DanielLeens commented on issue #10899:
URL: https://github.com/apache/seatunnel/issues/10899#issuecomment-4588880532

   Thanks for linking PR #10987 back here.
   
   I rechecked the fix direction against the issue path we discussed earlier, 
and the approach makes sense to me:
   - keep `startup.mode = timestamp` only for the initial bootstrap decision;
   - once the split has already been checkpointed with a concrete binlog 
file/position, restore from that persisted split offset instead of re-deriving 
from the original `startup.timestamp`;
   - skip the timestamp-based re-filter on the restore path when a concrete 
restored binlog position already exists.
   
   That is aligned with the core bug here: recovery should prioritize the 
checkpointed incremental split offset, while timestamp lookup should remain a 
first-start concern rather than the long-running recovery anchor.
   
   I would keep the issue open until PR #10987 is merged, but from the 
issue-analysis side this looks like the right fix direction. Please also make 
sure the PR includes a regression test that covers checkpoint restore after a 
timestamp bootstrap, so we do not regress this path again later.


-- 
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