Exemplars in Practice: Finding the Needle in Your Observability Haystack
summary
Exemplars link a metric point to a single trace so engineers can jump from dashboards to traces but they fail when requests are heterogeneous or multi dimensional and one trace hides larger issues The alternative shown queries traces across populations instead of sampling so you can reason over exemplar dots on metrics and then query traces across 100% of data The demo aggregates traces for demo-music-frontend play_track, filters by tag deployment=production, and locates relevant outliers effectively using Jaeger/Zipkin style traces
AI-written summary · about
chapters (6)
- 0:00Introduction
- 0:55Overview
- 1:36Exemplars Without
- 3:32Exemplars Overview
- 10:21Exemplars Limitations
- 15:21Conclusion
related talks
15mThe "Zen" of Python Exemplars
53mUnmask hidden issues with Observability superpowers!
39mThe Hidden Gems of Distributed Tracing
31mObservability for Developers: How to Get There
51mNavigating Distributed Traces
32mRules of Observability: The Data Edition
32mA Picture is Worth 1,000 Traces
35mAn Exemplary Path: Leveraging EBPFs and OpenTelemetry To Auto-instrument for Exemplars