merlimat opened a new pull request, #4769: URL: https://github.com/apache/bookkeeper/pull/4769
## Summary - Three tests in `BookieAutoRecoveryTest` were flaking in CI (`testOpenLedgers`, `testClosedLedgers`, `testEmptyLedgerLosesQuorumEventually`) when the bookie they killed happened to be the Auditor leader. The leader-failover chain — `AuditorElector.shutdown()`'s 10s timeout + ZK session expiry + new election + first audit cycle — routinely exceeded the tests' 20s/90s latch timeouts under CI load. - Make the bookie-to-kill choice deterministic w.r.t. the Auditor: pick a non-Auditor bookie from the ensemble, or for `testEmptyLedgerLosesQuorumEventually` (which needs to kill specific indices), pin the Auditor onto `ensemble[0]` by stopping `AutoRecovery` on the bookies the test will kill before the timed assertions. - Recent example failure: https://github.com/apache/bookkeeper/actions/runs/25072614471/job/73458460684 ## Test plan - [x] `mvn -pl bookkeeper-server test -Dtest='BookieAutoRecoveryTest'` passes locally on two consecutive runs (7/7 tests). -- 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]
