From our work
There is a backup – but has the restore ever been proven?
Recovery as code: proven first, then moved
Starting point
A central system on which every build depended had to move. There was a backup. Nobody had checked whether it would yield a complete system again in an emergency.
What we did
We built the entire migration with Ansible, including the restore. Recovery is therefore not a manual that someone works through in an emergency, but a procedure that runs the same way every time. And because it runs the same way every time, we ran it on the new environment and checked the result for completeness before switching over.
Result
The restore is proven, not assumed. The move only happened after that.
What remains
A repeatable restore procedure in code that can be proven again at any time.
Related service: Take over established software All case stories