Note: This post is an updated and reduced version of an old blog post if you want to know more about Karpenter you should check it https://nahuelhernandez.com/blog/karpenter_kubernetes_node_autoscaling/

Requirements:

• AWS Account
• Eksctl >= 0.99
• AWS cli >= 2.6
• Kubectl >= 1.23

Configuring cluster variables:

> export CLUSTER_NAME=eks-with-karpenter
> export VERSION=1.22
> export REGION=us-east-1

Creating the cluster using Eksctl:

> cat <<EOF | eksctl create cluster -f -
---
apiVersion: eksctl.io/v1alpha5
kind: ClusterConfig

metadata:
  name: $CLUSTERNAME
  region: $REGION
  version: $VERSION
  tags:
    karpenter.sh/discovery: $CLUSTERNAME
iam:
  withOIDC: true # required

karpenter:
  version: '0.9.0'

managedNodeGroups:
  - instanceType: t3.small
    name: managed-ng-1
    minSize: 1
    maxSize: 3
    desiredCapacity: 1
EOF

This task will create the EKS cluster with a managedNodeGroup and the Karpenter resources on Kubernetes. It will take approximately 20 minutes.

Check the Karpenter deployed resources on K8S:

> kubectl get all -n karpenter
NAME                             READY   STATUS    RESTARTS   AGE
pod/karpenter-768fc86b78-2zmv6   2/2     Running   0          39s

NAME                TYPE        CLUSTER-IP      EXTERNAL-IP   PORT(S)            AGE
service/karpenter   ClusterIP   10.100.204.73   <none>        8080/TCP,443/TCP   39s

NAME                        READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/karpenter   1/1     1            1           39s

NAME                                   DESIRED   CURRENT   READY   AGE
replicaset.apps/karpenter-768fc86b78   1         1         1       39s

Check the nodes:

> kubectl get nodes
NAME                            STATUS   ROLES    AGE     VERSION
ip-192-168-14-42.ec2.internal   Ready    <none>   7m18s   v1.22.6-eks-7d68063

Create the Karpenter provisioner:

> cat <<EOF | kubectl apply -f -
apiVersion: karpenter.sh/v1alpha5
kind: Provisioner
metadata:
  name: default
spec:
  requirements:
    - key: karpenter.sh/capacity-type
      operator: In
      values: ["spot"]
  limits:
    resources:
      cpu: 1000
  provider:
    subnetSelector:
      karpenter.sh/discovery: ${CLUSTER_NAME}
    securityGroupSelector:
      karpenter.sh/discovery: ${CLUSTER_NAME}
    instanceProfile: eksctl-KarpenterNodeInstanceProfile-${CLUSTER_NAME}
  ttlSecondsAfterEmpty: 30
EOF

Testing Node Autoscaling (optional)

Create a deployment:

> cat <<EOF | kubectl apply -f -
apiVersion: apps/v1
kind: Deployment
metadata:
  name: inflate
spec:
  replicas: 3
  selector:
    matchLabels:
      app: inflate
  template:
    metadata:
      labels:
        app: inflate
    spec:
      terminationGracePeriodSeconds: 0
      containers:
        - name: inflate
          image: public.ecr.aws/eks-distro/kubernetes/pause:3.2
          resources:
            requests:
              cpu: 1
EOF

Check the pods and the nodes

> kubectl get pod
NAME                           READY   STATUS    RESTARTS   AGE
pod/inflate-6b88c9fb68-kvr78   0/1     Pending   0          17s
pod/inflate-6b88c9fb68-qckpd   0/1     Pending   0          17s
pod/inflate-6b88c9fb68-wk4wk   0/1     Pending   0          17s

NAME                                 STATUS    ROLES    AGE   VERSION
node/ip-192-168-14-42.ec2.internal   Ready     <none>   12m   v1.22.6-eks-7d68063
node/ip-192-168-93-33.ec2.internal   Unknown   <none>   13s

After a couple of seconds the new node are “Ready” and the pods are not in pending status, they are running.

> kubectl get pod,nodes
NAME                           READY   STATUS    RESTARTS   AGE
pod/inflate-6b88c9fb68-kvr78   1/1     Running   0          64s
pod/inflate-6b88c9fb68-qckpd   1/1     Running   0          64s
pod/inflate-6b88c9fb68-wk4wk   1/1     Running   0          64s

NAME                                 STATUS   ROLES    AGE   VERSION
node/ip-192-168-14-42.ec2.internal   Ready    <none>   12m   v1.22.6-eks-7d68063
node/ip-192-168-93-33.ec2.internal   Ready    <none>   61s   v1.22.6-eks-7d68063