sdk/cwl/arvados_cwl/arvcontainer.py

   1 import logging
   2 import json
   3 import os
   4
   5 from cwltool.errors import WorkflowException
   6 from cwltool.process import get_feature, UnsupportedRequirement, shortname
   7 from cwltool.pathmapper import adjustFiles
   8
   9 import arvados.collection
  10
  11 from .arvdocker import arv_docker_get_image
  12 from . import done
  13 from .runner import Runner
  14
  15 logger = logging.getLogger('arvados.cwl-runner')
  16
  17 class ArvadosContainer(object):
  18     """Submit and manage a Crunch container request for executing a CWL CommandLineTool."""
  19
  20     def __init__(self, runner):
  21         self.arvrunner = runner
  22         self.running = False
  23         self.uuid = None
  24
  25     def update_pipeline_component(self, r):
  26         pass
  27
  28     def run(self, dry_run=False, pull_image=True, **kwargs):
  29         container_request = {
  30             "command": self.command_line,
  31             "owner_uuid": self.arvrunner.project_uuid,
  32             "name": self.name,
  33             "output_path": self.outdir,
  34             "cwd": self.outdir,
  35             "priority": 1,
  36             "state": "Committed"
  37         }
  38         runtime_constraints = {}
  39         mounts = {
  40             self.outdir: {
  41                 "kind": "tmp"
  42             }
  43         }
  44
  45         dirs = set()
  46         for f in self.pathmapper.files():
  47             _, p, tp = self.pathmapper.mapper(f)
  48             if tp == "Directory" and '/' not in p[6:]:
  49                 mounts[p] = {
  50                     "kind": "collection",
  51                     "portable_data_hash": p[6:]
  52                 }
  53                 dirs.add(p[6:])
  54         for f in self.pathmapper.files():
  55             _, p, tp = self.pathmapper.mapper(f)
  56             if p[6:].split("/")[0] not in dirs:
  57                 mounts[p] = {
  58                     "kind": "collection",
  59                     "portable_data_hash": p[6:]
  60                 }
  61
  62         if self.generatefiles["listing"]:
  63             raise UnsupportedRequirement("Generate files not supported")
  64
  65         container_request["environment"] = {"TMPDIR": self.tmpdir, "HOME": self.outdir}
  66         if self.environment:
  67             container_request["environment"].update(self.environment)
  68
  69         if self.stdin:
  70             raise UnsupportedRequirement("Stdin redirection currently not suppported")
  71
  72         if self.stderr:
  73             raise UnsupportedRequirement("Stderr redirection currently not suppported")
  74
  75         if self.stdout:
  76             mounts["stdout"] = {"kind": "file",
  77                                 "path": "%s/%s" % (self.outdir, self.stdout)}
  78
  79         (docker_req, docker_is_req) = get_feature(self, "DockerRequirement")
  80         if not docker_req:
  81             docker_req = {"dockerImageId": "arvados/jobs"}
  82
  83         container_request["container_image"] = arv_docker_get_image(self.arvrunner.api,
  84                                                                      docker_req,
  85                                                                      pull_image,
  86                                                                      self.arvrunner.project_uuid)
  87
  88         resources = self.builder.resources
  89         if resources is not None:
  90             runtime_constraints["vcpus"] = resources.get("cores", 1)
  91             runtime_constraints["ram"] = resources.get("ram") * 2**20
  92
  93         container_request["mounts"] = mounts
  94         container_request["runtime_constraints"] = runtime_constraints
  95
  96         try:
  97             response = self.arvrunner.api.container_requests().create(
  98                 body=container_request
  99             ).execute(num_retries=self.arvrunner.num_retries)
 100
 101             self.arvrunner.processes[response["container_uuid"]] = self
 102
 103             logger.info("Container %s (%s) request state is %s", self.name, response["uuid"], response["state"])
 104
 105             if response["state"] == "Final":
 106                 self.done(response)
 107         except Exception as e:
 108             logger.error("Got error %s" % str(e))
 109             self.output_callback({}, "permanentFail")
 110
 111     def done(self, record):
 112         try:
 113             if record["state"] == "Complete":
 114                 rcode = record["exit_code"]
 115                 if self.successCodes and rcode in self.successCodes:
 116                     processStatus = "success"
 117                 elif self.temporaryFailCodes and rcode in self.temporaryFailCodes:
 118                     processStatus = "temporaryFail"
 119                 elif self.permanentFailCodes and rcode in self.permanentFailCodes:
 120                     processStatus = "permanentFail"
 121                 elif rcode == 0:
 122                     processStatus = "success"
 123                 else:
 124                     processStatus = "permanentFail"
 125             else:
 126                 processStatus = "permanentFail"
 127
 128             try:
 129                 outputs = {}
 130                 if record["output"]:
 131                     outputs = done.done(self, record, "/tmp", self.outdir, "/keep")
 132             except WorkflowException as e:
 133                 logger.error("Error while collecting container outputs:\n%s", e, exc_info=(e if self.arvrunner.debug else False))
 134                 processStatus = "permanentFail"
 135             except Exception as e:
 136                 logger.exception("Got unknown exception while collecting job outputs:")
 137                 processStatus = "permanentFail"
 138
 139             self.output_callback(outputs, processStatus)
 140         finally:
 141             del self.arvrunner.processes[record["uuid"]]
 142
 143
 144 class RunnerContainer(Runner):
 145     """Submit and manage a container that runs arvados-cwl-runner."""
 146
 147     def arvados_job_spec(self, dry_run=False, pull_image=True, **kwargs):
 148         """Create an Arvados container request for this workflow.
 149
 150         The returned dict can be used to create a container passed as
 151         the +body+ argument to container_requests().create().
 152         """
 153
 154         workflowmapper = super(RunnerContainer, self).arvados_job_spec(dry_run=dry_run, pull_image=pull_image, **kwargs)
 155
 156         with arvados.collection.Collection(api_client=self.arvrunner.api) as jobobj:
 157             with jobobj.open("cwl.input.json", "w") as f:
 158                 json.dump(self.job_order, f, sort_keys=True, indent=4)
 159             jobobj.save_new(owner_uuid=self.arvrunner.project_uuid)
 160
 161         workflowname = os.path.basename(self.tool.tool["id"])
 162         workflowpath = "/var/lib/cwl/workflow/%s" % workflowname
 163         workflowcollection = workflowmapper.mapper(self.tool.tool["id"])[1]
 164         workflowcollection = workflowcollection[5:workflowcollection.index('/')]
 165         jobpath = "/var/lib/cwl/job/cwl.input.json"
 166
 167         container_image = arv_docker_get_image(self.arvrunner.api,
 168                                                {"dockerImageId": "arvados/jobs"},
 169                                                pull_image,
 170                                                self.arvrunner.project_uuid)
 171
 172         return {
 173             "command": ["arvados-cwl-runner", "--local", "--api=containers", workflowpath, jobpath],
 174             "owner_uuid": self.arvrunner.project_uuid,
 175             "name": self.name,
 176             "output_path": "/var/spool/cwl",
 177             "cwd": "/var/spool/cwl",
 178             "priority": 1,
 179             "state": "Committed",
 180             "container_image": container_image,
 181             "mounts": {
 182                 "/var/lib/cwl/workflow": {
 183                     "kind": "collection",
 184                     "portable_data_hash": "%s" % workflowcollection
 185                 },
 186                 jobpath: {
 187                     "kind": "collection",
 188                     "portable_data_hash": "%s/cwl.input.json" % jobobj.portable_data_hash()
 189                 },
 190                 "stdout": {
 191                     "kind": "file",
 192                     "path": "/var/spool/cwl/cwl.output.json"
 193                 },
 194                 "/var/spool/cwl": {
 195                     "kind": "collection",
 196                     "writable": True
 197                 }
 198             },
 199             "runtime_constraints": {
 200                 "vcpus": 1,
 201                 "ram": 1024*1024*256,
 202                 "API": True
 203             }
 204         }
 205
 206     def run(self, *args, **kwargs):
 207         kwargs["keepprefix"] = "keep:"
 208         job_spec = self.arvados_job_spec(*args, **kwargs)
 209         job_spec.setdefault("owner_uuid", self.arvrunner.project_uuid)
 210
 211         response = self.arvrunner.api.container_requests().create(
 212             body=job_spec
 213         ).execute(num_retries=self.arvrunner.num_retries)
 214
 215         self.uuid = response["uuid"]
 216         self.arvrunner.processes[response["container_uuid"]] = self
 217
 218         logger.info("Submitted container %s", response["uuid"])
 219
 220         if response["state"] in ("Complete", "Failed", "Cancelled"):
 221             self.done(response)