Skip to main content

Deployments

You just saw that a bare Pod, once deleted, stays gone. A Deployment fixes that. You tell it "I want 3 copies of this Pod running at all times," and it relentlessly makes that true - restarting crashed pods, replacing ones on failed nodes, and keeping the count exactly where you asked.

Under the hood, a Deployment manages a ReplicaSet, which in turn manages the Pods:

🚀 Create a Deployment​

Open k8s/deployment.yaml. Notice three important parts:

  • replicas: 3 - the desired number of Pods
  • selector - how the Deployment knows which Pods belong to it (by the label app: web)
  • template - the Pod blueprint it stamps out, including a readinessProbe so Kubernetes only sends traffic to pods that are actually ready

  1. Apply the Deployment:

    kubectl apply -f k8s/deployment.yaml

  2. Watch the pods appear. Run this a few times until all 3 are Running and READY 1/1:

    kubectl get pods

    Notice the pod names now have random suffixes like web-7d9c8b6f5-abc12 - the Deployment generated them for you.

  3. See the Deployment's own view of the world:

    kubectl get deployment web

    READY 3/3 means desired state matches actual state.

  4. Peek at the ReplicaSet the Deployment created to manage those pods:

    kubectl get replicaset

🩹 Watch it heal itself​

This is the magic moment. Delete a pod and watch Kubernetes notice and replace it - instantly.

  1. Delete exactly one pod (this command grabs the name of the first one for you):

    kubectl delete pod $(kubectl get pods -l app=web -o jsonpath='{.items[0].metadata.name}')
tip
You can also copy a specific pod name from `kubectl get pods` and run `kubectl delete pod <name>` yourself.

  1. Immediately list the pods again:

    kubectl get pods

    You'll still see 3 pods - one is brand new (check its AGE), already replacing the one you deleted. The Deployment's reconciliation loop noticed the count dropped to 2 and created a replacement to get back to 3. You never told it to; it just does that. 🩹

🧠 The big idea​

You declared a desired state ("3 healthy pods running this image") and Kubernetes takes ongoing responsibility for keeping it true. This is what makes Kubernetes resilient - and it's why you'll almost never create bare Pods in real life.

But there's a catch: those 3 pods each have their own IP, and those IPs change every time a pod is replaced. How does anything reliably reach your app? That's the job of a Service - up next. 🔌