01: First Deployment

Status: completed · Cluster: kind v1.37.0 (single control-plane node) · Tooling: kubectl, Docker Desktop

Objective

Run two replicas of an nginx web server with a Deployment, then verify that Kubernetes restores the desired state automatically when a Pod is deleted.

By the end of this exercise you should be able to explain:

1. How a Deployment, a ReplicaSet, and Pods relate to each other.
2. Why the selector and the Pod template labels must match.
3. What "desired state" means and how Kubernetes enforces it.

Concepts

Object hierarchy

• A Deployment declares the desired state.
• A ReplicaSet (created automatically) enforces the replica count.
• Pods run on Nodes. A Node is the machine; the Pod runs inside it.

Self-healing loop

The controller compares desired (2) with actual (1) and closes the gap. This loop never stops running.

Files

Reading the manifest

apiVersion: apps/v1          # API group and version that defines Deployment
kind: Deployment             # object type
metadata:
  name: hello-nginx          # unique name in the namespace
spec:                        # desired state
  replicas: 2                # keep exactly 2 Pods running
  selector:
    matchLabels:
      app: hello-nginx       # how the Deployment finds the Pods it owns
  template:                  # blueprint for every Pod
    metadata:
      labels:
        app: hello-nginx     # label stamped on each Pod (must match selector)
    spec:
      containers:            # list of containers in each Pod
        - name: nginx
          image: nginx:1.27  # pinned tag; avoid "latest"
          ports:
            - containerPort: 80   # informational; does not publish the port

Syntax rules used here

• Indentation uses spaces (2 per level). Tabs are invalid.
• key: value is a mapping. A leading - marks a list item.
• # starts a comment.
• Each key may appear only once per mapping. A duplicate key is an error.

Steps

# 1. Start the cluster (Docker Desktop must be running)
kind create cluster

# 2. Confirm the node is Ready
kubectl get nodes

# 3. Change into the exercise folder, then apply the manifest
cd /Users/kuroko/Desktop/Nikan/k8s-practice/01-first-deployment
kubectl apply -f deployment.yaml

# 4. List the Pods
kubectl get pods

# 5. Delete one Pod and observe the replacement
kubectl delete pod <pod-name>
kubectl get pods

Results

Step 2: node

NAME                 STATUS   ROLES           AGE     VERSION
kind-control-plane   Ready    control-plane   3m44s   v1.37.0

Step 4: initial Pods

NAME                           READY   STATUS    RESTARTS   AGE
hello-nginx-79788c7c7d-2sxmr   1/1     Running   0          40s
hello-nginx-79788c7c7d-7mrj5   1/1     Running   0          40s

Step 5: after deleting 2sxmr

NAME                           READY   STATUS    RESTARTS   AGE
hello-nginx-79788c7c7d-7mrj5   1/1     Running   0          2m12s
hello-nginx-79788c7c7d-f6584   1/1     Running   0          0s

The deleted Pod is gone. A new Pod (f6584) was created within a second, and the count is back to 2.

Observations

1. Name anatomy. hello-nginx-79788c7c7d-7mrj5 is three parts: the Deployment name, the ReplicaSet hash (79788c7c7d, which changes when the Pod template changes), and a unique Pod suffix.
2. Self-healing works. Deleting a Pod does not reduce the replica count. The controller recreates it immediately.
3. READY 1/1 means the one container in the Pod passed its readiness check. RESTARTS 0 means it has not crashed.
4. ContainerCreating is normal on first run. The image has to be pulled before the container starts.
5. Two Deployments can share a cluster as long as their selector labels differ (see below).

Common mistakes

Extra: a second Deployment

A second Deployment (nikan) was created from its own file with a separate label:

selector:
  matchLabels:
    app: nikan

Result: four Pods, two from each Deployment, running side by side without interfering with each other.

Interview phrasing

"A Deployment declares how many replicas I want. Its ReplicaSet makes sure that many Pods are running, and if one fails, it's replaced automatically."

Clean up

kubectl delete -f deployment.yaml
kubectl delete -f deployment-nikan.yaml
kind delete cluster

Next

Exercise 02: Service. The Pods above have no stable address and cannot be reached from outside the cluster. A Service gives them one address and load-balances traffic across the replicas.