SEZ9 commented on issue #11735: URL: https://github.com/apache/seatunnel/issues/11735#issuecomment-5770662634
@goutamadwant thanks for the two rounds of updates. The v1 boundary reads much tighter now: pipeline-attempt grouping as the primary model, atomic deduplication, explicit 4 KiB / 64 KiB limits for message and stack trace, a single bounded finished-job snapshot, no stored worker-address attribution, and the existing `errorMessage` contract left intact. A few points from the earlier review I'd still like spelled out in the STIP text before treating the design as settled: 1. **Dedupe key.** The first update used `SubPlan.pipelineRestoreNum` as the single HA-durable attempt counter; the latest one says the diagnostic attempt identity is now separate from the retry counter. Please restate the exact dedupe key composition on top of that identity, and confirm the behaviour on repeated delivery of the same terminal state is still "first write wins, later identical writes are no-ops". 2. **Truncation metadata.** Please confirm the snapshot carries explicit truncated flags / original lengths for message and stack trace, not just the clipped text. 3. **Finished-job snapshot persistence.** Please confirm persistence of the bounded snapshot is best-effort and cannot interfere with the original fail / restore path, and note how the storage stays bounded for finished-job backends. Once those are in, I think the acceptance criteria are sufficient for a v1. Assignment and the feature / design labels are up to the maintainers. Agreed this stays a design proposal until there is agreement here and the implementation lands. <!-- streview-comment:1227 --> -- 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]
