Simplifying Instrumentation in complex systems at scale with Open Telemetry
summary
Shows how teams get zero‑code auto‑instrumentation with OpenTelemetry SDKs and the OpenTelemetry Kubernetes Operator injecting agents across many deployments, and how eBPF provides kernel‑level traces (requires Linux 5.8+ and careful permissions). Describes OTLP + Collector pipelines with head/tail/hybrid sampling, batching and load‑balanced collectors to cut telemetry egress (example: 1 GB raw → ~200 MB after sampling/batching). Notes Go often needs manual traces or eBPF, AWS has ADOT distro, and cites Skyscanner migration as a working case.
AI-written summary · about
chapters (15)
- 0:00OpenTelemetry introduction
- 1:09Observability pains
- 4:57What is OpenTelemetry
- 9:11OTel architecture basics
- 11:39Instrumentation strategies
- 15:54Zero code with SDKs
- 20:03K8s auto-instrumentation
- 24:14eBPF runtime instrumentation
- 29:58Instrumentation comparison
- 31:25Collector pipelines
- 33:34Sampling strategies
- 37:30Scaling collectors
- 38:54OpenTelemetry in cloud environments
- 40:38Key takeaways
- 42:43Q&A
related talks
23mOpenTelemetry Auto Instrumentation for Effortless Observability
30mOpenTelemetry 101
01h 28mUnderstanding Distributed Systems with OpenTelemetry
36mObservability For You and Me with OpenTelemetry
50mGuide to Observability with OpenTelemetry
40mInstalling OpenTelemetry, Today and Tomorrow
32mDistributed Tracing in Action: Our Journey with OpenTelemetry
18mOpenTelemetry and Observability: What, Why, and Why Now?