Kubeadm Multi-master K8s Cluster with Kube-VIP

Sat, 01 Nov 2025 15:18:00 +0300

Introduction

I had turned inter-webs upside down back in late 2019 - early 2020 (pre-COVID) to find a tutorial that I could follow on multi-master Kubernetes clusters on bare-metal to no avail. Well, I have revisited that exercise, flock around! This tutorial will use VMs (on PVE as usual) for ease of use though, especially for backups.

CloudInit Debian 13 AMD64 machines are used (remember to increase disk size first :D). Checkout Techno Tim’s video which was linked in proxmox-cloudinit-arm64-vm for details or just use NoCloud/manual install Debian.
Bridge network is used on the host to better simulate bare-metal experience. However, for those who would try this on BM, adjust interface names, which are conveniently all eth0 in this post.
I would checkout Talos for bare-metal, Kairos (on Yocto) for embedded scenarios. If you work for government, nobody ever got fired for choosing IBM [or SUSE].
I am yet to test Cilium out for CNI. It is praised for its observability. But one step at a time, Calico is used instead.
Unfortunately, redundant storage (e.g. Rook, Longhorn) is not within scope of this exercise either. Hopefully, there will be follow up posts.

Aim

A self-hosted 3-node kubernetes cluster where each node is part of control plane.
Ability to take down any one of the nodes and keep having operational API server and Services.
Use of simple network equipment, as those are not cheap in where I currently am.

For the bullet points above, this should be threated as just an educational resource.

K8s on cbugk's blog

Kubeadm Multi-master K8s Cluster with Kube-VIP

Introduction

Aim