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         api_req, _ = get_feature(self, "http://arvados.org/cwl#APIRequirement")
  94         if api_req:
  95             runtime_constraints["API"] = True
  96
  97         runtime_req, _ = get_feature(self, "http://arvados.org/cwl#RuntimeConstraints")
  98         if runtime_req:
  99             logger.warn("RuntimeConstraints not yet supported by container API")
 100
 101         container_request["mounts"] = mounts
 102         container_request["runtime_constraints"] = runtime_constraints
 103
 104         try:
 105             response = self.arvrunner.api.container_requests().create(
 106                 body=container_request
 107             ).execute(num_retries=self.arvrunner.num_retries)
 108
 109             self.arvrunner.processes[response["container_uuid"]] = self
 110
 111             logger.info("Container %s (%s) request state is %s", self.name, response["uuid"], response["state"])
 112
 113             if response["state"] == "Final":
 114                 self.done(response)
 115         except Exception as e:
 116             logger.error("Got error %s" % str(e))
 117             self.output_callback({}, "permanentFail")
 118
 119     def done(self, record):
 120         try:
 121             if record["state"] == "Complete":
 122                 rcode = record["exit_code"]
 123                 if self.successCodes and rcode in self.successCodes:
 124                     processStatus = "success"
 125                 elif self.temporaryFailCodes and rcode in self.temporaryFailCodes:
 126                     processStatus = "temporaryFail"
 127                 elif self.permanentFailCodes and rcode in self.permanentFailCodes:
 128                     processStatus = "permanentFail"
 129                 elif rcode == 0:
 130                     processStatus = "success"
 131                 else:
 132                     processStatus = "permanentFail"
 133             else:
 134                 processStatus = "permanentFail"
 135
 136             try:
 137                 outputs = {}
 138                 if record["output"]:
 139                     outputs = done.done(self, record, "/tmp", self.outdir, "/keep")
 140             except WorkflowException as e:
 141                 logger.error("Error while collecting container outputs:\n%s", e, exc_info=(e if self.arvrunner.debug else False))
 142                 processStatus = "permanentFail"
 143             except Exception as e:
 144                 logger.exception("Got unknown exception while collecting job outputs:")
 145                 processStatus = "permanentFail"
 146
 147             self.output_callback(outputs, processStatus)
 148         finally:
 149             del self.arvrunner.processes[record["uuid"]]
 150
 151
 152 class RunnerContainer(Runner):
 153     """Submit and manage a container that runs arvados-cwl-runner."""
 154
 155     def arvados_job_spec(self, dry_run=False, pull_image=True, **kwargs):
 156         """Create an Arvados container request for this workflow.
 157
 158         The returned dict can be used to create a container passed as
 159         the +body+ argument to container_requests().create().
 160         """
 161
 162         workflowmapper = super(RunnerContainer, self).arvados_job_spec(dry_run=dry_run, pull_image=pull_image, **kwargs)
 163
 164         with arvados.collection.Collection(api_client=self.arvrunner.api) as jobobj:
 165             with jobobj.open("cwl.input.json", "w") as f:
 166                 json.dump(self.job_order, f, sort_keys=True, indent=4)
 167             jobobj.save_new(owner_uuid=self.arvrunner.project_uuid)
 168
 169         workflowname = os.path.basename(self.tool.tool["id"])
 170         workflowpath = "/var/lib/cwl/workflow/%s" % workflowname
 171         workflowcollection = workflowmapper.mapper(self.tool.tool["id"])[1]
 172         workflowcollection = workflowcollection[5:workflowcollection.index('/')]
 173         jobpath = "/var/lib/cwl/job/cwl.input.json"
 174
 175         container_image = arv_docker_get_image(self.arvrunner.api,
 176                                                {"dockerImageId": "arvados/jobs"},
 177                                                pull_image,
 178                                                self.arvrunner.project_uuid)
 179
 180         return {
 181             "command": ["arvados-cwl-runner", "--local", "--api=containers", workflowpath, jobpath],
 182             "owner_uuid": self.arvrunner.project_uuid,
 183             "name": self.name,
 184             "output_path": "/var/spool/cwl",
 185             "cwd": "/var/spool/cwl",
 186             "priority": 1,
 187             "state": "Committed",
 188             "container_image": container_image,
 189             "mounts": {
 190                 "/var/lib/cwl/workflow": {
 191                     "kind": "collection",
 192                     "portable_data_hash": "%s" % workflowcollection
 193                 },
 194                 jobpath: {
 195                     "kind": "collection",
 196                     "portable_data_hash": "%s/cwl.input.json" % jobobj.portable_data_hash()
 197                 },
 198                 "stdout": {
 199                     "kind": "file",
 200                     "path": "/var/spool/cwl/cwl.output.json"
 201                 },
 202                 "/var/spool/cwl": {
 203                     "kind": "collection",
 204                     "writable": True
 205                 }
 206             },
 207             "runtime_constraints": {
 208                 "vcpus": 1,
 209                 "ram": 1024*1024*256,
 210                 "API": True
 211             }
 212         }
 213
 214     def run(self, *args, **kwargs):
 215         kwargs["keepprefix"] = "keep:"
 216         job_spec = self.arvados_job_spec(*args, **kwargs)
 217         job_spec.setdefault("owner_uuid", self.arvrunner.project_uuid)
 218
 219         response = self.arvrunner.api.container_requests().create(
 220             body=job_spec
 221         ).execute(num_retries=self.arvrunner.num_retries)
 222
 223         self.uuid = response["uuid"]
 224         self.arvrunner.processes[response["container_uuid"]] = self
 225
 226         logger.info("Submitted container %s", response["uuid"])
 227
 228         if response["state"] in ("Complete", "Failed", "Cancelled"):
 229             self.done(response)