Github user kayousterhout commented on a diff in the pull request:
https://github.com/apache/spark/pull/5636#discussion_r29525425
--- Diff: core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala
---
@@ -1085,6 +1085,10 @@ class DAGScheduler(
if (disallowStageRetryForTest) {
abortStage(failedStage, "Fetch failure will not retry stage due
to testing config")
+ } else if (failedStage.failAndShouldAbort()) {
--- End diff --
Ohh this is a really ugly issue and I think something we should change. I
think this happens because we don't cancel other tasks in the stage when we get
a FetchFailed exception, so they can keep running. I'm guessing this can
happen if the FetchFailed exception happens for some weird non-deterministic
reason (or possibly because only a small number of the shuffle blocks are too
large), in which case the rest of the tasks in the failed stage will keep
running until completion.
I wonder if we should fix that issue before fixing this one? This is tricky
with the release timing.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
