Expose a service with Ingress

If your cluster was provisioned with the optional ingress bundle, it already runs Traefik (the default traefik IngressClass) and cert-manager with a letsencrypt-prod ClusterIssuer.

This guide uses the Ingress API, which is the simplest path. For more control (header routing, traffic splitting, multiple backends) see Expose a service with the Gateway API.

We start over plain HTTP to confirm routing works, then add a domain and HTTPS by overwriting the same Ingress.

Deploy a sample app

cat <<'EOF' | kubectl apply -f -
apiVersion: apps/v1
kind: Deployment
metadata:
  name: hello-ingress
spec:
  replicas: 1
  selector:
    matchLabels:
      app: hello-ingress
  template:
    metadata:
      labels:
        app: hello-ingress
    spec:
      containers:
        - name: hello-ingress
          image: nginx
          ports:
            - containerPort: 80
---
apiVersion: v1
kind: Service
metadata:
  name: hello-ingress
spec:
  selector:
    app: hello-ingress
  ports:
    - port: 80
      targetPort: 80
EOF

Expose over plain HTTP

Caution

Plain HTTP is unencrypted: traffic, including anything sensitive, travels in the clear and can be read or tampered with in transit. Use this only to confirm routing. For real traffic, add a domain and HTTPS.

Create an Ingress. This rule matches any request, so Traefik serves it to anyone hitting the node:

cat <<'EOF' | kubectl apply -f -
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  name: hello-ingress
spec:
  ingressClassName: traefik
  rules:
    - http:
        paths:
          - path: /
            pathType: Prefix
            backend:
              service:
                name: hello-ingress
                port:
                  number: 80
EOF

Reach it directly at your node IP (the same IP as your API server, https://<your-node-ip>:6443):

curl http://<your-node-ip>

You should get the nginx welcome page.

Add a domain and HTTPS

To serve real traffic over HTTPS you need a hostname that resolves to the cluster. Certificates are issued by Let’s Encrypt over an HTTP-01 challenge, so DNS must resolve before you apply.

In your DNS provider, create an A record pointing your hostname at your cluster’s node IP:

ingress.example.com.  A  <your-node-ip>

Confirm it resolves before continuing:

dig +short ingress.example.com

Now overwrite the Ingress with your real hostname, the cert-manager.io/cluster-issuer annotation, and a tls block. cert-manager issues a certificate into the hello-ingress-tls secret; Traefik then serves it.

Caution

Replace ingress.example.com with your own hostname before applying.

cat <<'EOF' | kubectl apply -f -
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  name: hello-ingress
  annotations:
    cert-manager.io/cluster-issuer: letsencrypt-prod
spec:
  ingressClassName: traefik
  tls:
    - hosts:
        - ingress.example.com
      secretName: hello-ingress-tls
  rules:
    - host: ingress.example.com
      http:
        paths:
          - path: /
            pathType: Prefix
            backend:
              service:
                name: hello-ingress
                port:
                  number: 80
EOF

The certificate takes a minute or two on first issuance. Watch it become ready:

kubectl get certificate hello-ingress-tls -w

Once READY is True, your app is live over HTTPS:

curl https://ingress.example.com

If issuance stalls, describe the certificate to see the challenge status:

kubectl describe certificate hello-ingress-tls

The usual cause is DNS that doesn’t yet resolve to your node IP.

Clean up

kubectl delete ingress hello-ingress
kubectl delete service hello-ingress
kubectl delete deployment hello-ingress
kubectl delete secret hello-ingress-tls --ignore-not-found