Distributed tracing doesn't work in Django

[keattang]

Specifying DISTRIBUTED_TRACING=True in the Django Datadog settings doesn't work because the propagator looks for the x-datadog-trace-id header in the Django request.META object which converts the header name to HTTP_X_DATADOG_TRACE_ID. For more details on this see here.

In order to work around this I've had to monkey patch the library so that it looks for the headers with the correct keys:

from ddtrace.propagation import http
http.HTTP_HEADER_TRACE_ID = 'HTTP_X_DATADOG_TRACE_ID'
http.HTTP_HEADER_PARENT_ID = 'HTTP_X_DATADOG_PARENT_ID'
http.HTTP_HEADER_SAMPLING_PRIORITY = 'HTTP_X_DATADOG_SAMPLING_PRIORITY'

[palazzem]

Thanks @keattang ! We'll investigate on that as soon the current 0.12 is out! I'll keep you updated.

[praseodym]

The monkeypatch works for us, but it would be great to see this fixed in the library.

[Kyle-Verhoog]

Hi all, looking into this issue now.

From what I can tell this behaviour is defined in the WSGI spec: https://www.python.org/dev/peps/pep-3333/#environ-variables so this shouldn't be an issue with just Django.

I think it makes sense to change our HTTP propagator to check for the converted headers as well when extracting.

[Kyle-Verhoog]

@praseodym @keattang we'll be including the fix for this issue in our upcoming release! 😄