Why Logs, Metrics and Traces Still Don’t Give You Real Observability
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
The three-pillar observability model (logs, metrics, traces) is a solid foundation but has meaningful gaps in modern distributed systems. Key blind spots include: the 'known unknowns' problem where uninstrumented failure modes leave teams blind during incidents; high-cardinality limitations in traditional metrics systems that prevent slicing by user, region, or tenant combinations; context collapse in async workflows, third-party calls, and serverless environments; and business logic blind spots where technically healthy services return wrong results. The post advocates moving toward structured events (JSON with rich fields), high-cardinality querying tools like Honeycomb or ClickHouse, continuous verification of business correctness in production, and post-incident instrumentation reviews. OpenTelemetry is recommended as the vendor-neutral standard for emitting all signal types.
