Automatic context propagation via instrumentation

[naumnaum]

Hey there everyone!
Maybe you can share some materials on how to use python auto-instrumentation for context propagation?
I'm working with a quite big number of separate FastAPI services running in docker, and naive
opentelemetry-instrument uvicorn app:app ...
doesn't provide any context propagation.
In result i get a bunch of unrelated traces.
I've succeeded to make a context propagation manually via extracting/injecting trace context from/in headers https://github.com/naumnaum/fastapi-opentelemetry-tracing
However it would be really painful to integrate this manually to the main project i'm working on.
Thanks everyone in advance!

[lariel-fernandes]

Don't know if this problem was solved since then, but here's my 2 cents.

At least for the part of propagating the context from an incoming request into your FastAPI app, you can do this:

from opentelemetry import context, propagate

@app.middleware("http")
async def propagate_context(request: fastapi.Request, call_next):
    context.attach(propagate.get_global_textmap().extract(request.headers))
    return await call_next(request)

Don't forget to also export the environment variable OTEL_PROPAGATORS (e.g.: tracecontext,baggage)