How do you use Sentry?

Sentry Saas (sentry.io)

Version

1.25.1

Steps to Reproduce

While using Python sentry_sdk version 1.25.1 together with multiprocessing.Manager that has some shared values, the process hangs after receiving an exception in subprocess.

The problem occurs only when manager.Value and Sentry are initialized at the same time. Commenting out any of them releases the execution. You can see that the values doesn't have to be used by subprocess. The multiprocessing creates a process called resource tracked which might be conflicting with closing the Sentry watcher correctly.

You can recreate the issue using the snippet below.

• SENTRY_DSN env variable is set to proper value
• produced on MacOS, so "spawn" context is used by default

import multiprocessing
import os
import sentry_sdk
import time


def _check_workers(workers):
    if any([worker.exitcode != 0 for worker in workers]):
        raise Exception("Hanging with this error")


def wait_for_processes(workers, sleep_seconds=5):
    while any([worker.is_alive() for worker in workers]):
        _check_workers(workers=workers)

        time.sleep(sleep_seconds)

    _check_workers(workers=workers)


def failing_function():
    raise ValueError("Test error")


def execute():
    manager = multiprocessing.Manager()
    this_hangs = manager.Value("i", 10)  # presence of this resource hangs the process

    worker = multiprocessing.Process(
        target=failing_function,
    )
    worker.start()

    wait_for_processes([worker])


if __name__ == "__main__":
    sentry_sdk.init(dsn=os.environ.get("SENTRY_DSN"))
    execute()

Expected Result

Sentry handles the exception received in the main process, reraises it and the program execution ends with non-zero exit code.

Actual Result

The program receives the exception but waits infinitely for something to end it.