Ngone51 commented on a change in pull request #23871:
[SPARK-23433][SPARK-25250] [CORE] Later created TaskSet should learn about the
finished partitions
URL: https://github.com/apache/spark/pull/23871#discussion_r261237163
##########
File path:
core/src/main/scala/org/apache/spark/scheduler/TaskSchedulerImpl.scala
##########
@@ -292,13 +298,16 @@ private[spark] class TaskSchedulerImpl(
* given TaskSetManager have completed, so state associated with the
TaskSetManager should be
* cleaned up.
*/
- def taskSetFinished(manager: TaskSetManager): Unit = synchronized {
+ def taskSetFinished(manager: TaskSetManager, success: Boolean): Unit =
synchronized {
Review comment:
When a zombie TaskSetManager finished, it could also lead to empty
TaskSetManager for the stage(just see `taskSetFinished()`). So, if we clean up
the `stageIdToFinishedPartitions` when there is no more task set manager for
the stage, the active TaskSetManager which is created later would lost finished
partitions. WDYT ? @cloud-fan
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
