tgravescs commented on a change in pull request #27636:
[SPARK-30873][CORE][YARN]Handling Node Decommissioning for Yarn cluster manger
in Spark
URL: https://github.com/apache/spark/pull/27636#discussion_r389694668
##########
File path: core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala
##########
@@ -1525,10 +1539,27 @@ private[spark] class DAGScheduler(
s" ${task.stageAttemptId} and there is a more recent attempt for
that stage " +
s"(attempt ${failedStage.latestInfo.attemptNumber}) running")
} else {
+ // Gracefully handling the stage abort due to fetch failure in the
+ // decommission nodes
+ if
(!event.reason.asInstanceOf[FetchFailed].countTowardsStageFailures) {
+ // Ignore stage attempts due to fetch failed only
+ // once per attempt
+ if (!failedStage.failedAttemptIds.contains(task.stageAttemptId)) {
+ failedStage.ignoredFailedStageAttempts += 1
+ DecommissionTracker.incrFetchFailIgnoreCnt()
+ failedStage.latestInfo.stageFailureIgnored(true)
+ }
+ logInfo(s"""Ignoring stage failure due to fetch failed from the
decommissioned""" +
Review comment:
should this log message be inside the if statement above. you are just
inside a conditional for fetch failure !countTowardsStageFailures. That doesn't
mean its from decomissioning. I haven't seen anything specified here that
marks it for decommissioning case
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to 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]
