I'm currently setting up a test Kubernetes environment using KIND on my Contabo VPS, and I'm looking for the best way to expose services externally without resorting to NodePort hacks. I have come across two main options: MetalLB and cloud-provider-kind. My goal is to mimic real Kubernetes networking as closely as possible for CI and testing purposes, not for production traffic. For anyone with experience on VPS providers like Contabo or Hetzner, which approach worked for you? Also, are there any potential issues with using MetalLB on a single-node KIND cluster?
3 Answers
You might want to consider alternatives like k3s with metallb and traefik, or k0s paired with cilium. I'm currently doing a proof of concept with k0s and cilium for multi-cluster management, and it works surprisingly well while keeping costs down. Might be worth looking into!
I've had great success with MetalLB in single-node setups for testing like yours. It really makes the cluster behave more like a real Kubernetes environment. Also, if you want to test across different regions, check out Lightnode's global locations—they're pretty handy!
Thanks for the insight!
This might sound off the wall, but using ngrok's gateway option has been a blast for quick ephemeral clusters! It's super easy to get started with. Check it out [here](https://ngrok.com/docs/k8s).
Ngrok definitely works for quick demos. My concern is that it doesn't really follow Kubernetes service semantics; it feels more like just a tunnel.
Are you using this setup for your production environment?