The state of each server node is inconsistent

### Overview of the Issue

Node Az2-1 thinks that the state of node AZ2-0 is leaving, but the state of node AZ2-0 in the master node or other nodes except Az2-1 is alive. What is the reason for this? What should be done?

- Az2-0 query:


- But Az2-1 query:

### Log Fragments

1. AZ2-0:

```
2020/09/15 11:26:05 [ERR] yamux: Failed to write header: write tcp 10.0.0.9:8300->10.0.0.7:58123: use of closed network connection
2020/09/22 01:09:54 [ERR] raft-net: Failed to flush response: write tcp 10.0.0.9:8300->10.0.0.11:60527: write: broken pipe
```

2. AZ2-1：

```
2020/09/10 19:22:16 [ERR] yamux: Failed to write header: write tcp 10.0.0.8:8300->10.0.0.13:41374: use of closed network connection
2020/09/22 01:09:55 [ERR] raft-net: Failed to flush response: write tcp 10.0.0.8:8300->10.0.0.11:38522: write: broken pipe
2020/09/27 10:48:14 [ERR] consul.rpc: multiplex conn accept failed: read tcp 10.0.0.8:8300->10.0.0.6:47987: read: connection reset by peer from=10.0.0.6:47987

```

3. AZ1-1 master node

```
2020/09/22 01:09:54 [ERR] raft: Failed to AppendEntries to {Voter 67317135-7676-3372-7474-676f616f3231 10.0.0.8:8300}: read tcp 10.0.0.11:55225->10.0.0.8:8300: i/o timeout
```