mjsax commented on PR #22595: URL: https://github.com/apache/kafka/pull/22595#issuecomment-4748952386
While digging into more details, I realized that I was a little bit too conservative. The existing log to estimate "offsets" and use "zero" (ie, omit) an unknown partition is actually sufficient. In the end, we can re-use the existing `StateDirectory` cache after all -- it's refreshed regularly, and it's a concurrent-map, so it's already thread-safe. Accessing it directly should be fine, as we don't need "snapshot consistency". It also contains data for all tasks. -- The only thing I change from "classic" is active-running tasks; as we report them via "currently assigned tasks" anyway, and they are caught up be definition, there is no need to include them in task-offset-sum IMHO (in the "classic" case, they are included with surrogate `offset-sum=-2` ). Thoughts? -- 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]
