5 lessons you can learn from OpenTelemetry Python

OpenTelemetry Python is a SIG (Special Interest Group) that develops and manages the Python implementation of OpenTelemetry, the second-highest velocity project in the CNCF landscape (only behind Kubernetes). This group delivers a vendor-neutral observability solution, building over 80 Python packages and counting more than 500M downloads per month. This makes it a peculiar project with interesting technical challenges and scales to share the lessons learned. Presented with Emidio Neto at Python Brasil 2025.

Transcript

1. 5 Lessons You Can Learn From OpenTelemetry Python Emídio Neto & Riccardo Magliocchetti
2. 2025 - Python Brasil 2 Who are we Emídio Neto Software Engineer, Independent Approver OpenTelemetry Python Riccardo Magliocchetti Sr. Software Engineer, Elastic Maintainer OpenTelemetry Python
3. Observability in 1min
4. 2025 - Python Brasil 4 Observability in 1min Observability shows what’s going on so you can find and fix problems faster Why is this issue happening? “Silent issue”, but Latency Spiked It’s always the Database DNS
5. 2025 - Python Brasil 5 Observability in 1min Metrics explains what Traces explains where Logs explains why
6. OpenTelemetry TL;DR
7. OpenTelemetry - Observabilidade - Tracing
8. 2025 - Python Brasil 8 OpenTelemetry OpenTelemetry is the framework in which software is instrumented to become observable. Probably graduated at Kubecon NA 2025
9. 2025 - Python Brasil 9 standards, specifications, and conventions for how telemetry data is generated and managed OpenTelemetry API / SDK Semantic Conventions OpenTelemetry Line Protocol Instrumentation Libraries OpenTelemetry
10. 2025 - Python Brasil 10 Code-based via APIs and SDK for application programming language + instrumentation libs Zero-code (auto-instrumentation) OpenTelemetry You can use both methods together!
11. OTel Python SDK
12. 2025 - Python Brasil 12 Python SDK Provides OTel API, SDK and libs you need to instrument Python applications! Traces, Metrics and Logs
13. 2025 - Python Brasil 13 Python SDK API, SDK, OTLP, Semantic Conventions, exporters grpc and http Core Repo and Contrib Repo (instrumentation and custom libs)
14. 2025 - Python Brasil 14 Python SDK - Packages, installation
15. Auto-instrumentation OTel Python SDK
16. Lesson 1: Monkey patch beyond just patching bugs
17. 2025 - Python Brasil 17 Monkey patching is a technique used to dynamically change the behavior of some code at runtime.
18. 2025 - Python Brasil 18
19. 2025 - Python Brasil 19
20. 2025 - Python Brasil 20 But sometimes it can be hard… Keep metadata Makes patching reversible and consistent Works transparently even if library reloads or is imported late
21. 2025 - Python Brasil 21 Just use wrapt! Documentation https://wrapt.readthedocs.io/en/master/
22. 2025 - Python Brasil 22 wrapt example: asgiref toy instrumentation
23. 2025 - Python Brasil 23 wrapt example: botocore instrumentation
24. Lesson 2: sitecustomize
25. 2025 - Python Brasil 25 sitecustomize?
26. 2025 - Python Brasil 26 By default the site module of your Python installation tries to load from your PYTHONPATH a module named sitecustomize. Reference https://docs.python.org/3/library/site.html#module-sitecustomize sitecustomize
27. 2025 - Python Brasil 27 sitecustomize.py $ python3 -c 'print("world")' world
28. 2025 - Python Brasil 28 sitecustomize.py $ python3 -c 'print("world")' world $ PYTHONPATH=. python3 -c 'print("world")' hello world
29. 2025 - Python Brasil 29 sitecustomize.py $ PYTHONPATH=. python3 -c 'print("world")' hello world $ cat sitecustomize.py print("hello", end=" ")
30. Lesson 3: Python Entry Points
31. 2025 - Python Brasil 31 Entry points are a mechanism that packages use to provide discoverability over some of their code.
32. 2025 - Python Brasil 32 Entry points [project.entry-points.opentelemetry_configurator] # group # name = code reference configurator = "opentelemetry.distro:OpenTelemetryConfigurator"
33. 2025 - Python Brasil 33 Entry points from opentelemetry.util._importlib_metadata import entry_points for entry_point in entry_points(group="opentelemetry_configurator"): klass = entry_point.load() # call configure method of OpenTelemetryConfigurator instance klass().configure() print(fˮConfigured: {entry_point.name}ˮ)
34. Lesson 4: Testing and Releasing
35. 2025 - Python Brasil 35 Lesson 4: Testing and Releasing
36. 2025 - Python Brasil 36 Lesson 4: Testing and Releasing
37. 2025 - Python Brasil 37 Lesson 4: Testing and Releasing
38. 2025 - Python Brasil 38 Lesson 4: Testing and Releasing
39. 2025 - Python Brasil 39
40. 2025 - Python Brasil 40 Releasing → PEP 440 • Release: X.Y.Z → e.g., 1.37.0 • Pre-releases: bN (beta) → e.g., 0.58b0 • Dev releases: .dev → e.g., 1.37.0.dev (at git only) e.g., • opentelemetry-api → 1.34.0, 1.34.1 (patch release), • opentelemetry-instrumentation → 0.58b0, 0.58b1 (patch) Lesson 4: Testing and Releasing
41. 2025 - Python Brasil 41 Lesson 4: Testing and Releasing
42. Lesson 5: Speeding up things with uv
43. 2025 - Python Brasil 43 Lesson 5: Speeding up things with uv • At testing level → tox-uv, which add UV support to tox while installing packages. • For local development → uv sync will setup opentelemetry-python packages for you
44. 2025 - Python Brasil 44 Takeaways • Monkey patching is powerful • Sitecustomize acts like magic • Entry points allow extensibility • Automations in CI • UV to install things faster
45. 2025 - Python Brasil 45 Obrigado! It’s time for Q&A Join OTel community emdneto xrmx