# Restore tests now run on a schedule

> Stonebridge 3.5 restores a random sample of every backup set each week and tells you when one fails.

- URL: https://nginx-example.notra.site/blog/scheduled-restore-tests
- Published: 2026-10-01
- Authors: Ines Hartmann, Paul Mbeki
- Tags: product

A backup you have never restored is a guess. Most teams know this and still test restores once a year, usually after something has gone wrong. Stonebridge 3.5 does it for you.

## What a restore test does

Once a week, at a time you pick, Stonebridge picks a random snapshot from each backup set and restores part of it into a scratch directory on the server:

- 200 random files, or 2 GB, whichever comes first
- the full catalog for database backups, plus one table restored and checksummed
- every file’s hash compared against the hash recorded at backup time

The scratch directory is deleted afterwards. Nothing touches your live data.

## Turn it on

1. Open the backup set

   Go to **Backup sets** and pick the set you want to test.

2. Add a schedule

   Under **Restore tests**, click **Add schedule** and pick a day and an hour. Off-peak hours keep the extra read load away from your users.

3. Pick who gets told

   Failures go to the alert channels you already use for backup jobs. Passing tests only show up in the report.

> **Note**
>
> Restore tests read from the same storage as a real restore, so they count against egress on cloud targets. On S3 with 1 TB of backups, expect about 2 GB of egress per set per week.

## What we found in the beta

Across 85 beta installs, restore tests caught 11 broken backup sets in the first month. Nine were expired credentials on an offsite target, where the backup job itself still reported success because the local copy was fine. Two were disks with silent bit rot, which the checksum caught.
