Desired state vs actual state reconciliation in Kubernetes - Hands-On Comparison

Desired State vs Actual State Reconciliation in Kubernetes

📖 Scenario: You are managing a small web application running on Kubernetes. You want to ensure that the number of running pods always matches the number you specify in your deployment configuration. This is called reconciliation, where Kubernetes compares the desired state you set with the actual state running in the cluster and makes changes to match them.

🎯 Goal: Learn how to define a Kubernetes Deployment manifest with a specific number of replicas (desired state), check the actual number of running pods, and understand how Kubernetes reconciles the difference automatically.

📋 What You'll Learn

Create a Kubernetes Deployment manifest with exactly 3 replicas

Add a label selector to the Deployment to identify pods

Use kubectl commands to check the actual number of running pods

Observe how Kubernetes reconciles the actual state to match the desired state

💡 Why This Matters

🌍 Real World

Kubernetes operators and DevOps engineers use reconciliation to keep applications running as expected, automatically fixing differences between desired and actual states.

💼 Career

Understanding reconciliation is essential for managing Kubernetes workloads reliably and troubleshooting deployment issues in real-world cloud environments.

1

Create a Deployment manifest with 3 replicas

Create a YAML file named deployment.yaml with a Deployment resource. Set replicas to 3. Use app: mywebapp as the label under metadata.labels and spec.selector.matchLabels. Use the container image nginx:latest.

Kubernetes

# Create deployment.yaml with Deployment, replicas: 3, labels: app: mywebapp, image: nginx:latest
# Your code here

Need a hint?

Make sure the replicas field is set to 3 and the labels match exactly app: mywebapp.

2

Add a label selector to the Deployment

Ensure the Deployment manifest has a spec.selector.matchLabels field with app: mywebapp to select pods correctly. This helps Kubernetes know which pods belong to this Deployment.

Kubernetes

apiVersion: apps/v1
kind: Deployment
metadata:
  name: mywebapp-deployment
  labels:
    app: mywebapp
spec:
  replicas: 3
  # Add selector.matchLabels: app: mywebapp below
  selector:
    matchLabels:
      app: mywebapp
  template:
    metadata:
      labels:
        app: mywebapp
    spec:
      containers:
      - name: nginx-container
        image: nginx:latest

Need a hint?

The selector.matchLabels must exactly match the pod template labels.

3

Check the actual number of running pods

After applying the Deployment with kubectl apply -f deployment.yaml, run the command kubectl get pods -l app=mywebapp to list pods with the label app=mywebapp. Count how many pods are in the Running state.

Kubernetes

apiVersion: apps/v1
kind: Deployment
metadata:
  name: mywebapp-deployment
  labels:
    app: mywebapp
spec:
  replicas: 3
  selector:
    matchLabels:
      app: mywebapp
  template:
    metadata:
      labels:
        app: mywebapp
    spec:
      containers:
      - name: nginx-container
        image: nginx:latest

# Run kubectl get pods -l app=mywebapp to check running pods
# Your code here

Need a hint?

Use the label selector -l app=mywebapp to filter pods belonging to your Deployment.

4

Observe reconciliation by scaling replicas

Run kubectl scale deployment mywebapp-deployment --replicas=5 to change the desired state to 5 replicas. Then run kubectl get pods -l app=mywebapp --no-headers | wc -l to count how many pods are running. Kubernetes will create new pods to match the desired state.

Kubernetes

apiVersion: apps/v1
kind: Deployment
metadata:
  name: mywebapp-deployment
  labels:
    app: mywebapp
spec:
  replicas: 3
  selector:
    matchLabels:
      app: mywebapp
  template:
    metadata:
      labels:
        app: mywebapp
    spec:
      containers:
      - name: nginx-container
        image: nginx:latest

# Scale deployment to 5 replicas
kubectl scale deployment mywebapp-deployment --replicas=5
# Count running pods
kubectl get pods -l app=mywebapp --no-headers | wc -l
# Your code here

Need a hint?

Scaling changes the desired state. Kubernetes will create or delete pods to match it.