def collect_output(self, schema, builder, outdir):
r = []
if "outputBinding" in schema:
binding = schema["outputBinding"]
globpatterns = []
#revmap = partial(revmap_file, builder, outdir)
if "glob" in binding:
for gb in aslist(binding["glob"]):
gb = builder.do_eval(gb)
#_logger.debug("gb evaluated to: '{}'".format(gb))
if gb:
globpatterns.extend(aslist(gb))
for gb in globpatterns:
g = os.path.join(outdir, gb)
cls = "File" if schema["type"] == "File" else "Directory"
r.extend([
{"location": g,
"path": g,
"basename": os.path.basename(g),
"nameroot": os.path.splitext(os.path.basename(g))[0],
"nameext": os.path.splitext(os.path.basename(g))[1],
"class": cls}
])
optional = False
single = False
if isinstance(schema["type"], list):
if "null" in schema["type"]:
optional = True
if "File" in schema["type"] or "Directory" in schema["type"]:
single = True
elif schema["type"] == "File" or schema["type"] == "Directory":
single = True
if "outputEval" in binding:
r = builder.do_eval(binding["outputEval"], context=r)
if single:
if not r and not optional:
raise WorkflowException("Did not find output file with glob pattern: '{}'".format(globpatterns))
elif not r and optional:
pass
elif isinstance(r,list):
if len(r) > 1:
raise WorkflowException("Multiple matches for output item that is a single file")
else:
r = r[0]
return r
def simplify_list(maybe_list):
"""Turn a length one list loaded by cwltool into a scalar.
Anything else is passed as-is, by reference."""
if isinstance(maybe_list, MutableSequence):
is_list = aslist(maybe_list)
if len(is_list) == 1:
return is_list[0]
return maybe_list
def simplify_list(l):
"""Turn a length one list loaded by cwltool into a scalar.
Anything else is passed as-is, by reference."""
if isinstance(l, CommentedSeq):
l = aslist(l)
if len(l) == 1:
return l[0]
return l
def upload_job_order(arvrunner, name, tool, job_order):
"""Upload local files referenced in the input object and return updated input
object with 'location' updated to the proper keep references.
"""
# Make a copy of the job order and set defaults.
builder_job_order = copy.copy(job_order)
# fill_in_defaults throws an error if there are any
# missing required parameters, we don't want it to do that
# so make them all optional.
inputs_copy = copy.deepcopy(tool.tool["inputs"])
for i in inputs_copy:
if "null" not in i["type"]:
i["type"] = ["null"] + aslist(i["type"])
fill_in_defaults(inputs_copy,
builder_job_order,
arvrunner.fs_access)
# Need to create a builder object to evaluate expressions.
builder = make_builder(builder_job_order,
tool.hints,
tool.requirements,
ArvRuntimeContext(),
tool.metadata)
# Now update job_order with secondaryFiles
discover_secondary_files(arvrunner.fs_access,
builder,
tool.tool["inputs"],
job_order)
jobmapper = upload_dependencies(arvrunner,
name,
tool.doc_loader,
job_order,
job_order.get("id", "#"),
False)
if "id" in job_order:
del job_order["id"]
# Need to filter this out, gets added by cwltool when providing
# parameters on the command line.
if "job_order" in job_order:
del job_order["job_order"]
return job_order
def run(self, dry_run=False, pull_image=True, **kwargs):
container_request = {
"command": self.command_line,
"owner_uuid": self.arvrunner.project_uuid,
"name": self.name,
"output_path": self.outdir,
"cwd": self.outdir,
"priority": 1,
"state": "Committed",
"properties": {},
}
runtime_constraints = {}
resources = self.builder.resources
if resources is not None:
runtime_constraints["vcpus"] = resources.get("cores", 1)
runtime_constraints["ram"] = resources.get("ram") * 2**20
mounts = {
self.outdir: {
"kind": "tmp",
"capacity": resources.get("outdirSize", 0) * 2**20
},
self.tmpdir: {
"kind": "tmp",
"capacity": resources.get("tmpdirSize", 0) * 2**20
}
}
scheduling_parameters = {}
rf = [
self.pathmapper.mapper(f) for f in self.pathmapper.referenced_files
]
rf.sort(key=lambda k: k.resolved)
prevdir = None
for resolved, target, tp, stg in rf:
if not stg:
continue
if prevdir and target.startswith(prevdir):
continue
if tp == "Directory":
targetdir = target
else:
targetdir = os.path.dirname(target)
sp = resolved.split("/", 1)
pdh = sp[0][5:] # remove "keep:"
mounts[targetdir] = {
"kind": "collection",
"portable_data_hash": pdh
}
if len(sp) == 2:
if tp == "Directory":
path = sp[1]
else:
path = os.path.dirname(sp[1])
if path and path != "/":
mounts[targetdir]["path"] = path
prevdir = targetdir + "/"
with Perf(metrics, "generatefiles %s" % self.name):
if self.generatefiles["listing"]:
vwd = arvados.collection.Collection(
api_client=self.arvrunner.api,
keep_client=self.arvrunner.keep_client,
num_retries=self.arvrunner.num_retries)
generatemapper = NoFollowPathMapper([self.generatefiles],
"",
"",
separateDirs=False)
with Perf(metrics, "createfiles %s" % self.name):
for f, p in generatemapper.items():
if not p.target:
pass
elif p.type in ("File", "Directory"):
source, path = self.arvrunner.fs_access.get_collection(
p.resolved)
vwd.copy(path, p.target, source_collection=source)
elif p.type == "CreateFile":
with vwd.open(p.target, "w") as n:
n.write(p.resolved.encode("utf-8"))
with Perf(metrics, "generatefiles.save_new %s" % self.name):
vwd.save_new()
for f, p in generatemapper.items():
if not p.target:
continue
mountpoint = "%s/%s" % (self.outdir, p.target)
mounts[mountpoint] = {
"kind": "collection",
"portable_data_hash": vwd.portable_data_hash(),
"path": p.target
}
container_request["environment"] = {
"TMPDIR": self.tmpdir,
"HOME": self.outdir
}
if self.environment:
container_request["environment"].update(self.environment)
if self.stdin:
sp = self.stdin[6:].split("/", 1)
mounts["stdin"] = {
"kind": "collection",
"portable_data_hash": sp[0],
"path": sp[1]
}
if self.stderr:
mounts["stderr"] = {
"kind": "file",
"path": "%s/%s" % (self.outdir, self.stderr)
}
if self.stdout:
mounts["stdout"] = {
"kind": "file",
"path": "%s/%s" % (self.outdir, self.stdout)
}
(docker_req, docker_is_req) = get_feature(self, "DockerRequirement")
if not docker_req:
docker_req = {"dockerImageId": "arvados/jobs"}
container_request["container_image"] = arv_docker_get_image(
self.arvrunner.api, docker_req, pull_image,
self.arvrunner.project_uuid)
api_req, _ = get_feature(self, "http://arvados.org/cwl#APIRequirement")
if api_req:
runtime_constraints["API"] = True
runtime_req, _ = get_feature(
self, "http://arvados.org/cwl#RuntimeConstraints")
if runtime_req:
if "keep_cache" in runtime_req:
runtime_constraints[
"keep_cache_ram"] = runtime_req["keep_cache"] * 2**20
if "outputDirType" in runtime_req:
if runtime_req["outputDirType"] == "local_output_dir":
# Currently the default behavior.
pass
elif runtime_req["outputDirType"] == "keep_output_dir":
mounts[self.outdir] = {
"kind": "collection",
"writable": True
}
partition_req, _ = get_feature(
self, "http://arvados.org/cwl#PartitionRequirement")
if partition_req:
scheduling_parameters["partitions"] = aslist(
partition_req["partition"])
intermediate_output_req, _ = get_feature(
self, "http://arvados.org/cwl#IntermediateOutput")
if intermediate_output_req:
self.output_ttl = intermediate_output_req["outputTTL"]
else:
self.output_ttl = self.arvrunner.intermediate_output_ttl
if self.output_ttl < 0:
raise WorkflowError(
"Invalid value %d for output_ttl, cannot be less than zero" %
container_request["output_ttl"])
container_request["output_ttl"] = self.output_ttl
container_request["mounts"] = mounts
container_request["runtime_constraints"] = runtime_constraints
container_request["use_existing"] = kwargs.get("enable_reuse", True)
container_request["scheduling_parameters"] = scheduling_parameters
if kwargs.get("runnerjob", "").startswith("arvwf:"):
wfuuid = kwargs["runnerjob"][6:kwargs["runnerjob"].index("#")]
wfrecord = self.arvrunner.api.workflows().get(uuid=wfuuid).execute(
num_retries=self.arvrunner.num_retries)
if container_request["name"] == "main":
container_request["name"] = wfrecord["name"]
container_request["properties"]["template_uuid"] = wfuuid
try:
response = self.arvrunner.api.container_requests().create(
body=container_request).execute(
num_retries=self.arvrunner.num_retries)
self.uuid = response["uuid"]
self.arvrunner.processes[self.uuid] = self
if response["state"] == "Final":
logger.info("%s reused container %s",
self.arvrunner.label(self),
response["container_uuid"])
self.done(response)
else:
logger.info("%s %s state is %s", self.arvrunner.label(self),
response["uuid"], response["state"])
except Exception as e:
logger.error("%s got error %s" %
(self.arvrunner.label(self), str(e)))
self.output_callback({}, "permanentFail")
def run(self, fileStore):
cwljob = resolve_indirect(self.cwljob)
# `promises` dict
# from: each parameter (workflow input or step output)
# that may be used as a "source" for a step input workflow output
# parameter
# to: the job that will produce that value.
promises = {}
# `jobs` dict from step id to job that implements that step.
jobs = {}
for inp in self.cwlwf.tool["inputs"]:
promises[inp["id"]] = SelfJob(self, cwljob)
alloutputs_fufilled = False
while not alloutputs_fufilled:
# Iteratively go over the workflow steps, scheduling jobs as their
# dependencies can be fufilled by upstream workflow inputs or
# step outputs. Loop exits when the workflow outputs
# are satisfied.
alloutputs_fufilled = True
for step in self.cwlwf.steps:
if step.tool["id"] not in jobs:
stepinputs_fufilled = True
for inp in step.tool["inputs"]:
if "source" in inp:
for s in aslist(inp["source"]):
if s not in promises:
stepinputs_fufilled = False
if stepinputs_fufilled:
jobobj = {}
for inp in step.tool["inputs"]:
key = shortname(inp["id"])
if "source" in inp:
if inp.get("linkMerge") or len(aslist(inp["source"])) > 1:
linkMerge = inp.get("linkMerge", "merge_nested")
if linkMerge == "merge_nested":
jobobj[key] = (
MergeInputsNested([(shortname(s), promises[s].rv())
for s in aslist(inp["source"])]))
elif linkMerge == "merge_flattened":
jobobj[key] = (
MergeInputsFlattened([(shortname(s), promises[s].rv())
for s in aslist(inp["source"])]))
else:
raise validate.ValidationException(
"Unsupported linkMerge '%s'", linkMerge)
else:
jobobj[key] = (shortname(inp["source"]),
promises[inp["source"]].rv())
elif "default" in inp:
d = copy.copy(inp["default"])
jobobj[key] = ("default", {"default": d})
if "valueFrom" in inp and "scatter" not in step.tool:
if key in jobobj:
jobobj[key] = StepValueFrom(inp["valueFrom"],
jobobj[key],
self.cwlwf.requirements)
else:
jobobj[key] = StepValueFrom(inp["valueFrom"],
("None", {"None": None}),
self.cwlwf.requirements)
if "scatter" in step.tool:
wfjob = CWLScatter(step, IndirectDict(jobobj), **self.executor_options)
followOn = CWLGather(step, wfjob.rv())
wfjob.addFollowOn(followOn)
else:
(wfjob, followOn) = makeJob(step.embedded_tool, IndirectDict(jobobj),
step_inputs=step.tool["inputs"],
**self.executor_options)
jobs[step.tool["id"]] = followOn
connected = False
for inp in step.tool["inputs"]:
for s in aslist(inp.get("source", [])):
if not promises[s].hasChild(wfjob):
promises[s].addChild(wfjob)
connected = True
if not connected:
# workflow step has default inputs only, isn't connected to other jobs,
# so add it as child of workflow.
self.addChild(wfjob)
for out in step.tool["outputs"]:
promises[out["id"]] = followOn
for inp in step.tool["inputs"]:
for s in aslist(inp.get("source", [])):
if s not in promises:
alloutputs_fufilled = False
# may need a test
for out in self.cwlwf.tool["outputs"]:
if "source" in out:
if out["source"] not in promises:
alloutputs_fufilled = False
outobj = {}
for out in self.cwlwf.tool["outputs"]:
outobj[shortname(out["id"])] = (shortname(out["outputSource"]), promises[out["outputSource"]].rv())
return IndirectDict(outobj)
def run(self, runtimeContext):
# ArvadosCommandTool subclasses from cwltool.CommandLineTool,
# which calls makeJobRunner() to get a new ArvadosContainer
# object. The fields that define execution such as
# command_line, environment, etc are set on the
# ArvadosContainer object by CommandLineTool.job() before
# run() is called.
runtimeContext = self.job_runtime
container_request = {
"command": self.command_line,
"name": self.name,
"output_path": self.outdir,
"cwd": self.outdir,
"priority": runtimeContext.priority,
"state": "Committed",
"properties": {},
}
runtime_constraints = {}
if runtimeContext.project_uuid:
container_request["owner_uuid"] = runtimeContext.project_uuid
if self.arvrunner.secret_store.has_secret(self.command_line):
raise WorkflowException(
"Secret material leaked on command line, only file literals may contain secrets"
)
if self.arvrunner.secret_store.has_secret(self.environment):
raise WorkflowException(
"Secret material leaked in environment, only file literals may contain secrets"
)
resources = self.builder.resources
if resources is not None:
runtime_constraints["vcpus"] = math.ceil(resources.get("cores", 1))
runtime_constraints["ram"] = math.ceil(
resources.get("ram") * 2**20)
mounts = {
self.outdir: {
"kind": "tmp",
"capacity": math.ceil(resources.get("outdirSize", 0) * 2**20)
},
self.tmpdir: {
"kind": "tmp",
"capacity": math.ceil(resources.get("tmpdirSize", 0) * 2**20)
}
}
secret_mounts = {}
scheduling_parameters = {}
rf = [
self.pathmapper.mapper(f) for f in self.pathmapper.referenced_files
]
rf.sort(key=lambda k: k.resolved)
prevdir = None
for resolved, target, tp, stg in rf:
if not stg:
continue
if prevdir and target.startswith(prevdir):
continue
if tp == "Directory":
targetdir = target
else:
targetdir = os.path.dirname(target)
sp = resolved.split("/", 1)
pdh = sp[0][5:] # remove "keep:"
mounts[targetdir] = {
"kind": "collection",
"portable_data_hash": pdh
}
if pdh in self.pathmapper.pdh_to_uuid:
mounts[targetdir]["uuid"] = self.pathmapper.pdh_to_uuid[pdh]
if len(sp) == 2:
if tp == "Directory":
path = sp[1]
else:
path = os.path.dirname(sp[1])
if path and path != "/":
mounts[targetdir]["path"] = path
prevdir = targetdir + "/"
with Perf(metrics, "generatefiles %s" % self.name):
if self.generatefiles["listing"]:
vwd = arvados.collection.Collection(
api_client=self.arvrunner.api,
keep_client=self.arvrunner.keep_client,
num_retries=self.arvrunner.num_retries)
generatemapper = NoFollowPathMapper(
self.generatefiles["listing"], "", "", separateDirs=False)
sorteditems = sorted(generatemapper.items(),
key=lambda n: n[1].target)
logger.debug("generatemapper is %s", sorteditems)
with Perf(metrics, "createfiles %s" % self.name):
for f, p in sorteditems:
if not p.target:
continue
if p.target.startswith("/"):
dst = p.target[len(self.outdir) +
1:] if p.target.startswith(
self.outdir +
"/") else p.target[1:]
else:
dst = p.target
if p.type in ("File", "Directory", "WritableFile",
"WritableDirectory"):
if p.resolved.startswith("_:"):
vwd.mkdirs(dst)
else:
source, path = self.arvrunner.fs_access.get_collection(
p.resolved)
vwd.copy(path or ".",
dst,
source_collection=source)
elif p.type == "CreateFile":
if self.arvrunner.secret_store.has_secret(
p.resolved):
mountpoint = p.target if p.target.startswith(
"/") else os.path.join(
self.outdir, p.target)
secret_mounts[mountpoint] = {
"kind":
"text",
"content":
self.arvrunner.secret_store.retrieve(
p.resolved)
}
else:
with vwd.open(dst, "w") as n:
n.write(p.resolved)
def keepemptydirs(p):
if isinstance(p, arvados.collection.RichCollectionBase):
if len(p) == 0:
p.open(".keep", "w").close()
else:
for c in p:
keepemptydirs(p[c])
keepemptydirs(vwd)
if not runtimeContext.current_container:
runtimeContext.current_container = arvados_cwl.util.get_current_container(
self.arvrunner.api, self.arvrunner.num_retries, logger)
info = arvados_cwl.util.get_intermediate_collection_info(
self.name, runtimeContext.current_container,
runtimeContext.intermediate_output_ttl)
vwd.save_new(name=info["name"],
owner_uuid=runtimeContext.project_uuid,
ensure_unique_name=True,
trash_at=info["trash_at"],
properties=info["properties"])
prev = None
for f, p in sorteditems:
if (not p.target or self.arvrunner.secret_store.has_secret(
p.resolved) or
(prev is not None and p.target.startswith(prev))):
continue
if p.target.startswith("/"):
dst = p.target[len(self.outdir) +
1:] if p.target.startswith(
self.outdir + "/") else p.target[1:]
else:
dst = p.target
mountpoint = p.target if p.target.startswith(
"/") else os.path.join(self.outdir, p.target)
mounts[mountpoint] = {
"kind": "collection",
"portable_data_hash": vwd.portable_data_hash(),
"path": dst
}
if p.type.startswith("Writable"):
mounts[mountpoint]["writable"] = True
prev = p.target + "/"
container_request["environment"] = {
"TMPDIR": self.tmpdir,
"HOME": self.outdir
}
if self.environment:
container_request["environment"].update(self.environment)
if self.stdin:
sp = self.stdin[6:].split("/", 1)
mounts["stdin"] = {
"kind": "collection",
"portable_data_hash": sp[0],
"path": sp[1]
}
if self.stderr:
mounts["stderr"] = {
"kind": "file",
"path": "%s/%s" % (self.outdir, self.stderr)
}
if self.stdout:
mounts["stdout"] = {
"kind": "file",
"path": "%s/%s" % (self.outdir, self.stdout)
}
(docker_req, docker_is_req) = self.get_requirement("DockerRequirement")
if not docker_req:
docker_req = {"dockerImageId": "arvados/jobs:" + __version__}
container_request["container_image"] = arv_docker_get_image(
self.arvrunner.api, docker_req, runtimeContext.pull_image,
runtimeContext.project_uuid, runtimeContext.force_docker_pull,
runtimeContext.tmp_outdir_prefix)
network_req, _ = self.get_requirement("NetworkAccess")
if network_req:
runtime_constraints["API"] = network_req["networkAccess"]
api_req, _ = self.get_requirement(
"http://arvados.org/cwl#APIRequirement")
if api_req:
runtime_constraints["API"] = True
runtime_req, _ = self.get_requirement(
"http://arvados.org/cwl#RuntimeConstraints")
if runtime_req:
if "keep_cache" in runtime_req:
runtime_constraints["keep_cache_ram"] = math.ceil(
runtime_req["keep_cache"] * 2**20)
if "outputDirType" in runtime_req:
if runtime_req["outputDirType"] == "local_output_dir":
# Currently the default behavior.
pass
elif runtime_req["outputDirType"] == "keep_output_dir":
mounts[self.outdir] = {
"kind": "collection",
"writable": True
}
partition_req, _ = self.get_requirement(
"http://arvados.org/cwl#PartitionRequirement")
if partition_req:
scheduling_parameters["partitions"] = aslist(
partition_req["partition"])
intermediate_output_req, _ = self.get_requirement(
"http://arvados.org/cwl#IntermediateOutput")
if intermediate_output_req:
self.output_ttl = intermediate_output_req["outputTTL"]
else:
self.output_ttl = self.arvrunner.intermediate_output_ttl
if self.output_ttl < 0:
raise WorkflowException(
"Invalid value %d for output_ttl, cannot be less than zero" %
container_request["output_ttl"])
if self.timelimit is not None and self.timelimit > 0:
scheduling_parameters["max_run_time"] = self.timelimit
extra_submit_params = {}
if runtimeContext.submit_runner_cluster:
extra_submit_params[
"cluster_id"] = runtimeContext.submit_runner_cluster
container_request["output_name"] = "Output for step %s" % (self.name)
container_request["output_ttl"] = self.output_ttl
container_request["mounts"] = mounts
container_request["secret_mounts"] = secret_mounts
container_request["runtime_constraints"] = runtime_constraints
container_request["scheduling_parameters"] = scheduling_parameters
enable_reuse = runtimeContext.enable_reuse
if enable_reuse:
reuse_req, _ = self.get_requirement("WorkReuse")
if reuse_req:
enable_reuse = reuse_req["enableReuse"]
reuse_req, _ = self.get_requirement(
"http://arvados.org/cwl#ReuseRequirement")
if reuse_req:
enable_reuse = reuse_req["enableReuse"]
container_request["use_existing"] = enable_reuse
if runtimeContext.runnerjob.startswith("arvwf:"):
wfuuid = runtimeContext.runnerjob[6:runtimeContext.runnerjob.
index("#")]
wfrecord = self.arvrunner.api.workflows().get(uuid=wfuuid).execute(
num_retries=self.arvrunner.num_retries)
if container_request["name"] == "main":
container_request["name"] = wfrecord["name"]
container_request["properties"]["template_uuid"] = wfuuid
self.output_callback = self.arvrunner.get_wrapped_callback(
self.output_callback)
try:
if runtimeContext.submit_request_uuid:
response = self.arvrunner.api.container_requests().update(
uuid=runtimeContext.submit_request_uuid,
body=container_request,
**extra_submit_params).execute(
num_retries=self.arvrunner.num_retries)
else:
response = self.arvrunner.api.container_requests().create(
body=container_request, **extra_submit_params).execute(
num_retries=self.arvrunner.num_retries)
self.uuid = response["uuid"]
self.arvrunner.process_submitted(self)
if response["state"] == "Final":
logger.info("%s reused container %s",
self.arvrunner.label(self),
response["container_uuid"])
else:
logger.info("%s %s state is %s", self.arvrunner.label(self),
response["uuid"], response["state"])
except Exception as e:
logger.exception("%s error submitting container\n%s",
self.arvrunner.label(self), e)
logger.debug("Container request was %s", container_request)
self.output_callback({}, "permanentFail")
def run(self, file_store):
cwljob = resolve_indirect(self.cwljob)
# `promises` dict
# from: each parameter (workflow input or step output)
# that may be used as a "source" for a step input workflow output
# parameter
# to: the job that will produce that value.
promises = {}
# `jobs` dict from step id to job that implements that step.
jobs = {}
for inp in self.cwlwf.tool["inputs"]:
promises[inp["id"]] = SelfJob(self, cwljob)
alloutputs_fufilled = False
while not alloutputs_fufilled:
# Iteratively go over the workflow steps, scheduling jobs as their
# dependencies can be fufilled by upstream workflow inputs or
# step outputs. Loop exits when the workflow outputs
# are satisfied.
alloutputs_fufilled = True
for step in self.cwlwf.steps:
if step.tool["id"] not in jobs:
stepinputs_fufilled = True
for inp in step.tool["inputs"]:
if "source" in inp:
for s in aslist(inp["source"]):
if s not in promises:
stepinputs_fufilled = False
if stepinputs_fufilled:
jobobj = {}
for inp in step.tool["inputs"]:
key = shortname(inp["id"])
if "source" in inp:
if inp.get("linkMerge") \
or len(aslist(inp["source"])) > 1:
linkMerge = inp.get(
"linkMerge", "merge_nested")
if linkMerge == "merge_nested":
jobobj[key] = (
MergeInputsNested(
[(shortname(s),
promises[s].rv())
for s in aslist(
inp["source"])]))
elif linkMerge == "merge_flattened":
jobobj[key] = (
MergeInputsFlattened(
[(shortname(s),
promises[s].rv())
for s in aslist(
inp["source"])]))
else:
raise validate.ValidationException(
"Unsupported linkMerge '%s'" %
linkMerge)
else:
inpSource = inp["source"]
if isinstance(inpSource, MutableSequence):
# It seems that an input source with a
# '#' in the name will be returned as a
# CommentedSeq list by the yaml parser.
inpSource = str(inpSource[0])
jobobj[key] = (shortname(inpSource),
promises[inpSource].rv())
if "default" in inp:
if key in jobobj:
if isinstance(jobobj[key][1], Promise):
d = copy.copy(inp["default"])
jobobj[key] = DefaultWithSource(
d, jobobj[key])
else:
if jobobj[key][1][
jobobj[key][0]] is None:
d = copy.copy(inp["default"])
jobobj[key] = (
"default", {"default": d})
else:
d = copy.copy(inp["default"])
jobobj[key] = ("default", {"default": d})
if "valueFrom" in inp \
and "scatter" not in step.tool:
if key in jobobj:
jobobj[key] = StepValueFrom(
inp["valueFrom"], jobobj[key],
self.cwlwf.requirements)
else:
jobobj[key] = StepValueFrom(
inp["valueFrom"], (
"None", {"None": None}),
self.cwlwf.requirements)
if "scatter" in step.tool:
wfjob = CWLScatter(step, IndirectDict(jobobj),
self.runtime_context)
followOn = CWLGather(step, wfjob.rv())
wfjob.addFollowOn(followOn)
else:
(wfjob, followOn) = makeJob(
step.embedded_tool, IndirectDict(jobobj),
step.tool["inputs"],
self.runtime_context)
jobs[step.tool["id"]] = followOn
connected = False
for inp in step.tool["inputs"]:
for s in aslist(inp.get("source", [])):
if (isinstance(
promises[s], (CWLJobWrapper, CWLGather)
) and
not promises[s].hasFollowOn(wfjob)):
promises[s].addFollowOn(wfjob)
connected = True
if (not isinstance(
promises[s], (CWLJobWrapper, CWLGather)
) and
not promises[s].hasChild(wfjob)):
promises[s].addChild(wfjob)
connected = True
if not connected:
# the workflow step has default inputs only & isn't
# connected to other jobs, so add it as child of
# this workflow.
self.addChild(wfjob)
for out in step.tool["outputs"]:
promises[out["id"]] = followOn
for inp in step.tool["inputs"]:
for source in aslist(inp.get("source", [])):
if source not in promises:
alloutputs_fufilled = False
# may need a test
for out in self.cwlwf.tool["outputs"]:
if "source" in out:
if out["source"] not in promises:
alloutputs_fufilled = False
outobj = {}
for out in self.cwlwf.tool["outputs"]:
key = shortname(out["id"])
if out.get("linkMerge") or len(aslist(out["outputSource"])) > 1:
link_merge = out.get("linkMerge", "merge_nested")
if link_merge == "merge_nested":
outobj[key] = (
MergeInputsNested(
[(shortname(s), promises[s].rv())
for s in aslist(out["outputSource"])]))
elif link_merge == "merge_flattened":
outobj[key] = (
MergeInputsFlattened([
(shortname(s), promises[s].rv())
for s in aslist(out["source"])]))
else:
raise validate.ValidationException(
"Unsupported linkMerge '{}'".format(link_merge))
else:
# A CommentedSeq of length one still appears here rarely -
# not clear why from the CWL code. When it does, it breaks
# the execution by causing a non-hashable type exception.
# We simplify the list into its first (and only) element.
src = simplify_list(out["outputSource"])
outobj[key] = (shortname(src), promises[src].rv())
return IndirectDict(outobj)
def run(self, dry_run=False, pull_image=True, **kwargs):
container_request = {
"command": self.command_line,
"owner_uuid": self.arvrunner.project_uuid,
"name": self.name,
"output_path": self.outdir,
"cwd": self.outdir,
"priority": 1,
"state": "Committed",
"properties": {}
}
runtime_constraints = {}
resources = self.builder.resources
if resources is not None:
runtime_constraints["vcpus"] = resources.get("cores", 1)
runtime_constraints["ram"] = resources.get("ram") * 2**20
mounts = {
self.outdir: {
"kind": "tmp",
"capacity": resources.get("outdirSize", 0) * 2**20
},
self.tmpdir: {
"kind": "tmp",
"capacity": resources.get("tmpdirSize", 0) * 2**20
}
}
scheduling_parameters = {}
rf = [self.pathmapper.mapper(f) for f in self.pathmapper.referenced_files]
rf.sort(key=lambda k: k.resolved)
prevdir = None
for resolved, target, tp, stg in rf:
if not stg:
continue
if prevdir and target.startswith(prevdir):
continue
if tp == "Directory":
targetdir = target
else:
targetdir = os.path.dirname(target)
sp = resolved.split("/", 1)
pdh = sp[0][5:] # remove "keep:"
mounts[targetdir] = {
"kind": "collection",
"portable_data_hash": pdh
}
if len(sp) == 2:
if tp == "Directory":
path = sp[1]
else:
path = os.path.dirname(sp[1])
if path and path != "/":
mounts[targetdir]["path"] = path
prevdir = targetdir + "/"
with Perf(metrics, "generatefiles %s" % self.name):
if self.generatefiles["listing"]:
vwd = arvados.collection.Collection(api_client=self.arvrunner.api,
keep_client=self.arvrunner.keep_client,
num_retries=self.arvrunner.num_retries)
generatemapper = NoFollowPathMapper([self.generatefiles], "", "",
separateDirs=False)
with Perf(metrics, "createfiles %s" % self.name):
for f, p in generatemapper.items():
if not p.target:
pass
elif p.type in ("File", "Directory"):
source, path = self.arvrunner.fs_access.get_collection(p.resolved)
vwd.copy(path, p.target, source_collection=source)
elif p.type == "CreateFile":
with vwd.open(p.target, "w") as n:
n.write(p.resolved.encode("utf-8"))
with Perf(metrics, "generatefiles.save_new %s" % self.name):
vwd.save_new()
for f, p in generatemapper.items():
if not p.target:
continue
mountpoint = "%s/%s" % (self.outdir, p.target)
mounts[mountpoint] = {"kind": "collection",
"portable_data_hash": vwd.portable_data_hash(),
"path": p.target}
container_request["environment"] = {"TMPDIR": self.tmpdir, "HOME": self.outdir}
if self.environment:
container_request["environment"].update(self.environment)
if self.stdin:
sp = self.stdin[6:].split("/", 1)
mounts["stdin"] = {"kind": "collection",
"portable_data_hash": sp[0],
"path": sp[1]}
if self.stderr:
mounts["stderr"] = {"kind": "file",
"path": "%s/%s" % (self.outdir, self.stderr)}
if self.stdout:
mounts["stdout"] = {"kind": "file",
"path": "%s/%s" % (self.outdir, self.stdout)}
(docker_req, docker_is_req) = get_feature(self, "DockerRequirement")
if not docker_req:
docker_req = {"dockerImageId": "arvados/jobs"}
container_request["container_image"] = arv_docker_get_image(self.arvrunner.api,
docker_req,
pull_image,
self.arvrunner.project_uuid)
api_req, _ = get_feature(self, "http://arvados.org/cwl#APIRequirement")
if api_req:
runtime_constraints["API"] = True
runtime_req, _ = get_feature(self, "http://arvados.org/cwl#RuntimeConstraints")
if runtime_req:
if "keep_cache" in runtime_req:
runtime_constraints["keep_cache_ram"] = runtime_req["keep_cache"] * 2**20
if "outputDirType" in runtime_req:
if runtime_req["outputDirType"] == "local_output_dir":
# Currently the default behavior.
pass
elif runtime_req["outputDirType"] == "keep_output_dir":
mounts[self.outdir]= {
"kind": "collection",
"writable": True
}
partition_req, _ = get_feature(self, "http://arvados.org/cwl#PartitionRequirement")
if partition_req:
scheduling_parameters["partitions"] = aslist(partition_req["partition"])
container_request["mounts"] = mounts
container_request["runtime_constraints"] = runtime_constraints
container_request["use_existing"] = kwargs.get("enable_reuse", True)
container_request["scheduling_parameters"] = scheduling_parameters
if kwargs.get("runnerjob", "").startswith("arvwf:"):
wfuuid = kwargs["runnerjob"][6:kwargs["runnerjob"].index("#")]
wfrecord = self.arvrunner.api.workflows().get(uuid=wfuuid).execute(num_retries=self.arvrunner.num_retries)
if container_request["name"] == "main":
container_request["name"] = wfrecord["name"]
container_request["properties"]["template_uuid"] = wfuuid
try:
response = self.arvrunner.api.container_requests().create(
body=container_request
).execute(num_retries=self.arvrunner.num_retries)
self.uuid = response["uuid"]
self.arvrunner.processes[self.uuid] = self
if response["state"] == "Final":
logger.info("%s reused container %s", self.arvrunner.label(self), response["container_uuid"])
self.done(response)
else:
logger.info("%s %s state is %s", self.arvrunner.label(self), response["uuid"], response["state"])
except Exception as e:
logger.error("%s got error %s" % (self.arvrunner.label(self), str(e)))
self.output_callback({}, "permanentFail")
def run(self, dry_run=False, pull_image=True, **kwargs):
container_request = {
"command": self.command_line,
"owner_uuid": self.arvrunner.project_uuid,
"name": self.name,
"output_path": self.outdir,
"cwd": self.outdir,
"priority": 1,
"state": "Committed",
"properties": {}
}
runtime_constraints = {}
mounts = {self.outdir: {"kind": "tmp"}}
scheduling_parameters = {}
dirs = set()
for f in self.pathmapper.files():
_, p, tp = self.pathmapper.mapper(f)
if tp == "Directory" and '/' not in p[6:]:
mounts[p] = {"kind": "collection", "portable_data_hash": p[6:]}
dirs.add(p[6:])
for f in self.pathmapper.files():
_, p, tp = self.pathmapper.mapper(f)
if p[6:].split("/")[0] not in dirs:
mounts[p] = {"kind": "collection", "portable_data_hash": p[6:]}
if self.generatefiles["listing"]:
raise UnsupportedRequirement(
"InitialWorkDirRequirement not supported with --api=containers"
)
container_request["environment"] = {
"TMPDIR": self.tmpdir,
"HOME": self.outdir
}
if self.environment:
container_request["environment"].update(self.environment)
if self.stdin:
raise UnsupportedRequirement(
"Stdin redirection currently not suppported")
if self.stderr:
raise UnsupportedRequirement(
"Stderr redirection currently not suppported")
if self.stdout:
mounts["stdout"] = {
"kind": "file",
"path": "%s/%s" % (self.outdir, self.stdout)
}
(docker_req, docker_is_req) = get_feature(self, "DockerRequirement")
if not docker_req:
docker_req = {"dockerImageId": arvados_jobs_image(self.arvrunner)}
container_request["container_image"] = arv_docker_get_image(
self.arvrunner.api, docker_req, pull_image,
self.arvrunner.project_uuid)
resources = self.builder.resources
if resources is not None:
runtime_constraints["vcpus"] = resources.get("cores", 1)
runtime_constraints["ram"] = resources.get("ram") * 2**20
api_req, _ = get_feature(self, "http://arvados.org/cwl#APIRequirement")
if api_req:
runtime_constraints["API"] = True
runtime_req, _ = get_feature(
self, "http://arvados.org/cwl#RuntimeConstraints")
if runtime_req:
if "keep_cache" in runtime_req:
runtime_constraints["keep_cache_ram"] = runtime_req[
"keep_cache"]
partition_req, _ = get_feature(
self, "http://arvados.org/cwl#PartitionRequirement")
if partition_req:
scheduling_parameters["partitions"] = aslist(
partition_req["partition"])
container_request["mounts"] = mounts
container_request["runtime_constraints"] = runtime_constraints
container_request["use_existing"] = kwargs.get("enable_reuse", True)
container_request["scheduling_parameters"] = scheduling_parameters
if kwargs.get("runnerjob", "").startswith("arvwf:"):
wfuuid = kwargs["runnerjob"][6:kwargs["runnerjob"].index("#")]
wfrecord = self.arvrunner.api.workflows().get(uuid=wfuuid).execute(
num_retries=self.arvrunner.num_retries)
if container_request["name"] == "main":
container_request["name"] = wfrecord["name"]
container_request["properties"]["template_uuid"] = wfuuid
try:
response = self.arvrunner.api.container_requests().create(
body=container_request).execute(
num_retries=self.arvrunner.num_retries)
self.uuid = response["uuid"]
self.arvrunner.processes[self.uuid] = self
logger.info("Container request %s (%s) state is %s", self.name,
response["uuid"], response["state"])
if response["state"] == "Final":
self.done(response)
except Exception as e:
logger.error("Got error %s" % str(e))
self.output_callback({}, "permanentFail")
def run(self, dry_run=False, pull_image=True, **kwargs):
container_request = {
"command": self.command_line,
"owner_uuid": self.arvrunner.project_uuid,
"name": self.name,
"output_path": self.outdir,
"cwd": self.outdir,
"priority": 1,
"state": "Committed",
"properties": {}
}
runtime_constraints = {}
mounts = {self.outdir: {"kind": "tmp"}}
scheduling_parameters = {}
dirs = set()
for f in self.pathmapper.files():
pdh, p, tp = self.pathmapper.mapper(f)
if tp == "Directory" and '/' not in pdh:
mounts[p] = {
"kind": "collection",
"portable_data_hash": pdh[5:]
}
dirs.add(pdh)
for f in self.pathmapper.files():
res, p, tp = self.pathmapper.mapper(f)
if res.startswith("keep:"):
res = res[5:]
elif res.startswith("/keep/"):
res = res[6:]
else:
continue
sp = res.split("/", 1)
pdh = sp[0]
if pdh not in dirs:
mounts[p] = {"kind": "collection", "portable_data_hash": pdh}
if len(sp) == 2:
mounts[p]["path"] = sp[1]
with Perf(metrics, "generatefiles %s" % self.name):
if self.generatefiles["listing"]:
vwd = arvados.collection.Collection(
api_client=self.arvrunner.api,
keep_client=self.arvrunner.keep_client,
num_retries=self.arvrunner.num_retries)
generatemapper = NoFollowPathMapper([self.generatefiles],
"",
"",
separateDirs=False)
with Perf(metrics, "createfiles %s" % self.name):
for f, p in generatemapper.items():
if not p.target:
pass
elif p.type in ("File", "Directory"):
source, path = self.arvrunner.fs_access.get_collection(
p.resolved)
vwd.copy(path, p.target, source_collection=source)
elif p.type == "CreateFile":
with vwd.open(p.target, "w") as n:
n.write(p.resolved.encode("utf-8"))
with Perf(metrics, "generatefiles.save_new %s" % self.name):
vwd.save_new()
for f, p in generatemapper.items():
if not p.target:
continue
mountpoint = "%s/%s" % (self.outdir, p.target)
mounts[mountpoint] = {
"kind": "collection",
"portable_data_hash": vwd.portable_data_hash(),
"path": p.target
}
container_request["environment"] = {
"TMPDIR": self.tmpdir,
"HOME": self.outdir
}
if self.environment:
container_request["environment"].update(self.environment)
if self.stdin:
raise UnsupportedRequirement(
"Stdin redirection currently not suppported")
if self.stderr:
raise UnsupportedRequirement(
"Stderr redirection currently not suppported")
if self.stdout:
mounts["stdout"] = {
"kind": "file",
"path": "%s/%s" % (self.outdir, self.stdout)
}
(docker_req, docker_is_req) = get_feature(self, "DockerRequirement")
if not docker_req:
docker_req = {"dockerImageId": "arvados/jobs"}
container_request["container_image"] = arv_docker_get_image(
self.arvrunner.api, docker_req, pull_image,
self.arvrunner.project_uuid)
resources = self.builder.resources
if resources is not None:
runtime_constraints["vcpus"] = resources.get("cores", 1)
runtime_constraints["ram"] = resources.get("ram") * 2**20
api_req, _ = get_feature(self, "http://arvados.org/cwl#APIRequirement")
if api_req:
runtime_constraints["API"] = True
runtime_req, _ = get_feature(
self, "http://arvados.org/cwl#RuntimeConstraints")
if runtime_req:
if "keep_cache" in runtime_req:
runtime_constraints[
"keep_cache_ram"] = runtime_req["keep_cache"] * 2**20
partition_req, _ = get_feature(
self, "http://arvados.org/cwl#PartitionRequirement")
if partition_req:
scheduling_parameters["partitions"] = aslist(
partition_req["partition"])
container_request["mounts"] = mounts
container_request["runtime_constraints"] = runtime_constraints
container_request["use_existing"] = kwargs.get("enable_reuse", True)
container_request["scheduling_parameters"] = scheduling_parameters
if kwargs.get("runnerjob", "").startswith("arvwf:"):
wfuuid = kwargs["runnerjob"][6:kwargs["runnerjob"].index("#")]
wfrecord = self.arvrunner.api.workflows().get(uuid=wfuuid).execute(
num_retries=self.arvrunner.num_retries)
if container_request["name"] == "main":
container_request["name"] = wfrecord["name"]
container_request["properties"]["template_uuid"] = wfuuid
try:
response = self.arvrunner.api.container_requests().create(
body=container_request).execute(
num_retries=self.arvrunner.num_retries)
self.uuid = response["uuid"]
self.arvrunner.processes[self.uuid] = self
if response["state"] == "Final":
logger.info("%s reused container %s",
self.arvrunner.label(self),
response["container_uuid"])
self.done(response)
else:
logger.info("%s %s state is %s", self.arvrunner.label(self),
response["uuid"], response["state"])
except Exception as e:
logger.error("%s got error %s" %
(self.arvrunner.label(self), str(e)))
self.output_callback({}, "permanentFail")
def arv_executor(self,
updated_tool,
job_order,
runtimeContext,
logger=None):
self.debug = runtimeContext.debug
workbench1 = self.api.config()["Services"]["Workbench1"]["ExternalURL"]
workbench2 = self.api.config()["Services"]["Workbench2"]["ExternalURL"]
controller = self.api.config()["Services"]["Controller"]["ExternalURL"]
logger.info("Using cluster %s (%s)",
self.api.config()["ClusterID"], workbench2 or workbench1
or controller)
updated_tool.visit(self.check_features)
self.project_uuid = runtimeContext.project_uuid
self.pipeline = None
self.fs_access = runtimeContext.make_fs_access(runtimeContext.basedir)
self.secret_store = runtimeContext.secret_store
self.trash_intermediate = runtimeContext.trash_intermediate
if self.trash_intermediate and self.work_api != "containers":
raise Exception(
"--trash-intermediate is only supported with --api=containers."
)
self.intermediate_output_ttl = runtimeContext.intermediate_output_ttl
if self.intermediate_output_ttl and self.work_api != "containers":
raise Exception(
"--intermediate-output-ttl is only supported with --api=containers."
)
if self.intermediate_output_ttl < 0:
raise Exception(
"Invalid value %d for --intermediate-output-ttl, cannot be less than zero"
% self.intermediate_output_ttl)
if runtimeContext.submit_request_uuid and self.work_api != "containers":
raise Exception(
"--submit-request-uuid requires containers API, but using '{}' api"
.format(self.work_api))
default_storage_classes = ",".join([
k for k, v in self.api.config().get("StorageClasses", {
"default": {
"Default": True
}
}).items() if v.get("Default") is True
])
if runtimeContext.storage_classes == "default":
runtimeContext.storage_classes = default_storage_classes
if runtimeContext.intermediate_storage_classes == "default":
runtimeContext.intermediate_storage_classes = default_storage_classes
if not runtimeContext.name:
runtimeContext.name = self.name = updated_tool.tool.get(
"label") or updated_tool.metadata.get(
"label") or os.path.basename(updated_tool.tool["id"])
# Upload local file references in the job order.
job_order = upload_job_order(self, "%s input" % runtimeContext.name,
updated_tool, job_order)
# the last clause means: if it is a command line tool, and we
# are going to wait for the result, and always_submit_runner
# is false, then we don't submit a runner process.
submitting = (runtimeContext.update_workflow
or runtimeContext.create_workflow or
(runtimeContext.submit
and not (updated_tool.tool["class"] == "CommandLineTool"
and runtimeContext.wait
and not runtimeContext.always_submit_runner)))
loadingContext = self.loadingContext.copy()
loadingContext.do_validate = False
if submitting:
loadingContext.do_update = False
# Document may have been auto-updated. Reload the original
# document with updating disabled because we want to
# submit the document with its original CWL version, not
# the auto-updated one.
tool = load_tool(updated_tool.tool["id"], loadingContext)
else:
tool = updated_tool
# Upload direct dependencies of workflow steps, get back mapping of files to keep references.
# Also uploads docker images.
merged_map = upload_workflow_deps(self, tool)
# Recreate process object (ArvadosWorkflow or
# ArvadosCommandTool) because tool document may have been
# updated by upload_workflow_deps in ways that modify
# inheritance of hints or requirements.
loadingContext.loader = tool.doc_loader
loadingContext.avsc_names = tool.doc_schema
loadingContext.metadata = tool.metadata
tool = load_tool(tool.tool, loadingContext)
existing_uuid = runtimeContext.update_workflow
if existing_uuid or runtimeContext.create_workflow:
# Create a pipeline template or workflow record and exit.
if self.work_api == "containers":
uuid = upload_workflow(
self,
tool,
job_order,
self.project_uuid,
uuid=existing_uuid,
submit_runner_ram=runtimeContext.submit_runner_ram,
name=runtimeContext.name,
merged_map=merged_map,
submit_runner_image=runtimeContext.submit_runner_image)
self.stdout.write(uuid + "\n")
return (None, "success")
self.apply_reqs(job_order, tool)
self.ignore_docker_for_reuse = runtimeContext.ignore_docker_for_reuse
self.eval_timeout = runtimeContext.eval_timeout
runtimeContext = runtimeContext.copy()
runtimeContext.use_container = True
runtimeContext.tmpdir_prefix = "tmp"
runtimeContext.work_api = self.work_api
if self.work_api == "containers":
if self.ignore_docker_for_reuse:
raise Exception(
"--ignore-docker-for-reuse not supported with containers API."
)
runtimeContext.outdir = "/var/spool/cwl"
runtimeContext.docker_outdir = "/var/spool/cwl"
runtimeContext.tmpdir = "/tmp"
runtimeContext.docker_tmpdir = "/tmp"
if runtimeContext.priority < 1 or runtimeContext.priority > 1000:
raise Exception("--priority must be in the range 1..1000.")
if self.should_estimate_cache_size:
visited = set()
estimated_size = [0]
def estimate_collection_cache(obj):
if obj.get("location", "").startswith("keep:"):
m = pdh_size.match(obj["location"][5:])
if m and m.group(1) not in visited:
visited.add(m.group(1))
estimated_size[0] += int(m.group(2))
visit_class(job_order, ("File", "Directory"),
estimate_collection_cache)
runtimeContext.collection_cache_size = max(
((estimated_size[0] * 192) // (1024 * 1024)) + 1, 256)
self.collection_cache.set_cap(
runtimeContext.collection_cache_size * 1024 * 1024)
logger.info("Using collection cache size %s MiB",
runtimeContext.collection_cache_size)
runnerjob = None
if runtimeContext.submit:
# Submit a runner job to run the workflow for us.
if self.work_api == "containers":
if submitting:
tool = RunnerContainer(
self,
updated_tool,
tool,
loadingContext,
runtimeContext.enable_reuse,
self.output_name,
self.output_tags,
submit_runner_ram=runtimeContext.submit_runner_ram,
name=runtimeContext.name,
on_error=runtimeContext.on_error,
submit_runner_image=runtimeContext.submit_runner_image,
intermediate_output_ttl=runtimeContext.
intermediate_output_ttl,
merged_map=merged_map,
priority=runtimeContext.priority,
secret_store=self.secret_store,
collection_cache_size=runtimeContext.
collection_cache_size,
collection_cache_is_default=self.
should_estimate_cache_size)
else:
runtimeContext.runnerjob = tool.tool["id"]
if runtimeContext.cwl_runner_job is not None:
self.uuid = runtimeContext.cwl_runner_job.get('uuid')
jobiter = tool.job(job_order, self.output_callback, runtimeContext)
if runtimeContext.submit and not runtimeContext.wait:
runnerjob = next(jobiter)
runnerjob.run(runtimeContext)
self.stdout.write(runnerjob.uuid + "\n")
return (None, "success")
current_container = arvados_cwl.util.get_current_container(
self.api, self.num_retries, logger)
if current_container:
logger.info("Running inside container %s",
current_container.get("uuid"))
self.poll_api = arvados.api('v1', timeout=runtimeContext.http_timeout)
self.polling_thread = threading.Thread(target=self.poll_states)
self.polling_thread.start()
self.task_queue = TaskQueue(self.workflow_eval_lock, self.thread_count)
try:
self.workflow_eval_lock.acquire()
# Holds the lock while this code runs and releases it when
# it is safe to do so in self.workflow_eval_lock.wait(),
# at which point on_message can update job state and
# process output callbacks.
loopperf = Perf(metrics, "jobiter")
loopperf.__enter__()
for runnable in jobiter:
loopperf.__exit__()
if self.stop_polling.is_set():
break
if self.task_queue.error is not None:
raise self.task_queue.error
if runnable:
with Perf(metrics, "run"):
self.start_run(runnable, runtimeContext)
else:
if (self.task_queue.in_flight + len(self.processes)) > 0:
self.workflow_eval_lock.wait(3)
else:
logger.error(
"Workflow is deadlocked, no runnable processes and not waiting on any pending processes."
)
break
if self.stop_polling.is_set():
break
loopperf.__enter__()
loopperf.__exit__()
while (self.task_queue.in_flight + len(self.processes)) > 0:
if self.task_queue.error is not None:
raise self.task_queue.error
self.workflow_eval_lock.wait(3)
except UnsupportedRequirement:
raise
except:
if sys.exc_info()[0] is KeyboardInterrupt or sys.exc_info(
)[0] is SystemExit:
logger.error("Interrupted, workflow will be cancelled")
elif isinstance(sys.exc_info()[1], WorkflowException):
logger.error(
"Workflow execution failed:\n%s",
sys.exc_info()[1],
exc_info=(sys.exc_info()[1] if self.debug else False))
else:
logger.exception("Workflow execution failed")
if self.pipeline:
self.api.pipeline_instances().update(
uuid=self.pipeline["uuid"], body={
"state": "Failed"
}).execute(num_retries=self.num_retries)
if self.work_api == "containers" and not current_container:
# Not running in a crunch container, so cancel any outstanding processes.
for p in self.processes:
try:
self.api.container_requests().update(
uuid=p, body={
"priority": "0"
}).execute(num_retries=self.num_retries)
except Exception:
pass
finally:
self.workflow_eval_lock.release()
self.task_queue.drain()
self.stop_polling.set()
self.polling_thread.join()
self.task_queue.join()
if self.final_status == "UnsupportedRequirement":
raise UnsupportedRequirement("Check log for details.")
if self.final_output is None:
raise WorkflowException("Workflow did not return a result.")
if runtimeContext.submit and isinstance(tool, Runner):
logger.info("Final output collection %s", tool.final_output)
if workbench2 or workbench1:
logger.info("Output at %scollections/%s", workbench2
or workbench1, tool.final_output)
else:
if self.output_name is None:
self.output_name = "Output of %s" % (shortname(
tool.tool["id"]))
if self.output_tags is None:
self.output_tags = ""
storage_classes = ""
storage_class_req, _ = tool.get_requirement(
"http://arvados.org/cwl#OutputStorageClass")
if storage_class_req and storage_class_req.get(
"finalStorageClass"):
storage_classes = aslist(
storage_class_req["finalStorageClass"])
else:
storage_classes = runtimeContext.storage_classes.strip().split(
",")
self.final_output, self.final_output_collection = self.make_output_collection(
self.output_name, storage_classes, self.output_tags,
self.final_output)
self.set_crunch_output()
if runtimeContext.compute_checksum:
adjustDirObjs(self.final_output,
partial(get_listing, self.fs_access))
adjustFileObjs(self.final_output,
partial(compute_checksums, self.fs_access))
if self.trash_intermediate and self.final_status == "success":
self.trash_intermediate_output()
return (self.final_output, self.final_status)
def run(self, dry_run=False, pull_image=True, **kwargs):
container_request = {
"command": self.command_line,
"owner_uuid": self.arvrunner.project_uuid,
"name": self.name,
"output_path": self.outdir,
"cwd": self.outdir,
"priority": 1,
"state": "Committed"
}
runtime_constraints = {}
mounts = {self.outdir: {"kind": "tmp"}}
dirs = set()
for f in self.pathmapper.files():
_, p, tp = self.pathmapper.mapper(f)
if tp == "Directory" and '/' not in p[6:]:
mounts[p] = {"kind": "collection", "portable_data_hash": p[6:]}
dirs.add(p[6:])
for f in self.pathmapper.files():
_, p, tp = self.pathmapper.mapper(f)
if p[6:].split("/")[0] not in dirs:
mounts[p] = {"kind": "collection", "portable_data_hash": p[6:]}
if self.generatefiles["listing"]:
raise UnsupportedRequirement("Generate files not supported")
container_request["environment"] = {
"TMPDIR": self.tmpdir,
"HOME": self.outdir
}
if self.environment:
container_request["environment"].update(self.environment)
if self.stdin:
raise UnsupportedRequirement(
"Stdin redirection currently not suppported")
if self.stderr:
raise UnsupportedRequirement(
"Stderr redirection currently not suppported")
if self.stdout:
mounts["stdout"] = {
"kind": "file",
"path": "%s/%s" % (self.outdir, self.stdout)
}
(docker_req, docker_is_req) = get_feature(self, "DockerRequirement")
if not docker_req:
docker_req = {"dockerImageId": "arvados/jobs"}
container_request["container_image"] = arv_docker_get_image(
self.arvrunner.api, docker_req, pull_image,
self.arvrunner.project_uuid)
resources = self.builder.resources
if resources is not None:
runtime_constraints["vcpus"] = resources.get("cores", 1)
runtime_constraints["ram"] = resources.get("ram") * 2**20
api_req, _ = get_feature(self, "http://arvados.org/cwl#APIRequirement")
if api_req:
runtime_constraints["API"] = True
runtime_req, _ = get_feature(
self, "http://arvados.org/cwl#RuntimeConstraints")
if runtime_req:
logger.warn(
"RuntimeConstraints not yet supported by container API")
partition_req, _ = get_feature(
self, "http://arvados.org/cwl#PartitionRequirement")
if partition_req:
runtime_constraints["partition"] = aslist(
partition_req["partition"])
container_request["mounts"] = mounts
container_request["runtime_constraints"] = runtime_constraints
try:
response = self.arvrunner.api.container_requests().create(
body=container_request).execute(
num_retries=self.arvrunner.num_retries)
self.arvrunner.processes[response["container_uuid"]] = self
logger.info("Container %s (%s) request state is %s", self.name,
response["uuid"], response["state"])
if response["state"] == "Final":
self.done(response)
except Exception as e:
logger.error("Got error %s" % str(e))
self.output_callback({}, "permanentFail")
def collect_output(
self,
schema, # type: Dict[Text, Any]
builder, # type: Builder
outdir, # type: Text
fs_access, # type: StdFsAccess
compute_checksum=True # type: bool
):
# type: (...) -> Optional[Union[Dict[Text, Any], List[Union[Dict[Text, Any], Text]]]]
result = [] # type: List[Any]
empty_and_optional = False
debug = LOGGER.isEnabledFor(logging.DEBUG)
if "outputBinding" in schema:
binding = schema["outputBinding"]
globpatterns = [] # type: List[Text]
revmap = partial(command_line_tool.revmap_file, builder, outdir)
if "glob" in binding:
with SourceLine(binding, "glob", WorkflowException, debug):
for glob in aslist(binding["glob"]):
glob = builder.do_eval(glob)
if glob:
globpatterns.extend(aslist(glob))
for glob in globpatterns:
if glob.startswith(outdir):
glob = glob[len(outdir) + 1:]
elif glob == ".":
glob = outdir
elif glob.startswith("/"):
raise WorkflowException(
"glob patterns must not start with '/'")
try:
prefix = fs_access.glob(outdir)
key = cmp_to_key(
cast(Callable[[Text, Text], int],
locale.strcoll))
# In case of stdout.log or stderr.log file not created
if "stdout" in self.tool and "stderr" in self.tool \
and glob in (self.tool["stdout"], self.tool["stderr"]):
filepath = Path(fs_access.join(outdir, glob))
if not filepath.is_file():
Path(filepath).touch()
result.extend([{
"location":
g,
"path":
fs_access.join(builder.outdir,
g[len(prefix[0]) + 1:]),
"basename":
os.path.basename(g),
"nameroot":
os.path.splitext(os.path.basename(g))[0],
"nameext":
os.path.splitext(os.path.basename(g))[1],
"class":
"File" if fs_access.isfile(g) else "Directory"
} for g in sorted(fs_access.glob(
fs_access.join(outdir, glob)),
key=key)])
except (OSError, IOError) as exc:
LOGGER.warning(Text(exc))
except Exception:
LOGGER.exception("Unexpected error from fs_access")
raise
for files in result:
rfile = files.copy()
# TODO This function raise an exception and seems to be related to docker (which is not used here)
# revmap(rfile)
if files["class"] == "Directory":
load_listing = builder.loadListing or (
binding and binding.get("loadListing"))
if load_listing and load_listing != "no_listing":
get_listing(fs_access, files,
(load_listing == "deep_listing"))
else:
with fs_access.open(rfile["location"], "rb") as f:
contents = b""
if binding.get("loadContents") or compute_checksum:
contents = f.read(CONTENT_LIMIT)
if binding.get("loadContents"):
files["contents"] = contents.decode("utf-8")
if compute_checksum:
checksum = hashlib.sha1() # nosec: B303
while contents != b"":
checksum.update(contents)
contents = f.read(1024 * 1024)
files[
"checksum"] = "sha1$%s" % checksum.hexdigest(
)
f.seek(0, 2)
file_size = f.tell()
files["size"] = file_size
optional = False
single = False
if isinstance(schema["type"], list):
if "null" in schema["type"]:
optional = True
if "File" in schema["type"] or "Directory" in schema["type"]:
single = True
elif schema["type"] == "File" or schema["type"] == "Directory":
single = True
if "outputEval" in binding:
with SourceLine(binding, "outputEval", WorkflowException,
debug):
result = builder.do_eval(binding["outputEval"],
context=result)
if single:
if not result and not optional:
with SourceLine(binding, "glob", WorkflowException, debug):
raise WorkflowException(
"Did not find output file with glob pattern: '{}'".
format(globpatterns))
elif not result and optional:
pass
elif isinstance(result, list):
if len(result) > 1:
raise WorkflowException(
"Multiple matches for output item that is a single file."
)
result = result[0]
if "secondaryFiles" in schema:
with SourceLine(schema, "secondaryFiles", WorkflowException,
debug):
for primary in aslist(result):
if isinstance(primary, dict):
primary.setdefault("secondaryFiles", [])
pathprefix = primary["path"][0:primary["path"].
rindex("/") + 1]
for file in aslist(schema["secondaryFiles"]):
if isinstance(
file,
dict) or "$(" in file or "${" in file:
sfpath = builder.do_eval(file,
context=primary)
subst = False
else:
sfpath = file
subst = True
for sfitem in aslist(sfpath):
if isinstance(sfitem, str):
if subst:
sfitem = {
"path":
substitute(
primary["path"], sfitem)
}
else:
sfitem = {
"path": pathprefix + sfitem
}
if "path" in sfitem and "location" not in sfitem:
revmap(sfitem)
if fs_access.isfile(sfitem["location"]):
sfitem["class"] = "File"
primary["secondaryFiles"].append(
sfitem)
elif fs_access.isdir(sfitem["location"]):
sfitem["class"] = "Directory"
primary["secondaryFiles"].append(
sfitem)
if "format" in schema:
for primary in aslist(result):
primary["format"] = builder.do_eval(schema["format"],
context=primary)
# Ensure files point to local references outside of the run environment
# TODO: Again removing revmap....
# adjustFileObjs(result, revmap)
if not result and optional:
return None
if not empty_and_optional and isinstance(
schema["type"], dict) and schema["type"]["type"] == "record":
out = {}
for f in schema["type"]["fields"]:
out[shortname(
f["name"])] = self.collect_output( # type: ignore
f,
builder,
outdir,
fs_access,
compute_checksum=compute_checksum)
return out
return result
def collect_output(
self,
schema, # type: Dict[Text, Any]
builder, # type: Builder
outdir, # type: Text
fs_access, # type: StdFsAccess
compute_checksum=True # type: bool
):
# type: (...) -> Optional[Union[Dict[Text, Any], List[Union[Dict[Text, Any], Text]]]]
"""
Collect outputs from the step :term:`Process` following its execution.
.. note:
When :term:`CWL` runner tries to forward ``step(i) outputs -> step(i+1) inputs``
using :meth:`collect_outputs`, it expects exact ``outputBindings`` locations to be matched.
In other words, a definition like ``outputBindings: {glob: outputs/*.txt}`` will generate results located
in ``step(i)`` as ``"<tmp-workdir>/outputs/file.txt"`` and ``step(i+1)`` will look explicitly
in ``"<tmp-workdir>/outputs`` using the ``glob`` pattern. Because each of our :term:`Process` in
the workflow are distinct/remote entities, each one stages its outputs at different URL locations,
not sharing the same *root directory*. When we stage intermediate results locally, the sub-dirs are lost.
Therefore, they act like individual :term:`CWL` runner calls where the *final results* are moved back
to the local directory for convenient access, but our *local directory* is the URL WPS-outputs location.
To let :term:`CWL` :term:`Workflow` inter-steps mapping work as intended, we must remap the locations
ignoring any nested dirs where the modified *outputBindings* definition will be able to match as if each
step :term:`Process` outputs were generated locally.
"""
result = [] # type: List[Any]
empty_and_optional = False
debug = LOGGER.isEnabledFor(logging.DEBUG)
if "outputBinding" in schema:
binding = schema["outputBinding"]
globpatterns = [] # type: List[Text]
revmap = partial(command_line_tool.revmap_file, builder, outdir)
if "glob" in binding:
with SourceLine(binding, "glob", WorkflowException, debug):
for glob in aslist(binding["glob"]):
glob = builder.do_eval(glob)
if glob:
globpatterns.extend(aslist(glob))
# rebase glob pattern as applicable (see note)
for glob in list(globpatterns):
if not any(
glob.startswith(part)
for part in [".", "/", "~"]) and "/" in glob:
glob = builder.do_eval(glob.split("/")[-1])
if glob:
globpatterns.extend(aslist(glob))
for glob in globpatterns:
if glob.startswith(outdir):
glob = glob[len(outdir) + 1:]
elif glob == ".":
glob = outdir
elif glob.startswith("/"):
raise WorkflowException(
"glob patterns must not start with '/'")
try:
prefix = fs_access.glob(outdir)
key = cmp_to_key(
cast(Callable[[Text, Text], int],
locale.strcoll))
# In case of stdout.log or stderr.log file not created
if "stdout" in self.tool and "stderr" in self.tool \
and glob in (self.tool["stdout"], self.tool["stderr"]):
filepath = Path(fs_access.join(outdir, glob))
if not filepath.is_file():
Path(filepath).touch()
result.extend([{
"location":
g,
"path":
fs_access.join(builder.outdir,
g[len(prefix[0]) + 1:]),
"basename":
os.path.basename(g),
"nameroot":
os.path.splitext(os.path.basename(g))[0],
"nameext":
os.path.splitext(os.path.basename(g))[1],
"class":
"File" if fs_access.isfile(g) else "Directory"
} for g in sorted(fs_access.glob(
fs_access.join(outdir, glob)),
key=key)])
except (OSError, IOError) as exc:
LOGGER.warning(Text(exc))
except Exception:
LOGGER.exception("Unexpected error from fs_access")
raise
for files in result:
rfile = files.copy()
# TODO This function raise an exception and seems to be related to docker (which is not used here)
# revmap(rfile)
if files["class"] == "Directory":
load_listing = builder.loadListing or (
binding and binding.get("loadListing"))
if load_listing and load_listing != "no_listing":
get_listing(fs_access, files,
(load_listing == "deep_listing"))
else:
with fs_access.open(rfile["location"], "rb") as f:
contents = b""
if binding.get("loadContents") or compute_checksum:
contents = f.read(CONTENT_LIMIT)
if binding.get("loadContents"):
files["contents"] = contents.decode("utf-8")
if compute_checksum:
checksum = hashlib.sha1() # nosec: B303
while contents != b"":
checksum.update(contents)
contents = f.read(1024 * 1024)
files[
"checksum"] = f"sha1${checksum.hexdigest()}"
f.seek(0, 2)
file_size = f.tell()
files["size"] = file_size
optional = False
single = False
if isinstance(schema["type"], list):
if "null" in schema["type"]:
optional = True
if "File" in schema["type"] or "Directory" in schema["type"]:
single = True
elif schema["type"] == "File" or schema["type"] == "Directory":
single = True
if "outputEval" in binding:
with SourceLine(binding, "outputEval", WorkflowException,
debug):
result = builder.do_eval(binding["outputEval"],
context=result)
if single:
if not result and not optional:
with SourceLine(binding, "glob", WorkflowException, debug):
raise WorkflowException(
f"Did not find output file with glob pattern: '{globpatterns}'"
)
elif not result and optional:
pass
elif isinstance(result, list):
if len(result) > 1:
raise WorkflowException(
"Multiple matches for output item that is a single file."
)
result = result[0]
if "secondaryFiles" in schema:
with SourceLine(schema, "secondaryFiles", WorkflowException,
debug):
for primary in aslist(result):
if isinstance(primary, dict):
primary.setdefault("secondaryFiles", [])
pathprefix = primary["path"][0:primary["path"].
rindex("/") + 1]
for file in aslist(schema["secondaryFiles"]):
if isinstance(
file,
dict) or "$(" in file or "${" in file:
sfpath = builder.do_eval(file,
context=primary)
subst = False
else:
sfpath = file
subst = True
for sfitem in aslist(sfpath):
if isinstance(sfitem, str):
if subst:
sfitem = {
"path":
substitute(
primary["path"], sfitem)
}
else:
sfitem = {
"path": pathprefix + sfitem
}
if "path" in sfitem and "location" not in sfitem:
revmap(sfitem)
if fs_access.isfile(sfitem["location"]):
sfitem["class"] = "File"
primary["secondaryFiles"].append(
sfitem)
elif fs_access.isdir(sfitem["location"]):
sfitem["class"] = "Directory"
primary["secondaryFiles"].append(
sfitem)
if "format" in schema:
for primary in aslist(result):
primary["format"] = builder.do_eval(schema["format"],
context=primary)
# Ensure files point to local references outside of the run environment
# TODO: Again removing revmap....
# adjustFileObjs(result, revmap)
if not result and optional:
return None
if not empty_and_optional and isinstance(
schema["type"], dict) and schema["type"]["type"] == "record":
out = {}
for f in schema["type"]["fields"]:
out[shortname(
f["name"])] = self.collect_output( # type: ignore
f,
builder,
outdir,
fs_access,
compute_checksum=compute_checksum)
return out
return result
def set_secondary(fsaccess, builder, inputschema, secondaryspec, primary,
discovered):
if isinstance(inputschema,
Sequence) and not isinstance(inputschema, basestring):
# union type, collect all possible secondaryFiles
for i in inputschema:
set_secondary(fsaccess, builder, i, secondaryspec, primary,
discovered)
return
if isinstance(inputschema, basestring):
sd = search_schemadef(inputschema,
reversed(builder.hints + builder.requirements))
if sd:
inputschema = sd
else:
return
if "secondaryFiles" in inputschema:
# set secondaryFiles, may be inherited by compound types.
secondaryspec = inputschema["secondaryFiles"]
if (isinstance(inputschema["type"], (Mapping, Sequence))
and not isinstance(inputschema["type"], basestring)):
# compound type (union, array, record)
set_secondary(fsaccess, builder, inputschema["type"], secondaryspec,
primary, discovered)
elif (inputschema["type"] == "record" and isinstance(primary, Mapping)):
#
# record type, find secondary files associated with fields.
#
for f in inputschema["fields"]:
p = primary.get(shortname(f["name"]))
if p:
set_secondary(fsaccess, builder, f, secondaryspec, p,
discovered)
elif (inputschema["type"] == "array" and isinstance(primary, Sequence)):
#
# array type, find secondary files of elements
#
for p in primary:
set_secondary(fsaccess, builder, {"type": inputschema["items"]},
secondaryspec, p, discovered)
elif (inputschema["type"] == "File" and secondaryspec
and isinstance(primary, Mapping) and primary.get("class") == "File"
and "secondaryFiles" not in primary):
#
# Found a file, check for secondaryFiles
#
specs = []
primary["secondaryFiles"] = secondaryspec
for i, sf in enumerate(aslist(secondaryspec)):
if builder.cwlVersion == "v1.0":
pattern = builder.do_eval(sf, context=primary)
else:
pattern = builder.do_eval(sf["pattern"], context=primary)
if pattern is None:
continue
if isinstance(pattern, list):
specs.extend(pattern)
elif isinstance(pattern, dict):
specs.append(pattern)
elif isinstance(pattern, str):
if builder.cwlVersion == "v1.0":
specs.append({"pattern": pattern, "required": True})
else:
specs.append({
"pattern": pattern,
"required": sf.get("required")
})
else:
raise SourceLine(
primary["secondaryFiles"], i,
validate.ValidationException).makeError(
"Expression must return list, object, string or null")
found = []
for i, sf in enumerate(specs):
if isinstance(sf, dict):
if sf.get("class") == "File":
pattern = None
if sf.get("location") is None:
raise SourceLine(
primary["secondaryFiles"], i,
validate.ValidationException).makeError(
"File object is missing 'location': %s" % sf)
sfpath = sf["location"]
required = True
else:
pattern = sf["pattern"]
required = sf.get("required")
elif isinstance(sf, str):
pattern = sf
required = True
else:
raise SourceLine(
primary["secondaryFiles"], i,
validate.ValidationException).makeError(
"Expression must return list, object, string or null")
if pattern is not None:
sfpath = substitute(primary["location"], pattern)
required = builder.do_eval(required, context=primary)
if fsaccess.exists(sfpath):
if pattern is not None:
found.append({"location": sfpath, "class": "File"})
else:
found.append(sf)
elif required:
raise SourceLine(
primary["secondaryFiles"], i,
validate.ValidationException).makeError(
"Required secondary file '%s' does not exist" % sfpath)
primary["secondaryFiles"] = cmap(found)
if discovered is not None:
discovered[primary["location"]] = primary["secondaryFiles"]
elif inputschema["type"] not in primitive_types_set:
set_secondary(fsaccess, builder, inputschema["type"], secondaryspec,
primary, discovered)
def run(self, runtimeContext):
# ArvadosCommandTool subclasses from cwltool.CommandLineTool,
# which calls makeJobRunner() to get a new ArvadosContainer
# object. The fields that define execution such as
# command_line, environment, etc are set on the
# ArvadosContainer object by CommandLineTool.job() before
# run() is called.
runtimeContext = self.job_runtime
container_request = {
"command": self.command_line,
"name": self.name,
"output_path": self.outdir,
"cwd": self.outdir,
"priority": runtimeContext.priority,
"state": "Committed",
"properties": {},
}
runtime_constraints = {}
if runtimeContext.project_uuid:
container_request["owner_uuid"] = runtimeContext.project_uuid
if self.arvrunner.secret_store.has_secret(self.command_line):
raise WorkflowException("Secret material leaked on command line, only file literals may contain secrets")
if self.arvrunner.secret_store.has_secret(self.environment):
raise WorkflowException("Secret material leaked in environment, only file literals may contain secrets")
resources = self.builder.resources
if resources is not None:
runtime_constraints["vcpus"] = math.ceil(resources.get("cores", 1))
runtime_constraints["ram"] = math.ceil(resources.get("ram") * 2**20)
mounts = {
self.outdir: {
"kind": "tmp",
"capacity": math.ceil(resources.get("outdirSize", 0) * 2**20)
},
self.tmpdir: {
"kind": "tmp",
"capacity": math.ceil(resources.get("tmpdirSize", 0) * 2**20)
}
}
secret_mounts = {}
scheduling_parameters = {}
rf = [self.pathmapper.mapper(f) for f in self.pathmapper.referenced_files]
rf.sort(key=lambda k: k.resolved)
prevdir = None
for resolved, target, tp, stg in rf:
if not stg:
continue
if prevdir and target.startswith(prevdir):
continue
if tp == "Directory":
targetdir = target
else:
targetdir = os.path.dirname(target)
sp = resolved.split("/", 1)
pdh = sp[0][5:] # remove "keep:"
mounts[targetdir] = {
"kind": "collection",
"portable_data_hash": pdh
}
if pdh in self.pathmapper.pdh_to_uuid:
mounts[targetdir]["uuid"] = self.pathmapper.pdh_to_uuid[pdh]
if len(sp) == 2:
if tp == "Directory":
path = sp[1]
else:
path = os.path.dirname(sp[1])
if path and path != "/":
mounts[targetdir]["path"] = path
prevdir = targetdir + "/"
with Perf(metrics, "generatefiles %s" % self.name):
if self.generatefiles["listing"]:
vwd = arvados.collection.Collection(api_client=self.arvrunner.api,
keep_client=self.arvrunner.keep_client,
num_retries=self.arvrunner.num_retries)
generatemapper = NoFollowPathMapper(self.generatefiles["listing"], "", "",
separateDirs=False)
sorteditems = sorted(generatemapper.items(), key=lambda n: n[1].target)
logger.debug("generatemapper is %s", sorteditems)
with Perf(metrics, "createfiles %s" % self.name):
for f, p in sorteditems:
if not p.target:
pass
elif p.type in ("File", "Directory", "WritableFile", "WritableDirectory"):
if p.resolved.startswith("_:"):
vwd.mkdirs(p.target)
else:
source, path = self.arvrunner.fs_access.get_collection(p.resolved)
vwd.copy(path or ".", p.target, source_collection=source)
elif p.type == "CreateFile":
if self.arvrunner.secret_store.has_secret(p.resolved):
secret_mounts["%s/%s" % (self.outdir, p.target)] = {
"kind": "text",
"content": self.arvrunner.secret_store.retrieve(p.resolved)
}
else:
with vwd.open(p.target, "w") as n:
n.write(p.resolved)
def keepemptydirs(p):
if isinstance(p, arvados.collection.RichCollectionBase):
if len(p) == 0:
p.open(".keep", "w").close()
else:
for c in p:
keepemptydirs(p[c])
keepemptydirs(vwd)
if not runtimeContext.current_container:
runtimeContext.current_container = arvados_cwl.util.get_current_container(self.arvrunner.api, self.arvrunner.num_retries, logger)
info = arvados_cwl.util.get_intermediate_collection_info(self.name, runtimeContext.current_container, runtimeContext.intermediate_output_ttl)
vwd.save_new(name=info["name"],
owner_uuid=runtimeContext.project_uuid,
ensure_unique_name=True,
trash_at=info["trash_at"],
properties=info["properties"])
prev = None
for f, p in sorteditems:
if (not p.target or self.arvrunner.secret_store.has_secret(p.resolved) or
(prev is not None and p.target.startswith(prev))):
continue
mountpoint = "%s/%s" % (self.outdir, p.target)
mounts[mountpoint] = {"kind": "collection",
"portable_data_hash": vwd.portable_data_hash(),
"path": p.target}
if p.type.startswith("Writable"):
mounts[mountpoint]["writable"] = True
prev = p.target + "/"
container_request["environment"] = {"TMPDIR": self.tmpdir, "HOME": self.outdir}
if self.environment:
container_request["environment"].update(self.environment)
if self.stdin:
sp = self.stdin[6:].split("/", 1)
mounts["stdin"] = {"kind": "collection",
"portable_data_hash": sp[0],
"path": sp[1]}
if self.stderr:
mounts["stderr"] = {"kind": "file",
"path": "%s/%s" % (self.outdir, self.stderr)}
if self.stdout:
mounts["stdout"] = {"kind": "file",
"path": "%s/%s" % (self.outdir, self.stdout)}
(docker_req, docker_is_req) = self.get_requirement("DockerRequirement")
if not docker_req:
docker_req = {"dockerImageId": "arvados/jobs"}
container_request["container_image"] = arv_docker_get_image(self.arvrunner.api,
docker_req,
runtimeContext.pull_image,
runtimeContext.project_uuid)
api_req, _ = self.get_requirement("http://arvados.org/cwl#APIRequirement")
if api_req:
runtime_constraints["API"] = True
runtime_req, _ = self.get_requirement("http://arvados.org/cwl#RuntimeConstraints")
if runtime_req:
if "keep_cache" in runtime_req:
runtime_constraints["keep_cache_ram"] = math.ceil(runtime_req["keep_cache"] * 2**20)
if "outputDirType" in runtime_req:
if runtime_req["outputDirType"] == "local_output_dir":
# Currently the default behavior.
pass
elif runtime_req["outputDirType"] == "keep_output_dir":
mounts[self.outdir]= {
"kind": "collection",
"writable": True
}
partition_req, _ = self.get_requirement("http://arvados.org/cwl#PartitionRequirement")
if partition_req:
scheduling_parameters["partitions"] = aslist(partition_req["partition"])
intermediate_output_req, _ = self.get_requirement("http://arvados.org/cwl#IntermediateOutput")
if intermediate_output_req:
self.output_ttl = intermediate_output_req["outputTTL"]
else:
self.output_ttl = self.arvrunner.intermediate_output_ttl
if self.output_ttl < 0:
raise WorkflowException("Invalid value %d for output_ttl, cannot be less than zero" % container_request["output_ttl"])
if self.timelimit is not None:
scheduling_parameters["max_run_time"] = self.timelimit
extra_submit_params = {}
if runtimeContext.submit_runner_cluster:
extra_submit_params["cluster_id"] = runtimeContext.submit_runner_cluster
container_request["output_name"] = "Output for step %s" % (self.name)
container_request["output_ttl"] = self.output_ttl
container_request["mounts"] = mounts
container_request["secret_mounts"] = secret_mounts
container_request["runtime_constraints"] = runtime_constraints
container_request["scheduling_parameters"] = scheduling_parameters
enable_reuse = runtimeContext.enable_reuse
if enable_reuse:
reuse_req, _ = self.get_requirement("http://arvados.org/cwl#ReuseRequirement")
if reuse_req:
enable_reuse = reuse_req["enableReuse"]
container_request["use_existing"] = enable_reuse
if runtimeContext.runnerjob.startswith("arvwf:"):
wfuuid = runtimeContext.runnerjob[6:runtimeContext.runnerjob.index("#")]
wfrecord = self.arvrunner.api.workflows().get(uuid=wfuuid).execute(num_retries=self.arvrunner.num_retries)
if container_request["name"] == "main":
container_request["name"] = wfrecord["name"]
container_request["properties"]["template_uuid"] = wfuuid
self.output_callback = self.arvrunner.get_wrapped_callback(self.output_callback)
try:
if runtimeContext.submit_request_uuid:
response = self.arvrunner.api.container_requests().update(
uuid=runtimeContext.submit_request_uuid,
body=container_request,
**extra_submit_params
).execute(num_retries=self.arvrunner.num_retries)
else:
response = self.arvrunner.api.container_requests().create(
body=container_request,
**extra_submit_params
).execute(num_retries=self.arvrunner.num_retries)
self.uuid = response["uuid"]
self.arvrunner.process_submitted(self)
if response["state"] == "Final":
logger.info("%s reused container %s", self.arvrunner.label(self), response["container_uuid"])
else:
logger.info("%s %s state is %s", self.arvrunner.label(self), response["uuid"], response["state"])
except Exception:
logger.exception("%s got an error", self.arvrunner.label(self))
self.output_callback({}, "permanentFail")
def run(self, fileStore):
cwljob = resolve_indirect(self.cwljob)
# `promises` dict
# from: each parameter (workflow input or step output)
# that may be used as a "source" for a step input workflow output
# parameter
# to: the job that will produce that value.
promises = {}
# `jobs` dict from step id to job that implements that step.
jobs = {}
for inp in self.cwlwf.tool["inputs"]:
promises[inp["id"]] = SelfJob(self, cwljob)
alloutputs_fufilled = False
while not alloutputs_fufilled:
# Iteratively go over the workflow steps, scheduling jobs as their
# dependencies can be fufilled by upstream workflow inputs or
# step outputs. Loop exits when the workflow outputs
# are satisfied.
alloutputs_fufilled = True
for step in self.cwlwf.steps:
if step.tool["id"] not in jobs:
stepinputs_fufilled = True
for inp in step.tool["inputs"]:
if "source" in inp:
for s in aslist(inp["source"]):
if s not in promises:
stepinputs_fufilled = False
if stepinputs_fufilled:
jobobj = {}
for inp in step.tool["inputs"]:
key = shortname(inp["id"])
if "source" in inp:
if inp.get("linkMerge") or len(aslist(inp["source"])) > 1:
linkMerge = inp.get("linkMerge", "merge_nested")
if linkMerge == "merge_nested":
jobobj[key] = (
MergeInputsNested([(shortname(s), promises[s].rv())
for s in aslist(inp["source"])]))
elif linkMerge == "merge_flattened":
jobobj[key] = (
MergeInputsFlattened([(shortname(s), promises[s].rv())
for s in aslist(inp["source"])]))
else:
raise validate.ValidationException(
"Unsupported linkMerge '%s'", linkMerge)
else:
jobobj[key] = (
shortname(inp["source"]), promises[inp["source"]].rv())
elif "default" in inp:
d = copy.copy(inp["default"])
jobobj[key] = ("default", {"default": d})
if "valueFrom" in inp and "scatter" not in step.tool:
if key in jobobj:
jobobj[key] = StepValueFrom(inp["valueFrom"],
jobobj[key],
self.cwlwf.requirements)
else:
jobobj[key] = StepValueFrom(inp["valueFrom"],
("None", {"None": None}),
self.cwlwf.requirements)
if "scatter" in step.tool:
wfjob = CWLScatter(step, IndirectDict(jobobj), **self.executor_options)
followOn = CWLGather(step, wfjob.rv())
wfjob.addFollowOn(followOn)
else:
(wfjob, followOn) = makeJob(step.embedded_tool, IndirectDict(jobobj),
**self.executor_options)
jobs[step.tool["id"]] = followOn
connected = False
for inp in step.tool["inputs"]:
for s in aslist(inp.get("source", [])):
if not promises[s].hasChild(wfjob):
promises[s].addChild(wfjob)
connected = True
if not connected:
# workflow step has default inputs only, isn't connected to other jobs,
# so add it as child of workflow.
self.addChild(wfjob)
for out in step.tool["outputs"]:
promises[out["id"]] = followOn
for inp in step.tool["inputs"]:
for s in aslist(inp.get("source", [])):
if s not in promises:
alloutputs_fufilled = False
# may need a test
for out in self.cwlwf.tool["outputs"]:
if "source" in out:
if out["source"] not in promises:
alloutputs_fufilled = False
outobj = {}
for out in self.cwlwf.tool["outputs"]:
outobj[shortname(out["id"])] = (shortname(out["outputSource"]), promises[out["outputSource"]].rv())
return IndirectDict(outobj)
