The Lighthouse Keeper's First Lesson: A Single, Meaningful Check

We talk a lot about coverage in our work. We want to monitor everything, to have a thousand tiny sensors reporting back, ensuring no dark corner of our service goes unwatched. It’s a natural impulse, born from a fear of the unknown. But I want to suggest a different starting point, one borrowed from an older tradition of watchfulness: that of the lighthouse keeper.

His duty wasn’t to report on every wave, every cloud, or every ship. His duty was to ensure the light was lit. That single, powerful beam was the ultimate health check. If it was on, it meant the keeper was awake, the fuel was stocked, the lens was clean, and the mechanism was sound. One check implied the health of the entire system.

We can apply this same principle. Instead of beginning with a sprawling suite of synthetic transactions that test every API endpoint and UI button, start by defining your lighthouse.

What is the one request that, if it succeeds, implies a vast amount of underlying health? It’s not just a ping to the root URL. It’s a carefully crafted call that travels through the heart of your service. It should touch your primary database, validate a connection to your core cache, exercise a critical piece of business logic, and return a predictable, verifiable result. For an e-commerce site, it might be a call that fetches a specific, known product’s details. For a SaaS application, it might be an authentication handshake with a test user.

The magic of this single check is in its depth, not its breadth. By designing one check that must traverse the same crucial paths as real user traffic, its success becomes a powerful signal. Its failure is an immediate and unambiguous siren. There’s no noise, no flapping from a minor ancillary service being down for maintenance. If the lighthouse goes dark, you know something is fundamentally wrong.

This isn’t to say you should stop at one check forever. But it should be your first. It becomes your canary, your ground truth. When your elaborate dashboard of a thousand metrics turns red, your first question should be: “But is the lighthouse still lit?” It cuts through the complexity and tells you the one thing you truly need to know in that moment: is the core of the service alive and functioning for users?

In an age of overwhelming observability data, the oldest form of watchfulness still has the most to teach us. Build your lighthouse first. Light that one, great beam. Then, and only then, begin watching the waves.

Notes & further reading

A few pages I came back to while writing this: