SD cards fail. Updates occasionally go wrong. A power outage at the exact wrong moment can corrupt a filesystem. None of these have to be catastrophic — they only become "start over from scratch" events if you don't have a current backup. Here's a backup strategy that takes about ten minutes to set up and then runs itself.

What a Home Assistant backup actually contains

A full backup includes your configuration files, automations, dashboards, add-ons, and the history/logbook database (optionally). Restoring one puts your instance back exactly as it was at the time the backup was taken — a different physical SD card or a brand-new Raspberry Pi can be brought back to your exact setup from a backup file.

Step 1: Turn on automatic backups

Home Assistant OS and Supervisor-based installs include a built-in backup system. Go to Settings → System → Backups and set up an automatic schedule — daily or weekly, depending on how often your configuration changes. Keep at least 3–4 recent backups rather than just one, so a backup taken right after a mistake doesn't leave you without a clean earlier copy.

Step 2: Get backups off the device

This is the step that actually matters and the one most people skip. A backup stored only on the same SD card it's protecting doesn't help you at all when that SD card fails — the backup dies with it.

  • Network storage integration: Home Assistant supports backing up directly to network locations (e.g. Samba/CIFS shares, or cloud storage via community add-ons) so backups land on a different device automatically.
  • Manual download routine: At minimum, download the latest backup file from Settings → System → Backups to your computer every so often, especially after making significant configuration changes.
  • 3-2-1 principle, scaled down: Ideally keep backups in at least two places — for example, on a NAS on your network and a periodic copy in cloud storage you already use. You don't need anything elaborate; consistency matters more than sophistication.
Rule of thumb: if your only backup lives on the same physical device it's backing up, you don't have a real backup — you have a false sense of security.
Advertisement

Step 3: Back up before every update

Home Assistant's update process typically prompts you to create a backup before applying a core update — don't skip this. The overwhelming majority of updates go fine, but the one time it doesn't is exactly when a fresh backup saves you hours.

Step 4: Know your restore process before you need it

Don't wait for a failure to learn how restoring works. The general process:

  1. Flash a fresh copy of Home Assistant OS onto new or repaired media, following the same install process as a new setup.
  2. During onboarding, choose the "restore from backup" option instead of setting up a new instance.
  3. Upload or select your most recent off-device backup file.
  4. Wait for the restore to complete — the system will reboot with your full previous configuration, automations, and history intact.

If you've never done a test restore, consider doing one on spare hardware (even a cheap secondary SD card) just to confirm your backup file is valid and you're comfortable with the steps. A backup you've never tested is a backup you're only assuming works.

Step 5: Keep a copy of what's not in the automatic backup

A few things are worth noting down separately, outside the Home Assistant backup system entirely:

  • Your Zigbee/Z-Wave coordinator's pairing information — some coordinator firmware stores network keys on the dongle itself, so keep track of which physical USB stick is paired to your setup.
  • Any custom scripts or files stored outside the main configuration directory (rare, but check add-on documentation for anything add-on-specific).
  • Login credentials for third-party services connected via cloud integrations.

A simple schedule that works for most homes

Frequency Action
DailyAutomatic on-device backup (built-in scheduler)
WeeklySync latest backup to network storage or download manually
Before any updateManual backup, confirmed before proceeding

None of this needs to be complicated. The goal is simply that a failed SD card becomes a twenty-minute inconvenience instead of a weekend spent rebuilding automations from memory.