li3zhi4 commented on PR #11835: URL: https://github.com/apache/seatunnel/pull/11835#issuecomment-5351167108
@DanielLeens — thanks for the precise trace, and for the approval. Both failures you identified are addressed by the fix on head `22405c9e0`: 1. **`testSourceKafka` → `testKafkaSpecificOffsetsToConsole`** (the `id=49` at offset 50): the `startUp()` warm-up produced a throwaway record at offset 0 of `test_topic_source` and then `deleteRecords(beforeOffset(1))`, permanently consuming offset 0 so the real records landed at offsets 1..100. Fixed by dropping the warm-up call from `startUp()` entirely — the static topics are written by `generateTestData` right after, whose produce itself forces end-to-end leader propagation before any job is submitted, so nothing was lost by removing it. 2. **`testKafkaToKafkaExactlyOnceOnBatch`** (on `kafka_topic_exactly_batch_once_1/2`): same root cause — those two topics are also static topics created in `startUp()`, so they were equally affected by the warm-up's offset shift (the test asserts against `endOffsetOnP0`/absolute offsets). The same fix resolves them; the CI rerun on `22405c9e0` (`kafka-connector-it (8)`) is now fully green, which covers both test methods across all containers. `warmUpKafkaTopics` remains in `AbstractKafkaIT` for test classes whose topics are empty at job-submission time (the scenario it was designed for), but it is no longer invoked on static, absolute-offset-read topics in `KafkaIT`. -- 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]
