Github user squito commented on a diff in the pull request:
https://github.com/apache/spark/pull/14079#discussion_r72284207
--- Diff: core/src/main/scala/org/apache/spark/TaskEndReason.scala ---
@@ -204,6 +213,7 @@ case object TaskResultLost extends TaskFailedReason {
@DeveloperApi
case object TaskKilled extends TaskFailedReason {
override def toErrorString: String = "TaskKilled (killed intentionally)"
+ override def countTowardsTaskFailures: Boolean = false
--- End diff --
the default in `TaskFailedReason` is true, so it will be true for
`ExceptionFailure`, `UnknownReason`, `ExecutorLost` if caused by app (eg., an
OOM), `TaskResultLost`, and `Resubmitted`. I admit I'm not certain if those
last two should count or not, but changing that behavior should probably be
separate from this change. I'm not changing the end behavior here, just trying
to refactor it to make it easier to follow and handle correctly with the
blacklist (it was special-cased before:
https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/scheduler/TaskSetManager.scala#L787)
Or do you mean we should change it to a `val` instead, since its a
constant? then I'd change `TaskCommitDenied` to use a `val` as well.
---
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]
