Kubernetes without the weight
We started with a Kubernetes emulator for the DevOps course. It turned out to be easier to build our own small version.
In the DevOps course at TU-Sofia we get to Kubernetes towards the end of the semester. The lab machines are weak, and a full cluster does not run well on each of them. So we started writing an emulator: something that behaves like Kubernetes in front of the students without being a real cluster.
The more we emulated, the closer we got to the real thing. Every behaviour the students need to see (replicas that come back on their own, rolling updates, rollbacks) meant we had to actually build it. At some point it became clear that it was easier to make our own small version of Kubernetes than to keep imitating it. That is how Royak was born.
Kubernetes is often taught as a list of objects: Pod, Deployment, Service, Role. The list is correct, but it does not explain why each of them exists. It is easier to understand if you start from one container and add only what the next problem forces on you.
From one container to an orchestrator
The heart of it all: the controller
Under all of this sits one simple idea. You say what you want, and the controller keeps looking at what there is and closes the gap. Instead of “start” and “stop” commands, you describe a state.
Try it below. Change the number of replicas, kill a pod or roll out a new version and watch how the controller reacts. With a new version, pay attention to the order: a new replica starts first, and only once it is ready does an old one stop.
Fail fast while it is cheap
In DevOps the most important rule is to fail fast. The earlier a mistake is stopped, the cheaper it is.
Press “Broken version” in the simulator. The new replica starts but never becomes ready. The controller does not stop a single old replica until the new one is ready, so users keep getting answers from the old version. When the new one does not become ready in time, the update stops and rolls itself back. The mistake stays with one replica that nobody ever saw.
Without the readiness check, the same version would have replaced every replica one by one and the service would have gone down for everyone. The check costs a few seconds on every update. Skipping it costs an outage.
Royak itself went down the same road. When the emulator started to grow, we stopped early and changed direction instead of piling up more imitation.
When this is enough
Royak is out in its first public beta. Update, end of September 2026: Royak received its first approval to run at clients and is now running in their internal cloud. It fits best where it started: learning, demos, home labs and small machines at the edge of the network. For a large system with many teams, Kubernetes remains the right choice. Every known limitation is described in the repository, together with a table of which parts of Kubernetes work and how they are tested.
If you want to see with your own eyes what an orchestrator does without setting up a whole cluster, it takes under two minutes:
royak apply examples/nginx.yaml # 2 nginx replicas
royak get pods
royak scale web 5
royak get deployment/web -o yaml # get, edit, apply, just like in Kubernetes
The code is open, under the MIT licence: github.com/evotech-bg/royak.