Github user squito commented on a diff in the pull request: https://github.com/apache/spark/pull/21758#discussion_r209652176 --- Diff: core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala --- @@ -1349,6 +1339,29 @@ class DAGScheduler( s"longer running") } + if (mapStage.rdd.isBarrier()) { --- End diff -- I'd also like to see a test case for what happens if you do not have enough compute resources after a stage failure. After all, one of the key reasons for multiple stage attempts is hardware failure -- so you might have had just enough resources to run your barrier stage with 20 executors on the first attempt, but on the second-attempt you're down to 19, and can't run it anymore.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org