← All homelab projects

Two older servers, many possibilities

My Proxmox homelab

My homelab does not run on small, shiny mini PCs, but on two older servers with a role of their own. With Proxmox, I use them as a flexible environment for containers, virtual machines, storage, backups and experiments that I want to understand and manage at home.

Active
  • Proxmox
  • Linux
  • LXC
  • Virtual machines
  • Storage
  • Backups
Illustration of two older large servers, network storage and virtual systems in a domestic homelab environment.

At a glance

Technology and key information

Active Tony and JarvosProxmox clusterLXC and virtual machinesRAID1 and Proxmox Backup Server

Tony and Jarvos

The two machines are named Tony and Jarvos. Tony is mainly where services and experiments run. Jarvos takes on more storage and backup work. Dividing their roles lets me try new things without making everything depend on one system.

Together they form a Proxmox cluster. This provides one central environment for managing containers, virtual machines and storage, while both physical machines remain useful on their own.

What runs on them

The environment hosts several personal services, including ESPHome, a Cloudflare tunnel, OpenClaw and a local AI server. Some components run in lightweight Linux containers; other applications can use a full virtual machine when they need more isolation or their own operating system.

That mixture makes the homelab interesting: every application requires a fresh choice between simplicity, performance, isolation and maintenance.

Storage and backups

Older hardware can still be very useful, but storage needs extra attention. Both servers combine their system drive with two hard drives in RAID1. That protects against one drive failing, but does not replace a real backup.

I therefore use Proxmox Backup Server and scheduled backups with different retention periods. This lets me restore a container or virtual machine after a failed change instead of preserving only an individual file.

Problems are part of it

A homelab is never entirely finished. During the build I dealt with cluster settings, quorum, storage choices and a lingering cluster node. Such problems are frustrating when they appear, but they reveal how the underlying infrastructure really works.

I therefore keep improving my change notes, checking backups and treating recovery as a practical task rather than a theoretical option.

Building further

Next steps

  1. Refine the division of tasks between Tony and Jarvos.
  2. Keep testing backups and recovery procedures regularly.
  3. Maintain documentation so experiments remain reproducible.
  4. Add new services only when they serve a clear purpose.
← All homelab projects
Illustration of an older computer with a graphics card powering a network of compact AI models locally.
Experiment

Local AI with llama.cpp

An experiment with llama.cpp, compact language models and an older NVIDIA Quadro to discover what local AI can do on limited hardware.

View project →
Illustration of a self-hosted server connecting containers and local services through a gateway to chat messages, notifications and status checks.
In development

OpenClaw as a personal gateway

A self-hosted OpenClaw gateway that brings chat, local services, automation and health checks together in my personal homelab.

View project →