Percona XtraDB Cluster on Kubernetes mysql operator failover setup HA/DR

MySQL & MariaDB Percona XtraDB Cluster 5.x
1

Hi team,
I have a query is there any tools available for failover setup on Percona XtraDB Cluster on Kubernetes MySQL operator ?
usually while using percona MySQL server on VM we usually use tools like mha for auto failover , re-route query /load balancing sprit read/write using proxysql and for static endpoint using vip IP for HA so is any there any similar tools available for Kubernetes MySQL percona operator ?? since I looking forward to setup DR/HA with Percona operator MySQL on Kubernetes environment

1 Like
2

Hello @mohamedkashifuddin ,

please have a look at Set up Percona XtraDB Cluster cross-site replication.

Also the blog post about it is going to be published pretty soon.
It is not HA, but more of a DR deployment which minimizes the recovery time objective.

1 Like
3

Yes it looks a DR approach can this be used within kubernetes by having multiple percona MySQL operator in different namespaces ? Plus in this scenario it doesn’t need to be expose to internet-facing will this approach work within kubernetes environment ?

1 Like
4

Hello @mohamedkashifuddin,
The Percona K8S Operator for PXC includes HA. When you launch a cluster you will have a 4th pod which has ProxySQL already configured for local HA. If one of your PXC pods goes down, ProxySQL will auto-redirect traffic.

1 Like
5

Hi matthewb,

Yes its a local ha but its not a optimal way for less or no down-time approach, I m looking for having multiple percona K8 MySQL operator on different region/zones node pools in K8S deployed on different namespace, so in case of a fail-over the traffic can to directed to another operator using similar tools like mha ? so is there any way to implement the traditional vm way of HA/DR (active/passive) approach in K8S with Percona XtraDB Cluster on Kubernetes MySQL operator ?

1 Like
6

It is optimal if you have K8S configured correctly. By default, the operator won’t put multiple PXC pods on the same node. You can further enhance that with additional restrictions for pod placement.

K8S itself is extremely HA. If a pod dies, K8S/Operator recreates it. Isn’t that the point of K8S in the first place?

The PXC operator only manages HA of a single PXC. if you want to use traditional async for DR, you’ll have to write up something yourself.

1 Like
7

@matthewb it is possible to use Set up Percona XtraDB Cluster cross-site replication to setup async DR between two PXC clusters with Operators.

@mohamedkashifuddin to answer your question - it is possible to use this feature to have async replication between two clusters is different namespaces. Just expose PXC nodes with ClusterIP services.

@matthewb has a valid point though, that probably it would be more efficient to rely on PXC replication if your workloads run in one Kubernetes cluster. I’m not sure I understand the use case of having two operators in the same cluster for HA.

1 Like
8

Hi spronin,

Thanks for detail on DR with operator we will try out and check if it works for us.
any idea on when this blog is coming out on DR deployment which minimizes the recovery time objective ???

1 Like