A rolling update gradually replaces running application instances with a newer container image, preserving service availability while the deployment moves toward the desired state. If a new pod encounters errors, the update pauses instead of replacing every old instance. section 3.3.3
A Kubernetes Deployment coordinates the update:
- The Deployment manages ReplicaSets, which manage pods containing the application container. section 3.3.3
- Changing the image from hello-joker:v1 to hello-joker:v2 causes Kubernetes to scale the new ReplicaSet up while scaling the old ReplicaSet down. section 3.3.3
Updating hello-joker
Apply a Deployment with three replicas, change its image to v2, and run kubectl apply -f deployment.yaml. Inspect readiness and Deployment events to observe the new and old ReplicaSets being scaled in turn.
