Why Kubernetes?

Learn the advantages Kubernetes provides.

Why Kubernetes?

On a single server, Compose is enough. As the application grows, these questions come up and Kubernetes answers all of them out of the box:

Self-healing: the workload on a crashed container or server is restarted elsewhere.
Scaling: the replica count changes by hand or automatically with load.
Zero-downtime updates: a new version rolls out gradually and is rolled back if there's a problem.
Service discovery and load balancing: replicas are gathered behind a single address.

kubectl scale deployment web --replicas=10

Seeing self-healing

If you delete a deployment's pod by hand, the deployment immediately creates a new one to keep the desired replica count. -w (watch) follows the changes live:

$ kubectl delete pod web-7c5b9f6d8-abcde
pod "web-7c5b9f6d8-abcde" deleted
$ kubectl get pods -w
web-7c5b9f6d8-pqrst   0/1   ContainerCreating   0   1s
web-7c5b9f6d8-pqrst   1/1   Running             0   3s

Exercise Scale and watch

Write two commands: scale the web deployment to 5 replicas; then watch the pods live.

Show example solution

kubectl scale deployment web --replicas=5
kubectl get pods -w

Test yourself

When can Kubernetes be unnecessary complexity?
- For a small application running on one server
- For hundreds of microservices
- For multi-region deployment
- When autoscaling is needed
Show answer

For a small application running on one server — Kubernetes has an operational cost; without the need, Compose is simpler.

Scale the deployment to 10 replicas.

kubectl ____ deployment web --replicas=10

Show answer

kubectl scale deployment web --replicas=10

Match the features.

self-healing

Restarts what crashed
scaling

Changes the replica count
rolling update

Zero-downtime update
service discovery

Replicas behind one address

Next lesson: Cluster architecture

We Are Enhancing Your Experience

Cookie Preferences

Strictly Necessary Cookies

Analytical Cookies