MET : Our vision (1/6)
How a web agency decided to take back control of its cloud infrastructure by building a modern, reproducible, and documented Kubernetes platform.
For years, we used traditional hosting solutions: shared servers, manually configured VPS, PaaS platforms like Jelastic or Platform.sh. Each solution had its advantages, but also its frustrations.
The reality check that changed everything
Here are the problems we regularly encountered:
- Rising costs: PaaS platforms charge by usage, and the bill quickly adds up with multiple clients
- Lack of control: Impossible to customize certain critical aspects (cache, server configuration, security)
- Vendor lock-in: Migrating from one platform to another = a project in itself
- Manual configuration: Each new server = hours of configuration, with risk of oversight
- No reproducibility: “It worked on my server” is not a viable strategy
We needed infrastructure that was both powerful, economical, and fully automated.
Our vision: Infrastructure as Code
We decided to build our own cloud platform, based on modern principles:
“Everything that can be automated should be. Everything that is configured must be versioned.”
Concretely, this means:
- Declarative infrastructure: Servers are defined in code (Terraform), not created manually
- Automated configuration: Ansible configures everything, from OS to applications
- Modern orchestration: Kubernetes manages the application lifecycle
- Continuous deployment: A Git push automatically triggers deployment
- Integrated backup: Automatic backups to the cloud, without intervention
The architecture we built
Our infrastructure relies on two high-performance servers hosted at Hetzner Cloud (German datacenter, GDPR-compliant):
| Server | Role |
|---|---|
| Primary Server | Control Plane + Worker |
| Secondary Server | Worker |
These two servers form a Kubernetes cluster capable of hosting dozens of applications with high availability.
The tech stack at a glance
Here are the technologies we use, which we’ll detail in upcoming articles:
🏗️ Provisioning
- Terraform: Server and network creation
- Hetzner Cloud: High-performance German hosting
⚙️ Configuration
- Ansible: Configuration automation
- K3s: Lightweight Kubernetes distribution
🚀 Orchestration
- Kubernetes: Container orchestration
- Rancher: Intuitive management interface
- Helm: Application packaging
🔄 Deployment
- GitLab: Code hosting and CI/CD
- ArgoCD: Automated GitOps deployment
🔒 Security
- Let’s Encrypt: Automatic SSL certificates
- WireGuard VPN: Secure access to internal services
- Sealed Secrets: Encrypted secrets in Git
💾 Storage & Backup
- Longhorn: Distributed storage with replication
- Velero: Backup to S3
What this infrastructure brings us
After several months of production use, here are the concrete benefits:
💰 Significant savings
For the same budget as a PaaS hosting 3-4 sites, we host over 20 applications with far superior resources.
⚡ Deployment in minutes
A new WordPress site? 5 minutes of configuration, automatic deployment, SSL included.
🔄 Total reproducibility
The entire infrastructure is versioned. In case of disaster, we can rebuild everything in less than an hour.
📊 Complete visibility
Integrated monitoring, centralized logs, automatic alerts. We know exactly what’s happening.
🛡️ Enhanced security
Client isolation by namespace, VPN for admin access, encrypted secrets, automatic backup.
What’s next in this series
In the upcoming articles, we’ll detail each component of this infrastructure:
- Terraform: How to provision your cloud infrastructure in a few lines of code
- Ansible: Configure 100 servers as easily as one
- Kubernetes & Rancher: Orchestration demystified
- Helm Charts: Deploy any application in 5 minutes
- GitOps & CI/CD: Git push = Automatic deployment
Each article will contain concrete code examples and best practices from our production experience.
