dajac opened a new pull request, #14120: URL: https://github.com/apache/kafka/pull/14120
This patch implements the OffsetFetch API in the new group coordinator. I found out that implementing the `RequireStable` flag is hard (to not say impossible) in the current model. For the context, the flag is here to ensure that an OffsetRequest request does not return stale offsets if there are pending offsets to be committed. In the scala code, we basically check the pending offsets data structure and if they are any pending offsets, we return the `UNSTABLE_OFFSET_COMMIT` error. This tells the consumer to retry. In our new model, we don't have the pending offsets data structure. Instead, we use a timeline data structure to handle all the pending/uncommitted changes. Because of this we don't know whether offsets are pending for a particular group. Instead of doing this, I propose to not return the `UNSTABLE_OFFSET_COMMIT` error anymore. Instead, when `RequireStable` is set, we use a write operation to ensure that we read the latest offsets. If they are uncommitted offsets, the write operation ensures that the response is only return when they are committed. This gives a similar behaviour in the end. Note that this PR does not adds the MemberId and MemberEpoch fields to the request. This will be done in the next PR. ### Committer Checklist (excluded from commit message) - [ ] Verify design and implementation - [ ] Verify test coverage and CI build status - [ ] Verify documentation (including upgrade notes) -- 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: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
