eskabetxe commented on PR #29422: URL: https://github.com/apache/flink/pull/29422#issuecomment-6200668084
> @eskabetxe How do you expect `stop-with-savepoint --drain` to help when switching back? My understanding is that drain closes the open files, but the savepoint is still taken before the commit, so the committer state still holds the commit recoverables of the final files, which the Hadoop serializer can't read. Is there a case I'm missing where that state ends up empty? You are right, drain doesn't help by itself, my doc sentence was wrong, removed it. drain does get all the data committed, then restore that savepoint with the sink state dropped, by changing the sink uids and --allowNonRestoredState, sources keep their offsets. Plain restart with no savepoint also works but replays/loses depending on source startup. if you want, I can put that recipe in the docs, or just keep "not supported" as it is now -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
