How I Deployed this Website to Kubernetes

Introduction

Hi! I deployed my personal website safademirdag.com to production using modern DevOps practices. In this post, I'll share which technologies I used and how I set up the deployment process without diving too deep into technical details.

---

Technologies Used

Application

• Next.js 14 + TypeScript
• Docker

Infrastructure

• 3 Servers (Master, Worker, Runner)
• Kubernetes (RKE2)
• GitLab CI/CD
• NGINX Ingress + Let's Encrypt

---

Architecture

I used three servers:

• Master Node: Kubernetes brain, all decisions made here
• Worker Node: My application runs here
• Runner Node: GitLab CI/CD pipelines run here

---

Deployment Process

1. Preparing Dockerfile

I converted Next.js into a small Docker image using multi-stage build. With output: 'standalone' setting in next.config.js, only necessary files go to production.

2. Kubernetes Manifests

I created three YAML files:

• Deployment: Runs 2 copies of the application
• Service: Provides internal network access to pods
• Ingress: Routes safademirdag.com domain to the application

3. GitLab CI/CD Pipeline

Two stages that run automatically on every git push:

1. Build: Create Docker image and upload to Registry
2. Deploy: Deploy new image to Kubernetes

4. DNS and SSL

• DNS A record: safademirdag.com → Master Node IP
• cert-manager + Let's Encrypt: Automatic SSL certificate

---

Flow

When I make a code change:

1. Git push
2. GitLab Runner builds Docker image (~5 minutes)
3. Image uploaded to Registry
4. Kubernetes starts new pods
5. Old pods shut down (zero-downtime)
6. Site live with new version!

Entire process takes 7-10 minutes and is fully automated.

---

Lessons Learned

Best Practices

• Small images with multi-stage Docker build
• Isolation with Kubernetes namespaces
• Auto-restart with health checks
• Uninterrupted deployment with rolling updates

Problems I Encountered

• Need to use node server.js instead of npm start in Dockerfile
• Need to create imagePullSecret for GitLab Registry
• Kubeconfig IP settings for runner to access master node

---

Conclusion

Now I can deploy to production with a single push. Thanks to Kubernetes, CI/CD, and containerization:

• Automatic deployment
• Zero-downtime updates
• Easily scalable
• Infrastructure as Code

I can easily focus on writing my code without worrying deployment issues!

---