API call status issue

[dibrother]

Description:

• Version: 3.0.30
• MySQL 8.0.42

Hi，I encountered a problem while calling the API interface to determine.

There is currently a structure in the master slave environment that, when events such as power failure or shutting down all programs occur.

stop mysql-master
stop mysql-slave
stop replication-manager

After restarting, the replication manager will encounter an error that will disappear after the master writes data.

At this point, when calling the/app/clusters/cluster1/status interface, the status will be "error". I wrote a VIP manager tool to check and obtain the API information of the replication manager to determine the health status and who is the master. I first checked the health status of the middleware, but since there was no data written, it remained in an error state.Therefore, the program will not execute adding VIP to the master.

May I directly call/app/clusters/cluster1/topology/master to confirm who is the master without checking the status of the middleware?
I'm worried that not judging the middleware status and only judging the master may cause brain cracking.

[caffeinated92]

/app/clusters/cluster1/topology/master will return the current master of the cluster
it will return the master data if the cluster has the master

[caffeinated92]

For trying, you can enable the API swagger. Go to the global Settings tab (outside clusters), and enable the API. To go there you can click the API button and it will show the current swagger

[svaroqui]

I would advice to let replication-manager move the vip via instructing him with post failover script , because this is a hard process to do it in the correct timing, need to wait slave to catch, need to put contraints when it's not desirable to failover avoid split brain etc
The reason you get this error is probably because you do have configure replication heartbeat , you can enable one replication-manager heartbeat if wanted ( called traffic in gui) it uses the famous pseudo gtid create view to simulate heartbeat.

[svaroqui]

@dibrother we would advice as well that the replication-manager does run in an other DC form the infra that runs your DB this could be dangerous in case of split brain if not done at least if you plan to use auto failover