Is a Small Cluster with Only Two Worker Nodes and One Manager Worth It?

Regarding those tutorials that keep it simple with one physical host running multiple VMs, that’s really just for ease of demonstration. Most people don’t have numerous nodes at home for testing. During learning, it's okay to tolerate downtime, but it’s definitely not something you'd rely on for production. A single machine can fail, taking down every VM along with it, so in a real-world scenario, you wouldn't want to do that.

As others mentioned, your approach is solid for learning or edge deployments. Just remember, if you don’t have three control plane nodes, any OS or Kubernetes upgrades might lead to downtime, since you’ll lack a proper high availability setup. It’s all about balancing your needs with the resources you have!

Having just two nodes might not be ideal for high availability, especially since you won't have backup control plane nodes. Raspberry Pis are great for home setups, but mixing architectures—like a Raspberry Pi with ARM and a laptop with x86—can get tricky. You'd need to make sure that any services can run on both types of hardware or else they'll only work on the architecture that supports them.

It really boils down to what you want to achieve. For learning and experimenting, your setup totally makes sense! If you're thinking of self-hosting non-critical apps, it could work—though it may not be the most efficient setup unless that’s all you’ve got right now. But if you want a robust environment or high availability, then you should aim for at least three control plane nodes. Oh, and steer clear of SD cards for important tasks like etcd; they wear out fast.