Github user liyinan926 commented on the issue: https://github.com/apache/spark/pull/21748 > I wonder if we want to have the pod name owner reference still be a thing, if you will, in client mode. For example what if the pod name that is given is accidentally one that is assigned to a different running pod in that namespace? Seems simpler to only require the pod name to be set in cluster mode, and in client mode, the user is responsible for their own cleanup. I think we should still keep the OwnerReference if there's a driver pod. But we need to document clearly that an `OwnerReference` will be added to the executor pods so it's important to set the right driver pod name. We already have an unsolved problem of cleanup of executors in out-cluster client mode. Making the problem even bigger to also include in-cluster client mode is much less desirable than risking having user errors.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org