- images = arvados.commands.keepdocker.list_images_in_arv(api_client, 3,
- image_name=image_name,
- image_tag=image_tag)
-
- if not images:
- imageId = cwltool.docker.get_image(dockerRequirement, pull_image)
- args = [image_name]
- if image_tag:
- args.append(image_tag)
- logger.info("Uploading Docker image %s", ":".join(args))
- arvados.commands.keepdocker.main(args)
-
- return dockerRequirement["dockerImageId"]
-
-
-class CollectionFsAccess(cwltool.process.StdFsAccess):
- def __init__(self, basedir):
- self.collections = {}
- self.basedir = basedir
-
- def get_collection(self, path):
- p = path.split("/")
- if p[0].startswith("keep:") and arvados.util.keep_locator_pattern.match(p[0][5:]):
- pdh = p[0][5:]
- if pdh not in self.collections:
- self.collections[pdh] = arvados.collection.CollectionReader(pdh)
- return (self.collections[pdh], "/".join(p[1:]))
- else:
- return (None, path)
-
- def _match(self, collection, patternsegments, parent):
- if not patternsegments:
- return []
-
- if not isinstance(collection, arvados.collection.RichCollectionBase):
- return []
-
- ret = []
- # iterate over the files and subcollections in 'collection'
- for filename in collection:
- if patternsegments[0] == '.':
- # Pattern contains something like "./foo" so just shift
- # past the "./"
- ret.extend(self._match(collection, patternsegments[1:], parent))
- elif fnmatch.fnmatch(filename, patternsegments[0]):
- cur = os.path.join(parent, filename)
- if len(patternsegments) == 1:
- ret.append(cur)
- else:
- ret.extend(self._match(collection[filename], patternsegments[1:], cur))
- return ret
-
- def glob(self, pattern):
- collection, rest = self.get_collection(pattern)
- patternsegments = rest.split("/")
- return self._match(collection, patternsegments, "keep:" + collection.manifest_locator())
-
- def open(self, fn, mode):
- collection, rest = self.get_collection(fn)
- if collection:
- return collection.open(rest, mode)
- else:
- return open(self._abs(fn), mode)
-
- def exists(self, fn):
- collection, rest = self.get_collection(fn)
- if collection:
- return collection.exists(rest)
- else:
- return os.path.exists(self._abs(fn))
-
-class ArvadosJob(object):
- def __init__(self, runner):
- self.arvrunner = runner
- self.running = False
-
- def run(self, dry_run=False, pull_image=True, **kwargs):
- script_parameters = {
- "command": self.command_line
- }
- runtime_constraints = {}
-
- if self.generatefiles:
- vwd = arvados.collection.Collection()
- script_parameters["task.vwd"] = {}
- for t in self.generatefiles:
- if isinstance(self.generatefiles[t], dict):
- src, rest = self.arvrunner.fs_access.get_collection(self.generatefiles[t]["path"].replace("$(task.keep)/", "keep:"))
- vwd.copy(rest, t, source_collection=src)
- else:
- with vwd.open(t, "w") as f:
- f.write(self.generatefiles[t])
- vwd.save_new()
- for t in self.generatefiles:
- script_parameters["task.vwd"][t] = "$(task.keep)/%s/%s" % (vwd.portable_data_hash(), t)
-
- script_parameters["task.env"] = {"TMPDIR": "$(task.tmpdir)"}
- if self.environment:
- script_parameters["task.env"].update(self.environment)
-
- if self.stdin:
- script_parameters["task.stdin"] = self.pathmapper.mapper(self.stdin)[1]
-
- if self.stdout:
- script_parameters["task.stdout"] = self.stdout
-
- (docker_req, docker_is_req) = get_feature(self, "DockerRequirement")
- if docker_req and kwargs.get("use_container") is not False:
- runtime_constraints["docker_image"] = arv_docker_get_image(self.arvrunner.api, docker_req, pull_image)
-
- resources = self.builder.resources
- if resources is not None:
- keys = resources.keys()
- if "coresMin" in keys:
- try:
- runtime_constraints["min_cores_per_node"] = int(resources["coresMin"])
- except:
- runtime_constraints["min_cores_per_node"] = None
- if "ramMin" in keys:
- try:
- runtime_constraints["min_ram_mb_per_node"] = int(resources["ramMin"])
- except:
- runtime_constraints["min_ram_mb_per_node"] = None
- if "tmpdirMin" in keys:
- try:
- runtime_constraints["min_scratch_mb_per_node"] = int(resources["tmpdirMin"])
- except:
- runtime_constraints["min_scratch_mb_per_node"] = None
-
- try:
- response = self.arvrunner.api.jobs().create(body={
- "script": "crunchrunner",
- "repository": "arvados",
- "script_version": "8488-cwl-crunchrunner-collection",
- "script_parameters": {"tasks": [script_parameters], "crunchrunner": crunchrunner_pdh+"/crunchrunner"},
- "runtime_constraints": runtime_constraints
- }, find_or_create=kwargs.get("enable_reuse", True)).execute(num_retries=self.arvrunner.num_retries)
-
- self.arvrunner.jobs[response["uuid"]] = self
-
- self.arvrunner.pipeline["components"][self.name] = {"job": response}
- self.arvrunner.pipeline = self.arvrunner.api.pipeline_instances().update(uuid=self.arvrunner.pipeline["uuid"],
- body={
- "components": self.arvrunner.pipeline["components"]
- }).execute(num_retries=self.arvrunner.num_retries)
-
- logger.info("Job %s (%s) is %s", self.name, response["uuid"], response["state"])
-
- if response["state"] in ("Complete", "Failed", "Cancelled"):
- self.done(response)
- except Exception as e:
- logger.error("Got error %s" % str(e))
- self.output_callback({}, "permanentFail")
-
- def update_pipeline_component(self, record):
- self.arvrunner.pipeline["components"][self.name] = {"job": record}
- self.arvrunner.pipeline = self.arvrunner.api.pipeline_instances().update(uuid=self.arvrunner.pipeline["uuid"],
- body={
- "components": self.arvrunner.pipeline["components"]
- }).execute(num_retries=self.arvrunner.num_retries)
-
- def done(self, record):
- try:
- self.update_pipeline_component(record)
- except:
- pass
-
- try:
- if record["state"] == "Complete":
- processStatus = "success"
- else:
- processStatus = "permanentFail"
-
- try:
- outputs = {}
- if record["output"]:
- logc = arvados.collection.Collection(record["log"])
- log = logc.open(logc.keys()[0])
- tmpdir = None
- outdir = None
- keepdir = None
- for l in log.readlines():
- g = tmpdirre.match(l)
- if g:
- tmpdir = g.group(1)
- g = outdirre.match(l)
- if g:
- outdir = g.group(1)
- g = keepre.match(l)
- if g:
- keepdir = g.group(1)
- if tmpdir and outdir and keepdir:
- break
-
- self.builder.outdir = outdir
- self.builder.pathmapper.keepdir = keepdir
- outputs = self.collect_outputs("keep:" + record["output"])
- except Exception as e:
- logger.exception("Got exception while collecting job outputs:")
- processStatus = "permanentFail"
-
- self.output_callback(outputs, processStatus)
- finally:
- del self.arvrunner.jobs[record["uuid"]]
-
-
-class ArvPathMapper(cwltool.pathmapper.PathMapper):
- def __init__(self, arvrunner, referenced_files, basedir, **kwargs):
- self._pathmap = arvrunner.get_uploaded()
- uploadfiles = []
-
- pdh_path = re.compile(r'^keep:[0-9a-f]{32}\+\d+/.+')
-
- for src in referenced_files:
- if isinstance(src, basestring) and pdh_path.match(src):
- self._pathmap[src] = (src, "$(task.keep)/%s" % src[5:])
- if src not in self._pathmap:
- ab = cwltool.pathmapper.abspath(src, basedir)
- st = arvados.commands.run.statfile("", ab, fnPattern="$(task.keep)/%s/%s")
- if kwargs.get("conformance_test"):
- self._pathmap[src] = (src, ab)
- elif isinstance(st, arvados.commands.run.UploadFile):
- uploadfiles.append((src, ab, st))
- elif isinstance(st, arvados.commands.run.ArvFile):
- self._pathmap[src] = (ab, st.fn)
- else:
- raise cwltool.workflow.WorkflowException("Input file path '%s' is invalid" % st)
-
- if uploadfiles:
- arvados.commands.run.uploadfiles([u[2] for u in uploadfiles],
- arvrunner.api,
- dry_run=kwargs.get("dry_run"),
- num_retries=3,
- fnPattern="$(task.keep)/%s/%s")
-
- for src, ab, st in uploadfiles:
- arvrunner.add_uploaded(src, (ab, st.fn))
- self._pathmap[src] = (ab, st.fn)
-
- self.keepdir = None
-
- def reversemap(self, target):
- if target.startswith("keep:"):
- return target
- elif self.keepdir and target.startswith(self.keepdir):
- return "keep:" + target[len(self.keepdir)+1:]
- else:
- return super(ArvPathMapper, self).reversemap(target)
-
-
-class ArvadosCommandTool(cwltool.draft2tool.CommandLineTool):
- def __init__(self, arvrunner, toolpath_object, **kwargs):
- super(ArvadosCommandTool, self).__init__(toolpath_object, **kwargs)
- self.arvrunner = arvrunner
-
- def makeJobRunner(self):
- return ArvadosJob(self.arvrunner)
-
- def makePathMapper(self, reffiles, input_basedir, **kwargs):
- return ArvPathMapper(self.arvrunner, reffiles, input_basedir, **kwargs)