← All guides

Your backup isn't a rollback plan (until you've tested a restore)

Published 19 October 2026 · 7 min read

If you look after WordPress sites, "we have backups" is usually the answer to "what if an update breaks something?" It's a reasonable answer. It's also an untested one, unless you've actually restored a backup and checked the result.

This isn't an attack on backup plugins. The ones below were all handled by developers who tried to help. It's that a restore is a different operation from a backup, it runs under different conditions, and it tends to be needed on the worst day.

What public support threads show

We read public WordPress.org support threads for a research project. These three show different ways a restore, or the recovery around it, goes wrong. We don't know how common any of them is.

A restore that said it failed at the last step. A user of one backup plugin reported "restore failed" even though the progress bar reached 100%, and said they had tried five times. The developer found that the server returned a 503 error during the final clean-up of roughly 24,000 temporary files, put down to server overload. The restore itself had in fact completed, but the user still had to reactivate plugins one at a time. Even a restore that technically works can look like a failure, and leave you with a half-finished site.

A restore that left the site unreachable. A user of another backup plugin restored a backup after an update and then couldn't reach the site or its admin area. The developer said it didn't look related to the new plugin version, and suggested testing the backup on a fresh staging install. The thread ends unresolved. Whatever the cause, "restore the backup" didn't get the site back.

An update that failed halfway, with no check. After a WooCommerce update in June 2024, one store owner found that many of the plugin's folders were missing on the live site, so WooCommerce fatally errored. They asked why the update didn't verify itself and roll back. WooCommerce's support team replied that WooCommerce has no built-in self-verification or automatic rollback. The fix that worked for the store owner wasn't a full restore: they went back to the previous WooCommerce version, 9.0.1, and the site came back.

What to take from it

How to test a restore

  1. Restore to a staging copy, not the live site. Create a fresh staging site on the same host if you can, because that's where a real restore would run.
  2. Time it. Note how long it takes and whether it hits a time or memory limit. A restore that takes 40 minutes is a different plan from one that takes 4.
  3. Check what matters afterwards: the home page, a product page, logging in to the admin, and the checkout if there's a shop. Our checkout testing checklist covers the last one.
  4. Check the plugins are active after the restore. In the first thread above, plugins failed to reactivate on their own.
  5. Write down the steps that worked. On the day you need it, you'll want a short list, not a memory.
  6. Repeat it after big changes, and at least a couple of times a year.

When to roll back one plugin instead

If an update has broken the site and you know which plugin did it, rolling back just that plugin is usually faster than restoring the whole site, and it doesn't throw away orders or content added since the backup. You need the previous version's files ready (the plugin's page on WordPress.org keeps older versions under "Advanced View"). A full restore is the right tool when the damage is wider, or you don't know the cause.

This is the kind of problem PatchSure is being built to avoid. The plan: take a snapshot before each update, run the update, check the site, including the checkout on WooCommerce shops, and roll back automatically if something has broken, then tell you in plain English what happened. PatchSure is still in development; join the list to hear when the private beta opens.

Join the early-access list

Summary

Sources