Infrastructure
The honest story of where this homelab's backup posture stands today — and the roadmap to get to a true 3-2-1. Including the hardware failure that made this a non-negotiable priority.
The Framework
The 3-2-1 rule is the gold standard for backup strategy — simple enough to remember, robust enough to survive almost any failure scenario. It's not a homelab concept; it's the same framework enterprise IT has used for decades. The principle is straightforward: never rely on a single copy, never rely on a single medium, and never rely on a single location.
At least three copies of your data exist at all times. The original plus two backups. If one copy fails — hardware failure, corruption, accidental deletion — you still have two more.
Copies live on at least two different types of storage — a dedicated backup server, a NAS, external drives. A single RAID array is not a backup; it's redundancy against drive failure, not data loss.
At least one copy lives somewhere physically separate. Fire, flood, theft, power surge — any event that hits your home hits all your on-premise copies simultaneously. Offsite is the last line of defense.
Real World
Backup strategy is easy to de-prioritize when everything is working. It took a hardware failure to make it non-negotiable in this lab — and the lesson stuck.
The Intel NUC That Didn't Make It
After three years in continuous production, the original Intel NUC that ran this homelab failed catastrophically. Everything on that node was gone. The VMs, the containers, the configurations — years of tinkering, of incremental improvements, of things finally working the way they should. The kind of loss that forces you to ask: what would it actually take to rebuild this from scratch?
The answer was: Proxmox Backup Server running on its own dedicated bare metal saved the lab. Because PBS was isolated on a separate physical machine — a Beelink NUC — it was completely unaffected by the primary node's failure. The backups were intact, the restore process worked, and the lab was back up faster than rebuilding from scratch would have taken by an order of magnitude.
The lesson is not complicated: be prepared to lose everything, because you eventually will. A small investment in dedicated backup hardware is not optional — it's the difference between an inconvenient afternoon and losing years of work.
Honest Assessment
The current backup posture is 2 copies, 2 media types, both on-premise. That covers hardware failure on any single node. It does not cover a site-level event. Progress, not completion.
Power Protection
A backup strategy protects against data loss. A UPS protects against the event that causes it. Power outages are among the most common causes of data corruption and unplanned downtime in a homelab — and they're entirely preventable with the right hardware and configuration.
The lab runs a UniFi UPS Tower as the primary power protection device, with all critical infrastructure connected on battery-backed outlets. But hardware alone is only half the solution — the other half is NUT (Network UPS Tools), which ensures every host on the network shuts down gracefully before the battery runs out.
An unexpected power cut mid-write is one of the fastest ways to corrupt a filesystem or lose a running VM's state. A UPS provides enough battery runtime to either ride through a short outage or — more importantly — trigger a controlled shutdown of every service before the battery is exhausted.
The UniFi UPS Tower acts as an embedded NUT server. Every host in the lab — the Proxmox node, the PBS Beelink, Home Assistant — runs a NUT client that polls battery runtime. When runtime drops to a configured threshold, each host triggers a graceful shutdown sequence: VMs first, then the hypervisor. No data mid-write. No corruption.
The Proxmox host polls battery runtime every 5 seconds. At 900 seconds (15 minutes) remaining, it begins graceful VM shutdown followed by a host halt. LOWBATT signal from the UPS acts as an emergency backstop if runtime degrades faster than expected. The result: power fails, the lab shuts down cleanly, and comes back up when power is restored — as if nothing happened.
All core infrastructure is on battery-backed outlets: the ASUS NUC (Proxmox VE), the Beelink NUC (PBS), the UNAS-2, the UDM-Pro, and the core switch. The UNAS-2 uses UniFi's native paired shutdown. Home Assistant uses its built-in NUT integration. Everything shuts down in the right order, automatically.
The Missing Piece
The on-premise backup posture is solid. The gap is geographic. A house fire, a flood, or a theft event would affect the Proxmox node, the PBS Beelink, and the UNAS-2 simultaneously — all three copies of the data gone at once. Adding an offsite destination closes that exposure and completes the 3-2-1.
Several options are being evaluated. The priority criteria: strong privacy posture (encrypted at rest and in transit), reasonable cost at homelab scale, and ideally native integration with the existing UniFi / Proxmox stack.
Bottom Line
If you run a homelab and don't have a backup strategy, you have a question of when, not if. Hardware fails. Power surges happen. Files get corrupted. The cost of a dedicated backup node — even something as modest as a Beelink NUC running Proxmox Backup Server — is trivial compared to the cost of rebuilding a lab from scratch.
Three things worth doing, in order of impact:
Get a UPS and configure NUT. Graceful shutdown is non-negotiable. Every host in your lab should know how to shut itself down cleanly before the battery runs out.
Run PBS on dedicated bare metal. It costs the price of a mini NUC. It saved this lab. Keep it isolated from the node it's backing up — that's the entire point.
Add an offsite copy before you need it. On-premise redundancy is not a complete strategy. One copy needs to be somewhere your house isn't.