Github user HeartSaVioR commented on the issue: https://github.com/apache/spark/pull/20675 Looks like the patch is outdated, and when continuous query supports shuffled stateful operators, implementing task level retry is not that trivial. To get correct result of aggregation, when one of task fails at epoch N, all the tasks and states should be restored to epoch N. I definitely agree that it would be ideal to have stable task level retry, just wondering this patch would work with follow-up tasks for continuous mode.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org