DanielLeens commented on issue #12382:
URL: https://github.com/apache/seatunnel/issues/12382#issuecomment-5832128081

   Classification: A / Connector-V2 PostgreSQL CDC restore-delivery integrity.
   
   Thanks for running the real Zeta savepoint -> restore -> post-reattachment 
sequence and for being explicit that the passing experimental-resume captures 
neither reproduce the loss nor compare old and new code. That is the right 
conclusion: there is not yet evidence for a production fix.
   
   The original failing scheduled run is available at 
https://github.com/apache/seatunnel/actions/runs/35244548330/job/105282609961, 
on `dev` commit `c7304ace6e18d350314e92480df1fd3c0962f1f2`. Current `dev` 
already contains merged #11864 (`5af8d789aa9ae3d94df4a9cc0f03cee3c0a6d0e6`), 
which changes the committed-LSN and equal-watermark boundary. It does not by 
itself prove or disprove a later handoff loss between WAL decoding, reader 
emission, and JDBC-sink delivery.
   
   Please use those two revisions as the comparison boundary. First make the 
existing `id=15` restore assertion deterministic at the failing revision and 
current `dev`; then add test-only, correlation-safe evidence for the single 
inserted row at these three points: `PostgresWalFetchTask` handoff, reader 
emission/checkpoint-offset advancement, and JDBC-sink receipt/commit. The 
evidence needs to show whether the row is absent before emission, after 
emission, or only after the sink boundary. Record the exact test revision and 
any experimental-resume changes separately.
   
   Please do not open a production-fix PR or add temporary production tracing 
until that boundary is demonstrated. A focused regression plus the smallest 
corresponding fix is the next acceptable path; there is no open PR for this 
exact signature today.
   


-- 
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