Uptime monitoring is the practice of regularly checking that a website, an API or another online service is available, and of alerting someone the moment it is not. It answers a simple question: is my site up right now? And it answers that question from outside, the same way your visitors see the site.

Why it matters

Every website goes down sometimes. Servers crash, certificates expire, domains are not renewed, deploys go wrong and hosting providers have outages. The real problem is not that something breaks, but how long it stays broken before anyone notices.

Without monitoring, you usually learn about an outage from a customer, a colleague or a drop in sales, often hours later. With monitoring, you learn about it within minutes, and you can start fixing it right away.

How it works

1. A monitoring service sends a request to your URL at a regular interval, for example every minute.
2. It checks the response: did the server answer, how long did it take, and what status code did it return?
3. If the check fails, the service usually confirms the failure, often from a second location, to rule out a network glitch.
4. When the failure is confirmed, it sends an alert by email, SMS, a chat app or another channel.
5. When the site recovers, it reports that too and records how long the outage lasted.

What to check

• Availability – the server responds at all.
• HTTP status code – the response is 200 OK, not 404 Not Found or 500 Internal Server Error.
• Response time – the page loads within a reasonable time. A very slow site is almost as bad as one that is down.
• Content – the page contains the text it should. This catches pages that respond with status 200 but show an error message or an empty template.
• SSL/TLS certificate – the certificate is valid and not about to expire.
• Domain – the domain registration is not about to expire.

Uptime percentage

Availability is often expressed as a percentage of time. The numbers look similar, but the difference in downtime is large:

Uptime	Downtime per year (approx.)	Downtime per month (approx.)
99%	3.65 days	7.3 hours
99.9%	8.8 hours	44 minutes
99.99%	53 minutes	4.4 minutes
99.999%	5.3 minutes	26 seconds

This is plain arithmetic: one year has about 8,760 hours, and 1% of that is about 87.6 hours, or 3.65 days. When a hosting provider promises a certain uptime in its service level agreement (SLA), this table shows what the promise actually means.

Choosing what to monitor

• Start with the homepage.
• Add the pages that make money or bring contacts: checkout, cart, contact form, login, sign-up.
• Monitor APIs that your site or apps depend on.
• Don't forget pages served by other systems, such as a blog on a separate platform or a payment page.

Good practices

• Monitor from outside your own network. If the monitoring runs on the same server as the website, it goes down together with it.
• Use content checks, not only status checks, to catch broken but responding pages.
• Send alerts to the people who can fix the problem, and avoid sending every alert to everyone.
• Avoid alert fatigue. Too many false alarms teach people to ignore alerts. Confirm failures before alerting, and tune the checks.
• Keep a history. Outage records help you find recurring problems and hold providers to their SLA.
• Have a plan. Decide in advance who reacts to an alert, and how.

Related topics

• Testomato – a website monitoring service with uptime and content checks.