How to configure the CoreDNS Autoscaler in a Rancher Kubernetes Engine (RKE) CLI or Rancher v2.x provisioned Kubernetes cluster

Task

During the life of a cluster, you may need to adjust the scaling parameters for the CoreDNS autoscaler. The autoscaler runs as an independent Deployment in the cluster, using the cluster-proportional-autoscaler container to scale up and down the related CoreDNS Deployment, using a linear or ladder pattern.

Pre-requisites

• A Rancher Kubernetes Engine (RKE) CLI or Rancher v2.x provisioned Kubernetes cluster
• The cluster is configured with the coredns addon (enabled by default)

Note When running rke up commands, ensure the .rkestate file for the cluster is present in the working directory as per the documentation here.

Steps

Note: When making the changes, the coredns-autoscaler pod will be restarted with updated command arguments, this will not cause any disruption to DNS resolution. Check the logs of the coredns-autoscaler pod after making changes to confirm they have taken effect.

Rancher provisioned cluster managed by Rancher

1. Navigate to Cluster Management in the Rancher UI and click 'Edit Config' for the cluster.
2. Click 'Edit as YAML'.
3. Locate or add the dns field, using the below as an example to add the desired parameters below:

4. rancher_kubernetes_engine_config:
     [...]
     dns:
       linear_autoscaler_params:
         cores_per_replica: 128
         max: 5
         min: 1
         nodes_per_replica: 4
         prevent_single_point_failure: true

5. Click 'Save' to update the cluster with the new configuration.

RKE provisioned cluster managed by RKE

1. Edit the cluster configuration YAML file to configure the dns field, using the below as an example to add the desired parameters below:

   dns:
     linear_autoscaler_params:
       cores_per_replica: 128
       max: 5
       min: 1
       nodes_per_replica: 4
       prevent_single_point_failure: true

2. Invoke rke up --config <cluster configuration YAML file> to update the cluster.