Why your server goes down and nobody notices until a client calls

There is a specific kind of phone call every business owner learns to dread: a customer saying "your website is not working". It is dreadful not because of the outage, but because of what it reveals — you found out last.

Uptime checks are not monitoring

Most hosting providers offer a ping check: every five minutes something asks your server if it is alive. If it answers, green light. The problem is that a server can answer perfectly while the site behind it is completely broken.

A full disk, an expired certificate, a database refusing connections, a PHP process pool exhausted — in all of those cases the server answers the ping and your monitoring stays green while customers see an error.

What you should actually be watching

Real monitoring checks the thing the customer cares about, not the machine. It loads a real page and verifies the content is there. It tries the checkout. It watches disk space, memory and certificate expiry before they become outages.

# What a useful check looks like
curl -sf https://site.com/health | grep -q "ok" || alert
 
# What most people have instead
ping site.com

The part nobody sets up

Even good monitoring is useless if the alert goes to an inbox nobody reads at 3am. Alerts need a destination someone actually looks at, and a rule about what deserves waking a person up.

An alert that fires every day for something nobody fixes is not an alert. It is noise, and it trains your team to ignore the one that matters.

Where to start

Pick the three things that would cost you money if they broke — the site loading, the form sending, the payment going through. Check those specifically, every minute, from outside your own network. Send failures somewhere with a notification sound. That is 80% of the value, and it takes an afternoon.