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     # Timestamps are added by crunch-job, so don't print redundant timestamps.
  30     arvados.log_handler.setFormatter(logging.Formatter('%(name)s %(levelname)s: %(message)s'))
  31
  32     # Print package versions
  33     logger.info(arvados_cwl.versionstring())
  34
  35     api = arvados.api("v1")
  36
  37     arvados_cwl.add_arv_hints()
  38
  39     runner = None
  40     try:
  41         job_order_object = arvados.current_job()['script_parameters']
  42         toolpath = "file://%s/%s" % (os.environ['TASK_KEEPMOUNT'], job_order_object.pop("cwl:tool"))
  43
  44         pdh_path = re.compile(r'^[0-9a-f]{32}\+\d+(/.+)?$')
  45
  46         def keeppath(v):
  47             if pdh_path.match(v):
  48                 return "keep:%s" % v
  49             else:
  50                 return v
  51
  52         def keeppathObj(v):
  53             v["location"] = keeppath(v["location"])
  54
  55         for k,v in job_order_object.items():
  56             if isinstance(v, basestring) and arvados.util.keep_locator_pattern.match(v):
  57                 job_order_object[k] = {
  58                     "class": "File",
  59                     "location": "keep:%s" % v
  60                 }
  61
  62         adjustFileObjs(job_order_object, keeppathObj)
  63         adjustDirObjs(job_order_object, keeppathObj)
  64         normalizeFilesDirs(job_order_object)
  65         adjustDirObjs(job_order_object, functools.partial(getListing, arvados_cwl.fsaccess.CollectionFsAccess("", api_client=api)))
  66
  67         output_name = None
  68         output_tags = None
  69         enable_reuse = True
  70         if "arv:output_name" in job_order_object:
  71             output_name = job_order_object["arv:output_name"]
  72             del job_order_object["arv:output_name"]
  73
  74         if "arv:output_tags" in job_order_object:
  75             output_tags = job_order_object["arv:output_tags"]
  76             del job_order_object["arv:output_tags"]
  77
  78         if "arv:enable_reuse" in job_order_object:
  79             enable_reuse = job_order_object["arv:enable_reuse"]
  80             del job_order_object["arv:enable_reuse"]
  81
  82         if "arv:on_error" in job_order_object:
  83             on_error = job_order_object["arv:on_error"]
  84             del job_order_object["arv:on_error"]
  85
  86         runner = arvados_cwl.ArvCwlRunner(api_client=arvados.api('v1', model=OrderedJsonModel()),
  87                                           output_name=output_name, output_tags=output_tags)
  88
  89         t = load_tool(toolpath, runner.arv_make_tool)
  90
  91         args = argparse.Namespace()
  92         args.project_uuid = arvados.current_job()["owner_uuid"]
  93         args.enable_reuse = enable_reuse
  94         args.on_error = on_error
  95         args.submit = False
  96         args.debug = False
  97         args.quiet = False
  98         args.ignore_docker_for_reuse = False
  99         args.basedir = os.getcwd()
 100         args.name = None
 101         args.cwl_runner_job={"uuid": arvados.current_job()["uuid"], "state": arvados.current_job()["state"]}
 102         outputObj = runner.arv_executor(t, job_order_object, **vars(args))
 103     except Exception as e:
 104         if isinstance(e, WorkflowException):
 105             logging.info("Workflow error %s", e)
 106         else:
 107             logging.exception("Unhandled exception")
 108         if runner and runner.final_output_collection:
 109             outputCollection = runner.final_output_collection.portable_data_hash()
 110         else:
 111             outputCollection = None
 112         api.job_tasks().update(uuid=arvados.current_task()['uuid'],
 113                                              body={
 114                                                  'output': outputCollection,
 115                                                  'success': False,
 116                                                  'progress':1.0
 117                                              }).execute()