State of the Cluster: Embracing Talos Linux

If you’ve followed my philosophy for this website—no trackers, no cookies, no unnecessary bloat—it shouldn’t come as a surprise that I hold my home infrastructure to the same standard. Recently, I have finally managed to finish the basic setup of my k8s cluster using Talos Linux.

Why Talos?

There is a chance you may ask yourself why I am using Talos Linux instead of something like OKD (which is like the Fedora of Openshift), or something like MicroK8s. The reason is actually quite simple as OKD is overkill and has much more out of the box than I require on my somewhat limited homelab. Talos has the advantage over full Linux Distros like Ubuntu or OpenSuse that is is tailored for using it as a k8s distribution.

Talos is immutable, minimal, and secure by default. There is no SSH, no console, and no shell. Everything is managed via an API. You for example use the talosctl command to patch/update your nodes. The command for this basically is talosctl upgrade --nodes 10.20.30.40 --image ghcr.io/siderolabs/installer:v1.8.0. You of course need to adjust the node IP and image version. This is much more comfortable than a fully fledged Linux distribution.

Current Cluster Status

Used Hardware

I am using Proxmox Cluster (that also runs a few more services like a DNS server and VPN) to provide 6 VMs for the Talos cluster.

• Control Plane: 3 dedicated nodes to ensure high availability. Each is 2 CPU cores and 8 GB RAM
• Worker Nodes: 6 nodes handling the actual compute workloads. Each is 6 CPU cores and 16 GB RAM

As for the hardware nodes, I am using Mini PCs, specifically with a Ryzen 7 8745hs, to provide the compute power. The CPU usage on my Cluster is not too much and they also provide a reasonably strong iGPU for tasks that need it. All in all, this does provide more than enough compute power for what I need. It also runs quite a lot of services for the eSports association Classic Tetromino-Games Germany e.V.

Admin Services

I of course am using a few services to make managing the cluster and services reasonable. A non-conclusive list is the following:

• The official Kubernetes web UI Dashboard.
• MetalLB for VIP (currently using the .120-.140 range).
• csi-driver-nfs to access network shares, like my fileshare server.
• Nginx as the Ingress.

Examples of services

I also am of course running quite a lot of services on my k8s cluster, the main reason why I am operating it. All of these are currently only accessible via my local network and a wireguard VPN, some may be opened up in the future.

The services I am running can be categorized in two main categories, editors and viewers.

• Editors
  • bentopdf for PDF editing.
  • minipaint for a basic paint tool wherever I am.
  • penpot as I want to get into designing.
• Viewers
  • Frontends
    • Invidious to access YouTube with a locally running subscription (synced between devices as the database/logins are on the server side).
    • Redlib to access Reddit wherever I am. Here the subreddits I follow are stored client side and need to be imported per device when I follow a new one.
    • Rimgo to access Imgur from anywhere.

Small tease

The next step is sharing my terraform setup I used to deploy these machines.