait")
 560
 561     exgroup = parser.add_mutually_exclusive_group()
 562     exgroup.add_argument("--log-timestamps", action="store_true", help="Prefix logging lines with timestamp",
 563                         default=True, dest="log_timestamps")
 564     exgroup.add_argument("--no-log-timestamps", action="store_false", help="No timestamp on logging lines",
 565                         default=True, dest="log_timestamps")
 566
 567     parser.add_argument("--api", type=str,
 568                         default=None, dest="work_api",
 569                         help="Select work submission API, one of 'jobs' or 'containers'. Default is 'jobs' if that API is available, otherwise 'containers'.")
 570
 571     parser.add_argument("--compute-checksum", action="store_true", default=False,
 572                         help="Compute checksum of contents while collecting outputs",
 573                         dest="compute_checksum")
 574
 575     parser.add_argument("--submit-runner-ram", type=int,
 576                         help="RAM (in MiB) required for the workflow runner job (default 1024)",
 577                         default=1024)
 578
 579     parser.add_argument("--name", type=str,
 580                         help="Name to use for workflow execution instance.",
 581                         default=None)
 582
 583     parser.add_argument("--on-error", type=str,
 584                         help="Desired workflow behavior when a step fails.  One of 'stop' or 'continue'. "
 585                         "Default is 'continue'.", default="continue", choices=("stop", "continue"))
 586
 587     parser.add_argument("workflow", type=str, nargs="?", default=None, help="The workflow to execute")
 588     parser.add_argument("job_order", nargs=argparse.REMAINDER, help="The input object to the workflow.")
 589
 590     return parser
 591
 592 def add_arv_hints():
 593     cache = {}
 594     res = pkg_resources.resource_stream(__name__, 'arv-cwl-schema.yml')
 595     cache["http://arvados.org/cwl"] = res.read()
 596     res.close()
 597     document_loader, cwlnames, _, _ = cwltool.process.get_schema("v1.0")
 598     _, extnames, _, _ = schema_salad.schema.load_schema("http://arvados.org/cwl", cache=cache)
 599     for n in extnames.names:
 600         if not cwlnames.has_name("http://arvados.org/cwl#"+n, ""):
 601             cwlnames.add_name("http://arvados.org/cwl#"+n, "", extnames.get_name(n, ""))
 602         document_loader.idx["http://arvados.org/cwl#"+n] = {}
 603
 604 def main(args, stdout, stderr, api_client=None, keep_client=None):
 605     parser = arg_parser()
 606
 607     job_order_object = None
 608     arvargs = parser.parse_args(args)
 609
 610     if arvargs.version:
 611         print versionstring()
 612         return
 613
 614     if arvargs.update_workflow:
 615         if arvargs.update_workflow.find('-7fd4e-') == 5:
 616             want_api = 'containers'
 617         elif arvargs.update_workflow.find('-p5p6p-') == 5:
 618             want_api = 'jobs'
 619         else:
 620             want_api = None
 621         if want_api and arvargs.work_api and want_api != arvargs.work_api:
 622             logger.error('--update-workflow arg {!r} uses {!r} API, but --api={!r} specified'.format(
 623                 arvargs.update_workflow, want_api, arvargs.work_api))
 624             return 1
 625         arvargs.work_api = want_api
 626
 627     if (arvargs.create_workflow or arvargs.update_workflow) and not arvargs.job_order:
 628         job_order_object = ({}, "")
 629
 630     add_arv_hints()
 631
 632     try:
 633         if api_client is None:
 634             api_client=arvados.api('v1', model=OrderedJsonModel())
 635         if keep_client is None:
 636             keep_client = arvados.keep.KeepClient(api_client=api_client, num_retries=4)
 637         runner = ArvCwlRunner(api_client, work_api=arvargs.work_api, keep_client=keep_client,
 638                               num_retries=4, output_name=arvargs.output_name,
 639                               output_tags=arvargs.output_tags)
 640     except Exception as e:
 641         logger.error(e)
 642         return 1
 643
 644     if arvargs.debug:
 645         logger.setLevel(logging.DEBUG)
 646
 647     if arvargs.quiet:
 648         logger.setLevel(logging.WARN)
 649         logging.getLogger('arvados.arv-run').setLevel(logging.WARN)
 650
 651     if arvargs.metrics:
 652         metrics.setLevel(logging.DEBUG)
 653         logging.getLogger("cwltool.metrics").setLevel(logging.DEBUG)
 654
 655     if arvargs.log_timestamps:
 656         arvados.log_handler.setFormatter(logging.Formatter(
 657             '%(asctime)s %(name)s %(levelname)s: %(message)s',
 658             '%Y-%m-%d %H:%M:%S'))
 659     else:
 660         arvados.log_handler.setFormatter(logging.Formatter('%(name)s %(levelname)s: %(message)s'))
 661
 662     arvargs.conformance_test = None
 663     arvargs.use_container = True
 664     arvargs.relax_path_checks = True
 665     arvargs.validate = None
 666
 667     return cwltool.main.main(args=arvargs,
 668                              stdout=stdout,
 669                              stderr=stderr,
 670                              executor=runner.arv_executor,
 671                              makeTool=runner.arv_make_tool,
 672                              versionfunc=versionstring,
 673                              job_order_object=job_order_object,
 674                              make_fs_access=partial(CollectionFsAccess,
 675                                                     api_client=api_client,
 676                                                     keep_client=keep_client),
 677                              fetcher_constructor=partial(CollectionFetcher,
 678                                                          api_client=api_client,
 679                                                          keep_client=keep_client),
 680                              resolver=partial(collectionResolver, api_client),
 681                              logger_handler=arvados.log_handler)