Monitoring your clients’ sites when you run an agency

Last updated: 9 November 2026

Monitoring one site is easy. Monitoring forty sites for fifteen clients, without missing one when you add another, is where it gets complicated.

What deserves an alert

The site stops responding, or responds with an error.
The SSL certificate is about to expire.
The server is running out of disk or memory.
A container has stopped.

Everything else (a slow page once, a two-minute CPU spike) deserves a trace in the report, not a text message at three in the morning.

Every minute?

The further apart the checks, the easier it is for an outage to go unnoticed. At five-minute intervals, a four-minute outage may never be seen. At one minute, you see almost everything. StatOps checks every site every minute.

Avoiding false alerts

An alert on a single failure produces many: a lost packet, a restart of a few seconds. After a week, you ignore alerts, and you miss the real one. Require several failures in a row. We wait for three, about three minutes, before warning.

Not forgetting a site

This is the point that hurts in practice: a site launched on a Friday, never added to the tool. It goes down, nobody knows. The remedy is to add nothing by hand. The StatOps agent reads the web server configuration (Nginx, Apache, Caddy, Traefik, Docker, Cloudflare Tunnel) and reports every domain served. A new site appears by itself, and you only have to attach it to its client.

Organise by client, not by server

You invoice by client, you report by client. Group sites by client from the start, and the monthly report becomes a simple consequence.