zhangshenghang opened a new issue, #13321:
URL: https://github.com/apache/seatunnel/issues/13321

   ### What happened
   
   One corrupted stored checkpoint aborts the entire restore, even when older 
healthy checkpoints exist — there is no fallback.
   
   Chain (dev @ 89f91a89af):
   
   1. `CheckpointManager.getLatestCheckpointStateByType` 
(`CheckpointManager.java:220-236`) streams **all** stored `PipelineState` 
candidates through a filter that calls `deserializeCheckpoint(state)` on each.
   2. `deserializeCheckpoint` (`CheckpointManager.java:238-244`) rethrows any 
deserialization failure as `RuntimeException`.
   3. The stream predicate throws on the corrupted candidate, the constructor's 
`parallelStream` (`CheckpointManager.java:157-217`) sneaky-throws it, and 
`CheckpointManager` construction fails — job submission / failover restore of 
the whole job fails.
   
   So a single damaged newest (or middle) checkpoint file makes the job 
unrestorable despite healthy older checkpoints (Flink, by comparison, skips 
corrupt savepoints and falls back).
   
   ### Trigger
   
   Corrupted/truncated checkpoint file in storage (partial write, disk issue, 
interrupted upload).
   
   ### Impact
   
   Job cannot be restored from any checkpoint; requires manual storage surgery.
   
   ### Suggested fix
   
   Catch per candidate: log the corruption, skip the candidate, and continue 
with the remaining (older) ones; only fail when no healthy candidate remains.
   
   ### SeaTunnel Version
   
   dev (89f91a89af), Zeta engine
   


-- 
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