Configure Karpenter with Rancher on the Linode Cluster

This topic describes the steps to deploy Karpenter with Rancher on the Linode cluster.

Deploy Karpenter

The following are the steps to deploy Karpenter:

1. SSH to Rancher worker node.

2. Use the following command to get the RKE controller server and the token:

   # cat /etc/rancher/rke2/config.yaml.d/50-rancher.yaml 
   {
     "node-label": [
       "rke.cattle.io/machine=d459c182-ba22-46e4-a1a9-97400437eace"
     ],
     "private-registry": "/etc/rancher/rke2/registries.yaml",
     "protect-kernel-defaults": false,
     "server": "https://[IP_ADDR]:9345",
     "token": "TOKEN"

3. Update the charts.yaml or values.yaml with the RKE controller URL and the token.

   rancherLinode:
     cluster:
       server: "" #base64 of "https://[IP_ADDR]:9345"
       token: "" #base64 of the RKE. See /etc/rancher/rke2/config.yaml.d/50-rancher.yaml of an existing worker node
       rootPass: "" #base64 of the root Password on the nodes. Please make sure the password meets linode requiremets
       linodeToken: "" #base64 of the linode token. Make sure the Linode token has enough rights.

4. Use the following command to deploy Karpenter:

   helm upgrade --install --debug karpenter charts --namespace smart-scaler --create-namespace

Create a NodeClass

Use the following command to create a NodeClass:

cat <<EOF |  kubectl apply -f -
apiVersion: karpenter.linode.sh/v1alpha1
kind: LinodeNodeClass
metadata:
  name: default
spec:
  imageName: linode/ubuntu20.04
  rke2Version: "v1.30.7+rke2r1"
  privateIP: True
  #scriptUrl: "https://get.rke2.io"
  #subnetID: 132000
  #firewallID: 1509699
EOF

Create a NodePool

Use the following command to create a NodePool:

cat <<EOF |  kubectl apply -f -
apiVersion: karpenter.sh/v1
kind: NodePool
metadata:
  name: default
spec:
  template:
    spec:
      requirements:
        - key: kubernetes.io/arch
          operator: In
          values: ["amd64"]
        - key: kubernetes.io/os
          operator: In
          values: ["linux"]
        - key: topology.kubernetes.io/zone
          operator: In
          #values: ["eu-west", "eu-central", "se-sto", "de-fra-2"]
          values: ["de-fra-2"]
      nodeClassRef:
        name: default
        kind: LinodeNodeClass
        group: karpenter.linode.sh
      expireAfter: 720h # 30 * 24h = 720h
  limits:
    cpu: 1000
  disruption:
    consolidationPolicy: WhenEmptyOrUnderutilized
    consolidateAfter: 2m
EOF

Update the System Agent Upgrader

Rancher System Agent Upgrader continuously deploys short-live pods on Karpenter's nodes, and Karpenter cannot disrupt those nodes even where no user pods are running. The nodes seem never to be empty to Karpenter.

To stop the Rancher System Agent Upgrader from deploying short-lived pods on Karpenter-managed nodes, follow these steps:

1. Use the following command to edit the plan:

   kubectl -n cattle-system edit plan system-agent-upgrader

2. Use the following command to add add the node.kubernetes.io/instance-type node selector:

   spec:
     concurrency: 10
     nodeSelector:
       matchExpressions:
       - key: kubernetes.io/os
         operator: In
         values:
         - linux
       - key: node.kubernetes.io/instance-type
         operator: In
         values:
         - rke2