kirktrue commented on PR #19886: URL: https://github.com/apache/kafka/pull/19886#issuecomment-2964612221
Thanks for the PR, @ShivsundarR. Since this change affects all consumer use cases, we need to tread carefully 😄 The console consumer is closing the underlying consumer with the default `close()` API, which uses a timeout of 30 seconds. That means the consumer is within its right to use up all 30 seconds to clean up. The change as is looks a little too broad because it assumes that because a node isn't immediately available that it should abort the request. This doesn't allow for the case where a node later becomes available within the timeout. There is a dependency with, e.g. `OFFSET_COMMIT` and `FIND_COORDINATOR`, so if the user is trying to commit their offsets, IMO we should try exhaustively to find the coordinator. That said, the `close()` process is rife with edge cases and I'm sure there are things to fix. Is there an approach by which we can "prove" that continuing to make requests is pointless? -- 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
