Uptime monitoring

About this project

If you run software for other people, you should find out it is down before they do. We watch well over a hundred checks across everything we operate: our own products, the storefronts, and every customer site we look after.

The unusual part is that none of it is clicked together in a dashboard. The full inventory lives in our repository as reviewed files: which accounts exist, what each monitor checks, and which of them may appear on a public status page. A command reconciles that file with the live monitoring system, so a change is a reviewed commit rather than someone's afternoon in a settings screen, and nothing silently drifts.

New hosts are picked up automatically and parked as candidates until someone confirms them, so growth does not quietly go unwatched. And because not everything worth checking is exposed to the internet, the monitoring host joins a private network of its own: we can watch things behind a customer's firewall without putting them on the open web to do it.

Highlights

• Well over a hundred checks across products, storefronts, and customer sites
• The whole inventory reviewed in git, then reconciled onto the live system
• A status page per customer, with monitors explicitly marked public or internal
• New hosts discovered automatically and held as candidates until confirmed
• A private network so services behind a firewall can be watched too