Operations · Monitoring
Observability
The ability to answer new questions about a system's behaviour from its telemetry, without shipping new instrumentation.
AdvancedUpdated 2026-09-01
Overview
Observability builds on metrics, logs and traces with high cardinality context. Distributed tracing links a single user request across services, which is what makes latency and failure attribution possible in distributed systems.
OpenTelemetry has standardised instrumentation, so telemetry is increasingly portable between backends, an important consideration given how quickly observability costs escalate.
Security considerations
- Scrub sensitive data before ingestion; traces and logs frequently capture tokens and personal data.
- Control retention and sampling deliberately, balancing cost against investigative need.