Hi HBase community, Could someone please help review the following three PRs addressing flaky tests?
HBASE-30093 / PR #8587: Stochastic load balancer timeout Start the stochastic search timeout after initialization, so initialization on a busy host does not consume the search budget and cause intermittent test failures. The patch also adds a regression test for the deadline ordering. HBASE-30096 / PR #8577: TestHRegion thread cleanup Ensure background threads are stopped and joined in finally blocks, including when tests fail or time out. This prevents leftover threads from interfering with subsequent tests and also covers TestHRegionWithInMemoryFlush. HBASE-30344 / PR #8578: Sync replication state synchronization Wait for the RegionServer serving the test table to observe DOWNGRADE_ACTIVE before verifying data in TestSyncReplicationStandbyKillMaster. During Master recovery, the Master-side state can be updated before the RegionServer refreshes its local state, causing reads to fail intermittently. I have added further details in the PR discussion in response to the earlier review question. Validation details are included in each PR. For HBASE-30344, the affected test passed five consecutive runs with Surefire retries disabled. Any feedback would be appreciated. If someone else would be better placed to review these changes, suggestions are welcome. Thanks for your time! Best regards, Ma Zhengxuan Pluto [email protected]
