X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/5b5be8dfac71c7db9d5a966485fa04f24fd20957..6aef698e8f1f99aa0511afd21db86d4c7cf8b5da:/sdk/cwl/arvados_cwl/__init__.py diff --git a/sdk/cwl/arvados_cwl/__init__.py b/sdk/cwl/arvados_cwl/__init__.py index e4f5ceab74..7affade073 100644 --- a/sdk/cwl/arvados_cwl/__init__.py +++ b/sdk/cwl/arvados_cwl/__init__.py @@ -17,6 +17,8 @@ import json import re from functools import partial import pkg_resources # part of setuptools +import Queue +import time from cwltool.errors import WorkflowException import cwltool.main @@ -38,6 +40,7 @@ from .arvworkflow import ArvadosWorkflow, upload_workflow from .fsaccess import CollectionFsAccess, CollectionFetcher, collectionResolver, CollectionCache from .perf import Perf from .pathmapper import NoFollowPathMapper +from .task_queue import TaskQueue from ._version import __version__ from cwltool.pack import pack @@ -62,11 +65,11 @@ class ArvCwlRunner(object): """ - def __init__(self, api_client, work_api=None, keep_client=None, output_name=None, output_tags=None, num_retries=4): + def __init__(self, api_client, work_api=None, keep_client=None, + output_name=None, output_tags=None, num_retries=4): self.api = api_client self.processes = {} - self.lock = threading.Lock() - self.cond = threading.Condition(self.lock) + self.workflow_eval_lock = threading.Condition(threading.RLock()) self.final_output = None self.final_status = None self.uploaded = {} @@ -82,6 +85,8 @@ class ArvCwlRunner(object): self.intermediate_output_ttl = 0 self.intermediate_output_collections = [] self.trash_intermediate = False + self.thread_count = 4 + self.poll_interval = 12 if keep_client is not None: self.keep_client = keep_client @@ -123,40 +128,58 @@ class ArvCwlRunner(object): return cwltool.workflow.defaultMakeTool(toolpath_object, **kwargs) def output_callback(self, out, processStatus): - if processStatus == "success": - logger.info("Overall process status is %s", processStatus) - if self.pipeline: - self.api.pipeline_instances().update(uuid=self.pipeline["uuid"], - body={"state": "Complete"}).execute(num_retries=self.num_retries) - else: - logger.warn("Overall process status is %s", processStatus) - if self.pipeline: - self.api.pipeline_instances().update(uuid=self.pipeline["uuid"], - body={"state": "Failed"}).execute(num_retries=self.num_retries) - self.final_status = processStatus - self.final_output = out + with self.workflow_eval_lock: + if processStatus == "success": + logger.info("Overall process status is %s", processStatus) + if self.pipeline: + self.api.pipeline_instances().update(uuid=self.pipeline["uuid"], + body={"state": "Complete"}).execute(num_retries=self.num_retries) + else: + logger.warn("Overall process status is %s", processStatus) + if self.pipeline: + self.api.pipeline_instances().update(uuid=self.pipeline["uuid"], + body={"state": "Failed"}).execute(num_retries=self.num_retries) + self.final_status = processStatus + self.final_output = out + self.workflow_eval_lock.notifyAll() + + + def start_run(self, runnable, kwargs): + self.task_queue.add(partial(runnable.run, **kwargs)) + + def process_submitted(self, container): + with self.workflow_eval_lock: + self.processes[container.uuid] = container + + def process_done(self, uuid): + with self.workflow_eval_lock: + if uuid in self.processes: + del self.processes[uuid] + + def wrapped_callback(self, cb, obj, st): + with self.workflow_eval_lock: + cb(obj, st) + self.workflow_eval_lock.notifyAll() + + def get_wrapped_callback(self, cb): + return partial(self.wrapped_callback, cb) def on_message(self, event): if "object_uuid" in event: if event["object_uuid"] in self.processes and event["event_type"] == "update": - if event["properties"]["new_attributes"]["state"] == "Running" and self.processes[event["object_uuid"]].running is False: - uuid = event["object_uuid"] - with self.lock: + uuid = event["object_uuid"] + if event["properties"]["new_attributes"]["state"] == "Running": + with self.workflow_eval_lock: j = self.processes[uuid] - logger.info("%s %s is Running", self.label(j), uuid) - j.running = True - j.update_pipeline_component(event["properties"]["new_attributes"]) + if j.running is False: + j.running = True + j.update_pipeline_component(event["properties"]["new_attributes"]) + logger.info("%s %s is Running", self.label(j), uuid) elif event["properties"]["new_attributes"]["state"] in ("Complete", "Failed", "Cancelled", "Final"): - uuid = event["object_uuid"] - try: - self.cond.acquire() + with self.workflow_eval_lock: j = self.processes[uuid] - logger.info("%s %s is %s", self.label(j), uuid, event["properties"]["new_attributes"]["state"]) - with Perf(metrics, "done %s" % j.name): - j.done(event["properties"]["new_attributes"]) - self.cond.notify() - finally: - self.cond.release() + self.task_queue.add(partial(j.done, event["properties"]["new_attributes"])) + logger.info("%s %s is %s", self.label(j), uuid, event["properties"]["new_attributes"]["state"]) def label(self, obj): return "[%s %s]" % (self.work_api[0:-1], obj.name) @@ -168,15 +191,19 @@ class ArvCwlRunner(object): """ try: + remain_wait = self.poll_interval while True: - self.stop_polling.wait(15) + if remain_wait > 0: + self.stop_polling.wait(remain_wait) if self.stop_polling.is_set(): break - with self.lock: - keys = self.processes.keys() + with self.workflow_eval_lock: + keys = list(self.processes.keys()) if not keys: + remain_wait = self.poll_interval continue + begin_poll = time.time() if self.work_api == "containers": table = self.poll_api.container_requests() elif self.work_api == "jobs": @@ -186,6 +213,7 @@ class ArvCwlRunner(object): proc_states = table.list(filters=[["uuid", "in", keys]]).execute(num_retries=self.num_retries) except Exception as e: logger.warn("Error checking states on API server: %s", e) + remain_wait = self.poll_interval continue for p in proc_states["items"]: @@ -196,12 +224,13 @@ class ArvCwlRunner(object): "new_attributes": p } }) + finish_poll = time.time() + remain_wait = self.poll_interval - (finish_poll - begin_poll) except: - logger.error("Fatal error in state polling thread.", exc_info=(sys.exc_info()[1] if self.debug else False)) - self.cond.acquire() - self.processes.clear() - self.cond.notify() - self.cond.release() + logger.exception("Fatal error in state polling thread.") + with self.workflow_eval_lock: + self.processes.clear() + self.workflow_eval_lock.notifyAll() finally: self.stop_polling.set() @@ -357,6 +386,7 @@ class ArvCwlRunner(object): collection_cache=self.collection_cache) self.fs_access = make_fs_access(kwargs["basedir"]) self.secret_store = kwargs.get("secret_store") + self.thread_count = kwargs.get("thread_count", 4) self.trash_intermediate = kwargs["trash_intermediate"] if self.trash_intermediate and self.work_api != "containers": @@ -377,11 +407,13 @@ class ArvCwlRunner(object): # Reload tool object which may have been updated by # upload_workflow_deps + # Don't validate this time because it will just print redundant errors. tool = self.arv_make_tool(tool.doc_loader.idx[tool.tool["id"]], makeTool=self.arv_make_tool, loader=tool.doc_loader, avsc_names=tool.doc_schema, - metadata=tool.metadata) + metadata=tool.metadata, + do_validate=False) # Upload local file references in the job order. job_order = upload_job_order(self, "%s input" % kwargs["name"], @@ -476,13 +508,15 @@ class ArvCwlRunner(object): logger.info("Pipeline instance %s", self.pipeline["uuid"]) if runnerjob and not kwargs.get("wait"): - runnerjob.run(wait=kwargs.get("wait")) + runnerjob.run(**kwargs) return (runnerjob.uuid, "success") self.poll_api = arvados.api('v1') self.polling_thread = threading.Thread(target=self.poll_states) self.polling_thread.start() + self.task_queue = TaskQueue(self.workflow_eval_lock, self.thread_count) + if runnerjob: jobiter = iter((runnerjob,)) else: @@ -493,10 +527,11 @@ class ArvCwlRunner(object): **kwargs) try: - self.cond.acquire() - # Will continue to hold the lock for the duration of this code - # except when in cond.wait(), at which point on_message can update - # job state and process output callbacks. + 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__() @@ -506,20 +541,25 @@ class ArvCwlRunner(object): 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"): - runnable.run(**kwargs) + self.start_run(runnable, kwargs) else: - if self.processes: - self.cond.wait(1) + if (self.task_queue.in_flight + len(self.processes)) > 0: + self.workflow_eval_lock.wait(3) else: - logger.error("Workflow is deadlocked, no runnable jobs and not waiting on any pending jobs.") + logger.error("Workflow is deadlocked, no runnable jobs and not waiting on any pendingjobs.") break loopperf.__enter__() loopperf.__exit__() - while self.processes: - self.cond.wait(1) + 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 @@ -535,9 +575,11 @@ class ArvCwlRunner(object): self.api.container_requests().update(uuid=runnerjob.uuid, body={"priority": "0"}).execute(num_retries=self.num_retries) finally: - self.cond.release() + 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.") @@ -679,6 +721,17 @@ def arg_parser(): # type: () -> argparse.ArgumentParser help="Workflow priority (range 1..1000, higher has precedence over lower, containers api only)", default=DEFAULT_PRIORITY) + parser.add_argument("--disable-validate", dest="do_validate", + action="store_false", default=True, + help=argparse.SUPPRESS) + + parser.add_argument("--disable-js-validation", + action="store_true", default=False, + help=argparse.SUPPRESS) + + parser.add_argument("--thread-count", type=int, + default=4, help="Number of threads to use for job submit and output collection.") + exgroup = parser.add_mutually_exclusive_group() exgroup.add_argument("--trash-intermediate", action="store_true", default=False, dest="trash_intermediate", @@ -735,7 +788,8 @@ def main(args, stdout, stderr, api_client=None, keep_client=None): try: if api_client is None: - api_client=arvados.api('v1', model=OrderedJsonModel()) + api_client = arvados.safeapi.ThreadSafeApiCache(api_params={"model": OrderedJsonModel()}, keep_params={"num_retries": 4}) + keep_client = api_client.keep if keep_client is None: keep_client = arvados.keep.KeepClient(api_client=api_client, num_retries=4) runner = ArvCwlRunner(api_client, work_api=arvargs.work_api, keep_client=keep_client,