Free server and website monitoring with instant downtime alerts

Free Server Monitoring: a practical guide

External uptime monitoring or a self-hosted tool?

Open source uptime monitoring tools such as Uptime Kuma are easy to self-host, and many teams install them on the same server that runs the website. When that server goes down, the monitor goes down with it and no alert is sent. The check has to run from a different network and data center, so a power cut, a dead link or a full disk on your main server cannot silence it.

An external check answers one question: does the service respond. It cannot see a disk filling up slowly or a memory leak. Those need an agent-based tool such as Zabbix or Netdata running inside the server.

Why does the location of the check matter?

A server hosted in Iran can be fully reachable for local users while a monitor abroad reports it as down during international link disruptions, and the reverse happens to servers hosted abroad. If most of your customers are inside the country, at least one check point should be local, with a second one abroad if the service is also used from outside.

Requiring confirmation from more than one location before an alert goes out keeps a single broken route from being logged as a full outage.

Which URL should website uptime monitoring check?

Home pages are often served from a CDN or page cache and keep returning 200 even when the database is down. A better target is a page that needs the backend, such as search or login, or a /health endpoint that also tests the database connection.

Sites behind Cloudflare or ArvanCloud may block monitor requests as bot traffic and trigger false alarms, so the check point's IP or user agent gets an exception in the firewall rules. Mail deserves its own check on port 587 or 465, since a healthy web server says nothing about email.

Check interval, timeouts and false downtime alerts

A shorter interval catches outages sooner but turns every brief network blip into an alert. The usual setup waits for several failed checks in a row before declaring an outage. The timeout also has to fit the site's normal response time. A page that normally takes two seconds will always look down with a one-second timeout.

Uptime reports are most useful next to your hosting agreement. If the provider committed to an availability level, the recorded outage history is what you use to hold them to it.