Enable horizontal pod autoscaling¶

Using MCP, you can adjust the number of pod replicas without using an external orchestrator by enabling the horizontal pod autoscaling feature in your MCP Kubernetes deployment. The feature is based on observed CPU and/or memory utilization and can be enabled using the metrics-server add-on.

To enable horizontal pod autoscaling:

1. While generating a deployment metadata model for your new MCP Kubernetes cluster as described in Create a deployment metadata model, select the Kubernetes metrics server enabled option in the Kubernetes Product parameters section of the Model Designer UI.

2. If you have already generated a deployment metadata model without the metrics-server parameter or to enable this feature on an existing Kubernetes cluster:

   1. Open your Reclass model Git project repository on the cluster level.

   2. In /kubernetes/control.yml, add the metrics-server parameters:

      parameters:
        kubernetes:
          common:
            addons:
              ...
              metrics-server:
                enabled: true
      

3. Select from the following options:

   • If you are performing an initial deployment of your cluster, proceed with further configuration as required. Pod autoscaling will be enabled during your Kubernetes cluster deployment.

   • If you are making changes to an existing cluster:

     1. Log in to the Salt Master node.

     2. Refresh your Reclass storage data:

        salt-call state.sls reclass.storage
        

     3. Apply the kube-addons state:

        salt -C 'I@kubernetes:master' state.sls kubernetes.master.kube-addons
        

4. On a running Kubernetes cluster, verify that autoscaling works successfully using the Official Kubernetes documentation.