Pankraz76 commented on PR #23947: URL: https://github.com/apache/kafka/pull/23947#issuecomment-6384734221
Thanks to Alyssa Huang for reporting this and to Artem Livshits for the analysis on the JIRA — this follows the approach sketched there, keeping the existing `Optional` chain style. I noticed #22680 targets the same JIRA. Happy to coordinate or close this one in favor of whichever reviewers prefer; the main addition here is an end-to-end integration test that reproduces the lost commit marker. A few questions for reviewers: - Is the integration test in `TransactionsTest` the right level and place, or would you prefer only the unit tests? - While a partition is offline, the parked marker is re-checked in a tight loop by the send thread (pre-existing unknown-broker queue behavior, also present before KAFKA-17573). Should I file a follow-up JIRA for that? - `MetadataCache.toCluster` seems to have a similar KAFKA-17573 regression: partitions without a live leader are now omitted from the `Cluster` instead of being reported with `Node.noNode()` as leader. Worth a separate JIRA? - Given the Blocker priority, should this be backported to 4.3 (and 4.4)? Thanks for taking a look! -- 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]
