request.bounded_stream is consumed

[sauravshanu1]

Our firm recently adopted Sentry. We use the default Falcon Integration for request tracing etc.

We are facing trouble while reading the request.bounded_stream data.

In

sentry-python/sentry_sdk/integrations/falcon.py

Line 81 in f62c83d

return self.request.media

,
request.bounded_stream is already consumed before the request comes to request handler.
So, this causes the variable data to be empty. See snippet below.

def on_post(self, request, response):
     data = request.bounded_stream.read()

It is true that the parsed data can be accessed by request.get_media(). However, I would like to access the raw data for purposes like signature verification etc.
Right now, we don't have any option other than disabling Sentry Falcon Integration altogether and writing our own integration to circumvent this problem.

Can Sentry media not consume the media if it is not consumed yet? Or, if it does, reset the bounded_stream with the payload data.

[sentrivana]

Heya @sauravshanu1, thank you for the report. This shouldn't happen. We need to see how we can retrieve the data without exhausting the stream.

In the meantime, you could set max_request_body_size to never in your SDK init() -- this should prevent the SDK from accessing the stream altogether I believe. (You won't get the request body in Sentry this way.)

[sauravshanu1]

Thanks @sentrivana for the fast response and your solution. It does work and the raw data can now be read from the bounded_stream exactly. I can live with no request.body in Sentry until this is fixed.

[szokeasaurusrex]

Hi @sauravshanu1, could you please provide a minimal reproduction of the error? I tried to reproduce your bug, but I observed different behavior with the following code:

# test.py

from wsgiref.simple_server import make_server
import falcon
import sentry_sdk

sentry_sdk.init(traces_sample_rate=1.0, debug=True)


class TestResource(object):
    def on_post(self, req: falcon.Request, resp: falcon.Response):
        print(req.bounded_stream.read())
        resp.status = falcon.HTTP_200
        resp.text = "Hello, World!\n"


app = falcon.App()
app.add_route("/", TestResource())


if __name__ == "__main__":
    with make_server("", 8000, app) as httpd:
        print("Serving on port 8000...")

        # Serve until process is killed
        httpd.serve_forever()

My pip freeze is:

# requirements.txt

certifi==2024.8.30
falcon==4.0.2
pip==24.3.1
sentry-sdk==2.18.0
urllib3==2.2.3

I am running the code as follows:

$ export SENTRY_DSN="my DSN here"
$ python test.py

And in a separate terminal:

$ curl -H "Content-Type: application/json" \
-X POST -d '{"value":"Hello"}' http://localhost:8000

The request data gets printed correctly to console:

b'{"value":"Hello"}'

The only strange behavior I observed is that on this line that you linked, we are not able to read the request.media in the Sentry SDK. However, if I comment out the print(req.bounded_stream.read()), the request.media is there.