[FLINK-29173] Upgrade curator

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Won't Fix
• Affects Version/s: 1.16.0, 1.17.0, 1.15.3
• Component/s: Runtime / Coordination
• Assignee: Unassigned
• Reporter: Matthias Pohl
• Created: 01/Sep/22 12:45
• Updated: 30/Jan/23 11:48
• Resolved: 30/Jan/23 11:48

Description

FLINK-28078 revealed a bug in curator's LeaderLatch implementation that is covered by CURATOR-645. FLINK-28078 came up with a workaround in ZooKeeperMultipleComponentLeaderElectionDriverTest.testLeaderElectionWithMultipleDrivers (see 655184c master and ed8700d for release-1.15) that should get reverted after the curator version was bumbed to a version that includes CURATOR-645.

There's a similar issue CURATOR-644 which we didn't run into but we might want to observe and include as well, probably.

Issue Links

• fixes: FLINK-30484 ZooKeeperLeaderElectionTest.testZooKeeperReelection timed out
• fixes: FLINK-28078 ZooKeeperMultipleComponentLeaderElectionDriverTest.testLeaderElectionWithMultipleDrivers runs into timeout
• is fixed by: FLINK-30772 Update Flink Shaded dependencies to the latest versions
• is related to: CURATOR-645 LeaderLatch generates infinite loop with two LeaderLatch instances competing for the leadership
• is related to: CURATOR-644 CLONE - Race conditions in LeaderLatch after reconnecting to ensemble