A readable error taxonomy beats a loud dashboard

Illustration for A readable error taxonomy beats a loud dashboard

Dashboards with fifty tiles teach anxiety, not action. In the Node service track we ask cohorts to agree on ten error families before writing handlers.

Each family maps to user-visible copy, retry hints, and whether support should be looped in. The exercise feels bureaucratic until someone’s “unknown_error” string appears in three services at once.

We also require a single request id field in structured logs for the capstone slice. Correlation beats volume.

This is opinionated. Your organisation may already have standards — bring them; we adapt the rubric rather than replacing wholesale.

observabilityapidx

← All notes