Enabling CPU and Memory Stats in k9s on Bare-Metal Kubernetes

k9s displays CPU and MEM columns for pods and nodes, but on bare-metal kubeadm clusters these show “N/A” by default. This occurs because k9s relies on the Kubernetes Metrics API, which requires metrics-server to be installed.

The Problem

Without metrics-server:

NAME             CPU   MEM
k8s-master01     N/A   N/A
k8s-worker01     N/A   N/A

Managed Kubernetes services (EKS, GKE, AKS) typically pre-install metrics-server. Bare-metal kubeadm clusters do not include it.

Solution

Install metrics-server via Helm with configuration for kubeadm’s self-signed certificates.

Directory Structure

infrastructure/metrics-server/
├── metrics-server-values.yaml
├── setup-metrics-server.sh
└── README.md

Helm Values

# metrics-server-values.yaml

args:
  # Required for kubeadm clusters with self-signed kubelet certificates
  - --kubelet-insecure-tls
  # Prefer internal IP for kubelet connections
  - --kubelet-preferred-address-types=InternalIP,ExternalIP,Hostname

# Run on control plane
nodeSelector:
  node-role.kubernetes.io/control-plane: ""

tolerations:
  - key: node-role.kubernetes.io/control-plane
    operator: Exists
    effect: NoSchedule

resources:
  requests:
    cpu: 100m
    memory: 200Mi
  limits:
    memory: 256Mi

# Linkerd integration
podAnnotations:
  linkerd.io/inject: enabled

The --kubelet-insecure-tls Flag

Kubeadm clusters use self-signed certificates for kubelet endpoints. Without this flag, metrics-server fails with:

unable to fully scrape metrics from node k8s-worker01:
x509: certificate signed by unknown authority

This flag skips TLS verification for kubelet connections. In a controlled homelab environment, this trade-off is acceptable compared to setting up proper PKI for kubelet serving certificates.

Setup Script

#!/bin/bash
set -euo pipefail

SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
NAMESPACE="kube-system"

helm repo add metrics-server https://kubernetes-sigs.github.io/metrics-server/ 2>/dev/null || true
helm repo update metrics-server

helm upgrade --install metrics-server metrics-server/metrics-server \
  -f "${SCRIPT_DIR}/metrics-server-values.yaml" \
  -n "${NAMESPACE}"

kubectl rollout status deployment/metrics-server -n "${NAMESPACE}" --timeout=120s

sleep 5
kubectl top nodes || echo "Note: Metrics may take 30-60 seconds to populate"

Installation

cd infrastructure/metrics-server/
chmod +x setup-metrics-server.sh
./setup-metrics-server.sh

Verification

Wait approximately 30 seconds after installation for metrics to populate:

kubectl top nodes

NAME             CPU(cores)   CPU%   MEMORY(bytes)   MEMORY%
k8s-master01     592m         14%    4502Mi          29%
k8s-worker01     197m         4%     3527Mi          22%
k8s-worker02     335m         8%     4053Mi          26%
polycephala      1750m        0%     23662Mi         9%

k9s now displays actual CPU and MEM values instead of N/A.

What metrics-server Enables

Troubleshooting

If metrics show N/A after installation:

# Check pod logs
kubectl logs -n kube-system -l app.kubernetes.io/name=metrics-server

# Verify API registration
kubectl get apiservice v1beta1.metrics.k8s.io

Metrics typically take 30-60 seconds to appear after the API aggregation layer registers.

Notes

• Headlamp and similar dashboards may show metrics even without metrics-server because they query Prometheus directly. k9s and kubectl top require the Metrics API.
• Running metrics-server on the control plane with appropriate tolerations keeps it co-located with other critical cluster components.

Configuration available at k8s-configs/infrastructure/metrics-server.