The Proxmox backup conversation has matured. A year ago it was "how do I install Proxmox Backup Server?" Now PBS is running and the interesting question is the one tutorials skip: what actually deserves an offsite copy, and how long do you keep it? Backing up everything, forever, to a second location is simple and wasteful. Backing up nothing important offsite is a disaster waiting for a fire. The answer is selective, tiered planning.
Local PBS is not offsite
First, the distinction that trips people up: a Proxmox Backup Server in the same room as the host it backs up protects you from a bad upgrade or a fat-fingered delete — not from theft, fire, flood, or ransomware that reaches both. The classic 3-2-1 rule still holds: at least one copy offsite, on different media. Local PBS is your fast-restore tier; offsite is your disaster tier. You need both, but not for everything.
Tier your VMs by recoverability, not by size
Sort every VM/CT into three buckets by how you'd recover it:
- Irreplaceable data (photos, documents, app databases) → local PBS and offsite. This is the data that's gone forever if both copies die.
- Reproducible-but-painful (configured app servers) → local PBS, plus the config in a repo so a from-scratch rebuild can recreate them. Offsite optional.
- Trivially reproducible (a stateless container you can redeploy from a template in minutes) → maybe no backup at all. Don't pay to back up what you can recreate.
Most homelabs discover the offsite set is small — which is exactly the point. Offsite the irreplaceable, reproduce the rest.
Plan retention deliberately
Keep-everything-forever fills disks and budgets. A workable scheme:
- Frequent recent backups (daily for a week), thinning to weekly and monthly.
- PBS's prune/retention settings encode this; set them so old backups age out automatically.
- Match retention to how far back a problem might go unnoticed — if a corruption could hide for a month, keep a month.
Verify, don't assume
A backup you've never restored is a guess. PBS can verify backup integrity on a schedule — enable it. Then, periodically, do a real restore of a tiered VM to a scratch target. The same restore-drill discipline that applies to Docker volumes applies to whole VMs, and it's what separates "I have backups" from "I can recover."
Planning checklist
- Every VM/CT tiered: irreplaceable / reproducible-painful / trivial.
- Irreplaceable data has a real offsite copy (3-2-1).
- Reproducible servers have config-as-code, not just images.
- Retention/prune configured so old backups age out.
- PBS verification on; a real restore rehearsed periodically.
Takeaway
Past the install, Proxmox backup is a planning problem, not a tooling one. Tier by recoverability, send only the irreplaceable offsite, let config-as-code handle the reproducible, set retention so it self-prunes, and prove restores work. That's the difference between a backup setup and a backup plan.
From across the StoicSoft network
Hand-curated reads on the same topic from sister sites in the StoicSoft family.
