sdk/cwl/arvados_cwl/crunch_script.py

   1 # Crunch script integration for running arvados-cwl-runner (importing
   2 # arvados_cwl module) inside a crunch job.
   3 #
   4 # This gets the job record, transforms the script parameters into a valid CWL
   5 # input object, then executes the CWL runner to run the underlying workflow or
   6 # tool.  When the workflow completes, record the output object in an output
   7 # collection for this runner job.
   8
   9 import arvados
  10 import arvados_cwl
  11 import arvados.collection
  12 import arvados.util
  13 import cwltool.main
  14 import logging
  15 import os
  16 import json
  17 import argparse
  18 import re
  19 import functools
  20
  21 from arvados.api import OrderedJsonModel
  22 from cwltool.process import shortname, adjustFileObjs, adjustDirObjs, getListing, normalizeFilesDirs
  23 from cwltool.load_tool import load_tool
  24 from cwltool.errors import WorkflowException
  25
  26 logger = logging.getLogger('arvados.cwl-runner')
  27
  28 def run():
  29     # Print package versions
  30     logger.info(arvados_cwl.versionstring())
  31
  32     api = arvados.api("v1")
  33
  34     arvados_cwl.add_arv_hints()
  35
  36     runner = None
  37     try:
  38         job_order_object = arvados.current_job()['script_parameters']
  39
  40         pdh_path = re.compile(r'^[0-9a-f]{32}\+\d+(/.+)?$')
  41
  42         def keeppath(v):
  43             if pdh_path.match(v):
  44                 return "keep:%s" % v
  45             else:
  46                 return v
  47
  48         def keeppathObj(v):
  49             v["location"] = keeppath(v["location"])
  50
  51         job_order_object["cwl:tool"] = "file://%s/%s" % (os.environ['TASK_KEEPMOUNT'], job_order_object["cwl:tool"])
  52
  53         for k,v in job_order_object.items():
  54             if isinstance(v, basestring) and arvados.util.keep_locator_pattern.match(v):
  55                 job_order_object[k] = {
  56                     "class": "File",
  57                     "location": "keep:%s" % v
  58                 }
  59
  60         adjustFileObjs(job_order_object, keeppathObj)
  61         adjustDirObjs(job_order_object, keeppathObj)
  62         normalizeFilesDirs(job_order_object)
  63         adjustDirObjs(job_order_object, functools.partial(getListing, arvados_cwl.fsaccess.CollectionFsAccess("", api_client=api)))
  64
  65         output_name = None
  66         enable_reuse = True
  67         if "arv:output_name" in job_order_object:
  68             output_name = job_order_object["arv:output_name"]
  69             del job_order_object["arv:output_name"]
  70
  71         if "arv:enable_reuse" in job_order_object:
  72             enable_reuse = job_order_object["arv:enable_reuse"]
  73             del job_order_object["arv:enable_reuse"]
  74
  75         runner = arvados_cwl.ArvCwlRunner(api_client=arvados.api('v1', model=OrderedJsonModel()),
  76                                           output_name=output_name)
  77
  78         t = load_tool(job_order_object, runner.arv_make_tool)
  79
  80         args = argparse.Namespace()
  81         args.project_uuid = arvados.current_job()["owner_uuid"]
  82         args.enable_reuse = enable_reuse
  83         args.submit = False
  84         args.debug = True
  85         args.quiet = False
  86         args.ignore_docker_for_reuse = False
  87         args.basedir = os.getcwd()
  88         args.cwl_runner_job={"uuid": arvados.current_job()["uuid"], "state": arvados.current_job()["state"]}
  89         outputObj = runner.arv_executor(t, job_order_object, **vars(args))
  90     except Exception as e:
  91         if isinstance(e, WorkflowException):
  92             logging.info("Workflow error %s", e)
  93         else:
  94             logging.exception("Unhandled exception")
  95         if runner and runner.final_output_collection:
  96             outputCollection = runner.final_output_collection.portable_data_hash()
  97         else:
  98             outputCollection = None
  99         api.job_tasks().update(uuid=arvados.current_task()['uuid'],
 100                                              body={
 101                                                  'output': outputCollection,
 102                                                  'success': False,
 103                                                  'progress':1.0
 104                                              }).execute()