Your Distributed System is Talking to You: OpenTelemetry
summary
Modern observability uses distributed tracing with OpenTelemetry to turn requests into correlated spans and full traces; a demo app with about 10 services in 8 languages sends spans to Jaeger or a vendor backend (Honeycomb mentioned) via the OpenTelemetry Collector (receivers, processors, exporters) over OTLP; automatic instrumentation covers web frameworks, custom attributes carry business fields, and storage cost forces tail sampling and batching; an open-source Refinery tail-sampler and semantic-convention debates (blocking .NET 1.0) are notable operational realities.
AI-written summary · about
chapters (20)
- 0:00Introduction to observability
- 1:27Why software can't explain itself
- 4:36What is telemetry?
- 6:23Logs vs metrics vs traces
- 10:16Modern observability pipelines
- 11:19Live demo of OpenTelemetry
- 18:00What are distributed traces?
- 21:33OpenTelemetry's origins
- 23:43OpenTelemetry Collector
- 31:25Instrumentation strategies
- 34:39Checking language support
- 35:00Sending telemetry to a backend
- 40:27OpenTelemetry weaknesses
- 43:17Alternatives to OpenTelemetry
- 44:22Getting started with OpenTelemetry
- 45:24Choosing a telemetry backend
- 49:18Sampling and storage costs
- 51:23Building an observability team
- 53:10The instrumentation loop
- 54:37Why observability matters
related talks
01h 28mUnderstanding Distributed Systems with OpenTelemetry
35mOpenTelemetry for Busy People
46mThe State of OpenTelemetry
42mFirst Steps with Observability: OpenTelemetry and You
43mWhat is this OpenTelemetry thing?
22mOpenTelemetry – An Observability Framework for Cloud-Native Software
50mGuide to Observability with OpenTelemetry
59mOpenTelemetry will save your day and night