[RATIS-1446] Avoid leader election for invalid conf

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Fix Version/s: 2.3.0
• Component/s: server
• Assignee: Xu Shao Hong
• Reporter: Xu Shao Hong
• Created: 29/Nov/21 06:06
• Updated: 15/Dec/21 14:23
• Resolved: 15/Dec/21 14:23

Description

 

1private void yieldLeaderToHigherPriorityPeer() {
2  if (!server.getInfo().isLeader()) {
3    return;
4  }

5  final RaftConfigurationImpl conf = server.getRaftConf();
6  int leaderPriority = conf.getPeer(server.getId()).getPriority(); 

During my test in k8s env, I met the corner case where the ozone progress met NPE and was shut down. The log shows it is thrown at line 6. It is possible in some cases configuration may not include the server itself. The case happened when bootstrapping the second SCM. Thus there should have a null check before getPriority.

-------

The attachment is the error log caught.

Issue Links

• links to: GitHub Pull Request #560