Metrics Server — Lightweight Core Metrics for Kubernetes Autoscaling

What it is

Kubernetes Metrics Server is a lightweight, cluster-wide aggregator of resource usage data. It collects CPU and memory metrics from kubelets and exposes them through the Kubernetes Metrics API. This API powers kubectl top, HorizontalPodAutoscaler (HPA), and VerticalPodAutoscaler (VPA) without requiring a full monitoring stack.

The tool targets cluster operators who need autoscaling and basic resource visibility without deploying Prometheus, Grafana, or other heavyweight monitoring solutions.

How it saves time or tokens

Metrics Server is a single deployment that replaces the need to configure a full monitoring pipeline just for autoscaling. Without it, kubectl top returns nothing and HPA cannot make scaling decisions. Installing it takes one command and provides immediate visibility into pod and node resource consumption. It stores only the latest data point in memory, so it adds minimal overhead to the cluster.

How to use

1. Install Metrics Server with a single command:

kubectl apply -f https://github.com/kubernetes-sigs/metrics-server/releases/latest/download/components.yaml

1. For development clusters with self-signed kubelet certificates:

kubectl patch deployment metrics-server -n kube-system \
  --type='json' \
  -p='[{"op": "add", "path": "/spec/template/spec/containers/0/args/-", "value": "--kubelet-insecure-tls"}]'

1. Verify metrics are available:

kubectl top nodes
kubectl top pods -A

Example

# HPA that uses Metrics Server data
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: web-app-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: web-app
  minReplicas: 2
  maxReplicas: 10
  metrics:
    - type: Resource
      resource:
        name: cpu
        target:
          type: Utilization
          averageUtilization: 70
    - type: Resource
      resource:
        name: memory
        target:
          type: Utilization
          averageUtilization: 80

Related on TokRepo

• DevOps tools directory -- Kubernetes management and infrastructure tools
• Monitoring tools -- Full observability stacks and metrics platforms

Common pitfalls

• On minikube and kind clusters, kubelet certificates are self-signed; Metrics Server fails to scrape without the --kubelet-insecure-tls flag
• Metrics Server only stores the latest data point; it is not a replacement for Prometheus or other time-series databases for historical analysis
• HPA requires resource requests to be set on containers; without them, utilization percentages cannot be calculated and autoscaling silently fails