Scaling Percona XtraDB Cluster with ProxySQL in Docker Swarm

In this post, we’ll look at scaling Percona XtraDB Cluster with ProxySQL in Docker Swarm.

In my previous post, I showed how to employ Percona XtraDB Cluster on multiple nodes in a Docker network.

The intention is to be able to start/stop nodes and increase/decrease the cluster size dynamically. This means that we should track running nodes, but also to have an easy way to connect to the cluster.

So there are two components we need: service discovery to register nodes and ProxySQL to handle incoming traffic.

The work with service discovery is already bundled with Percona XtraDB Cluster Docker images, and I have experimental images for ProxySQL https://hub.docker.com/r/perconalab/proxysql/.

For multi-node management, we also need some orchestration tool, and a good start is Docker Swarm. Docker Swarm is simple and only provides basic functionality, but it works for a good start. (For more complicated setups, consider Kubernetes.)

I assume you have Docker Swarm running, but if not here is some good material on how to get it rolling. You also need to have service discovery running (see http://chunqi.li/2015/11/09/docker-multi-host-networking/ and my previous post).

To start a cluster with ProxySQL, we need a docker-compose definition file docker-compose.yml.:

For convenience, both proxy and percona-xtradb-cluster share the same environment files (.env):

You can also get both files from https://github.com/percona/percona-docker/tree/master/pxc-56/swarm.

To start both the cluster node and proxy:

We can start as many Percona XtraDB Cluster nodes as we want:

The command above will make sure that five nodes are running.

We can check it with docker ps:

We can see that Docker schedules containers on two different nodes, the Proxy SQL container is smblade04/swarm_proxy_1, and the connection point is 10.20.2.66:6032.

To register Percona XtraDB Cluster in ProxySQL we can just execute the following:

The script will connect to the service discovery DISCOVERY_SERVICE (defined in .env file) and register nodes in ProxySQL.

To check that they are all running:

We can connect to a cluster using a ProxySQL endpoint:

We can see that we connect to a different node every time.

Now if we want to get crazy and make sure we have ten Percona XtraDB Cluster nodes running, we can execute the following:

And Docker Swarm will make sure ten nodes are running.

I hope this demonstrates that you can easily start playing with multi-nodes using Percona XtraDB Cluster. In the next post, I will show how to use Percona XtraDB Cluster with Kubernetes.