import sys
import threading
import hashlib
+import copy
+import json
from functools import partial
import pkg_resources # part of setuptools
from .arvworkflow import ArvadosWorkflow, upload_workflow
from .fsaccess import CollectionFsAccess
from .perf import Perf
-from cwltool.pack import pack
+from .pathmapper import FinalOutputPathMapper
+from cwltool.pack import pack
from cwltool.process import shortname, UnsupportedRequirement
-from cwltool.pathmapper import adjustFileObjs
+from cwltool.pathmapper import adjustFileObjs, adjustDirObjs
from cwltool.draft2tool import compute_checksums
from arvados.api import OrderedJsonModel
logger = logging.getLogger('arvados.cwl-runner')
+metrics = logging.getLogger('arvados.cwl-runner.metrics')
logger.setLevel(logging.INFO)
+
class ArvCwlRunner(object):
"""Execute a CWL tool or workflow, submit work (using either jobs or
containers API), wait for them to complete, and report output.
self.stop_polling = threading.Event()
self.poll_api = None
self.pipeline = None
+ self.final_output_collection = None
if self.work_api is None:
# todo: autodetect API to use.
self.cond.acquire()
j = self.processes[uuid]
logger.info("Job %s (%s) is %s", j.name, uuid, event["properties"]["new_attributes"]["state"])
- with Perf(logger, "done %s" % j.name):
+ with Perf(metrics, "done %s" % j.name):
j.done(event["properties"]["new_attributes"])
self.cond.notify()
finally:
for v in obj:
self.check_writable(v)
+ def make_output_collection(self, name, outputObj):
+ outputObj = copy.deepcopy(outputObj)
+
+ files = []
+ def capture(fileobj):
+ files.append(fileobj)
+
+ adjustDirObjs(outputObj, capture)
+ adjustFileObjs(outputObj, capture)
+
+ generatemapper = FinalOutputPathMapper(files, "", "", separateDirs=False)
+
+ final = arvados.collection.Collection()
+
+ srccollections = {}
+ for k,v in generatemapper.items():
+ sp = k.split("/")
+ srccollection = sp[0][5:]
+ if srccollection not in srccollections:
+ srccollections[srccollection] = arvados.collection.CollectionReader(srccollection)
+ reader = srccollections[srccollection]
+ try:
+ final.copy("/".join(sp[1:]), v.target, source_collection=reader, overwrite=False)
+ except IOError as e:
+ logger.warn("While preparing output collection: %s", e)
+
+ def rewrite(fileobj):
+ fileobj["location"] = generatemapper.mapper(fileobj["location"]).target
+ for k in ("basename", "size", "listing"):
+ if k in fileobj:
+ del fileobj[k]
+
+ adjustDirObjs(outputObj, rewrite)
+ adjustFileObjs(outputObj, rewrite)
+
+ with final.open("cwl.output.json", "w") as f:
+ json.dump(outputObj, f, sort_keys=True, indent=4)
+
+ final.save_new(name=name, owner_uuid=self.project_uuid, ensure_unique_name=True)
+
+ logger.info("Final output collection %s (%s)", final.portable_data_hash(), final.manifest_locator())
+
+ self.final_output_collection = final
+
def arv_executor(self, tool, job_order, **kwargs):
self.debug = kwargs.get("debug")
# except when in cond.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 runnable:
- with Perf(logger, "run"):
+ with Perf(metrics, "run"):
runnable.run(**kwargs)
else:
if self.processes:
else:
logger.error("Workflow is deadlocked, no runnable jobs and not waiting on any pending jobs.")
break
+ loopperf.__enter__()
+ loopperf.__exit__()
while self.processes:
self.cond.wait(1)
if self.final_output is None:
raise WorkflowException("Workflow did not return a result.")
+ if kwargs.get("submit"):
+ logger.info("Final output collection %s", runnerjob.final_output)
+ else:
+ self.make_output_collection("Output of %s" % (shortname(tool.tool["id"])),
+ self.final_output)
+
if kwargs.get("compute_checksum"):
adjustFileObjs(self.final_output, partial(compute_checksums, self.fs_access))
exgroup.add_argument("--quiet", action="store_true", help="Only print warnings and errors.")
exgroup.add_argument("--debug", action="store_true", help="Print even more logging")
+ parser.add_argument("--metrics", action="store_true", help="Print timing metrics")
+
parser.add_argument("--tool-help", action="store_true", help="Print command line help for tool")
exgroup = parser.add_mutually_exclusive_group()
def add_arv_hints():
cache = {}
res = pkg_resources.resource_stream(__name__, 'arv-cwl-schema.yml')
- cache["https://w3id.org/cwl/arv-cwl-schema.yml"] = res.read()
+ cache["http://arvados.org/cwl"] = res.read()
res.close()
_, cwlnames, _, _ = cwltool.process.get_schema("v1.0")
- _, extnames, _, _ = schema_salad.schema.load_schema("https://w3id.org/cwl/arv-cwl-schema.yml", cache=cache)
+ _, extnames, _, _ = schema_salad.schema.load_schema("http://arvados.org/cwl", cache=cache)
for n in extnames.names:
if not cwlnames.has_name("http://arvados.org/cwl#"+n, ""):
cwlnames.add_name("http://arvados.org/cwl#"+n, "", extnames.get_name(n, ""))
logger.setLevel(logging.WARN)
logging.getLogger('arvados.arv-run').setLevel(logging.WARN)
+ if arvargs.metrics:
+ metrics.setLevel(logging.DEBUG)
+ logging.getLogger("cwltool.metrics").setLevel(logging.DEBUG)
+
arvargs.conformance_test = None
arvargs.use_container = True
projects / arvados.git / blobdiff