Scaling

Hub › Kubernetes › 06

Goal

Scale the web Deployment from 1 replica to 4 and back to 2 to see how Kubernetes distributes pods across the cluster.

Why

One replica is a single point of failure. If that pod crashes or its node becomes unavailable, the app is down until the pod is rescheduled. Scaling to multiple replicas means the Service load-balances across healthy pods and Kubernetes can reschedule if a node fails.

Step 1 — Scale up

kubectl scale deployment/web --replicas=4

Watch the new pods come up:

kubectl get pods -l app=web

Expected output once all four are ready:

NAME                   READY   STATUS    RESTARTS   AGE
web-<hash>-<hash>      1/1     Running   0          2m
web-<hash>-<hash>      1/1     Running   0          15s
web-<hash>-<hash>      1/1     Running   0          15s
web-<hash>-<hash>      1/1     Running   0          15s

Kubernetes runs the readiness probe on each new pod before marking it READY.

Step 2 — Inspect the Deployment

kubectl get deployment web

Expected output:

NAME   READY   UP-TO-DATE   AVAILABLE   AGE
web    4/4     4            4           5m

READY 4/4 confirms all four replicas passed the readiness probe and are in the Service endpoints.

Step 3 — Scale back down

kubectl scale deployment/web --replicas=2

Kubernetes terminates two pods gracefully:

kubectl get pods -l app=web

Expected output:

NAME                   READY   STATUS    RESTARTS   AGE
web-<hash>-<hash>      1/1     Running   0          3m
web-<hash>-<hash>      1/1     Running   0          3m

Checkpoint

You scaled the web Deployment up to 4 replicas and back to 2. The Service automatically updated its endpoint list as pods came and went.

Next: 07 Wrap-up