Infrastructure, cloud & DevOps · Advanced

Backup and recovery: what isn't tested doesn't exist

In one line: A backup isn't a policy; a policy is how much data you may lose and how long you may be down — and proof you met it.

Two numbers

How far back you may lose data, and how long recovery may take. Both are set with the business side and not the infrastructure team, because they're a decision about risk and cost.

What a healthy backup looks like

Automatic, encrypted, in copies in several locations, and outside the account that can be deleted along with the system. A backup accessible with the same permissions as production isn't protection against malicious deletion.

Including what people forget: configuration, secrets, file storage, and dashboards.

A recovery drill

Once a quarter, restore to a separate environment and measure how long it took and what was missing. Something almost always turns up — a permission, an order, a dependency. Better to find it in a drill.

Going deeper

Separate backup from archive: a backup is for fast recovery, an archive is for long retention by obligation. And a soft delete has great value in human error — most data loss isn't an infrastructure disaster but a wrong command from a tired person.