unknowntpo commented on PR #23357: URL: https://github.com/apache/kafka/pull/23357#issuecomment-5542567794
Good catch. I do not think the zero wait is intentional. In both paths, `retry.backoff.ms` is used as a bound while the application thread is waiting for background progress; it does not mean that useful work is immediately available. If it is configured as 0, `FetchRequestManager.maximumTimeToWait()` can feed 0 directly into `pollForFetches()`, and the state-based clamps inside `pollForFetches()` can also reduce a positive timeout to 0. `fetchBuffer.awaitWakeup()` then returns immediately and the outer poll loop can repeat without progress. So the same positive floor seems appropriate. I would keep it separate from this PR, though: this change is specifically about the heartbeat-timer/request-in-flight state, whereas the fetch paths are in the KAFKA-20854 area and need their own focused tests. I suggest tracking that as a follow-up rather than broadening KAFKA-21031. -- 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]
