How to capture an easily serialisable exception event with Sentry?

Question

I'm using Sentry to monitor a Python worker process that is running under some restrictions. For example it can't open files or connect to the internet. For this reason I can't use sentry_sdk.capture_exception directly.

What I can do is put an object on a multiprocessing.Queue to send it to the parent process which can talk to the Sentry server. So I tried doing something like this:

from sentry_sdk.utils import event_from_exception

try:
    # failing code
except Exception:
    exc_info = sys.exc_info()
    sentry_event = event_from_exception(exc_info)
    queue.put(sentry_event)

and then in the parent process sentry_sdk.capture_event(sentry_event).

However it turns out that sentry_event still contains some raw Python objects (local variables) which cannot be pickled so putting it on the queue fails.

Is there a way to obtain an event where the data has already been extracted and it can easily be serialised, e.g. as JSON?

I've also asked about this here: https://github.com/getsentry/sentry-python/issues/654

Answer 1

Based on Markus' answer, without modifying global state:

import sentry_sdk


def get_exception_event():
    event = None

    def transport(e):
        nonlocal event
        event = e

    client = sentry_sdk.Client(transport=transport)
    hub = sentry_sdk.Hub(client)
    hub.capture_exception()

    assert event is not None
    return event

try:
    1 / 0
except Exception:
    print(get_exception_event())