Is your software working as intended in production? It may have worked when you tested it, but is it working now?

Too often monitoring is something developers leave to "the ops people". Too often we build systems that are hard to monitor. Too often our monitoring ends up focused on symptoms like "my application is not running" rather than business problems like "we are not making money"

Let's explore the idea of applying the extreme-programming concept of test-first development to our monitoring. We'll see how it helps us monitor the right things, build systems that are easier to run, and incrementally improve our architecture.

I'll also cover some of the techniques we can use to avoid common pitfalls such as brittle and unreliable checks, and how to stop side effects polluting our production data.