Describe your environment

OS: Python docker image (python:3.11.9-bullseye)
Python version: Python 3.11.9
SDK version: 1.22.0 and 1.25.0
API version: 1.22.0 and 1.25.0

What happened?

As mentioned in #3482, autoinstrumentation does not work when using an ASGI webserver with multiple worker processes. I thought that it would be productive to open a more specific issue for this bug but feel free to close this if it is simply a duplicate.

I have personally tested this on both of the specified API and SDK version (version 0.43b0 and 0.46b0 of the fastapi instrumentation module) with a FastAPI application using both uvicorn and hypercorn servers (auto-instrumentation with version 0.47.0b0 and 1.26.0 seems to not work even with a single process for which I have created a separate issue #4111).

This issue occurs both when using the Kubernetes operator init container auto-instrumentation as well as adding the instrumentation manually and auto-instrumenting the service like so:

opentelemetry-instrument --traces_exporter console uvicorn main:app --workers 4

and

opentelemetry-instrument --traces_exporter console hypercorn -w 4 main:app

Steps to Reproduce

Run the start.sh script in the repo for either branch (demonstrating the problem on either version) and open http://127.0.0.1:8000/ in a browser.

• https://github.com/emil-vdw/otel-asgi-issue/tree/emil-vdw/otel-1.22.0-issue
• https://github.com/emil-vdw/otel-asgi-issue/tree/emil-vdw/otel-1.25.0-issue

Expected Result

Traces are logged to the console.

Actual Result

No traces are logged to the console.

Additional context

In order to get autoinstrumentation to work on the aforementioned examples, simply drop the --worker 4 arg.

Would you like to implement a fix?

None