VPS Backup and Restore — 3-2-1 on Root Servers

Separate state from rebuildable, keep three copies on two media, and run the one restore test that decides whether any of it works.

长文正文为英文。页头、目录与结账已翻译。

Decide what is state

On most servers the irreplaceable set is small: databases, uploaded files, TLS keys, the configs that took a week to get right. The OS is reproducible — restore speed comes from automating the rebuild, and a short Ansible or cloud-init script you actually update beats a golden image you forgot existed. Draw that line first: what is state gets copies, what is rebuildable gets a script.

The pattern

• 3 copies: live data, an on-box snapshot for fast rollback, and an off-box backup that survives the box.
• 2 media: local NVMe for second-level recovery, plus a copy on separate storage — different account, provider, or cold.
• 1 tested restore: a scheduled run that restores into a scratch directory and diffs. Untested backup is a hope, not a backup.

On NulNet

1. Snapshots ride VPS and storage plans as a checkout add-on — right for pre-change rollback, too shallow to call a strategy alone.
2. For the off-box leg, a Depot 1 box (1 TB NVMe, $30 list, $24 on the annual term) as a Borg or Restic target keeps copies inside the same crypto-only, no-KYC rails.
3. Encrypt at the source — Borg and Restic do this natively — so the target never holds plaintext and a compromised target is not a breach.