#
# SPDX-License-Identifier: Apache-2.0
+from future import standard_library
+standard_library.install_aliases()
+from builtins import str
+
import logging
import json
import os
-import urllib
+import urllib.request, urllib.parse, urllib.error
import time
import datetime
import ciso8601
import uuid
+import math
+import arvados_cwl.util
import ruamel.yaml as yaml
from cwltool.errors import WorkflowException
from cwltool.process import UnsupportedRequirement, shortname
-from cwltool.pathmapper import adjustFileObjs, adjustDirObjs, visit_class
-from cwltool.utils import aslist
+from cwltool.utils import aslist, adjustFileObjs, adjustDirObjs, visit_class
from cwltool.job import JobBase
import arvados.collection
-from arvados.errors import ApiError
from .arvdocker import arv_docker_get_image
from . import done
-from .runner import Runner, arvados_jobs_image, packed_workflow, trim_anonymous_location, remove_redundant_fields
+from .runner import Runner, arvados_jobs_image, packed_workflow, trim_anonymous_location, remove_redundant_fields, make_builder
from .fsaccess import CollectionFetcher
from .pathmapper import NoFollowPathMapper, trim_listing
from .perf import Perf
+from ._version import __version__
logger = logging.getLogger('arvados.cwl-runner')
metrics = logging.getLogger('arvados.cwl-runner.metrics')
class ArvadosContainer(JobBase):
"""Submit and manage a Crunch container request for executing a CWL CommandLineTool."""
- def __init__(self, runner,
+ def __init__(self, runner, job_runtime,
builder, # type: Builder
joborder, # type: Dict[Text, Union[Dict[Text, Any], List, Text]]
make_path_mapper, # type: Callable[..., PathMapper]
):
super(ArvadosContainer, self).__init__(builder, joborder, make_path_mapper, requirements, hints, name)
self.arvrunner = runner
+ self.job_runtime = job_runtime
self.running = False
self.uuid = None
def update_pipeline_component(self, r):
pass
+ def _required_env(self):
+ env = {}
+ env["HOME"] = self.outdir
+ env["TMPDIR"] = self.tmpdir
+ return env
+
def run(self, runtimeContext):
# ArvadosCommandTool subclasses from cwltool.CommandLineTool,
# which calls makeJobRunner() to get a new ArvadosContainer
# ArvadosContainer object by CommandLineTool.job() before
# run() is called.
- container_request = {
- "command": self.command_line,
- "name": self.name,
- "output_path": self.outdir,
- "cwd": self.outdir,
- "priority": runtimeContext.priority,
- "state": "Committed",
- "properties": {},
- }
+ runtimeContext = self.job_runtime
+
+ if runtimeContext.submit_request_uuid:
+ container_request = self.arvrunner.api.container_requests().get(
+ uuid=runtimeContext.submit_request_uuid
+ ).execute(num_retries=self.arvrunner.num_retries)
+ else:
+ container_request = {}
+
+ container_request["command"] = self.command_line
+ container_request["name"] = self.name
+ container_request["output_path"] = self.outdir
+ container_request["cwd"] = self.outdir
+ container_request["priority"] = runtimeContext.priority
+ container_request["state"] = "Committed"
+ container_request.setdefault("properties", {})
+
runtime_constraints = {}
- if self.arvrunner.project_uuid:
- container_request["owner_uuid"] = self.arvrunner.project_uuid
+ 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")
resources = self.builder.resources
if resources is not None:
- runtime_constraints["vcpus"] = resources.get("cores", 1)
- runtime_constraints["ram"] = resources.get("ram") * 2**20
+ 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": resources.get("outdirSize", 0) * 2**20
+ "capacity": math.ceil(resources.get("outdirSize", 0) * 2**20)
},
self.tmpdir: {
"kind": "tmp",
- "capacity": resources.get("tmpdirSize", 0) * 2**20
+ "capacity": math.ceil(resources.get("tmpdirSize", 0) * 2**20)
}
}
secret_mounts = {}
"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]
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], "", "",
+ generatemapper = NoFollowPathMapper(self.generatefiles["listing"], "", "",
separateDirs=False)
- sorteditems = sorted(generatemapper.items(), None, key=lambda n: n[1].target)
+ 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"):
+ 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(p.target)
+ vwd.mkdirs(dst)
else:
source, path = self.arvrunner.fs_access.get_collection(p.resolved)
- vwd.copy(path, p.target, source_collection=source)
+ vwd.copy(path or ".", dst, source_collection=source)
elif p.type == "CreateFile":
if self.arvrunner.secret_store.has_secret(p.resolved):
- secret_mounts["%s/%s" % (self.outdir, p.target)] = {
+ 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(p.target, "w") as n:
- n.write(p.resolved.encode("utf-8"))
+ with vwd.open(dst, "w") as n:
+ n.write(p.resolved)
def keepemptydirs(p):
if isinstance(p, arvados.collection.RichCollectionBase):
keepemptydirs(vwd)
- info = self._get_intermediate_collection_info()
+ 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"])
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)
+ 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": p.target}
+ "path": dst}
if p.type.startswith("Writable"):
mounts[mountpoint]["writable"] = True
prev = p.target + "/"
"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,
- self.arvrunner.project_uuid)
+ 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_req, _ = self.get_requirement("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
+ 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.
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:
+
+ if self.arvrunner.api._rootDesc["revision"] >= "20210628":
+ storage_class_req, _ = self.get_requirement("http://arvados.org/cwl#OutputStorageClass")
+ if storage_class_req and storage_class_req.get("intermediateStorageClass"):
+ container_request["output_storage_classes"] = aslist(storage_class_req["intermediateStorageClass"])
+ else:
+ container_request["output_storage_classes"] = runtimeContext.intermediate_storage_classes.strip().split(",")
+
+ 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
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
+ properties_req, _ = self.get_requirement("http://arvados.org/cwl#ProcessProperties")
+ if properties_req:
+ for pr in properties_req["processProperties"]:
+ container_request["properties"][pr["propertyName"]] = self.builder.do_eval(pr["propertyValue"])
+
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 runtimeContext.submit_request_uuid:
response = self.arvrunner.api.container_requests().update(
uuid=runtimeContext.submit_request_uuid,
- body=container_request
+ body=container_request,
+ **extra_submit_params
).execute(num_retries=self.arvrunner.num_retries)
else:
response = self.arvrunner.api.container_requests().create(
- body=container_request
+ body=container_request,
+ **extra_submit_params
).execute(num_retries=self.arvrunner.num_retries)
self.uuid = response["uuid"]
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)))
+ 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 done(self, record):
else:
processStatus = "permanentFail"
- if processStatus == "permanentFail":
- logc = arvados.collection.CollectionReader(container["log"],
+ if processStatus == "permanentFail" and record["log_uuid"]:
+ logc = arvados.collection.CollectionReader(record["log_uuid"],
api_client=self.arvrunner.api,
keep_client=self.arvrunner.keep_client,
num_retries=self.arvrunner.num_retries)
- done.logtail(logc, logger.error, "%s (%s) error log:" % (self.arvrunner.label(self), record["uuid"]), maxlen=40)
+ label = self.arvrunner.label(self)
+ done.logtail(
+ logc, logger.error,
+ "%s (%s) error log:" % (label, record["uuid"]), maxlen=40)
if record["output_uuid"]:
if self.arvrunner.trash_intermediate or self.arvrunner.intermediate_output_ttl:
# Compute the trash time to avoid requesting the collection record.
- trash_at = ciso8601.parse_datetime_unaware(record["modified_at"]) + datetime.timedelta(0, self.arvrunner.intermediate_output_ttl)
+ trash_at = ciso8601.parse_datetime_as_naive(record["modified_at"]) + datetime.timedelta(0, self.arvrunner.intermediate_output_ttl)
aftertime = " at %s" % trash_at.strftime("%Y-%m-%d %H:%M:%S UTC") if self.arvrunner.intermediate_output_ttl else ""
orpart = ", or" if self.arvrunner.trash_intermediate and self.arvrunner.intermediate_output_ttl else ""
oncomplete = " upon successful completion of the workflow" if self.arvrunner.trash_intermediate else ""
if container["output"]:
outputs = done.done_outputs(self, container, "/tmp", self.outdir, "/keep")
except WorkflowException as e:
+ # Only include a stack trace if in debug mode.
+ # A stack trace may obfuscate more useful output about the workflow.
logger.error("%s unable to collect output from %s:\n%s",
self.arvrunner.label(self), container["output"], e, exc_info=(e if self.arvrunner.debug else False))
processStatus = "permanentFail"
- except Exception as e:
- logger.exception("%s while getting output object: %s", self.arvrunner.label(self), e)
+ except Exception:
+ logger.exception("%s while getting output object:", self.arvrunner.label(self))
processStatus = "permanentFail"
finally:
self.output_callback(outputs, processStatus)
- def _get_intermediate_collection_info(self):
- trash_time = None
- if self.arvrunner.intermediate_output_ttl > 0:
- trash_time = datetime.datetime.now() + datetime.timedelta(seconds=self.arvrunner.intermediate_output_ttl)
-
- current_container_uuid = None
- try:
- current_container = self.arvrunner.api.containers().current().execute(num_retries=self.arvrunner.num_retries)
- current_container_uuid = current_container['uuid']
- except ApiError as e:
- # Status code 404 just means we're not running in a container.
- if e.resp.status != 404:
- logger.info("Getting current container: %s", e)
- props = {"type": "Intermediate",
- "container": current_container_uuid}
-
- return {"name" : "Intermediate collection",
- "trash_at" : trash_time,
- "properties" : props}
-
class RunnerContainer(Runner):
"""Submit and manage a container that runs arvados-cwl-runner."""
},
"secret_mounts": secret_mounts,
"runtime_constraints": {
- "vcpus": 1,
- "ram": 1024*1024 * self.submit_runner_ram,
+ "vcpus": math.ceil(self.submit_runner_cores),
+ "ram": 1024*1024 * (math.ceil(self.submit_runner_ram) + math.ceil(self.collection_cache_size)),
"API": True
},
- "use_existing": self.enable_reuse,
+ "use_existing": False, # Never reuse the runner container - see #15497.
"properties": {}
}
- if self.tool.tool.get("id", "").startswith("keep:"):
- sp = self.tool.tool["id"].split('/')
+ if self.embedded_tool.tool.get("id", "").startswith("keep:"):
+ sp = self.embedded_tool.tool["id"].split('/')
workflowcollection = sp[0][5:]
workflowname = "/".join(sp[1:])
workflowpath = "/var/lib/cwl/workflow/%s" % workflowname
"portable_data_hash": "%s" % workflowcollection
}
else:
- packed = packed_workflow(self.arvrunner, self.tool, self.merged_map)
+ packed = packed_workflow(self.arvrunner, self.embedded_tool, self.merged_map)
workflowpath = "/var/lib/cwl/workflow.json#main"
container_req["mounts"]["/var/lib/cwl/workflow.json"] = {
"kind": "json",
"content": packed
}
- if self.tool.tool.get("id", "").startswith("arvwf:"):
- container_req["properties"]["template_uuid"] = self.tool.tool["id"][6:33]
+ if self.embedded_tool.tool.get("id", "").startswith("arvwf:"):
+ container_req["properties"]["template_uuid"] = self.embedded_tool.tool["id"][6:33]
+ properties_req, _ = self.embedded_tool.get_requirement("http://arvados.org/cwl#ProcessProperties")
+ if properties_req:
+ builder = make_builder(self.job_order, self.embedded_tool.hints, self.embedded_tool.requirements, runtimeContext, self.embedded_tool.metadata)
+ for pr in properties_req["processProperties"]:
+ container_req["properties"][pr["propertyName"]] = builder.do_eval(pr["propertyValue"])
# --local means execute the workflow instead of submitting a container request
# --api=containers means use the containers API
# --eval-timeout is the timeout for javascript invocation
# --parallel-task-count is the number of threads to use for job submission
# --enable/disable-reuse sets desired job reuse
+ # --collection-cache-size sets aside memory to store collections
command = ["arvados-cwl-runner",
"--local",
"--api=containers",
"--no-log-timestamps",
"--disable-validate",
+ "--disable-color",
"--eval-timeout=%s" % self.arvrunner.eval_timeout,
"--thread-count=%s" % self.arvrunner.thread_count,
- "--enable-reuse" if self.enable_reuse else "--disable-reuse"]
+ "--enable-reuse" if self.enable_reuse else "--disable-reuse",
+ "--collection-cache-size=%s" % self.collection_cache_size]
if self.output_name:
command.append("--output-name=" + self.output_name)
if runtimeContext.debug:
command.append("--debug")
- if runtimeContext.storage_classes != "default":
+ if runtimeContext.storage_classes != "default" and runtimeContext.storage_classes:
command.append("--storage-classes=" + runtimeContext.storage_classes)
+ if runtimeContext.intermediate_storage_classes != "default" and runtimeContext.intermediate_storage_classes:
+ command.append("--intermediate-storage-classes=" + runtimeContext.intermediate_storage_classes)
+
if self.on_error:
command.append("--on-error=" + self.on_error)
if self.arvrunner.project_uuid:
command.append("--project-uuid="+self.arvrunner.project_uuid)
+ if self.enable_dev:
+ command.append("--enable-dev")
+
command.extend([workflowpath, "/var/lib/cwl/cwl.input.json"])
container_req["command"] = command
if self.arvrunner.project_uuid:
job_spec["owner_uuid"] = self.arvrunner.project_uuid
+ extra_submit_params = {}
+ if runtimeContext.submit_runner_cluster:
+ extra_submit_params["cluster_id"] = runtimeContext.submit_runner_cluster
+
if runtimeContext.submit_request_uuid:
+ if "cluster_id" in extra_submit_params:
+ # Doesn't make sense for "update" and actually fails
+ del extra_submit_params["cluster_id"]
response = self.arvrunner.api.container_requests().update(
uuid=runtimeContext.submit_request_uuid,
- body=job_spec
+ body=job_spec,
+ **extra_submit_params
).execute(num_retries=self.arvrunner.num_retries)
else:
response = self.arvrunner.api.container_requests().create(
- body=job_spec
+ body=job_spec,
+ **extra_submit_params
).execute(num_retries=self.arvrunner.num_retries)
self.uuid = response["uuid"]
self.arvrunner.process_submitted(self)
- logger.info("%s submitted container %s", self.arvrunner.label(self), response["uuid"])
+ logger.info("%s submitted container_request %s", self.arvrunner.label(self), response["uuid"])
+
+ workbench1 = self.arvrunner.api.config()["Services"]["Workbench1"]["ExternalURL"]
+ workbench2 = self.arvrunner.api.config()["Services"]["Workbench2"]["ExternalURL"]
+ url = ""
+ if workbench2:
+ url = "{}processes/{}".format(workbench2, response["uuid"])
+ elif workbench1:
+ url = "{}container_requests/{}".format(workbench1, response["uuid"])
+ if url:
+ logger.info("Monitor workflow progress at %s", url)
+
def done(self, record):
try:
container = self.arvrunner.api.containers().get(
uuid=record["container_uuid"]
).execute(num_retries=self.arvrunner.num_retries)
- except Exception as e:
- logger.exception("%s while getting runner container: %s", self.arvrunner.label(self), e)
+ container["log"] = record["log_uuid"]
+ except Exception:
+ logger.exception("%s while getting runner container", self.arvrunner.label(self))
self.arvrunner.output_callback({}, "permanentFail")
else:
super(RunnerContainer, self).done(container)