Key takeaways
- Restore into an isolated environment on a schedule, not just when it breaks
- Record how long a restore actually takes; that number is usually wrong in the plan
- Back up media, redirects and configuration, not only the database
- A database-only backup restores a site with no images and no redirects
- Reapply deletion requests after a restore, or a recovery becomes a privacy incident
- Ask when a backup was last restored, not whether backups exist
Everyone has backups. Rather fewer have restores.
The distinction matters because the failure modes of a backup are mostly invisible until the moment you need it, which is also the moment you can least afford to discover them.
The ways backups quietly fail
- The job has been failing for weeks and the notification went to someone who left
- The database is backed up but the uploaded images are not
- The backup is on the same infrastructure as the thing it protects
- Nobody has the credentials to restore it
- The restore works, but takes eleven hours, and nobody knew that
The last one is the most common and the most damaging, because the plan assumed minutes.
A restore drill is the only real test
Restore into an isolated environment, on a schedule, and check:
- The database restores and the application starts
- Product images and uploaded media are present, not just referenced
- Redirects and URL history survived
- User accounts and permissions are intact
- How long the whole thing actually took
That fifth point should be recorded, because it is the number your recovery plan depends on and it is usually wrong.
Back up more than the database
An ecommerce site is not only its database. A complete recovery usually needs:
- The database
- Uploaded media, which is often much larger
- The redirect and URL history
- Configuration and environment variables
- Any locally stored certificates or keys
A database-only backup produces a site with no images and no redirects, which is not a working site.
Reapply deletions after a restore
This one is easy to miss and has real consequences. If a customer asked you to delete their data last month, and you restore a backup from two months ago, their data is back.
A restore procedure needs a step that reapplies deletion requests. Without it, a routine recovery becomes a data protection incident.
The question to ask
When someone tells you backups are handled, the useful follow-up is not whether they exist. It is:
When was the last time you restored one, and how long did it take?
If the answer is a date and a duration, you are in good hands. If the answer is that backups run nightly, you have learned that nobody has tested them.
Frequently asked
Questions readers ask
How often should I test restoring a backup?
On a schedule, quarterly at minimum, into an isolated environment. The only evidence a backup works is having used it, and the most common surprise is how long the restore actually takes compared with what the plan assumed.
What needs backing up besides the database?
Uploaded media, which is often much larger, plus the redirect and URL history, configuration and environment variables, and any locally stored certificates or keys. A database-only backup produces a site with no images and no redirects.
Can restoring a backup cause a data protection problem?
Yes. If a customer asked you to delete their data last month and you restore a backup from two months ago, their data is back. A restore procedure needs a step that reapplies deletion requests.





