[GH] (fluss/regain-leadership): Workflow run "CI" failed!

[GitBox <gi...@apache.org>]

The GitHub Actions job "CI" on fluss.git/regain-leadership has failed.
Run started by GitHub user Kaixuan-Duan (triggered by Kaixuan-Duan).

Head commit for run:
9196faed730fa97716d2ef9f30a3df26a14eb765 / duankaixuan <14...@qq.com>
[server] Recover coordinator leader election when ZK election parent is garbage-collected

When the ZooKeeper session of the coordinator expires, ZooKeeper deletes
the ephemeral LeaderLatch node and may then garbage-collect the empty
container election parent /coordinators/election. On reconnection, the
Curator 5.4.0 LeaderLatch silently ignores the NONODE result of listing
the election path and stays permanently idle, so a single-coordinator
cluster remains without a leader until it is restarted (upstream
CURATOR-724, fixed in Curator 5.8.0 which Fluss does not shade yet).

Recreate the election parent as a persistent znode on RECONNECTED before
the latch's own reconnection handling, so the latch can re-create its
election node and the coordinator regains leadership without a restart.
The recovery piggybacks on a LeaderLatch subclass since Curator notifies
connection-state listeners in ConcurrentHashMap iteration order, which
does not follow registration order.

Add a regression test that kills the ZK session and deletes the empty
election parent, verifying that leadership is regained without restart;
it fails with a timeout before the fix and passes after.

Co-Authored-By: Qoder <no...@qoder.com>
AI-Model: Qoder Auto
AI-Contributed/Feature: 73/73
AI-Contributed/UT: 93/93

Report URL: https://github.com/apache/fluss/actions/runs/35098269491

With regards,
GitHub Actions via GitBox