Skip to main content
Version: release-0.9

ApeCloud MySQL Proxy

Before you start​

  1. Install kbcli.

  2. Install Helm.

  3. Install KubeBlocks.

    You can run kbcli playground init to install a k3d cluster and KubeBlocks. For details, refer to Try KubeBlocks on your laptop or Try KubeBlocks on cloud.

    kbcli playground init

    # Use --version to specify a version
    kbcli playground init --version='x.y.z'

    Or if you already have a Kubernetes cluster, you can install KubeBlocks directly.

  4. Prepare an ApeCloud MySQL RaftGroup named mycluster for demonstrating how to enable the proxy function for an existing cluster. Refer to Create a MySQL cluster for details.

Create a Proxy Cluster​

It is recommended to use kbcli to create an ApeCloud MySQL Proxy Cluster.

  1. Add the KubeBlocks repository.

    helm repo add kubeblocks https://apecloud.github.io/helm-charts

  2. View the repository list to verify whether the KubeBlocks repository is added successfully.

    helm repo list

  3. Run the update command to make sure you have added the latest version.

    helm repo update

  4. Install etcd to create the external service reference.

    1. View all versions of etcd.

      helm search repo kubeblocks/etcd --devel --versions

    2. Install the etcd Addon.

      helm install etcd kubeblocks/etcd --version=v0.6.5

    3. Install the etcd cluster.

      helm install etcd-cluster kubeblocks/etcd-cluster

    4. view the status of the etcd cluster and make sure it is running.

      kubectl get cluster
      >
      NAME           CLUSTER-DEFINITION   VERSION       TERMINATION-POLICY   STATUS      AGE
      etcd-cluster   etcd                 etcd-v3.5.6   Halt                 Running     10s

    5. View the service address of this etcd clsuter.

      kubectl get service
      >
      NAME                             TYPE        CLUSTER-IP     EXTERNAL-IP   PORT(S)                                                  AGE
      etcd-cluster-etcd                ClusterIP   10.110.23.89   <none>        2379/TCP                                                 55s
      etcd-cluster-etcd-headless       ClusterIP   None           <none>        2379/TCP,2380/TCP,3501/TCP,50001/TCP                     55s
      kubernetes                       ClusterIP   10.96.0.1      <none>        443/TCP                                                  13m

      You can combine the service address to get the endpoint or you can use the IP of the service address as the access address.

      Here is an example of combining the service address.

      etcd-cluster-etcd.default.svc.cluster.local:2379

  5. Create an ApeCloud MySQL Proxy Cluster.

    1. View all versions of ApeCloud MySQL Proxy.

      helm search repo kubeblocks/apecloud-mysql --devel --versions

    2. (Optional) If you disable the apecloud-mysql Addon when installing KuebBlocks, run the command below to specify a version and install the cluster definition of ApeCloud MySQL. Skip this step if you install KubeBlocks with the default settings.

      helm install myproxy kubeblocks/apecloud-mysql --version=v0.9.0

    3. Create an ApeCloud MySQL Proxy Cluster.

      helm install myproxy kubeblocks/apecloud-mysql-cluster --version=v0.9.0 --set mode=raftGroup,proxyEnabled=true,etcd.serviceReference.endpoint="etcd-cluster-etcd.default.svc.cluster.local:2379"
note

If you only have one node for deploying a RaftGroup Cluster, set the extra.availabilityPolicy as none when creating a RaftGroup Cluster.

helm install myproxy kubeblocks/apecloud-mysql-cluster --version=v0.9.0 --set mode=raftGroup,proxyEnabled=true,etcd.serviceReference.endpoint="etcd-cluster-etcd.default.svc.cluster.local:2379" --set extra.availabilityPolicy=none

  1. Check the status of the clusters.

    kubectl get cluster

    kubectl get pods

    You can also enter the etcd container or wesql-scale container to view the configuration of wesql-scale or to check the availability of the etcd service.

    etcdctl --endpoints=http://etcd-cluster-etcd.default.svc.cluster.local:2379 get /vitess --prefix --keys-only

Enable/Disable Proxy dynamically​

As its name suggests, ApeCloud MySQL Proxy in nature is a database proxy. An ApeCloud MySQL RaftGroup Cluster that already exists can be switched to an ApeCloud MySQL Proxy Cluster by setting proxyEnabled=true.

helm upgrade mycluster kubeblocks/apecloud-mysql-cluster --set mode=raftGroup,proxyEnabled=true,etcd.serviceReference.endpoint="etcd-cluster-etcd.default.svc.cluster.local:2379"

If you want to disable proxy, run the command below.

helm upgrade mycluster kubeblocks/apecloud-mysql-cluster --set mode=raftGroup

Connect Proxy Cluster​

ApeCloud MySQL Proxy is routed through the vtgate component, and the way the MySQL Server accesses vtgate is similar to the way of accessing mysqld. The external SQL access address provided by ApeCloud MySQL Proxy is the vtgate address and port. The vtgate address created by KubeBlocks by default is myproxy-cluster-vtgate-headless, and the port number is 15306. You can visit ApeCloud MySQL Proxy through the MySQL Server in any pod under the same namespace as ApeCloud MySQL Proxy.

Connect Proxy Cluster by VTGate​

  1. Expose the port of VTGate to the localhost so that the localhost can access the Proxy.

    kubectl port-forward svc/vt-vtgate-headless 15306:15306

  2. Connect to the cluster.

    mysql -h 127.0.0.1 -P 15306

Connect Proxy Cluster by MySQL Server​

  1. Expose the port of the MySQL Server to the localhost so that the localhost can access the MySQL Server.

    kubectl port-forward svc/vt-mysql 3306:3306

  2. Connect to the cluster.

    mysql -h 127.0.0.1 -P 3306
note

If you need to test the failover of MySQL, you need to delete the Pod first and continue to port-forward the port, and you can also write a shell script. Here are examples.

For VTGate,

while true; do date; kubectl port-forward svc/vt-vtgate-headless 15306:15306; sleep 0.5; done

For the MySQL Server,

while true; do date; kubectl port-forward svc/vt-mysql 3306:3306; sleep 0.5; done

Configure Proxy Cluster parameters​

  1. Get the configuration file of this cluster.

    kubectl edit configurations.apps.kubeblocks.io myproxy-vtgate

  2. Configure parameters according to your needs. The example below adds the spec.configFileParams part to configure max_connections.

    spec:
      clusterRef: myproxy
      componentName: vtgate
      configItemDetails:
      - configFileParams:
          vtgate.cnf:
            parameters:
              healthcheck_timeout: "5s"
        configSpec:
          constraintRef: mysql-scale-vtgate-config-constraints
          name: vtgate-config
          namespace: kb-system
          templateRef: vtgate-config-template
          volumeName: mysql-scale-config
        name: vtgate-config
        payload: {}

  3. Connect to this cluster to verify whether the configuration takes effect.

    1. Expose the port of the MySQL Server to the localhost so that the localhost can access the MySQL Server.

      kubectl port-forward svc/vt-vtgate-headless 15306:15306

    2. Connect to this cluster and verify whether the parameters are configured as expected.

      mysql -h 127.0.0.1 -P 3306

      >
      mysql> show variables like 'healthcheck_timeout';
      +---------------------+-------+
      | Variable_name       | Value |
      +---------------------+-------+
      | healthcheck_timeout |  5s   |
      +---------------------+-------+
      1 row in set (0.00 sec)

Log​

You can view the log files of components, Pods, and containers by both kbcli and kubectl.

View the log of VTGate.

kubectl logs myproxy-cluster-vtgate-8659d5db95-4dzt5

View the log of VTTablet and -c is required.

kubectl logs myproxy-cluster-mysql-0 -c vttablet

Enter the container and view more logs of VTGate.

kubectl exec -it myproxy-cluster-vtgate-8659d5db95-4dzt5 -- bash
ls /vtdataroot

Enter the container and view more logs of VTTablet.

kubectl exec -it myproxy-cluster-mysql-0  -c vttablet -- bash
ls /vtdataroot

Read-write splitting​

You can enable the read-write splitting function.

  1. Get the configuration file of this cluster.

    kubectl edit configurations.apps.kubeblocks.io myproxy-vtgate

  2. Configure read_write_splitting_policy as random.

    spec:
      clusterRef: myproxy
      componentName: vtgate
      configItemDetails:
      - configFileParams:
          vtgate.cnf:
            parameters:
              read_write_splitting_policy: "random"
        configSpec:
          constraintRef: mysql-scale-vtgate-config-constraints
          name: vtgate-config
          namespace: kb-system
          templateRef: vtgate-config-template
          volumeName: mysql-scale-config
        name: vtgate-config
        payload: {}

You can also set the ratio for read-write splitting and here is an example of directing 70% flow to the read-only node.

spec:
   clusterRef: myproxy
   componentName: vtgate
   configItemDetails:
   - configFileParams:
      vtgate.cnf:
         parameters:
            read_write_splitting_rati: "70"
      configSpec:
      constraintRef: mysql-scale-vtgate-config-constraints
      name: vtgate-config
      namespace: kb-system
      templateRef: vtgate-config-template
      volumeName: mysql-scale-config
      name: vtgate-config
      payload: {}

Transparent failover​

  1. Get the configuration file of this cluster.

    kubectl edit configurations.apps.kubeblocks.io myproxy-vtgate

  2. Configure enable_buffer as true.

    spec:
      clusterRef: myproxy
      componentName: vtgate
      configItemDetails:
      - configFileParams:
          vtgate.cnf:
            parameters:
              enable_buffer: "true"
        configSpec:
          constraintRef: mysql-scale-vtgate-config-constraints
          name: vtgate-config
          namespace: kb-system
          templateRef: vtgate-config-template
          volumeName: mysql-scale-config
        name: vtgate-config
        payload: {}