Don't Panic — Synthetic Monitoring

Synthetic monitoring is a scheduled attempt to do one small, important thing before a customer has to discover that it stopped working. A probe asks an endpoint for an answer, follows an API workflow, or drives a browser through a journey. It records whether the expected interaction happened and how long it took. The arrangement is not glamorous. Neither is discovering a broken sign-in path from a customer email.

The central idea is the check contract: target, probe location, schedule, timeout, assertion, alert owner, and first investigation step. That list exists because a URL is not a promise. A health endpoint might reply while the response is wrong. A login page might load while sign-in is broken. The assertion has to prove the behavior that deserves an alert, which is annoyingly more specific than asking whether the internet is generally in a good mood.

A probe also changes the evidence. A public probe asks whether the public target works from that external location. A private probe asks whether it works from the network where that probe runs. Neither one speaks for every device, network, identity state, cache, or user journey. This is controlled evidence, not a tiny oracle with a clipboard.

Choose the least complicated check that proves the promise. HTTP, DNS, TCP, ICMP, TLS, and gRPC checks can answer narrow protocol questions. A multistep check carries state through a sequence of requests. A browser check drives an automated interaction. Each step up can prove more behavior, but also brings more state and more ways for a check to become brittle. The surprise is that a deeper check is not automatically a better one; it is only better when that extra behavior is the thing you are protecting.

A failed check starts an investigation. It does not announce the root cause with ceremonial authority. Put the failed assertion, location, time, and diagnostic detail in the alert, then compare it with logs, metrics, traces, and recent changes. Treat test identities and secrets as production-sensitive, avoid irreversible actions, and give any state-changing test a cleanup path.

Read the Intro for the full check path and the limits of synthetic evidence. Use Slides for the compact map of check types, locations, and decisions. Keep the Cheatsheet nearby when defining an assertion or alert. The Practice Reference turns one service promise into a check contract, and the Exercise lets you watch an assertion pass, then fail for a deliberate reason. That is a much more useful kind of reassurance.