Homelab. Part 3 : living with Proxmox.

A homelab is not a professional setup. It never will be, and that's not the point. It's a permanent negotiation between what you can spend, what you're willing to maintain, what services you actually need, and what you believe in technically. Every decision is a compromise. The honest ones are the ones where you know which compromise you're making.

Proxmox on a single Odroid H2+ is exactly that kind of decision. No HA, no live migration, one machine does everything. The constraints are real and accepted upfront. What you get in return is simplicity — one thing to monitor, one thing to update, one thing to understand deeply. On a limited budget and limited time, that trade is worth it.

The first practical question Proxmox asks is VM versus LXC. The answer depends on the workload. LXC for services that are self-contained and have no kernel requirements — low overhead, fast to operate, easy to snapshot. VMs when the workload has specific needs or when proper isolation justifies the cost. OpenMediaVault runs in a VM because direct HDD passthrough is cleaner that way. K3S runs in a VM because Kubernetes in a shared kernel environment creates friction that isn't worth the savings. Everything else gets evaluated case by case. There's no universal rule, and pretending otherwise wastes time.

Resource management on this hardware is about knowing the limits and working within them. The H2+ handles the daily load without issue. Under contention — simultaneous backups, heavy updates, a K3S node rebooting — every LXC on the host feels it. The answer isn't more hardware. It's scheduling and realistic expectations : backups at 3am, updates one service at a time, intensive operations given a quiet window. A homelab that fights its own constraints is one that becomes exhausting to run.

Updates are manual and deliberate. On a node where everything is interdependent, knowing what changed and when has real operational value. Automating routine updates is the next step — as is extending fail2ban to the Proxmox host itself, currently only running on the reverse proxy. The gap between what's done and what should be done is part of the homelab reality too. You close it progressively, on your own schedule.

Monitoring is Beszel — self-hosted, lightweight, alert-driven. The principle is simple : if it's quiet, it's running. Alerts surface when something needs attention. No dashboard-checking ritual, no daily manual verification. A setup that demands constant supervision stops being sustainable. The goal is something that runs in the background of your life, not something that competes with it.

The single point of failure is real. Weekly backups, one retention point. Some data loss is possible in a worst case. Most services here are either stateless or low enough stakes that rebuilding is a viable recovery path. The backup strategy is calibrated to the actual risk, not to an ideal that doesn't match the resources available.

That's the homelab compromise. You know what you have, you know what it costs you, and you make it work within those boundaries. Not because it's perfect — because it's yours, and it runs.

So be it, then.