Restore one to a separate location, time it, and list what is missing. An untested backup is an assumption rather than a protection.

The restore is the only test

A backup running nightly with a green tick is evidence that something ran, not that anything usable was produced.

The failures are specific and common: an incomplete database, a missing uploads folder, a file too large to have been included, or an archive that will not open.

None of those is visible from the backup log, and all of them are obvious within ten minutes of attempting a restore.

So the annual job is not checking that backups exist; it is performing one restore, properly, once.

The drill, step by step

The first is deliberate. Restoring last night's backup tests the most recent one; restoring one from a month ago tests whether your retention actually goes back that far, which is what you need after a problem you did not notice immediately.

Restore somewhere separate

Which is the safety requirement.

Never restore over the live site to test a backup, since a bad restore then becomes an outage.

Most hosting offers a staging area, and where it does not, a subdomain or a local environment works.

The point is that the test can fail completely without anybody noticing except you.

Ask your host how to do this if you do not know, which is a reasonable support question they answer routinely.

If they cannot answer it, that is itself the finding.

A worked example

A business restored a backup from five weeks earlier to a staging area.

The site came up and looked correct until they opened a product page and found no images.

The uploads directory had been excluded from the backup configuration since a hosting migration two years earlier.

Every backup they held contained a working site with no photographs, which for a shop is the same as no backup at all.

The fix took the host twenty minutes once it was reported.

Two years of backups had been giving them a green tick and no photographs.

Time it, because the time is the plan

Which is the output people overlook.

Knowing a restore takes forty minutes changes what you say to a customer during an outage.

Knowing it takes six hours because the files come from cold storage changes what you promise and possibly what you pay for.

Write the figure down beside the procedure, so that whoever is doing this under pressure has a number rather than a hope.

Most businesses have never been able to answer how long a full restore would take.

After one drill, they can.

Write the procedure while you do it

Which costs nothing extra and is most of the value.

As you work through the restore, write down each step, where you logged in, and what you clicked.

Half a page is enough, and it means the next restore can be done by somebody who has not done one, at a moment when nobody is thinking clearly.

Keep it with the domain list, the file set, and the supplier list, since those are all the same document in spirit.

A procedure written calmly in December is worth a great deal at nine on a Tuesday morning in March.

Check what is not in the backup at all

Which is the wider question.

The website is usually backed up and the email frequently is not.

Neither is the local listing, the accounting system, the customer records in a spreadsheet on somebody's laptop, or the photographs on a phone.

List what would be genuinely painful to lose and check that each is backed up somewhere, by somebody, on a schedule.

For most small businesses that exercise finds two or three things nobody has ever considered.

The customer list is the most common answer.

Keep one copy somewhere else

Which is the rule that covers the worst case.

A backup stored on the same server as the site protects against a mistake and not against losing the server or the account.

One copy somewhere unrelated, updated at least monthly, is what covers the account being suspended, a payment failing, or a supplier disappearing.

Downloading a monthly copy to a drive you control is unglamorous and sufficient.

Do it now while you are already logged in.

Know who would actually do it

Which is the question the drill answers incidentally.

If the site went down on a Tuesday morning, name the person who would perform the restore, and check they know that is their job.

For most small businesses the honest answer is the host's support team, in which case find out their response time, their hours, and whether that changes at a weekend.

Where the answer is a supplier, confirm they still hold the access they would need, which is the same question the permissions review asked.

An untested backup and an unnamed restorer are the same problem in two forms, and the drill exposes both at once.

The counter-case

Some sites lose very little.

A five-page brochure site can be rebuilt in a day from the text and the photographs, so an elaborate backup regime is more than the situation warrants.

Managed hosting also handles this competently for many businesses, and the restore is a support request rather than a procedure.

And a drill takes an hour that a very small business may reasonably decide to spend on something else.

Even then, ask your host what a restore involves and how long it takes, which costs one email.

Their answer is worth having written down whether or not you ever run a drill yourself.

Restore one backup from a month ago to a separate location, time it, list what is missing, and write the procedure down.

The hour

  1. Pick a backup from a month ago.
  2. Restore it to staging.
  3. Time it.
  4. Click around properly.
  5. List what is missing.
  6. Write the procedure down.
  7. Download one copy elsewhere.

Step four is where the failures appear, since a restored site that loads correctly can still be missing every photograph, and only opening pages will show it.

A prior version of this is at backups you have never restored.


Frequently asked questions

Why is a green tick not enough?

Because it shows something ran, not that anything usable was produced. An incomplete database or a missing uploads folder is invisible in the log.

Which backup should I test?

One from about a month ago, which tests whether your retention actually goes back that far. That is what you need after a problem you did not notice immediately.

Where should I restore it?

Somewhere separate: a staging area, a subdomain, or a local environment. Never over the live site, since a bad restore then becomes an outage.

Why time it?

Because the time is the plan. Knowing a restore takes forty minutes rather than six hours changes what you tell a customer during an outage.

What is usually missing from backups?

Email, the local listing, customer records in a spreadsheet on a laptop, and photographs on a phone. The customer list is the most common answer.

Where should a copy live?

Somewhere unrelated to the server, updated at least monthly, which covers the account being suspended or a supplier disappearing.

West Coast Media Solutions Inc. provides web design, web development, hosting, digital marketing, and business consulting to organisations across Canada, drawing on more than twenty-five years in the field.

Backups running with a green tick?

Restore one to a staging area and open a few pages. That is the only test that means anything.

Start a Conversation