Skip to content

Commit effb1ce

Browse files
committed
Move log message.
1 parent 49282b3 commit effb1ce

File tree

1 file changed

+4
-4
lines changed

1 file changed

+4
-4
lines changed

core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala

+4-4
Original file line numberDiff line numberDiff line change
@@ -1046,22 +1046,22 @@ class DAGScheduler(
10461046

10471047
case FetchFailed(bmAddress, shuffleId, mapId, reduceId) =>
10481048
val failedStage = stageIdToStage(task.stageId)
1049-
val mapStage = shuffleToMapStage(shuffleId)
10501049
// It is likely that we receive multiple FetchFailed for a single stage (because we have
10511050
// multiple tasks running concurrently on different executors). In that case, it is possible
10521051
// the fetch failure has already been handled by the scheduler.
10531052
if (runningStages.contains(failedStage)) {
10541053
markStageAsFinished(failedStage, Some("Fetch failure"))
10551054
runningStages -= failedStage
1056-
// TODO: Cancel running tasks in the stage
1057-
logInfo(s"Marking $failedStage (${failedStage.name}) for resubmision " +
1058-
s"due to a fetch failure from $mapStage (${mapStage.name}")
10591055
}
10601056

1057+
val mapStage = shuffleToMapStage(shuffleId)
10611058
if (failedStages.isEmpty && eventProcessActor != null) {
10621059
// Don't schedule an event to resubmit failed stages if failed isn't empty, because
10631060
// in that case the event will already have been scheduled. eventProcessActor may be
10641061
// null during unit tests.
1062+
// TODO: Cancel running tasks in the stage
1063+
logInfo(s"Marking $failedStage (${failedStage.name}) for resubmision " +
1064+
s"due to a fetch failure from $mapStage (${mapStage.name}")
10651065
import env.actorSystem.dispatcher
10661066
env.actorSystem.scheduler.scheduleOnce(
10671067
RESUBMIT_TIMEOUT, eventProcessActor, ResubmitFailedStages)

0 commit comments

Comments
 (0)