DanielLeens commented on issue #13322:
URL: https://github.com/apache/seatunnel/issues/13322#issuecomment-6983647120
I reproduced the two queue behaviors on
`89f91a89af15f33d97a942d8b65f4fc1b0c80bf5` (Java 11.0.27), using actual
coordinator methods with an injected storage failure.
Steps:
1. Set up a coordinator with an empty `completedCheckpointIds` queue and a
valid checkpoint ID 9.
2. Make the storage dependency's `storeCheckPoint` throw
`RuntimeException("injected storage write failure")`.
3. Call real `completePendingCheckpoint`. The exception propagates, but the
failed ID is already in the queue: `[9]`.
4. Call real `cleanPendingCheckpoint(CHECKPOINT_COORDINATOR_RESET)`. The
queue remains `[9]`.
[Executable Java probe and exact build/run
commands](https://github.com/apache/seatunnel/issues/13313#issuecomment-6983599115):
see `issue13322()`.
Scope: this confirms enqueue-before-success and retention of that queue
across RESET. I have **not** independently measured the claimed ~2x on-disk
retention, partial files, or subsequent deletion behavior. This is a focused
fault-injection test, not a whole-cluster restore test.
--
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]