Github user viirya commented on a diff in the pull request:
https://github.com/apache/spark/pull/21494#discussion_r193648185
--- Diff: core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala
---
@@ -1310,6 +1311,44 @@ class DAGScheduler(
}
}
+ case failure: TaskFailedReason if task.isBarrier =>
+ // Always fail the current stage and retry all the tasks when a
barrier task fail.
+ val failedStage = stageIdToStage(task.stageId)
+ logInfo(s"Marking $failedStage (${failedStage.name}) as failed " +
+ "due to a barrier task failed.")
+ val message = "Stage failed because a barrier task finished
unsuccessfully. " +
+ s"${failure.toErrorString}"
+ try { // cancelTasks will fail if a SchedulerBackend does not
implement killTask
+ taskScheduler.cancelTasks(stageId, interruptThread = false)
+ } catch {
+ case e: UnsupportedOperationException =>
+ logInfo(s"Could not cancel tasks for stage $stageId", e)
--- End diff --
Under barrier execution, will it be a problem if we can not cancel tasks?
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
