Experiments about the supporting to MySQL high avaliable cluster in Flink MySQL CDC

[luoyuxia]

Enviroments:
MySQL cluster with docker
https://github.com/docker-box/mysql-cluster

Case 1:
One master, two slaves assuming master, slave1, slave2. Flink CDC listens to slave2, and master fails, change slave1 to master.
1: change slave1 to master
excute the following sql:
stop slave;
GRANT REPLICATION SLAVE ON *.* TO 'mydb_slave_user'@'%' IDENTIFIED BY 'mydb_slave_pwd';
SHOW MASTER STATUS;
NOTE: SHOW MASTER STATUS; is aimed to get the 'File' and 'Position' which will be use when make slave2 follow slave1.

2: make slave2 follow slave1
excute the following sql:
stop slave;
CHANGE MASTER TO MASTER_HOST='172.20.0.4', MASTER_USER='mydb_slave_user', MASTER_PASSWORD='mydb_slave_pwd', MASTER_LOG_FILE='mysql-bin.000003',MASTER_LOG_POS=1207;
START SLAVE;

NOTE: you should replace MASTER_HOST, MASTER_USER, MASTER_PASSWORD, MASTER_LOG_FILE, MASTER_LOG_POS with proper value.

3: then insert data to slave1, you will find Flink CDC output the inserted data.

Case2:
One master, one slave assuming master, slave. Flink CDC listens to slave, and master fails, change slave to master.
1: change slave to master
excute the following sql:
stop slave;
2: then insert data to slave, you will find Flink CDC output the inserted data.

Case3:
One master, one slave assuming master, slave. Flink CDC listens to master, and master fails, change slave to master.

1: change slave to master
excute the following sql:
stop slave;
2: modify job and then restart
modify the job(change the source table's hostname/port ) to make it listen to slave,
3: then insert data to slave, you will find Flink CDC output the inserted data.

NOTE:
In production enviroments for case3, we can dynamiclly change the MySQL instance the job listens to in the help of DNS or VIP without modifying/restarting the job.