Why websites go down (and how to catch it fast)

A website almost never goes down for a mysterious, unknowable reason. Outages cluster around a short list of usual suspects, and once you have seen each one a few times, you start to recognise the cause from the symptom.

Here are the failures that take sites offline most often, roughly in the order you will meet them.

Expired SSL certificates

The classic 2 a.m. outage. The site was fine yesterday, nothing changed, and now every browser throws up a big red warning. Certificates expire on a fixed date, and if the auto-renewal quietly failed three weeks ago, you find out the moment it lapses. It is one of the most common outages there is, and one of the most preventable.

A bad deploy

You shipped a change and something in it was wrong: a syntax error, a missing environment variable, a migration that locked the database. The site was up at 14:32 and down at 14:33, and the timing points straight at the release. The upside is that this is also one of the fastest outages to fix, because you know exactly what changed.

DNS problems

DNS is the phone book that turns your domain into a server address. If a record gets edited wrong, a domain lapses, or a DNS provider has an incident, visitors simply cannot find you even though the server is running perfectly. These are sneaky, because from inside the server everything looks healthy.

Traffic spikes

You got featured somewhere, a campaign went out, or a bot decided to hammer you. Traffic climbs past what the server can handle, response times stretch, and eventually requests start timing out. The painful irony is that it happens at the exact moment the most people are trying to reach you.

The host itself

Sometimes it is not you at all. The hosting provider, the data centre, or an upstream network has a bad day and takes your site down with it. There is not much you can do in the moment except know about it, tell your users, and have a record for when you ask for that SLA credit.

Catching it fast is the whole game

Notice what these have in common: from the inside, the server often looks fine. The certificate is "installed", the process is "running", DNS is "configured". The only reliable way to know your site is actually reachable is to check it the way a visitor does, from outside, on a schedule.

That is the entire job of an uptime monitor. It loads your real URL every minute or few minutes from somewhere out on the internet, and the moment a check fails and a follow-up check confirms it, it tells you. You stop being the last person to find out, and the clock on every one of these outages starts when it begins, not whenever someone happens to notice.