Hi all, ExtensibleLoadManager tests are broken in branch-3.0 and branch-3.3.
Example build failures: pulsar-ci.yaml for branch-3.0 (ExtensibleLoadManagerImplTest is in broker group 1 in branch-3.0): https://github.com/apache/pulsar/actions/workflows/pulsar-ci.yaml?query=branch%3Abranch-3.0 In many cases, the failing ExtensibleLoadManagerImplTest triggered an OOME in branch-3.0 since the test framework improvement https://github.com/apache/pulsar/pull/21467 was missing from branch-3.0. I have cherry-picked that now, so OOME won't be an issue in the most recent build. pulsar-ci-flaky.yaml for branch-3.3 (ExtensibleLoadManagerImplTest is in the flaky group in branch-3.3): https://github.com/apache/pulsar/actions/workflows/pulsar-ci-flaky.yaml?query=branch%3Abranch-3.3 I've been triggering CI builds for maintenance branches after completing a batch of cherry-picking and before performing releases. The last successful build for branch-3.0 pulsar-ci.yaml workflow was on May 20th (before 3.0.12): https://github.com/apache/pulsar/actions/runs/15132614429 revision: https://github.com/apache/pulsar/commits/76e5479efdca225b138c0370989431731787cd90/ and the last successful build for branch-3.3 pulsar-ci-flaky.yaml workflow was on May 16th (one commit before 3.3.7 release): https://github.com/apache/pulsar/actions/runs/15062146251 revision: https://github.com/apache/pulsar/commits/529ed8e60cc1c8c2e12a58031dddc6babdebb182/ I haven't yet investigated the reason and whether there's a commit that introduces a regression. I'm sharing these details so that others are also aware of the current state of the branch-3.0 and branch-3.3 maintenance branches. -Lari
