Infrastructure Cloud

As we added several small "helper" projects, the need for more powerful machines showed up fast. So we brought an old PC back to life, and it is now the main host for our production projects.

Today we run our pods on genuinely distributed physical nodes, going beyond replica redundancy inside the Kube cluster. This physical redundancy is also geographic: the servers sit in separate locations (within a 100km radius) and run on different internet service providers (ISPs), adding a final layer of structural resilience to our cloud.

To improve the developer experience (DevEx), we isolated the deployment scripts and secrets management into a separate template project.

This approach lightened the structural load a lot. Each project no longer needs a complex deployment script: a simple Dockerfile and a reference pipeline are enough. This system standardizes our deployments, whether for Node.js web projects or Steam games.

For the game Speedrun Academy, we built a custom deployment pipeline to integrate Steam directly into our GitLab CI/CD flow.

In practice, when a change is pushed to the "Dev" branch, it triggers a seamless, automatic update of the Beta branch on Steam, without even requiring 2FA at the entry point. We kept the manual step only for the final deployment to production, giving maximum velocity during testing while protecting players.

Croco is our very first on-premise server. This is the machine where we installed Kubernetes and Traefik, officially making us the owners of our cloud solution, and marking the first member of our Tailscale VPS.

Croco has a founding double duty: it also acts as the GitLab runner we use for our deployment work, making it much easier to adopt smooth continuous deployment internally.