slachiewicz opened a new pull request, #313: URL: https://github.com/apache/flink-connector-kafka/pull/313
Under `read_committed`, `endOffsets` is the last stable offset, which trails the high watermark until the broker has written the markers of a transaction it already acknowledged. The helper snapshotted it once, right after the job finished, and cut off the final transaction in `rescaleListing` (run 31916828021, `[2->5]`). It now waits up to 10 s for the LSO to meet the `read_uncommitted` end offsets and logs the pinned offsets if it does not, so a truncated drain names its cause. Call sites that drain with transactions legitimately open (`testAbortTransactionsAfterScaleInBeforeFirstCheckpoint`, `ExactlyOnceKafkaWriterITCase`) keep their result and pay the bounded wait. The other failure mode in FLINK-40585 (run 30181683657, `[3->5]`) is a product bug, tracked separately as FLINK-40626. Verified: `KafkaSinkITCase#rescaleListing` → 10/10 locally. *This change was created with AI assistance.* -- 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]
