Free 20-point conversion & SEO audit for export stores — Free website audit — claim yours

hello@woo.one Mon–Fri · 9:00–18:00 (UTC+8)

Backups Are Not Recovery: Testing What You Have

Most stores have backups and no recovery. RPO and RTO, the four ways backups fail, the quarterly restore test, and the one-page document written before you need it.

Almost every store has backups. Very few have a recovery. The difference shows up on the worst day, when someone needs to restore a database and discovers the archive is incomplete, the files are from a different date, or the restore takes six hours instead of twenty minutes.

What a backup must contain

Three parts, and they must be consistent with each other:

  • Database — products, orders, customers, settings, plugin configuration
  • Files — themes, plugins, uploads, and any custom code
  • Configuration — wp-config.php, server config, cron entries, DNS records

The third part is the one usually missing. A restore that brings back a perfect database into a server with no cron jobs, no SSL configuration and different PHP settings is not a restore.

Consistency matters more than frequency. A database from 03:00 and files from 03:00 are useful. A database from 03:00 and files from last Tuesday produce a store where product images do not match products.

Frequency, decided by what you can lose

Two numbers define the policy:

RPO — how much data can you lose? A store taking 200 orders a day cannot afford a daily backup; that is 200 orders to reconstruct. Hourly for the database is the practical floor for an active shop.

RTO — how long can you be offline? A restore that takes four hours is four hours of lost sales, plus the support load.

A workable policy for a mid-sized store:

  • Database: hourly, retained 48 hours
  • Files: daily, retained 14 days
  • Full snapshot: weekly, retained 8 weeks
  • Offsite copy: everything, in a different provider than the host

Where backups fail

Stored on the same server. A disk failure or a compromised host takes the backups with it. If the only copy is in /backup on the same machine, you have a snapshot, not a backup.

Not encrypted, or publicly accessible. Backup files sitting in a web-accessible directory are a routine source of database leaks — they contain customer data in plain text.

Never tested. This is the one that matters. A backup is a hypothesis until you restore it.

Excluded tables. Some plugins and hosts exclude large tables by default. A restore missing the order meta table looks complete and is not.

Silent failure. The job runs, fails, and nothing reports it. Two months later nobody knows the last good backup was in August.

The restore test, done properly

Once a quarter, restore to a separate environment and check:

  1. The site loads, and the homepage renders
  2. A product page shows the correct price and stock
  3. An order from the backup date exists in the admin
  4. Images load — check three products and one category
  5. Checkout works with a test payment
  6. Cron jobs are scheduled
  7. SSL works, and no mixed-content warnings appear

Record how long the restore took. That is your real RTO, and it is usually longer than assumed — especially if the restore requires manual steps documented in someone’s memory.

What to write down

A one-page recovery document, stored outside the server, containing:

  • Where backups live, and how to access them
  • Which credentials are needed, and where they are kept
  • The exact restore steps, in order
  • What must be reconfigured after a restore: cron, DNS, email, SSL, cache
  • Who to contact at the host, with the account number

The last item sounds bureaucratic until the day the host’s status page says nothing and the support queue is four hours long.

Recovery is not only disasters

Most restores are small and specific:

Situation What you need
A bad plugin update broke layout Files only, or a single plugin folder rollback
Bulk price edit went wrong Targeted database restore, or a product-level export from before
A theme file edit caused a fatal error File-level restore of one directory
Customer deleted an order by accident Row-level recovery, which full restores make painful

Full-site restores are the wrong tool for the first three. Having file-level version history, or keeping the last few plugin versions, makes these ten-minute problems instead of hour-long ones.

The fourth case is the reason some teams keep short-interval database dumps: restoring yesterday’s full database to recover one order loses a day of newer orders. A staging copy of the recent dump lets you extract the single row and insert it.

The minimum viable setup

  1. Automated database backups, hourly, offsite, encrypted
  2. Daily file backups, offsite
  3. Retention that matches your RPO, and a documented policy
  4. Monitoring that alerts when a backup fails or is missing
  5. Quarterly restore test, timed and recorded
  6. One-page recovery document, stored outside the server

That is roughly a day of setup and about an hour a quarter afterwards. It is also the cheapest insurance in this entire domain — cheaper than a performance project, cheaper than a migration, and the only one that matters when something goes irreversibly wrong.

If you run a store and cannot answer “when was the last verified restore, and how long did it take”, that is the gap worth closing first. It sits alongside the operational work we cover under site maintenance, and it is always the first thing we check on a store we are taking over.

S

Infrastructure and security engineer at woo.one, based in Foshan, China. Works on WordPress, WooCommerce and OpenCart systems, self-hosted mail, and the operational side of running stores for exporters.

More about the team

No pressure · No sales script

Ready for a store that sells while you sleep?

Send us your current site or product list. You will receive a fixed-scope proposal, a 20-point audit and a realistic timeline within two business days.

Reply within 1 business day NDA on request 100% code ownership
Chat with us
WhatsApp Get a quote