[YARN-2946] DeadLocks in RMStateStore<->ZKRMStateStore

• Type: Bug
• Status: Closed
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 2.7.0
• Component/s: resourcemanager
• Target Version/s: 2.7.0
• Hadoop Flags: Reviewed
• Assignee: Rohith Sharma K S
• Reporter: Rohith Sharma K S
• Created: 10/Dec/14 13:57
• Updated: 10/Apr/15 20:09
• Resolved: 24/Dec/14 06:16

Description

Found one deadlock in ZKRMStateStore.

1. Initial stage zkClient is null because of zk disconnected event.
2. When ZKRMstatestore#runWithCheck() wait(zkSessionTimeout) for zkClient to re establish zookeeper connection either via synconnected or expired event, it is highly possible that any other thred can obtain lock on ZKRMStateStore.this from state machine transition events. This cause Deadlock in ZKRMStateStore.