DanielLeens commented on issue #12116: URL: https://github.com/apache/seatunnel/issues/12116#issuecomment-5552994705
Supporting evidence from the Flink translation layer (`dev`, `seatunnel-translation-flink-common/.../sink/FlinkSink.java` lines 78–88): `FlinkSink#createWriter` wraps `sink.createWriter(stContext)` in one `FlinkSinkWriter` per Flink subtask, with no per-table routing at that level — so every `AssertSinkWriter` instance in a Flink TaskManager receives whatever rows reach its subtask and accumulates into the same static `LONG_ACCUMULATOR`/`TABLE_NAMES`, then evaluates its rules in its own `close()`. Under Zeta each job has its own classloader, so the statics are effectively isolated there, which is why only Flink legs fail. Reproduction count so far on the same dev test: PR #11458 (Flink 1.18.0: 97/125/150), #11077 (Flink 1.15.3: 75/125/150), #11503 (Flink 1.15.3 + one more Flink leg: 87/125/150). -- 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]
