Sentry error reporting is dropping some local variables in favour of other - and no way to control this

[sam-ghosh]

How do you use Sentry?

Sentry Saas (sentry.io)

Version

1.39

Steps to Reproduce

I am using sentry with the following config

import sentry_sdk
from sentry_sdk import capture_exception
from sentry_sdk.integrations.logging import LoggingIntegration

sentry_sdk.init(
     dsn="https://2ca804fbf56741bf929d85fa2547b6c6@o176489.ingest.us.sentry.io/1260928",
    

    max_breadcrumbs=50,
    attach_stacktrace=True,
    max_value_length=100  # Example length, adjust as needed.
)

and then creating a test error with a few local vars large small

def generate_error():
    # Large nested dictionary
    small_a = 1
    small_b = 2
    # large_dict = {i: {j: "x" * 1000 for j in range(10)} for i in range(10)}
    
    # Large list
    large_list = ["y" * 1000 for _ in range(100)]

    try:
        result = 1 / 0  # This will cause a ZeroDivisionError
    except Exception as e:
        capture_exception(e)

generate_error()

when the error is generated i see that small_a and small_b are not captured, the entirely is filled up by large_list

how can i control this feature? I have seen my projects before that when the local vars have one large list and other small vars, I cannot control and tell sentry to show the smaller variables

Expected Result

small_ and small_b being shown in sentry

Actual Result

[antonpirker]

Hey @sam-ghosh thanks for writing in!

In the SDK there is something called the serializer that tries to trim events in a semi smart way to prevent events being too big.

Iirc it should remove data based on max_value_length as you tried. (But may the length applies to the list of all local vars, not every single one.)

Anyhow, what you can do:

• Use a event_scrubber to remove the big list. (The event scrubber is run BEFORE the serializer mentioned above.
• Use a before_send hook to change the data. (The before_send hook is run AFTER the serializer mentioned above)

Hope this helps,
Anton