Backups that restore: Restic/Kopia and the 3-2-1 rule

Every homelab has backups. Fewer have restores. The difference shows up exactly once, on the worst day. Restic and Kopia are the modern answer for Docker households: encrypted, deduplicated, snapshot-based backups that can target a local disk, a second NAS, or object storage — with integrity checks built in. This guide builds the strategy (3-2-1), picks the tool, and finishes with the step everyone skips: proving a restore works.

Published Nov 18, 2026

The strategy before the software

3-2-1: three copies of what matters, on two different media/systems, one of them off-site. For a Docker household, 'what matters' is smaller than you think: every container's config/data volumes (the folders your Compose files mount), your Compose files themselves, Home Assistant's configuration, and the irreplaceable payloads (photos, documents) — which usually have their own bulk-backup path. Media libraries you can re-download are not backup priorities; the Immich database that knows your albums is. Write the inventory down. The tool then just executes it.

Inventory what matters→Snapshot locally→Replicate off-site→Verify integrity→Test a restore

Restic vs Kopia: pick your fighter

The reference setup

  1. Create the repository on your second location first (an external drive, a second machine, or a Backblaze B2/S3 bucket): restic init (or Kopia's connect) with a strong repository password — stored in your password manager, because an encrypted backup whose password you lost is a sculpture.
  2. Snapshot the config tree: your entire Docker folder (configs, Compose files, .env files) nightly. It is small — deduplication makes nightly snapshots of config data cost almost nothing.
  3. Retention: keep 7 daily, 4 weekly, 6 monthly snapshots. Both tools prune by policy; set it once and let arithmetic be the janitor.
  4. Add the bulk payload (photos/documents) as a separate, less frequent job if it does not already have one — different data, different cadence, same repository family.
  5. Schedule integrity checks (restic check / Kopia verify) monthly. Backups rot silently; checks are how you find out in peacetime.

Databases: snapshot-aware beats lucky. For the handful of containers with real databases (Immich, Paperless), either stop them briefly during the snapshot window or use their built-in backup/dump first and back up the dump. Files copied mid-write usually work — 'usually' is doing heavy lifting in that sentence.

The restore test (the entire point)

  1. Pick a sacrificial service (Uptime Kuma is perfect). Restore its config from last night's snapshot into a scratch folder — not over the live one.
  2. Diff it against live, then actually run it: spin the container against the restored config on a spare port and confirm it behaves (history present, settings intact).
  3. Time it. Write the time down. That number — your real recovery time for one service — is the first honest backup metric your homelab has ever had.
  4. Repeat quarterly, rotating services. The households that do this are the ones whose 'disaster' stories end with 'and then I restored.'

Keep our Duplicati guide for simple file jobs if you like it — but let Restic or Kopia own the stack's configs. And if Home Assistant is the patient, its own backup system pairs with this: our HA backups guide covers the in-app side.

Gear for this build

Prices are editorial and change often; buttons show live prices. Some links may earn a commission at no cost to you — see the disclosure page.

Official documentation and references

Frequently asked questions

Restic or Kopia for a beginner?

Kopia, if a UI makes you more likely to set policies and check status; Restic, if you are comfortable with a small script and cron. Both produce encrypted, deduplicated snapshots to the same kinds of destinations — the restore confidence is equivalent.

Where should the off-site copy live?

Object storage (Backblaze B2 or S3) is the common answer for configs — the data is small, so it costs very little — or a drive at a family member's house synced periodically. The rule is 'different building, different failure domain,' not a specific vendor.

How often should I back up container configs?

Nightly is the sweet spot: configs change when you tinker, deduplication makes nightly snapshots nearly free, and a day of re-tinkering is the most anyone wants to redo.

Related guides