DanielLeens commented on issue #10926:
URL: https://github.com/apache/seatunnel/issues/10926#issuecomment-4597810345

   I checked the current Zeta-side restore path in the codebase, and this does 
look like a real capability gap rather than a connector-specific bug.
   
   Today the operator-facing restore entry is still explicitly modeled as 
savepoint-based restore by job id. The CLI already exposes `-r/--restore`, but 
that path is tied to an explicit restore job id, while checkpoint 
overview/history are mainly inspection utilities. I do not see a clear upstream 
contract yet for the different case described here: a streaming job is 
force-stopped externally, and the next run should resume from the latest 
successful checkpoint automatically or through a clearly supported restart 
workflow.
   
   So I think this is worth tracking as a Zeta feature request.
   
   The most useful way to narrow it would be:
   1. keep the first scope Zeta-only
   2. define how the latest successful checkpoint is discovered and bound to 
the next run
   3. make the fallback behavior explicit when the checkpoint has been cleaned 
up, is incompatible, or restore fails partway
   4. document the source / transform / sink restore expectations for this 
restart path, so operators are not guessing from implementation details
   
   We already have important building blocks in place on the state snapshot 
side, but the job-lifecycle-level recovery contract is still the missing piece 
here.
   
   We've kept this under `feature` + `help wanted` for now. Contributions are 
very welcome once the first scope is narrowed.


-- 
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