The master ip is still its own ip when the master is restared

[gengjun-git]

Steps to reproduce the behavior (Required)

Execute the following steps in a 3 follower cluster without any tasks

1. Hang the two non-master nodes in the cluster

kill -19 pid

2. Immediately execute create database operation on the master node

create database test;

3. Wait for the master node to exit
4. Restart the hung fe
5. After the hung fe started, start the old master node.
6. connect to the old master node, and execute the command:

show frontends;

Expected behavior (Required)

show frontends should be executed successfully.

Real behavior (Required)

show frontends is hung,

2022-05-20 14:53:25,691 INFO (thrift-server-pool-2978|3102) [MasterOpExecutor.forward():134] Forward statement 2980 to Master TNetworkAddress(hostname:172.26.92.155, port:29927)
2022-05-20 14:53:25,692 INFO (thrift-server-pool-2979|3103) [ThriftServerEventProcessor.createContext():98] create thrift context. client: TNetworkAddress(hostname:172.26.92.155, port:54664)
2022-05-20 14:53:25,692 INFO (thrift-server-pool-2979|3103) [FrontendServiceImpl.forward():649] receive forwarded stmt 2980 from FE: 172.26.92.155

StarRocks version (Required)

• You can get the StarRocks version by executing SQL select current_version()

[jaogoy]

@gengjun-git You can modify the title to make it more accurate to the description.