sdk/cwl/arvados_cwl/runner.py

   1 import os
   2 import urlparse
   3 from functools import partial
   4 import logging
   5 import json
   6 import subprocess
   7
   8 from StringIO import StringIO
   9
  10 from schema_salad.sourceline import SourceLine
  11
  12 import cwltool.draft2tool
  13 from cwltool.draft2tool import CommandLineTool
  14 import cwltool.workflow
  15 from cwltool.process import get_feature, scandeps, UnsupportedRequirement, normalizeFilesDirs, shortname
  16 from cwltool.load_tool import fetch_document
  17 from cwltool.pathmapper import adjustFileObjs, adjustDirObjs
  18 from cwltool.utils import aslist
  19 from cwltool.builder import substitute
  20 from cwltool.pack import pack
  21
  22 import arvados.collection
  23 import ruamel.yaml as yaml
  24
  25 from .arvdocker import arv_docker_get_image
  26 from .pathmapper import ArvPathMapper, trim_listing
  27 from ._version import __version__
  28 from . import done
  29
  30 logger = logging.getLogger('arvados.cwl-runner')
  31
  32 def trim_anonymous_location(obj):
  33     """Remove 'location' field from File and Directory literals.
  34
  35     To make internal handling easier, literals are assigned a random id for
  36     'location'.  However, when writing the record back out, this can break
  37     reproducibility.  Since it is valid for literals not have a 'location'
  38     field, remove it.
  39
  40     """
  41
  42     if obj.get("location", "").startswith("_:"):
  43         del obj["location"]
  44
  45 def upload_dependencies(arvrunner, name, document_loader,
  46                         workflowobj, uri, loadref_run, include_primary=True):
  47     """Upload the dependencies of the workflowobj document to Keep.
  48
  49     Returns a pathmapper object mapping local paths to keep references.  Also
  50     does an in-place update of references in "workflowobj".
  51
  52     Use scandeps to find $import, $include, $schemas, run, File and Directory
  53     fields that represent external references.
  54
  55     If workflowobj has an "id" field, this will reload the document to ensure
  56     it is scanning the raw document prior to preprocessing.
  57     """
  58
  59     loaded = set()
  60     def loadref(b, u):
  61         joined = document_loader.fetcher.urljoin(b, u)
  62         defrg, _ = urlparse.urldefrag(joined)
  63         if defrg not in loaded:
  64             loaded.add(defrg)
  65             # Use fetch_text to get raw file (before preprocessing).
  66             text = document_loader.fetch_text(defrg)
  67             if isinstance(text, bytes):
  68                 textIO = StringIO(text.decode('utf-8'))
  69             else:
  70                 textIO = StringIO(text)
  71             return yaml.safe_load(textIO)
  72         else:
  73             return {}
  74
  75     if loadref_run:
  76         loadref_fields = set(("$import", "run"))
  77     else:
  78         loadref_fields = set(("$import",))
  79
  80     scanobj = workflowobj
  81     if "id" in workflowobj:
  82         # Need raw file content (before preprocessing) to ensure
  83         # that external references in $include and $mixin are captured.
  84         scanobj = loadref("", workflowobj["id"])
  85
  86     sc = scandeps(uri, scanobj,
  87                   loadref_fields,
  88                   set(("$include", "$schemas", "location")),
  89                   loadref, urljoin=document_loader.fetcher.urljoin)
  90
  91     normalizeFilesDirs(sc)
  92
  93     if include_primary and "id" in workflowobj:
  94         sc.append({"class": "File", "location": workflowobj["id"]})
  95
  96     if "$schemas" in workflowobj:
  97         for s in workflowobj["$schemas"]:
  98             sc.append({"class": "File", "location": s})
  99
 100     mapper = ArvPathMapper(arvrunner, sc, "",
 101                            "keep:%s",
 102                            "keep:%s/%s",
 103                            name=name,
 104                            single_collection=True)
 105
 106     def setloc(p):
 107         if "location" in p and (not p["location"].startswith("_:")) and (not p["location"].startswith("keep:")):
 108             p["location"] = mapper.mapper(p["location"]).resolved
 109     adjustFileObjs(workflowobj, setloc)
 110     adjustDirObjs(workflowobj, setloc)
 111
 112     if "$schemas" in workflowobj:
 113         sch = []
 114         for s in workflowobj["$schemas"]:
 115             sch.append(mapper.mapper(s).resolved)
 116         workflowobj["$schemas"] = sch
 117
 118     return mapper
 119
 120
 121 def upload_docker(arvrunner, tool):
 122     """Uploads Docker images used in CommandLineTool objects."""
 123
 124     if isinstance(tool, CommandLineTool):
 125         (docker_req, docker_is_req) = get_feature(tool, "DockerRequirement")
 126         if docker_req:
 127             if docker_req.get("dockerOutputDirectory"):
 128                 # TODO: can be supported by containers API, but not jobs API.
 129                 raise SourceLine(docker_req, "dockerOutputDirectory", UnsupportedRequirement).makeError(
 130                     "Option 'dockerOutputDirectory' of DockerRequirement not supported.")
 131             arv_docker_get_image(arvrunner.api, docker_req, True, arvrunner.project_uuid)
 132     elif isinstance(tool, cwltool.workflow.Workflow):
 133         for s in tool.steps:
 134             upload_docker(arvrunner, s.embedded_tool)
 135
 136 def packed_workflow(arvrunner, tool):
 137     """Create a packed workflow.
 138
 139     A "packed" workflow is one where all the components have been combined into a single document."""
 140
 141     return pack(tool.doc_loader, tool.doc_loader.fetch(tool.tool["id"]),
 142                 tool.tool["id"], tool.metadata)
 143
 144 def tag_git_version(packed):
 145     if tool.tool["id"].startswith("file://"):
 146         path = os.path.dirname(tool.tool["id"][7:])
 147         try:
 148             githash = subprocess.check_output(['git', 'log', '--first-parent', '--max-count=1', '--format=%H'], stderr=subprocess.STDOUT, cwd=path).strip()
 149         except (OSError, subprocess.CalledProcessError):
 150             pass
 151         else:
 152             packed["http://schema.org/version"] = githash
 153
 154
 155 def upload_job_order(arvrunner, name, tool, job_order):
 156     """Upload local files referenced in the input object and return updated input
 157     object with 'location' updated to the proper keep references.
 158     """
 159
 160     for t in tool.tool["inputs"]:
 161         def setSecondary(fileobj):
 162             if isinstance(fileobj, dict) and fileobj.get("class") == "File":
 163                 if "secondaryFiles" not in fileobj:
 164                     fileobj["secondaryFiles"] = [{"location": substitute(fileobj["location"], sf), "class": "File"} for sf in t["secondaryFiles"]]
 165
 166             if isinstance(fileobj, list):
 167                 for e in fileobj:
 168                     setSecondary(e)
 169
 170         if shortname(t["id"]) in job_order and t.get("secondaryFiles"):
 171             setSecondary(job_order[shortname(t["id"])])
 172
 173     jobmapper = upload_dependencies(arvrunner,
 174                                     name,
 175                                     tool.doc_loader,
 176                                     job_order,
 177                                     job_order.get("id", "#"),
 178                                     False)
 179
 180     if "id" in job_order:
 181         del job_order["id"]
 182
 183     # Need to filter this out, gets added by cwltool when providing
 184     # parameters on the command line.
 185     if "job_order" in job_order:
 186         del job_order["job_order"]
 187
 188     return job_order
 189
 190 def upload_workflow_deps(arvrunner, tool, override_tools):
 191     # Ensure that Docker images needed by this workflow are available
 192
 193     upload_docker(arvrunner, tool)
 194
 195     document_loader = tool.doc_loader
 196
 197     def upload_tool_deps(deptool):
 198         if "id" in deptool:
 199             upload_dependencies(arvrunner,
 200                                 "%s dependencies" % (shortname(deptool["id"])),
 201                                 document_loader,
 202                                 deptool,
 203                                 deptool["id"],
 204                                 False,
 205                                 include_primary=False)
 206             document_loader.idx[deptool["id"]] = deptool
 207             override_tools[deptool["id"]] = json.dumps(deptool)
 208
 209     tool.visit(upload_tool_deps)
 210
 211 def arvados_jobs_image(arvrunner, img):
 212     """Determine if the right arvados/jobs image version is available.  If not, try to pull and upload it."""
 213
 214     try:
 215         arv_docker_get_image(arvrunner.api, {"dockerPull": img}, True, arvrunner.project_uuid)
 216     except Exception as e:
 217         raise Exception("Docker image %s is not available\n%s" % (img, e) )
 218     return img
 219
 220 def upload_workflow_collection(arvrunner, name, packed):
 221     collection = arvados.collection.Collection(api_client=arvrunner.api,
 222                                                keep_client=arvrunner.keep_client,
 223                                                num_retries=arvrunner.num_retries)
 224     with collection.open("workflow.cwl", "w") as f:
 225         f.write(json.dumps(packed, indent=2, sort_keys=True, separators=(',',': ')))
 226
 227     filters = [["portable_data_hash", "=", collection.portable_data_hash()],
 228                ["name", "like", name+"%"]]
 229     if arvrunner.project_uuid:
 230         filters.append(["owner_uuid", "=", arvrunner.project_uuid])
 231     exists = arvrunner.api.collections().list(filters=filters).execute(num_retries=arvrunner.num_retries)
 232
 233     if exists["items"]:
 234         logger.info("Using collection %s", exists["items"][0]["uuid"])
 235     else:
 236         collection.save_new(name=name,
 237                             owner_uuid=arvrunner.project_uuid,
 238                             ensure_unique_name=True,
 239                             num_retries=arvrunner.num_retries)
 240         logger.info("Uploaded to %s", collection.manifest_locator())
 241
 242     return collection.portable_data_hash()
 243
 244
 245 class Runner(object):
 246     """Base class for runner processes, which submit an instance of
 247     arvados-cwl-runner and wait for the final result."""
 248
 249     def __init__(self, runner, tool, job_order, enable_reuse,
 250                  output_name, output_tags, submit_runner_ram=0,
 251                  name=None, on_error=None, submit_runner_image=None):
 252         self.arvrunner = runner
 253         self.tool = tool
 254         self.job_order = job_order
 255         self.running = False
 256         self.enable_reuse = enable_reuse
 257         self.uuid = None
 258         self.final_output = None
 259         self.output_name = output_name
 260         self.output_tags = output_tags
 261         self.name = name
 262         self.on_error = on_error
 263         self.jobs_image = submit_runner_image or "arvados/jobs:"+__version__
 264
 265         if submit_runner_ram:
 266             self.submit_runner_ram = submit_runner_ram
 267         else:
 268             self.submit_runner_ram = 3000
 269
 270         if self.submit_runner_ram <= 0:
 271             raise Exception("Value of --submit-runner-ram must be greater than zero")
 272
 273     def update_pipeline_component(self, record):
 274         pass
 275
 276     def done(self, record):
 277         """Base method for handling a completed runner."""
 278
 279         try:
 280             if record["state"] == "Complete":
 281                 if record.get("exit_code") is not None:
 282                     if record["exit_code"] == 33:
 283                         processStatus = "UnsupportedRequirement"
 284                     elif record["exit_code"] == 0:
 285                         processStatus = "success"
 286                     else:
 287                         processStatus = "permanentFail"
 288                 else:
 289                     processStatus = "success"
 290             else:
 291                 processStatus = "permanentFail"
 292
 293             outputs = {}
 294
 295             if processStatus == "permanentFail":
 296                 logc = arvados.collection.CollectionReader(record["log"],
 297                                                            api_client=self.arvrunner.api,
 298                                                            keep_client=self.arvrunner.keep_client,
 299                                                            num_retries=self.arvrunner.num_retries)
 300                 done.logtail(logc, logger, "%s error log:" % self.arvrunner.label(self), maxlen=40)
 301
 302             self.final_output = record["output"]
 303             outc = arvados.collection.CollectionReader(self.final_output,
 304                                                        api_client=self.arvrunner.api,
 305                                                        keep_client=self.arvrunner.keep_client,
 306                                                        num_retries=self.arvrunner.num_retries)
 307             if "cwl.output.json" in outc:
 308                 with outc.open("cwl.output.json") as f:
 309                     if f.size() > 0:
 310                         outputs = json.load(f)
 311             def keepify(fileobj):
 312                 path = fileobj["location"]
 313                 if not path.startswith("keep:"):
 314                     fileobj["location"] = "keep:%s/%s" % (record["output"], path)
 315             adjustFileObjs(outputs, keepify)
 316             adjustDirObjs(outputs, keepify)
 317         except Exception as e:
 318             logger.exception("[%s] While getting final output object: %s", self.name, e)
 319             self.arvrunner.output_callback({}, "permanentFail")
 320         else:
 321             self.arvrunner.output_callback(outputs, processStatus)
 322         finally:
 323             if record["uuid"] in self.arvrunner.processes:
 324                 del self.arvrunner.processes[record["uuid"]]