Kernel picker does not show existing Jupyter kernel

[jeckst]

Applies To

• Notebooks (.ipynb files)
• Interactive Window and/or Cell Scripts (.py files with #%% markers)

What happened?

After opening a notebook, or workspace with some notebooks opened, the kernel picker is in the "Select kernel" state, clicking on it reveals menu offering only "python environments" and "existing jupyter server" options, so the Jupyter kernel option is missing.

This issue keeps appearing here and there and I've also seen similar issues being opened by other users in the past. Reloading window etc usually solves the issue.

Running jupyter kernelspec list in a terminal confirms that the expected kernel is there, although it is located in /home/jeckst/snap/code/179/.local/share/jupyter/kernels/, and the output does not list the original in /home/jeckst/.local/share/jupyter/kernels/ (which is still there). Does VS Code copy/move kernels around instead of using them as is?

The OS is Ubuntu 24.04.1 LTS

VS Code Version

Version: 1.96.2 Commit: fabdb6a30b49f79a7aba0f2ad9df9b399473380f Date: 2024-12-19T10:22:47.216Z Electron: 32.2.6 ElectronBuildId: 10629634 Chromium: 128.0.6613.186 Node.js: 20.18.1 V8: 12.8.374.38-electron.0 OS: Linux x64 6.8.0-51-generic snap

Jupyter Extension Version

2024.11.2024102401

Jupyter logs

Visual Studio Code (1.96.2, undefined, desktop)
Jupyter Extension Version: 2024.11.2024102401.
Python Extension Version: 2024.22.2.
Pylance Extension Version: 2024.12.1.
Platform: linux (x64).
Temp Storage folder ~/.config/Code/User/globalStorage/ms-toolsai.jupyter/version-2024.11.2024102401
Workspace folder ~/Code/repo_dir, Home = /home/jeckst
10:02:08.560 [debug] Start refreshing Kernel Picker (1736499728560)
10:02:08.678 [trace] Search for KernelSpecs in Interpreter /usr/bin/python3
10:02:08.678 [trace] Search for KernelSpecs in Interpreter /bin/python3.12
10:02:08.678 [trace] Search for KernelSpecs in Interpreter /usr/bin/python3.12
10:02:08.679 [trace] Search for KernelSpecs in Interpreter /bin/python3
10:02:08.681 [debug] Get Custom Env Variables, Class name = Em, completed in 122ms, has a truthy return value, Arg 1: undefined, Arg 2: "RunPythonCode"
10:02:08.681 [debug] Jupyter Paths /kernels: 
10:02:08.681 [debug] Kernel Spec Root Paths, /usr/share/jupyter/kernels, /usr/local/share/jupyter/kernels, ~/.local/share/jupyter/kernels
10:02:09.243 [debug] End refreshing Kernel Picker (1736499728560)
10:02:09.270 [trace] Conda file is conda
10:02:10.617 [debug] No interpreter for Pylance for Notebook URI "~/Code/repo_dir/my_ntbk.ipynb"
10:02:10.992 [debug] No interpreter for Pylance for Notebook URI "~/Code/repo_dir/my_ntbk.ipynb"
10:10:27.268 [trace] No controller, hence notebook communications cannot be initialized for editor ~/Code/repo_dir/my_ntbk.ipynb

Coding Language and Runtime Version

Python 3.12

Language Extension Version (if applicable)

No response

Anaconda Version (if applicable)

No response

Running Jupyter locally or remotely?

Local

[DonJayamanne]

Please can you run jupyter kernelspec list --debug in the terminal and share the full output.

/home/jeckst/.local/share/jupyter/kernels/ (which is still there).

Thats interesting, thats generally the expected location.

Does VS Code copy/move kernels around instead of using them as is?

No.

I think you may have changed some environment variables or the like (or installed a different Python installation) thats causing the kernels to be located in a different directory.

[jeckst]

Please can you run jupyter kernelspec list --debug in the terminal and share the full output.

As I was writing this response, I came across something fishy (excuse me if some parts of this post are a bit out of order).
Here is the output when executed in terminal inside VSCode, using source /home/jeckst/Code/repo-dir/venv/bin/activate environment:

[ListKernelSpecs] Searching ['/home/jeckst/Code/repo-dir/venv/etc/jupyter', '/home/jeckst/.jupyter', '/usr/local/etc/jupyter', '/etc/jupyter'] for config files
[ListKernelSpecs] Looking for jupyter_config in /etc/jupyter
[ListKernelSpecs] Looking for jupyter_config in /usr/local/etc/jupyter
[ListKernelSpecs] Looking for jupyter_config in /home/jeckst/.jupyter
[ListKernelSpecs] Looking for jupyter_config in /home/jeckst/Code/repo-dir/venv/etc/jupyter
[ListKernelSpecs] Found kernel python3 in /home/jeckst/Code/repo-dir/venv/share/jupyter/kernels
[ListKernelSpecs] Found kernel custom in /home/jeckst/snap/code/179/.local/share/jupyter/kernels
[ListKernelSpecs] Found kernel python3 in /home/jeckst/Code/repo-dir/venv/share/jupyter/kernels
[ListKernelSpecs] Found kernel custom in /home/jeckst/snap/code/179/.local/share/jupyter/kernels
Available kernels:
  python3    /home/jeckst/Code/repo-dir/venv/share/jupyter/kernels/python3
  custom        /home/jeckst/snap/code/179/.local/share/jupyter/kernels/custom

But here it is executed outside VSCode using the same environment:

[ListKernelSpecs] Searching ['/home/jeckst/Code/repo-dir/venv/etc/jupyter', '/home/jeckst/.jupyter', '/usr/local/etc/jupyter', '/etc/jupyter'] for config files
[ListKernelSpecs] Looking for jupyter_config in /etc/jupyter
[ListKernelSpecs] Looking for jupyter_config in /usr/local/etc/jupyter
[ListKernelSpecs] Looking for jupyter_config in /home/jeckst/.jupyter
[ListKernelSpecs] Looking for jupyter_config in /home/jeckst/Code/repo-dir/venv/etc/jupyter
[ListKernelSpecs] Found kernel python3 in /home/jeckst/Code/repo-dir/venv/share/jupyter/kernels
[ListKernelSpecs] Found kernel custom in /home/jeckst/.local/share/jupyter/kernels
[ListKernelSpecs] Found kernel general in /home/jeckst/.local/share/jupyter/kernels
[ListKernelSpecs] Found kernel python3 in /home/jeckst/Code/repo-dir/venv/share/jupyter/kernels
[ListKernelSpecs] Found kernel custom in /home/jeckst/.local/share/jupyter/kernels
[ListKernelSpecs] Found kernel general in /home/jeckst/.local/share/jupyter/kernels
Available kernels:
  python3    /home/jeckst/Code/repo-dir/venv/share/jupyter/kernels/python3
  general    /home/jeckst/.local/share/jupyter/kernels/general
  custom        /home/jeckst/.local/share/jupyter/kernels/custom

That's completely different! Could this be because it is a snap verison?

Thats interesting, thats generally the expected location.

Yeah, and as I've said, it's still there.

$ cat ~/.local/share/jupyter/kernels/custom/kernel.json 
{
 "argv": [
  "/home/jeckst/Code/repo-dir/venv/bin/python3.12",
  "-Xfrozen_modules=off",
  "-m",
  "ipykernel_launcher",
  "-f",
  "{connection_file}"
 ],
 "display_name": "custom",
 "language": "python",
 "metadata": {
  "debugger": true
 }

No.

Do you have any explanation for the kernel located in the snap folder (/home/jeckst/snap/code/179/.local/share/jupyter/kernels/) then? I am honestly starting to be very suspicious
of anything snap-related. Could it be that this is some snap-version-specific workaround because of the snap's
sandboxing? Especially seeing the difference between the "in VSCode" and "out of VSCode" outputs above.

I think you may have changed some environment variables or the like

Not that I know of, but that's obviously a poor guarantee. Would you have any leads which env vars to check?

[DonJayamanne]

Unfortunately I’m not familiar with snap
Out of corridor why do you need to use these kernels? Can you not just select the python environment instead?

[jeckst]

why do you need to use these kernels? Can you not just select the python environment instead?

The kernel specs enable defining env vars and are consistent for other Jupyter IDEs (AFAIK JupyterLab or the native Jupyter do not allow to just use Python venvs as kernels). As I don't have any env vars defined for this project, I guess I could do it in this case.

Unfortunately I’m not familiar with snap

I guess something I can do on my side is to try and install non-snap version, if that's something that Ubuntu (or VS Code) still supports. In any case, do you have any idea with who or where to address this issue if it turns out to be specific for the snap version?

[DonJayamanne]

At this stage the only solution I can offer odd to create the kernel in the following location

/home/jeckst/.local/share/jupyter/kernels/

As you mentioned those directors work in vscode
You can just copy the kernel spec folders into that directory and keep using your env variables

Optionally you can define env variables using ‘.env’ files in your workspace folders

Given this is specific to snap folders, and it’s not something we control I’m closing this issue for now, hopefully you can copy the lenses to get around this issue

Let me know if this doesn’t work for you

[jeckst]

There are some typos or I might be reading it wrong, but are you suggesting to create the kernels at
/home/jeckst/.local/share/jupyter/kernels/?
Because I've written above multiple times, that I do have them there, that is where I've created them, see e.g.

$ cat ~/.local/share/jupyter/kernels/custom/kernel.json 
{
 "argv": [
  "/home/jeckst/Code/repo-dir/venv/bin/python3.12",
  "-Xfrozen_modules=off",
  "-m",
  "ipykernel_launcher",
  "-f",
  "{connection_file}"
 ],
 "display_name": "custom",
 "language": "python",
 "metadata": {
  "debugger": true
 }

above. I haven't created the ones at /home/jeckst/snap/code/179/.local/share/jupyter/kernels/, either VS Code or the jupyter extension did.