MySQL High availability cluster

MySQL Operator

Install MySQL operator and create namespace

# version: 8.0.33-2.0.10
helm repo add mysql-operator https://mysql.github.io/mysql-operator/
helm repo update
helm install mysql-operator mysql-operator/mysql-operator --namespace mysql-cluster --create-namespace

MySQL InnoDBCluster need enough privilege to access namespace and manage router, backup, cluster modification.

kubectl apply -f rbac.yml

Create backup volume and InnoDBCluster. Remember to watch statefulset log and check whether it shows you dont have enough permission, normally i already gave enough privilege, but i am not sure if it will change in later version.

kubectl apply -f backup-pvc.yml
kubectl apply -f mysql.yml

Connect to InnoDBCluster

kubectl run --rm -it myshell --image=container-registry.oracle.com/mysql/community-operator -- mysqlsh root@mysql-innodb.mysql-cluster:3306  --sql
# If you don't see a command prompt, try pressing enter.
# ****************************** # <-pwd: <password-of-root-user-jecawe>

Conclusion

The main disadvantage of MySQL Operator is it needs long time to delete cluster. In some cases, you have to clean its finalizers even you had set correct settings. The reason is deletion of secret before pod terminating.

If you wanna deploy a middle size cluster this should be a good choice. Comparing to Elastic Operator, MySQL Operator is lacks of custom settings, it can not separate primary and secondary pod settings.

Notes

Can't delete MySQL objects even applying force delete - MySQL Operator

Not able to completely remove Kubernetes CustomResource

# check what kind of resource is stucking
kubectl api-resources --verbs=list --namespaced -o name | xargs -n 1 kubectl get --show-kind --ignore-not-found -n mysql-cluster

# if it shows `innodbclusters.mysql.oracle.com/mysql-innodb` is pending forever, then run the command below
kubectl patch innodbclusters.mysql.oracle.com/mysql-innodb  -p '{"metadata":{"finalizers":[]}}' --type=merge -n mysql-cluster

# if shows pod is terminating long time
kubectl patch pods mysql-innodb-0 -n mysql-cluster -p '{"metadata":{"finalizers":null}}'

You have to manually clean resource after you set finalizers to empty

You can also control how and when garbage collection deletes resources that have owner references using Kubernetes finalizers - Cascading deletion

This example does not simulate ClusterSet - MySQL InnoDB ClusterSet