MySQL cluster has comaster instances when I recreated it.

[juexun]

I deleted the mysql cluster using kubectl delete -f demo.yaml --cascade=false and reserved the pvc.
When i recreate the mysql cluster using kubectl apply -f demo.yaml
The cluster was in maintenance mode with two co-master instances.
How can I fix it?

[AMecea]

I need a little more details, we can chat further on the MySQL operator channel on k8s slack to debug this.

EDIT:
I managed to reproduce it. This is a bug and I will try to figure out a fix for it. Until then when deleted a cluster but want to keep the PVCs then make sure that cluster is scaled to 1 or 0, to keep only one PVC.

The problem is because the master is node-1 and when came up node-0 can't get from orchestrator who was the original master because the operator remove it from the orchestrator. Furthermore the node-0 knows that it has to connect to node-1 but the sidecar containers configure it as master. And when node-1 came up he is configure as slave for node-0 but previously was a master.

Thanks @juexun for reporting this bug!!

[calind]

This should be solved by #199.

[AMecea]

Also when deleting a cluster with --cascade=false the stateful set and pods should not be deleted. Updating controller-util fixes this, this is addressed by #221.

To reproduce this issue, in other words to delete the cluster but don't delete the PVCs, you should delete the cluster with --cascade=false and the related objects of the cluster, like stateful set, config maps, etc. Anyway when using cascade deletion you have to be more careful when recreating the cluster.