HA K8s Docs

[reavessm]

Documentation for running a highly available kubernetes cluster

This doc runs through a complex example using the kubernetes sysext. It involves multiple kubernetes configurations like swapping container runtimes and CNIs, as well as more advanced ignition configurations like running static pods via Quadlet and creating a hierarchy of configs. This also demonstrates how to use a make file to spin up multiple distinct flatcar VMs in a controlled manner.

This doc may need to be updated once flatcar/sysext-bakery#114 gets merged.

Also, this is my first time writing technical documentation of this nature so please feel free to critique this technically, grammatically, or in any other manner I may have missed.

[tormath1]

Thanks a lot for documenting this setup - I left some comments but it globally looks really good.

[tormath1]

I ran the tutorial until the Ignition generation and it went fine. I think it's worth to mention again Typhoon project: https://github.com/poseidon/typhoon for folks into Terraform deployments.

[ader1990]

Wow, great documentation, I am sure it will be extremely helpful for people who want to know what happening under the hood.

One question, kube-vip or metallb/cilium is usually the preferred way to have HA deployments. Any reason in particular for having HAProxy and KeepAlived?

[reavessm]

Thanks! I hope so!

Yes and no. It's mostly for educational purposes because I haven't seen a lot of documentation on running podman/quadlet with k8s. On the technical side of things, as far as I understand it, Cilium does kind of overlap with metallb with it's L2 aware stuff that's currently in beta, so this doc only installs the stuff that Cilium can't do. Also, it is documented in the kubeadm docs

[tormath1]

LGTM. Feel free to squash the fixup! commits and we're good to go.

[tormath1]

Thanks again @reavessm for this PR. Changes are live here: https://www.flatcar.org/docs/latest/container-runtimes/high-availability-kubernetes/ 🎉