Github user markhamstra commented on a diff in the pull request:
https://github.com/apache/spark/pull/11720#discussion_r57668558
--- Diff: core/src/main/scala/org/apache/spark/scheduler/DAGScheduler.scala
---
@@ -286,6 +286,7 @@ class DAGScheduler(
case None =>
// We are going to register ancestor shuffle dependencies
getAncestorShuffleDependencies(shuffleDep.rdd).foreach { dep =>
+ assert(!shuffleToMapStage.get(dep.shuffleId).isDefined)
--- End diff --
Yes, I tend to agree that the way what really should be the same Stage gets
recreated as a new Stage is undesirable in the DAGScheduler, but the more
important point is that, even though it is not optimal, it does produce correct
results, and we can't change that for the sake of code that seems more
desirable.
That's not to say that some of the efforts that were begun to fix this kind
of behavior while still generating correct results shouldn't be driven to
completion.
---
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]
