Github user squito commented on a diff in the pull request:
https://github.com/apache/spark/pull/6817#discussion_r32439981
--- Diff:
core/src/main/scala/org/apache/spark/ExecutorAllocationManager.scala ---
@@ -537,10 +537,19 @@ private[spark] class ExecutorAllocationManager(
}
}
+ override def onTaskResubmit(taskResubmit: SparkListenerTaskResubmit):
Unit = {
+ val stageId = taskResubmit.stageId
+ allocationManager.synchronized {
+ val num = stageIdToNumTasks.getOrElse(stageId, 0)
+ stageIdToNumTasks.update(stageId, num + 1)
+ }
+ }
+
--- End diff --
I don't see why this is necessary -- when an executor goes down, and stages
get resubmitted, [the stage gets
resubmitted](https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala#L1088)
which will [post another `StageSubmitted`
event](https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala#L874).
It looks to me like that will already handle this -- so won't this double
count?
---
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.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
