Install Karpenter on Amazon EKS and provision a NodePool with an EC2NodeClass

Steps

1. Install prerequisite tools (AWS CLI, kubectl, eksctl >= v0.202.0, helm) and create an EKS cluster with eksctl, tagging cluster resources with karpenter.sh/discovery=<CLUSTER_NAME> and provisioning the KarpenterNodeRole plus an IRSA/Pod Identity association for the karpenter service account via the getting-started cloudformation.yaml.
2. Install Karpenter with Helm from the public OCI registry, e.g. `helm registry logout public.ecr.aws && helm upgrade --install karpenter oci://public.ecr.aws/karpenter/karpenter --version <KARPENTER_VERSION> --namespace kube-system --create-namespace --set settings.clusterName=<CLUSTER_NAME> --set settings.interruptionQueue=<CLUSTER_NAME>`.
3. Create an EC2NodeClass (apiVersion karpenter.k8s.aws/v1) with spec.role (or spec.instanceProfile), a required spec.amiSelectorTerms entry (e.g. `alias: al2023@<version>`), and spec.subnetSelectorTerms / spec.securityGroupSelectorTerms matching the karpenter.sh/discovery tag.
4. Create a NodePool (apiVersion karpenter.sh/v1) whose spec.template.spec.nodeClassRef points at the EC2NodeClass (group: karpenter.k8s.aws, kind: EC2NodeClass, name: default), with spec.template.spec.requirements (kubernetes.io/arch, kubernetes.io/os, karpenter.sh/capacity-type, karpenter.k8s.aws/instance-category) and a spec.limits.cpu ceiling.
5. Apply both manifests with kubectl, then scale a test deployment (e.g. the pause-image 'inflate' deployment) to trigger provisioning and confirm nodes join via `kubectl get nodeclaims` and `kubectl get nodes`.
6. Tail controller logs with `kubectl logs -f -n kube-system -l app.kubernetes.io/name=karpenter -c controller` to confirm successful instance launch and EC2NodeClass Ready status.