Lightning Talk: Avoiding Chaotic Dysfunction: The ‘Don’t Do This’ Guide T...
summary
OpenTelemetry is running in production for about two years inside an API platform used by enterprise and financial customers, instrumenting a GraphQL pipeline and shipping spans through the Go collector into Grafana and Prometheus with Jaeger traces. Key actionable rules: don't start without a plan, don't collect everything (cardinality and storage costs balloon), implement sampling and exemplars, filter and sanitize attributes for privacy, prefer semantic conventions over ad hoc custom attributes, and ensure reliable context propagation across async and queueing hops.
AI-written summary · about
chapters (10)
- 0:00OpenTelemetry pitfalls
- 1:03Speaker background and context
- 3:53Five don'ts for OpenTelemetry
- 6:21Plan observability first
- 7:46Don't collect blindly
- 9:11Security and privacy matter
- 9:53Don't use custom attributes prematurely
- 11:16Context propagation matters
- 12:18Don't overlook users
- 13:41Q&A
related talks
11mLightning Talk: Beyond Billions: Operating Thanos, Prometheus & OpenTelemetry...
11m⚡ Lightning Talk: Don't Get Blown up! Avoiding Configuration Gotchas for Tetragon Newbies
10mLightning Talk: Is OpenTelemetry Too Complicated to Get Started?
11mLightning Talk: OTel Me How to Build a Data Pipeline for Observability
07mLightning Talk: Debugging OpenTelemetry
10mLightning Talk: Observability Debt: When Telemetry Stops Telling the Truth
55mBeginner's guide to observability
12mLightning Talk: With Great Traces Comes Great Costs: How to Reduce That Bill?