Github user mridulm commented on a diff in the pull request:
https://github.com/apache/spark/pull/214#discussion_r10949757
--- Diff:
core/src/main/scala/org/apache/spark/scheduler/TaskSchedulerImpl.scala ---
@@ -198,6 +201,13 @@ private[spark] class TaskSchedulerImpl(
*/
def resourceOffers(offers: Seq[WorkerOffer]): Seq[Seq[TaskDescription]]
= synchronized {
SparkEnv.set(sc.env)
+ // Make thread pool local for shutdown before the function returns
+ // This is for driver can exit normally which not call sc.stop or
sys.exit
+ val serializeWorkerPool = new ThreadPoolExecutor(
+ conf.getInt("spark.scheduler.task.serialize.threads.min", 20),
+ conf.getInt("spark.scheduler.task.serialize.threads.max", 60),
+ conf.getInt("spark.scheduler.task.serialize.threads.keepalive", 60),
TimeUnit.SECONDS,
+ new LinkedBlockingDeque[Runnable]())
--- End diff --
Maybe do this only if there is expected to be an overhead ? Say number of
tasks which can be submitted > some threshold ?
Also, move the pool out of the method as others have suggested.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
