Lucas Sant' Anna

Kubernetes Platform Architecture

Mon, 01 Jan 0001 00:00:00 +0000

Context #

Fortis Games was building game studios and had no internal platform. Every team that needed a service running needed networking, deployment, secrets, monitoring and DNS, and without somewhere shared to get those, each team would have solved them separately and differently. I joined as infrastructure technical lead and the platform was the first thing to design.

The requirement wasn’t “run Kubernetes.” It was that a studio team should be able to ship a service without having to learn how the infrastructure underneath it is wired.

Infrastructure Automation with an Agent

Mon, 01 Jan 0001 00:00:00 +0000

Context #

A new game environment at Fortis meant a repository, the infrastructure to run it, a baseline service on EKS, observability wired up, and SSO configured. Each piece was understood. The cost was the coordination between them: about two days, spread across more than one person, with waiting in between.

The second-order effect is what bothered me. At that cost, teams stop standing things up speculatively. Experiments that should be cheap get batched, deferred or skipped, and nobody files a ticket saying “I didn’t try something.”

Consolidating 38 AWS Accounts Without Losing Cost Visibility

Mon, 01 Jan 0001 00:00:00 +0000

Context #

Fortis ran 38 AWS accounts. Consolidating them into six shared accounts was the right move operationally, and it had a consequence that’s easy to miss until the first invoice: an account per game is also a cost report per game. The account boundary had been doing accounting work nobody designed it to do.

Once several games share an account, “what does this game cost” stops having an answer the billing console can give you.

About

Mon, 01 Jan 0001 00:00:00 +0000

I’m a platform and infrastructure engineer who moved into leading the function and kept the architecture. The work I care about sits where a system’s shape is still undecided: design it, take it to production, then run and evolve it with a team that owns it.

I live in New Westminster, British Columbia, and I’ve worked remotely for most of the last decade.

Leading the function #

At Fortis Games I built DevOps Engineering from nothing. By the end it ran through four area leaders, one each for CI/CD, SRE, platform and AI infrastructure, and I kept platform architecture and technical direction alongside that. The team was spread across Brazil, Portugal, England and Canada, an eight-hour spread, so almost nothing could depend on everyone being awake at once.