Restore drills

What you get

• An automatic restore test every week
• Every live production database restored from backup
• Each restored table checked against the live one
• A written report for every business
• An email to owners when a report is ready
• The last four drills shown in the dashboard
• Whisk's own team alerted when a drill fails

How it works

Once a week, Whisk restores every live production database from its WAL-G backups to the state it was in an hour earlier. The restored copy is made beside the live database, on the same server, and is deleted after the check. The same run takes a fresh restic backup of the files on each server.

Whisk counts the rows in every table, in the live database and in the restored copy. The check passes when every live table is in the copy with nearly all of its rows, allowing for recent deletes.

The report names no business, app or table, so every customer reads the same one. Nobody is emailed about a drill that passed: the Trust page in the dashboard lists the last four drills with the checks that passed. For day-to-day restores, see backups and restore, and for live platform health, the status page.

Example

A steel fabricator is asked by a large customer to show that its order system's backups are tested. Its operations manager opens the Trust page and sends the latest weekly drill report with the date and the checks that passed.

For your coding agent

whisk restore --at 1h
whisk db schema --database <restored database> --json

The drill needs nothing from the app. On a paid plan, an agent can check one app's restore itself with these commands. More on how Whisk protects data is on the security page.