Metadata: The Secret Sauce for Full Observability
summary
Observability becomes actionable by concrete metadata practices: use structured logs with trace ID in logs or even emit logs inside spans to guarantee precise trace–log correlation, add exemplars on metrics to jump from a latency spike to a sample trace, enrich metrics via Prometheus labels and the OpenTelemetry collector kubernetes_attributes processor for consistent node/namespace/deployment context, and avoid blind defaults (about 75% of default Kubernetes metrics go unused) to prevent cardinality blowup.
AI-written summary · about
chapters (13)
- 0:00Introduction and speaker intro
- 1:32What is observability
- 3:41Observability as data analytics
- 5:26Defining metadata
- 7:16Metadata in logs
- 11:13Metadata in traces
- 16:39Metadata in metrics
- 20:31Cross-signal correlation
- 24:53Standardizing metadata
- 29:50OpenTelemetry and metadata
- 30:52Summary and key takeaways
- 31:14Q&A: Cardinality concerns
- 33:47Q&A: OpenFeature for metadata
related talks
32mUnderstanding Observability in Kubernetes
36mObservability? It's a Data Analytics Problem, You Fool!
11mKubernetes Observability
40mDemystifying Observability and Monitoring
16mShrinking the Observability Bill
26mYou Can’t Fix What You Can’t See: Tales from the Observability Frontier
35mBeyond Logs, Metrics and Traces. The Data Analytics Paradigm Shift in Observability — Dotan Horovits
55mBeginner's guide to observability