CryoThrust commented on issue #12207: URL: https://github.com/apache/seatunnel/issues/12207#issuecomment-5594637332
Thanks for pointing to the STIP process. Before implementation is treated as accepted, I would suggest making a few contract points explicit in the STIP/design record:\n\n- Event identity and idempotency: define a stable key and how retries/restarts avoid duplicate START or terminal events.\n- Delivery semantics: document whether the backend is best-effort, bounded retry, or durable outbox; lineage reporting should not block job scheduling or checkpoint completion.\n- Failure isolation: backend/receiver timeouts, malformed responses, and queue saturation should be observable but must not fail an otherwise healthy data job.\n- Dataset identity: state how catalog, database/schema, table, and multi-table sink names map to OpenLineage namespace/name, including normalization and collision handling.\n- Statistics semantics: keep the distinction between committed and attempted output counts in the event facets, especially after retries or partial failures.\n- Configuration validation: r eject impossible combinations early, but keep fully inert with no receiver connection attempt.\n\nThese choices are important for Zeta and Flink to remain behaviorally consistent even though their lifecycle hooks differ. I am happy to review the STIP draft or take a focused test-contract slice once the proposal is accepted. -- 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]
