SEZ9 commented on PR #11932:
URL: https://github.com/apache/seatunnel/pull/11932#issuecomment-5659094855

   Quick status recap on this PR, since the head is still at `b6c96b97356d` 
with no new commit to re-review:
   
   **F7 (Collector.restoreSchema no-op on Flink/Spark)** — fine to close this 
one as a documented limitation rather than a blocker, on the condition that a 
follow-up issue for the Flink/Spark side is filed and linked here in the 
thread. Once that link is in place, F7 is fully closed from my side.
   
   Everything else remains open exactly as previously stated:
   
   1. **F1 / F2 — `IncrementalSourceReader`**: guard the schema restore against 
a non-null-but-empty `checkpointTables` (so we never roll the 
deserializer/collector back to an empty schema), and gate 
`restoreCheckpointProducedType` and the collector-side `restoreSchema` on the 
same condition so the two restores cannot diverge.
   2. **F3 / F6 / F8 — `JdbcSink`**: make schema selection in `restoreWriter` 
deterministic (or fail fast when merged states carry differing schemas), keep 
the primary-key/upsert path when a restored schema lacks a `PrimaryKey` instead 
of silently downgrading to insert-only, and reject or filter null-Xid states 
before they reach `JdbcExactlyOnceSinkWriter` after an `exactly_once` flip.
   3. **F4 / F5 — `JdbcSinkState` / `JdbcSinkWriter`**: add a docs/changelog 
note describing the new checkpoint state format (schema persisted in 
`JdbcSinkState`, non-exactly-once writer now emitting state), and confirm that 
the persisted `TableSchema`/column payloads are Serializable-safe so 
checkpoints that previously carried no writer state don't start failing.
   
   Please push a commit addressing the above (or reply inline on any point you 
disagree with) and drop the follow-up issue link for F7 — I'll do the full 
re-review as soon as that lands. Thanks for the work so far!
   
   <!-- streview-comment:1042 -->


-- 
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]

Reply via email to