What is Observability?
The ability to understand a system's internal state from its outputs — logs, metrics, and traces — especially for debugging unknown failures.
Definition
Observability is the degree to which you can explain what a system is doing purely from the telemetry it emits. It rests on three pillars — structured logs, metrics, and distributed traces — increasingly collected through the vendor-neutral OpenTelemetry standard and analyzed in tools like Grafana, Datadog, or Honeycomb. Where traditional monitoring answers known questions ("is CPU above 80%?"), observability supports exploring unknown failure modes: slicing high-cardinality data by customer, region, or release to find why one specific request path is slow. It is a prerequisite for operating microservices at scale.