Kernel Stop Fix

[rakhmaevao]

Hello. I found this floating problem. If you start sessions and stop them, then the 500th error occurs (see the tracing and video below).

Screencast.from.2025-03-08.14-37-26.mp4

  File "/home/rakhmaevao/Projects/jupyverse/plugins/kernels/fps_kernels/kernel_server/server.py", line 137, in start
    await self._wait_for_ready()
  File "/home/rakhmaevao/Projects/jupyverse/plugins/kernels/fps_kernels/kernel_server/server.py", line 233, in _wait_for_ready
    msg = await receive_message(self.iopub_channel, timeout=0.2)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/rakhmaevao/Projects/jupyverse/plugins/kernels/fps_kernels/kernel_driver/message.py", line 136, in receive_message
    ready = await sock.apoll(timeout).wait()
                  ^^^^^^^^^^^^^^^^^^^
  File "/home/rakhmaevao/.local/share/hatch/env/virtual/jupyverse/u_XNhuOG/dev.jupyterlab-auth/lib/python3.12/site-packages/zmq_anyio/_socket.py", line 503, in apoll
    self._check_started()
  File "/home/rakhmaevao/.local/share/hatch/env/virtual/jupyverse/u_XNhuOG/dev.jupyterlab-auth/lib/python3.12/site-packages/zmq_anyio/_socket.py", line 939, in _check_started
    self._task_group.start_soon(self._start)
  File "/home/rakhmaevao/.local/share/hatch/env/virtual/jupyverse/u_XNhuOG/dev.jupyterlab-auth/lib/python3.12/site-packages/anyio/_backends/_asyncio.py", line 880, in start_soon
    self._spawn(func, args, name)
  File "/home/rakhmaevao/.local/share/hatch/env/virtual/jupyverse/u_XNhuOG/dev.jupyterlab-auth/lib/python3.12/site-packages/anyio/_backends/_asyncio.py", line 832, in _spawn
    raise RuntimeError(
RuntimeError: This task group is not active; no new tasks can be started

I fixed this error (see plugins/kernels/fps_kernels/kernel_server/server.py). I guarantee that sockets will close. Now the problem is gone.

To eliminate this problem in the future, I wrote a test. This test is e2e. That's why I put it in a separate package. I want to expand the tests in the future. I have a few more cases that I found. But first I want to make sure that you like this way of fixing bugs. The method is to write tests.

I also added the CONTRIBUTING.md file. I think this will help the community.

I also deleted the .fileid.db file and moved it to .gitignore. It may have ended up in the index by accident.

[rakhmaevao]

@davidbrochart, Hi, I think this is an important fix. Maybe you could take a look?

[rakhmaevao]

If you don't like my tests, maybe we can at least merge the fix itself?

plugins/kernels/fps_kernels/kernel_server/server.py

@davidbrochart

[davidbrochart]

I didn't have time to look at this yet, please bear with me.

[davidbrochart]

How does this fix the bug?

[rakhmaevao]

Hi. Sorry for being pushy. I don't want to seem like a bore. But I really want to contribute to this project.

The solution is to ensure that sockets are closed when the function is finished. Otherwise, we would exit the taskgroup without waiting for it to finish. This will allow them to be opened correctly when restarting.

My test passes on my commit. It fails on the main branch.

[davidbrochart]

Thanks for your interest in the project, but I don't think your PR fixes the bug.
I'm working on a fix, I will open a PR soon.

[rakhmaevao]

Excellent. Then I will close this PR. And I will wait for yours. I will be able to run my test on it, so that you can be sure of the fact of the correction.