Github user jiangxb1987 commented on a diff in the pull request: https://github.com/apache/spark/pull/22112#discussion_r210449640 --- Diff: core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala --- @@ -1441,6 +1441,18 @@ class DAGScheduler( failedStages += failedStage failedStages += mapStage if (noResubmitEnqueued) { + if (!mapStage.rdd.isIdempotent) { + // The map stage is not idempotent, we have to rerun all the tasks for the + // failed stage to get expected result. + failedStage match { + case s: ShuffleMapStage => --- End diff -- Like we discussed, we shall also retry the partially finished succeeding stages.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org