python kernel still up after closing VSCode if extension host is on a linux machine

Environment data

• VS Code version: 1.54.3
• Jupyter Extension version (available under the Extensions sidebar): v2021.3.619093157
• Python Extension version (available under the Extensions sidebar): v2021.3.658691958
• OS (Windows | Mac | Linux distro) and version: 20.04
• Python and/or Anaconda version: Python 3.8.5 /conda 4.9.2
• Type of virtual environment used (N/A | venv | virtualenv | conda | …): conda
• Jupyter server running: Local

Expected behaviour

Close python kernel.

Actual behaviour

Python kernel’s still up.

Steps to Reproduce:

1. Open an ipython notebook
2. Type the following

import torch

a = torch.ones((1,)).cuda()

1. Close VSCode
2. Open terminal
3. nvidia-smi
4. Check if any python process is consuming GPU memory

Gif:

Logs

N/A

[Aleko2286]

The same thing is happening on my machine (Ubuntu 21.04), but it’s not limited to processes on the GPU.

I am only using interactive windows, but I suspect it’s the same issue.

Right now, when you run an interactive window, there will be 5 python processes, 2 of them being related to the kernel and 3 helpers. If you exit VS Code, it will shut down the 3 helper processes, but the 2 kernel processes will remain open. However, if you close the interactive window, VS code will shut down the kernel processes just fine (while the 3 helper processes remain). So the current work around for me is to first close the interactive window and only then close VS code.

[greazer]

@goncamateus, we’ve made some significant changes to our notebook support. Can you try this scenario in VS Code insiders to see if it still repros?