Caideyipi opened a new pull request, #18879:
URL: https://github.com/apache/iotdb/pull/18879

   ## Description
   
   `RecoverReadTest.consistentReadWithSlowApply` can pass its read assertions 
and then fail in teardown with `Unexpected exit`. The [Windows CI 
failure](https://github.com/apache/iotdb/actions/runs/38550272247/job/115806242034)
 shows `LeaderStateImpl.checkPeersForYieldingLeader()` reading a 
`SegmentedRaftLog` that shutdown has already closed.
   
   Pause elections on live mini-cluster peers and wait for leader stepdown 
before closing the servers. The stepdown runs through the leader event 
processor, allowing in-flight leadership checks to finish while the log is 
still open. Handle repeated stops and peers stopped earlier by a test, and 
retain cleanup in a `finally` block. The unexpected-exit assertion remains 
active.
   
   Add a regression that holds the leader's monitor during a simulated 
in-flight log check while another thread stops the cluster. It verifies that 
the state machine stays open until the check completes and that the log remains 
readable.
   
   ## Validation
   
   - The new regression fails against the original `MiniCluster.stop()` and 
passes with this change.
   - Windows / JDK 17: `mvn -nsu -B -ntp -Ddevelocity.off=true -pl 
iotdb-core/consensus clean test` — 78 tests passed, 0 failures/errors/skips.
   - Spotless formatting and Checkstyle.
   
   This PR has:
   
   - [x] been self-reviewed.
   - [x] added comments explaining the shutdown ordering.
   - [x] added a concurrency regression test.
   
   Key changed classes: `TestUtils.MiniCluster`, `RecoverReadTest`.
   


-- 
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