In Part 1, we covered Deployment fundamentals, scaling, and self-healing.

In Part 2, we go deeper into:

• Rolling Updates
• Rollbacks
• Blue/Green Deployments
• Canary Deployments
• Production Best Practices

Rolling Updates

Rolling Updates allow you to update your application without downtime.

Instead of deleting all old Pods at once, Kubernetes:

1. Creates new Pods with the new version
2. Gradually terminates old Pods
3. Maintains availability during the process

Rolling Update Configuration

strategy:
  type: RollingUpdate
  rollingUpdate:
    maxUnavailable: 1
    maxSurge: 1

maxUnavailable

Maximum number of Pods that can be unavailable during the update.

Example:

replicas: 4
maxUnavailable: 1

Kubernetes guarantees that at least 3 Pods will always be available during the update.

Why it matters:

• Controls availability
• Protects against downtime
• Critical for production systems
• Important for SLA-based applications

For high-availability systems (banking, checkout, auth services), set maxUnavailable: 0 to ensure zero downtime rollout.

maxSurge

Maximum extra Pods created above desired replicas.

You can read more about maxSurge and maxUnavailable use cases in the official docs.

Updating an Image Version

kubectl set image deployment/web-deployment nginx-container=nginx:1.26

Watch the rollout:

kubectl rollout status deployment/web-deployment

Rollbacks

If something goes wrong, Kubernetes allows instant rollback.

Check rollout history:

kubectl rollout history deployment/web-deployment

Rollback:

kubectl rollout undo deployment/web-deployment

This restores the previous ReplicaSet.

Blue/Green Deployment

Blue/Green means:

• Blue = current version
• Green = new version
• Traffic switches only after validation

Instead of updating Pods gradually, you deploy a separate environment.

How It Works

1. Deploy new version (Green)
2. Test internally
3. Switch Service selector to Green
4. Remove Blue after validation

This provides zero-downtime and safe testing.

Canary Deployment

Canary releases send traffic to a small subset of users first.

Example:

• 90% traffic → old version
• 10% traffic → new version

Canary requires:

• Multiple Deployments
• Traffic splitting (Ingress / Service Mesh)

Used when you want gradual risk reduction.

Production Best Practices

Always Use Probes

livenessProbe:
  httpGet:
    path: /health
    port: 80
  initialDelaySeconds: 10
  periodSeconds: 5

readinessProbe:
  httpGet:
    path: /ready
    port: 80
  initialDelaySeconds: 5
  periodSeconds: 5

• Liveness → restart unhealthy container
• Readiness → control traffic routing

Define Resource Limits

resources:
  requests:
    cpu: "100m"
    memory: "128Mi"
  limits:
    cpu: "500m"
    memory: "256Mi"

Prevents noisy neighbor problems.

Use Proper Labels & Versioning

Example:

labels:
  app: web
  version: v2

Essential for Blue/Green & Canary strategies.

What You Learned in Part 2

• How Rolling Updates maintain availability
• How to rollback safely
• When to use Blue/Green
• When to use Canary
• Production-ready configuration tips

In Part 3, we will explore advanced Deployment topics: - Deployment failure scenarios - HPA integration - Pod disruption budgets - Real production debugging strategies