Github user vanzin commented on a diff in the pull request: https://github.com/apache/spark/pull/17882#discussion_r120710864 --- Diff: resource-managers/yarn/src/main/scala/org/apache/spark/scheduler/cluster/YarnSchedulerBackend.scala --- @@ -307,6 +301,9 @@ private[spark] abstract class YarnSchedulerBackend( case RetrieveLastAllocatedExecutorId => context.reply(currentExecutorIdCounter) + + case RetrieveLastRequestExecutors => --- End diff -- Why make a new call for this? What I had in mind would re-use `RetrieveLastAllocatedExecutorId` adding a new properties that tell the AM what's the initial state. Might be good to change the name of the message at that point to (e.g. `GetAMInitialState` or something). That call is already made during initialization of `YarnAllocator`, so it should be a smaller change overall. So, you'd have one already existing RPC instead of one RPC that may cause the AM to send another RPC to signal the driver to send an RPC to the AM (my head is spinning now).
--- 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 infrastruct...@apache.org or file a JIRA ticket with INFRA. --- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org