You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@kafka.apache.org by GitBox <gi...@apache.org> on 2021/03/18 03:09:52 UTC

[GitHub] [kafka] C0urante edited a comment on pull request #10345: KAFKA-12477: dynamically upgrade rebalancing protocol based on selected assignor

C0urante edited a comment on pull request #10345:
URL: https://github.com/apache/kafka/pull/10345#issuecomment-801583332


   Thanks for the ping @ableegoldman! If I'm reading this correctly, the entire consumer group will throw an exception from `poll` whenever a new member joins with an outdated protocol, except for that new member, which doesn't know that anything is wrong since it hasn't detected the downgrade in protocol.
   
   If that's correct, I'm wondering if there's a way to soften the blow here... would it be viable for a consumer to permit and handle the downgrade in protocol by revoking all previously-held partitions and then claiming only the ones provided in the new assignment? Given the choice between a revert to eager rebalancing and having most of the group die the latter seems like it might be more user-friendly.
   
   I'm optimistic that if we do that in `onJoinComplete` we can avoid having multiple consumers process the same topic partition(s) at once, but haven't been able to confirm this just yet.


----------------------------------------------------------------
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org