Maintenance
Website Backups That Actually Restore: What to Back Up, Where and How Often
Most sites have backups; far fewer have ever restored one. Here is what a complete website backup contains, how often to take it, where to keep it safe from failures and attackers, and how to test that it really brings the site back.

Almost every website owner we speak to believes their site is backed up. Far fewer have ever restored one of those backups, and that is usually when the problems surface: the backup contains the files but not the database, the database copy is from three months ago, the only copy was on the server that has just failed, or the backup finished every night with an error nobody was told about. A backup is only as good as the last restore that worked.
Backups are a core part of our hosting and server management service, and we treat them as a restore plan rather than a setting. Here is what to back up, how often, where to keep it and how to prove it works.
What a Complete Backup Contains
To rebuild a site from nothing, you need more than a copy of the files:
- The database: pages, products, orders, customers, settings. For most sites this is the most valuable and the most frequently changing part.
- Uploaded files: images, documents, downloads and anything else users or editors have added.
- The code that is actually deployed, including the exact versions of themes, plugins and dependencies. A Git repository is a good start, but only if what is live matches what is committed.
- Configuration: environment files, wp-config.php, web server and PHP settings, scheduled tasks, and any custom firewall rules.
- Things outside the server: an export of the DNS zone, webhook settings with payment and email providers, and a note of which third-party services the site depends on.
Leave out what can be regenerated, such as caches, compiled assets and dependency folders. They make backups slower and larger without making restores easier.
How Often: Match It to What You Can Afford to Lose
The right frequency comes from one question: if the server disappeared now, how much recent work could you live without? For a brochure site edited once a week, a nightly backup is plenty. For an online store taking orders all day, losing even a few hours means orders that exist at the payment provider but not in your system.
- Low change rate: nightly database and file backups.
- Regular editing or bookings: database backups several times a day.
- Constant transactions: point-in-time recovery, which continuously archives the database's transaction log (binary logs in MySQL, WAL in PostgreSQL) so you can restore to a chosen minute. Managed databases on AWS, Azure and Google Cloud offer this as a setting.
Keep a range of ages, not just the latest copy. A common scheme is daily backups for two weeks, weekly ones for two months and monthly ones for a year. If a site is hacked and nobody notices for a month, the only clean backup is an old one; our hacked website cleanup work regularly depends on exactly that.
Where to Keep the Copies
The old 3-2-1 rule still holds: three copies of your data, on two different kinds of storage, with one kept somewhere else entirely. For a website that means:
- Off the server. A backup on the same machine is useless if the disk fails or the server is deleted.
- Outside the hosting account. Host snapshots are convenient for quick rollbacks, but they share the same account, the same billing and the same login. If the account is suspended or an attacker gets in, they go too.
- Protected from deletion. Object storage with versioning or object lock, such as S3 or Backblaze B2, can keep backups for a fixed period even if someone with the credentials tries to remove them. Give the server permission to write new backups but not to delete old ones.
- Encrypted, with the encryption key stored separately from the backups and known to more than one person.
Taking Backups That Are Consistent
Copying the raw database files of a running server can produce a backup that will not open, because the files changed during the copy. Use the database's own tools instead. For MySQL and MariaDB with InnoDB tables, mysqldump with the single-transaction option gives a consistent copy without locking the site; for PostgreSQL, pg_dump does the same. Larger databases are better served by physical backup tools such as Percona XtraBackup, Mariabackup or pgBackRest, which are faster to restore.
For files, tools such as restic and BorgBackup encrypt, compress and deduplicate, so frequent backups of a large media library only store what changed. Our database development team sets up the heavier database cases.
Test the Restore, on a Schedule
This is the step that makes the rest worth doing. We recommend:
- Automatic checks after every backup: did the job finish, is the file a sensible size, can the archive be listed? Alert a person if not, rather than writing to a log nobody reads.
- Automated restore tests: load last night's database dump into a temporary database and check something meaningful, such as the date of the most recent order or the number of published pages.
- A full restore rehearsal every quarter or so: rebuild the site on a separate server from backups alone, time how long it takes, and note every step you had to look up.
- A written runbook from that rehearsal, so the restore does not depend on one person's memory.
Plugins, Host Backups and Managed Databases
Backup plugins for WordPress are fine for small sites, but they run inside the site they are protecting, can time out on large media libraries, and often store archives on the same server unless configured otherwise. Host backups are a useful extra layer, not a replacement for an independent copy. Managed cloud databases take automated backups, but retention is limited and, by default, automated backups can be removed along with the database instance, so take a final snapshot or keep an independent export.
Backups belong on the same list as updates and monitoring; our website maintenance checklist shows where they fit, and our website maintenance and support plans include tested backups. Before any server move, a fresh, verified backup is step one, as our guide to moving hosting and DNS without downtime explains.
If you are not sure your backups would bring your site back, ask us for a restore test. We will rebuild your site from its current backups on a separate server and tell you exactly what was missing.
Frequently Asked Questions
Are my web host’s backups enough on their own?
They are a useful layer but should not be the only one. Host backups sit in the same account as the site, so an account suspension, billing problem, compromised login or provider outage can take the backups with it. Retention is often short too. Keep at least one independent copy with a different provider that you control.
How long should website backups be kept?
Long enough to cover problems that are noticed late. Hacks, accidental deletions and data corruption are often discovered weeks after they happen, so a mix of daily backups for a couple of weeks, weekly ones for a couple of months and monthly ones for a year suits most business sites. Legal or contractual rules on data retention may also apply.
How long does it take to restore a website from backup?
For a small site with a tested process, often under an hour. Large databases, big media libraries and undocumented server configuration can stretch it to many hours or days. The only reliable way to know your restore time is to rehearse it on a separate server and time each step, then fix whatever slowed it down.
Do backups need to follow GDPR rules?
Yes. Backups containing customer data are personal data and need the same protection as the live system: encryption, access control and storage in an appropriate location. Your retention policy should cover backups, and if someone asks for their data to be erased, you need a documented approach for copies that remain in older backups until they expire.
What is the difference between a snapshot and a backup?
A snapshot is a point-in-time image of a whole disk or server, usually stored by the same provider. It is quick to take and restore, which makes it ideal before an update. A backup is an independent copy of the data, ideally stored elsewhere, that can rebuild the site on any server. A good plan uses both.








