Python 3.12 support - RuntimeError

[Toaster192]

Problem Statement

Tried using sentry with python 3.12. Ran into newly implemented RuntimeError tests. Want to give a heads up for once there is 3.12 (hopefuly after 3.11 #1950 🤞 ) support that this is something that will probably need fixing(?)

Example traceback (source):

Exception ignored in atexit callback: <function AtexitIntegration.setup_once.<locals>._shutdown at 0x7fa77fa24e00>
Traceback (most recent call last):
  File "/builds/4704997395/software/venv/lib64/python3.12/site-packages/sentry_sdk/integrations/atexit.py", line 61, in _shutdown
    client.close(callback=integration.callback)
  File "/builds/4704997395/software/venv/lib64/python3.12/site-packages/sentry_sdk/client.py", line 580, in close
    self.flush(timeout=timeout, callback=callback)
  File "/builds/4704997395/software/venv/lib64/python3.12/site-packages/sentry_sdk/client.py", line 604, in flush
    self.transport.flush(timeout=timeout, callback=callback)
  File "/builds/4704997395/software/venv/lib64/python3.12/site-packages/sentry_sdk/transport.py", line 525, in flush
    self._worker.submit(lambda: self._flush_client_reports(force=True))
  File "/builds/4704997395/software/venv/lib64/python3.12/site-packages/sentry_sdk/worker.py", line 117, in submit
    self._ensure_thread()
  File "/builds/4704997395/software/venv/lib64/python3.12/site-packages/sentry_sdk/worker.py", line 42, in _ensure_thread
    self.start()
  File "/builds/4704997395/software/venv/lib64/python3.12/site-packages/sentry_sdk/worker.py", line 70, in start
    self._thread.start()
  File "/builds/4704997395/software/venv/lib64/python3.12/site-packages/sentry_sdk/integrations/threading.py", line 56, in sentry_start
    return old_start(self, *a, **kw)
           ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/lib64/python3.12/threading.py", line 971, in start
    _start_new_thread(self._bootstrap, ())
RuntimeError: can't create new thread at interpreter shutdown

cpython commit adding this "detection": python/cpython@ce558e6

Solution Brainstorm

Some sort of "graceful shutdown" on exit perhaps?

[Toaster192]

We got to work around this by closing sentry manually at the end of our programs 🤷

[sentrivana]

Thanks for the heads up @Toaster192!

[jnoordsij]

Would be great to have this fixed now that Python 3.12 is officially released.

[sentrivana]

Agreed @jnoordsij. We're currently figuring out what we need to fix for 3.12, this is definitely on the list.

[ThiefMaster]

Looking at #1950, is there some hope hat it will not take month(s) for 3.12 support? Even more so considering that 3.11 was just about (non-technical) metadata, while there's an actual issue with 3.12 right now which is not necessarily a blocker but at least a bit annoying when an application has CLI tools for example where this exception gets raised.

[antonpirker]

We have a PR open for supporting 3.12, but we can not give you an estimate on when it will be merged/released:

#2408

[robd003]

I'm also seeing this issue when terminating a Sanic server

^C[2023-10-14 15:38:55 -0400] [18739] [INFO] Received signal SIGINT. Shutting down.
[2023-10-14 15:38:55 -0400] [18739] [INFO] Server Stopped
Exception ignored in atexit callback: <function AtexitIntegration.setup_once.<locals>._shutdown at 0x103eb0c20>
Traceback (most recent call last):
  File "/Users/robd/code/allfactors-sanic/env/lib/python3.12/site-packages/sentry_sdk/integrations/atexit.py", line 61, in _shutdown
    client.close(callback=integration.callback)
  File "/Users/robd/code/allfactors-sanic/env/lib/python3.12/site-packages/sentry_sdk/client.py", line 618, in close
    self.flush(timeout=timeout, callback=callback)
  File "/Users/robd/code/allfactors-sanic/env/lib/python3.12/site-packages/sentry_sdk/client.py", line 646, in flush
    self.transport.flush(timeout=timeout, callback=callback)
  File "/Users/robd/code/allfactors-sanic/env/lib/python3.12/site-packages/sentry_sdk/transport.py", line 549, in flush
    self._worker.submit(lambda: self._flush_client_reports(force=True))
  File "/Users/robd/code/allfactors-sanic/env/lib/python3.12/site-packages/sentry_sdk/worker.py", line 117, in submit
    self._ensure_thread()
  File "/Users/robd/code/allfactors-sanic/env/lib/python3.12/site-packages/sentry_sdk/worker.py", line 42, in _ensure_thread
    self.start()
  File "/Users/robd/code/allfactors-sanic/env/lib/python3.12/site-packages/sentry_sdk/worker.py", line 70, in start
    self._thread.start()
  File "/Users/robd/code/allfactors-sanic/env/lib/python3.12/site-packages/sentry_sdk/integrations/threading.py", line 56, in sentry_start
    return old_start(self, *a, **kw)
           ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Cellar/python@3.12/3.12.0/Frameworks/Python.framework/Versions/3.12/lib/python3.12/threading.py", line 971, in start
    _start_new_thread(self._bootstrap, ())
RuntimeError: can't create new thread at interpreter shutdown
Traceback (most recent call last):
  File "/Users/robd/code/allfactors-sanic/src/af_events.py", line 1455, in <module>
    app.run(host='0.0.0.0', port=8081, workers=2, access_log=False, debug=False)
  File "/Users/robd/code/allfactors-sanic/env/lib/python3.12/site-packages/sanic/mixins/startup.py", line 209, in run
    serve(primary=self)  # type: ignore
    ^^^^^^^^^^^^^^^^^^^
  File "/Users/robd/code/allfactors-sanic/env/lib/python3.12/site-packages/sanic/mixins/startup.py", line 880, in serve
    sock.shutdown(SHUT_RDWR)
OSError: [Errno 57] Socket is not connected
Sentry is attempting to send 2 pending events

[ethrgeist]

This also occurs when running migrations in Django.

python .\src\manage.py migrate
2023-10-20 09:44:42,475 DEBUG __init__: couldn't import psycopg 'c' implementation: No module named 'psycopg_c'
Operations to perform:
  Apply all migrations: <bunch of apps>
Running migrations:
  No migrations to apply.
Exception ignored in atexit callback: <function AtexitIntegration.setup_once.<locals>._shutdown at 0x000001F382D9B380>
Traceback (most recent call last):
  File "C:\Users\user\.virtualenvs\sad-django\Lib\site-packages\sentry_sdk\integrations\atexit.py", line 61, in _shutdown
    client.close(callback=integration.callback)
  File "C:\Users\user\.virtualenvs\sad-django\Lib\site-packages\sentry_sdk\client.py", line 618, in close
    self.flush(timeout=timeout, callback=callback)
  File "C:\Users\user\.virtualenvs\sad-django\Lib\site-packages\sentry_sdk\client.py", line 646, in flush
    self.transport.flush(timeout=timeout, callback=callback)
  File "C:\Users\user\.virtualenvs\sad-django\Lib\site-packages\sentry_sdk\transport.py", line 549, in flush
    self._worker.submit(lambda: self._flush_client_reports(force=True))
  File "C:\Users\user\.virtualenvs\sad-django\Lib\site-packages\sentry_sdk\worker.py", line 117, in submit
    self._ensure_thread()
  File "C:\Users\user\.virtualenvs\sad-django\Lib\site-packages\sentry_sdk\worker.py", line 42, in _ensure_thread
    self.start()
  File "C:\Users\user\.virtualenvs\sad-django\Lib\site-packages\sentry_sdk\worker.py", line 70, in start
    self._thread.start()
  File "C:\Users\user\.virtualenvs\sad-django\Lib\site-packages\sentry_sdk\integrations\threading.py", line 56, in sentry_start
    return old_start(self, *a, **kw)
           ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "c:\users\user\.pyenv\pyenv-win\versions\3.12.0\Lib\threading.py", line 971, in start
    _start_new_thread(self._bootstrap, ())
RuntimeError: can't create new thread at interpreter shutdown

[ThiefMaster]

Yes, I think it always happens when the Python interpreter exits while the sentry SDK is 'active'...

[antonpirker]

Thanks for all the input! We will try to tackle this soon!

[antonpirker]

Hey everybody,
we just released an update that should fix this problem: https://github.com/getsentry/sentry-python/releases/tag/1.33.0

Could you give it a try @Toaster192 @robd003 and @ethrgeist ?

[ethrgeist]

Looks good, CLI and reload spam is gone, thank you very much!

[shikharvaish28]

Hi team, I am getting this error everytime a program has to exit:

I am using python 3.12 and sentry-sdk==1.14.0

Exception ignored in atexit callback: <function AtexitIntegration.setup_once.<locals>._shutdown at 0x7f1cc3949940>
Traceback (most recent call last):
  File "/opt/dataDumper/myenv/lib/python3.12/site-packages/sentry_sdk/integrations/atexit.py", line 62, in _shutdown
    client.close(callback=integration.callback)
  File "/opt/dataDumper/myenv/lib/python3.12/site-packages/sentry_sdk/client.py", line 482, in close
    self.flush(timeout=timeout, callback=callback)
  File "/opt/dataDumper/myenv/lib/python3.12/site-packages/sentry_sdk/client.py", line 504, in flush
    self.transport.flush(timeout=timeout, callback=callback)
  File "/opt/dataDumper/myenv/lib/python3.12/site-packages/sentry_sdk/transport.py", line 492, in flush
    self._worker.submit(lambda: self._flush_client_reports(force=True))
  File "/opt/dataDumper/myenv/lib/python3.12/site-packages/sentry_sdk/worker.py", line 113, in submit
    self._ensure_thread()
  File "/opt/dataDumper/myenv/lib/python3.12/site-packages/sentry_sdk/worker.py", line 42, in _ensure_thread
    self.start()
  File "/opt/dataDumper/myenv/lib/python3.12/site-packages/sentry_sdk/worker.py", line 70, in start
    self._thread.start()
  File "/opt/dataDumper/myenv/lib/python3.12/site-packages/sentry_sdk/integrations/threading.py", line 54, in sentry_start
    return old_start(self, *a, **kw)
           ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/lib/python3.12/threading.py", line 992, in start
    _start_new_thread(self._bootstrap, ())
RuntimeError: can't create new thread at interpreter shutdown

[ddelange]

the fix was released in 1.33.0