All glossary terms
Verify

Three pillars of observability

The three pillars of observability — logs, metrics, traces — are the foundational telemetry types that together let engineers reason about a system's runtime behaviour. Logs record discrete events with rich context, metrics aggregate numerical measurements over time, and traces follow a single request across distributed service boundaries.

The 'pillars' framing is contested — Charity Majors and the honeycomb.io school argue that the pillars are a vendor abstraction and the real primitive is structured wide events, with metrics and traces derivable from events. Both views agree on the substance: a production system needs the ability to answer 'what just happened?' (logs), 'how is the system trending?' (metrics), and 'where did this slow request spend its time?' (traces). OpenTelemetry has emerged as the cross-vendor instrumentation standard for all three. The pragmatic guidance: invest first in structured logs with high-cardinality context, derive metrics from those logs, and add tracing when service-graph complexity exceeds what logs alone can explain.

Related terms