Homelab Proxmox — Self-Hosted Infrastructure
On-premise virtualization cluster with GPU passthrough, distributed storage and operational automation
Context
Problem and context
The homelab was born from the need to have a realistic test environment to experiment with technologies before bringing them to production, host self-hosted services and develop skills on stacks that in enterprise environments would require significant time and budget.
The infrastructure is completely separate from the work network and managed as a production environment: high availability where possible, automated backups, active monitoring.
Solution
Implemented solution
Proxmox VE as the main hypervisor on high-end consumer hardware (Ryzen 7 5700G, 32 GB RAM, NVMe storage) with GPU passthrough to dedicated VMs for AI workloads. Network segmented in VLANs to separate lab traffic, IoT, exposed services and management.
Hardware and Architecture
Main host
AMD Ryzen 7 5700G, 32 GB DDR4 RAM, 2 TB NVMe SSD system + 4 TB storage, NVIDIA A10 24 GB GPU (AI passthrough), AMD RX580 (media passthrough).
Network
MikroTik RB5009 as main router/firewall with VLAN tagging on all segments: VLAN 10 Management, VLAN 20 Lab, VLAN 30 IoT, VLAN 40 Exposed Services.
Storage
Synology DS923+ (4 bay, 2x8TB RAID1) as NAS storage for backups and media. Proxmox backups go to PBS (Proxmox Backup Server) in a dedicated VM with deduplication. 3-2-1 strategy: local snapshot, NAS backup, offsite copy on Backblaze B2.
Production Services
15+ active VMs and LXC containers, including:
- danplab.com (LXC): Next.js site with Nginx reverse proxy
- n8n (Docker LXC): workflow automation and API integration
- Ollama + Open WebUI (VM with A10 GPU passthrough): self-hosted LLM
- Grafana + Prometheus + Netdata: monitoring stack
- Hermes Agent (VM): AI agent with Discord bridge
- Nextcloud (LXC): file sync and internal collaboration
- AdGuard Home (LXC): DNS ad-blocking + split-horizon
- WireGuard (LXC): VPN gateway for secure remote access
Automation and Maintenance
Automatic snapshots before updates via Proxmox hooks. Bash scripts for old snapshot cleanup and backup integrity verification. Uptime monitoring and alerting via Netdata with Discord notifications. Automated OS updates with unattended-upgrades on containers and manual maintenance window for critical VMs.
Outcomes