cloud-fan commented on a change in pull request #27752: [SPARK-30999][SQL]
Don't cancel a QueryStageExec which is already finished
URL: https://github.com/apache/spark/pull/27752#discussion_r386201503
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/adaptive/AdaptiveSparkPlanExec.scala
##########
@@ -158,25 +158,34 @@ case class AdaptiveSparkPlanExec(
var result = createQueryStages(currentPhysicalPlan)
val events = new LinkedBlockingQueue[StageMaterializationEvent]()
val errors = new mutable.ArrayBuffer[SparkException]()
+ val runningStages = util.Collections.synchronizedSet(new
util.HashSet[QueryStageExec]())
var stagesToReplace = Seq.empty[QueryStageExec]
while (!result.allChildStagesMaterialized) {
currentPhysicalPlan = result.newPlan
if (result.newStages.nonEmpty) {
stagesToReplace = result.newStages ++ stagesToReplace
executionId.foreach(onUpdatePlan)
- // Start materialization of all new stages.
- result.newStages.foreach { stage =>
+ // Start materialization of all new stages and fail fast if any
stages failed eagerly
+ var anyFail = false
+ val stages = result.newStages.toIterator
+ while (!anyFail && stages.hasNext) {
+ val stage = stages.next()
+ runningStages.add(stage)
try {
stage.materialize().onComplete { res =>
+ runningStages.remove(stage)
if (res.isSuccess) {
events.offer(StageSuccess(stage, res.get))
} else {
events.offer(StageFailure(stage, res.failed.get))
}
}(AdaptiveSparkPlanExec.executionContext)
} catch {
- case e: Throwable => events.offer(StageFailure(stage, e))
+ case e: Throwable =>
+ anyFail = true
+ runningStages.remove(stage)
Review comment:
This is the one we need to track, as `QueryStageExec` can't tell if it fails
when initializing lazy val.
Can we only track one stage which fails at the beginning? (as there can only
be one)
----------------------------------------------------------------
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]
