Flask Traces Sampler Function seems to have no effect

Environment

How do you use Sentry? Sentry SaaS (sentry.io)

Which SDK and version? sentry-sdk[flask]==1.4.1

Steps to Reproduce

1. Changed from previous code:

    sentry_sdk.init(
        dsn=sentry_dsn,
        traces_sample_rate=0.1,
        environment=sentry_environment,
        attach_stacktrace=True,
        integrations=[
            FlaskIntegration(),
            sentry_logging,
        ],
    )

to

    sentry_sdk.init(
        dsn=sentry_dsn,
        traces_sampler=sentry_traces_sampler,
        environment=sentry_environment,
        attach_stacktrace=True,
        integrations=[
            FlaskIntegration(),
            sentry_logging,
        ],
    )

where

def sentry_traces_sampler(sampling_context):
    try:
        # wsgi_environ is populated by sentry flask integration
        # https://docs.sentry.io/platforms/python/guides/flask/configuration/sampling/
        request_uri = sampling_context['wsgi_environ']['REQUEST_URI']
    except KeyError:
        return 0.1

    if request_uri == "/ping":
        # very low rate as these are extremely frequent and pretty useless
        return 0.001

    if request_uri in {
        "/api/xml/home_aaa_request",
        "/api/xml/home_bbb_request",
        "/api/xml/home_ccc_request",
        "/api/xml/home_ddd_request",
    }:
        # lower rate as these are very frequent
        return 0.01

    # default sample rate
    return 0.1

1. tested locally with Postman to confirm that sentry_traces_sampler goes down correct route for each endpoint
2. released to production

Expected Result

Number of traced requests to API endpoints should drop to < 10% of previous value.

Actual Result

Number of transactions stays roughly same as before. Screenshots showing the release which was roughly 10:00 (28/09/2021):

Assuming that requests haven’t spiked hugely, we should expect to see a dramatic drop in transactions per minute around that time, but it is roughly the same.

One hypothesis is that the try/except catch default value of 0.1 is being hit in production but not locally (I’ll test this now), but it seems possible that the sentry_traces_sampler return values are not being interpreted correctly?

[AbhiPrasad]

Hey @mfumagalli68 - could you open a new issue to track with code snippets, versions of sdk + fastapi you’re using, and your traces sampler config? Thanks!

[antonpirker]

Hey @AdamUrbanJungle Sorry for taking so long to reply. Can you confirm that this is still an Issue in our newest SDK? If yes, I will try to reproduce this. Thanks!