Github user squito commented on a diff in the pull request:
https://github.com/apache/spark/pull/5636#discussion_r29312685
--- Diff: core/src/main/scala/org/apache/spark/scheduler/Stage.scala ---
@@ -70,6 +70,26 @@ private[spark] abstract class Stage(
/** Pointer to the latest [StageInfo] object, set by DAGScheduler. */
var latestInfo: StageInfo = StageInfo.fromStage(this)
+ // The maximum number of times to retry a stage before aborting
+ val maxStageFailures = 4
+
+ // To avoid cyclical stage failures (see SPARK-5945) we limit the number
of times that a stage
+ // may be retried.
--- End diff --
I think "cyclical" is a little confusing here. I think you could probably
leave the comment out completely, or maybe something like: "Spark is resilient
to executors dying, by retrying stages on FetchFailures. However, we don't
want to retry stages indefinitely, so keep track of the number of failures so
we can eventually bail"
---
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]
