frankvicky opened a new pull request, #23663: URL: https://github.com/apache/kafka/pull/23663
JIRA: KAFKA-18905 The producer now tracks, per partition, how many batches it has sent after the oldest in-flight batch, including the ones that have already been acknowledged. It stops handing out new sequences for that partition once that number reaches 4, so the oldest batch can never fall out of the broker's window before it is resolved. Retries are not affected by the new check. On a healthy connection the broker answers requests in order, so this condition is never hit and nothing changes. It only kicks in during a leader failover, where new batches for the affected partition are held back until the stuck batch is resolved, at most `request.timeout.ms`. This trades a bounded delay on one partition for the batch no longer failing spuriously. -- 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]
