How Does a Kubernetes Cluster Work?

Let's take a simple example.

Suppose you want to run a web application with three replicas.

You define the desired state:

replicas: 3

Then you submit the configuration to Kubernetes.

Something like:

kubectl apply -f deployment.yaml

The flow is roughly:

Developer
    |
    v
kubectl
    |
    v
API Server
    |
    v
Scheduler / Controllers
    |
    v
Worker Node
    |
    v
Pod
    |
    v
Container

• The scheduler decides where the Pod should run.
• The kubelet on that node makes sure the Pod is running.
• The controllers continue watching the state of the cluster.
• If something changes, Kubernetes works to move the actual state back toward the desired state.