Observing Python applications with OpenTelemetry
summary
Compares three concrete ways to add OpenTelemetry to Python without code changes: Kubernetes operator that injects autoinstrumentation into pods and runs a collector (works but often needs a custom agent image because one operator image per Python/C-lib causes binary wheel and version breakages); otel-instrument wrapper / eBootstrap that scans the environment and prefixes the entrypoint to preload the agent; and programmatic instrumentation requiring a couple of lines early in startup. Also shows GenAI work: span attributes for chat completions and embeddings, metrics for duration and token counts, and existing instrumentations for OpenAI, AWS Bedrock and Google.
AI-written summary · about
chapters (7)
- 0:00Introduction and speaker intro
- 1:26What is observability
- 5:46Kubernetes operator for auto-instrumentation
- 14:51OpenTelemetry instrument wrapper
- 17:45Programmatic instrumentation
- 19:57OpenTelemetry and GenAI work
- 27:44Q&A session
related talks
27mObserving Python applications with OpenTelemetry
24m"Introduction to OpenTelemetry with Python"
17mEfficient Python Observability with OpenTelemetry
45mSimplifying Instrumentation in complex systems at scale with Open Telemetry
30mPractical introduction to OpenTelemetry Tracing
02h 02mHow To Troubleshoot and Monitor Applications using OpenTelemetry
32mUsing OpenTelemetry in a polyglot environment
38mEnhance your Python application observability