Karpenter Has a Free API — Automatic Kubernetes Node Scaling Without Node Groups

Karpenter is AWS's open-source Kubernetes node autoscaler that provisions just-right compute in seconds. Unlike Cluster Autoscaler, Karpenter doesn't need node groups — it directly provisions EC2 instances based on pod requirements.

Free, open source, now a CNCF sandbox project. Works with any cloud (not just AWS).

Why Use Karpenter?

• No node groups — provisions exactly the right instance type for pending pods
• Sub-minute scaling — 10x faster than Cluster Autoscaler
• Cost optimization — automatically uses Spot instances, ARM, and right-sized nodes
• Consolidation — removes underutilized nodes and repacks workloads
• Drift detection — automatically upgrades nodes when AMI changes

Quick Setup

1. Install Karpenter

helm repo add karpenter https://charts.karpenter.sh
helm install karpenter karpenter/karpenter \
  --namespace karpenter --create-namespace \
  --set clusterName=my-cluster \
  --set clusterEndpoint=$(aws eks describe-cluster --name my-cluster --query 'cluster.endpoint' --output text)

2. Create a NodePool

kubectl apply -f - <<EOF
apiVersion: karpenter.sh/v1
kind: NodePool
metadata:
  name: default
spec:
  template:
    spec:
      requirements:
      - key: kubernetes.io/arch
        operator: In
        values: [amd64, arm64]
      - key: karpenter.sh/capacity-type
        operator: In
        values: [spot, on-demand]
      - key: node.kubernetes.io/instance-type
        operator: In
        values: [m5.large, m5.xlarge, m6g.large, c5.large]
      nodeClassRef:
        group: karpenter.k8s.aws
        kind: EC2NodeClass
        name: default
  limits:
    cpu: "100"
    memory: 200Gi
  disruption:
    consolidationPolicy: WhenEmptyOrUnderutilized
    consolidateAfter: 30s
EOF

3. Check NodePool Status

# Via kubectl
kubectl get nodepools
kubectl get nodepools default -o json | jq '{cpu: .status.resources.cpu, memory: .status.resources.memory, nodes: .status.resources["nodes"]}'

# Get provisioned nodes
kubectl get nodeclaims
kubectl get nodeclaims -o json | jq '.items[] | {name: .metadata.name, type: .status.instanceType, zone: .status.zone, capacity: .status.capacity}'

4. Trigger Scaling

# Deploy a workload — Karpenter auto-provisions nodes
kubectl apply -f - <<EOF
apiVersion: apps/v1
kind: Deployment
metadata:
  name: inflate
spec:
  replicas: 50
  selector:
    matchLabels:
      app: inflate
  template:
    metadata:
      labels:
        app: inflate
    spec:
      containers:
      - name: inflate
        image: public.ecr.aws/eks-distro/kubernetes/pause:3.7
        resources:
          requests:
            cpu: "1"
            memory: 1Gi
EOF

# Watch Karpenter provision nodes
kubectl logs -n karpenter -l app.kubernetes.io/name=karpenter -f | grep 'launched'

Python Example

from kubernetes import client, config

config.load_kube_config()
api = client.CustomObjectsApi()

# List NodePools
nodepools = api.list_cluster_custom_object(
    group="karpenter.sh", version="v1", plural="nodepools")

for np in nodepools["items"]:
    name = np["metadata"]["name"]
    limits = np["spec"].get("limits", {})
    print(f"NodePool: {name} | CPU limit: {limits.get('cpu','unlimited')} | Memory: {limits.get('memory','unlimited')}")

# List NodeClaims (provisioned nodes)
nodeclaims = api.list_cluster_custom_object(
    group="karpenter.sh", version="v1", plural="nodeclaims")

for nc in nodeclaims["items"]:
    print(f"Node: {nc['metadata']['name']} | Type: {nc['status'].get('instanceType','pending')} | Zone: {nc['status'].get('zone','pending')}")

Key Resources

Resource	Description
NodePool	Defines constraints for node provisioning
EC2NodeClass	AWS-specific config (AMI, subnets, security groups)
NodeClaim	Represents a provisioned node

---

Need custom data extraction or scraping solution? I build production-grade scrapers for any website. Email: [e-mail] | My Apify Actors