Website down? How to know before your customers do

Nobody watches their own website. You built it, it works, you get on with the business. So when it goes down at eleven on a Friday night, because the hosting card expired or the disk filled up or an automatic update failed, nobody knows until Monday, or until a customer says “your site is not working” in a tone that suggests they tried more than once.

What uptime monitoring is

A service outside your hosting that requests your site every minute from several places in the world, and tells a person the moment it fails. Free tiers exist. Setting one up takes ten minutes. It is the cheapest insurance a website can have and most business sites do not have it.

The five checks worth running

1. The homepage returns normally. Status 200, within a few seconds. The basic check, from at least two locations so a local network problem does not cause a false alarm.
2. A keyword is on the page. Not just “the server answered”, but “the page contains the company name”. A site showing a database error or a blank white page still returns a response; this check catches that.
3. The contact form works. Once a day, a test submission, and a check that it arrived. Forms fail silently more than any other part of a site.
4. The SSL certificate is valid, with a warning two weeks before it expires. Expired certificates take sites down as surely as a crashed server, and they are entirely predictable.
5. The domain is not expiring. Registrars send emails to whichever address was used years ago. A domain lapse is the most complete outage there is, and the recovery can take days.

For a shop, add a check that a product page and the cart load. For a site with logins, the login page.

Who gets told, and how

The alert goes to someone who can act, by a channel they will see at eleven at night: a phone notification, not an email to the office inbox. If the answer is “there is no such person”, that is the gap a care plan fills, and it is the reason we monitor every site we look after ourselves.

The first ten minutes when the alert arrives

1. Confirm it. Open the site from your phone on mobile data. Monitoring services occasionally alert on their own network problems.
2. Check the host’s status page. If the host is down, there is nothing to fix on your side; note the time and wait.
3. Check the obvious. Hosting account paid? Domain current? Certificate valid? These cause a large share of outages and take minutes to fix.
4. Check what changed. An update, a new plugin, a change someone made that afternoon. Undo it.
5. Restore if needed. If the site is broken rather than the server, and the cause is not obvious, restoring last night’s backup gets it back up while you investigate. This is why backups and monitoring belong together.
6. Write down what happened, with times, once it is back. The pattern across a few incidents is what tells you whether the host, the software or the process is the problem.

What “99.9 percent uptime” means

It means about forty-three minutes of downtime a month, which a good host achieves. Ninety-nine percent, which sounds similar, is over seven hours a month. If your host’s promise is the second number, downtime is part of the deal.

The point

Downtime is not really the problem. Unnoticed downtime is. A site that is down for ten minutes and fixed is a non-event. A site that is down for two days and discovered by a customer is a story they tell. Ten minutes of setup moves you from the second to the first.