Lab: Deploy, expose and update an application with correction

What you will learn in this TP

• Deploy and expose an application
• Do a rolling update
• Roll back

Application lifecycle #

What you will learn in this section

• Deploy and expose an application
• Do a rolling update
• Roll back

Context. You must ship a new application version with no downtime and be able to roll back if it misbehaves. Kubernetes provides exactly that with Deployments: you practise on a disposable local cluster.

Prerequisites: Docker, kind, kubectl  ·  Estimated time: 40 min

From manifest to zero-downtime update.

1. 1. Create the cluster and the Deployment

   Why? A Deployment maintains N replicas: if a pod disappears the controller recreates it with no intervention, that is the reconciliation loop.

   3 nginx 1.26 replicas with requests and a readinessProbe.
   See the correction

   kind create cluster --name tp
   kubectl create deployment web --image=nginx:1.26 --replicas=3 --dry-run=client -o yaml > web.yaml
   kubectl apply -f web.yaml
   kubectl rollout status deployment/web
   kubectl get pods -o wide

2. 2. Expose with a Service

   Why? Pods change IP address: the Service provides a stable access point other applications can use.

   Create a ClusterIP Service and test it with an ephemeral pod.
   See the correction

   kubectl expose deployment web --port=80
   kubectl get svc,endpoints web
   kubectl run curl --rm -it --image=curlimages/curl --restart=Never -- curl -sI http://web

3. 3. Rolling update

   Why? The rolling update replaces pods one by one: the application stays available throughout the update.

   Move to nginx 1.27 and watch pods roll over.
   See the correction

   kubectl set image deployment/web nginx=nginx:1.27
   kubectl rollout status deployment/web
   kubectl get rs
   kubectl rollout history deployment/web

4. 4. Rollback and scaling

   Why? Rollback is possible because Kubernetes keeps old ReplicaSets: a bad version is fixed with one command.

   Undo the update then scale to 5 replicas.
   See the correction

   kubectl rollout undo deployment/web
   kubectl scale deployment web --replicas=5
   kubectl get pods -l app=web
   kind delete cluster --name tp

Difficulty level : ●●○○○ (2/5)