[Linux] "Connect to New Server" option disappears and requires manual extension folder deletion to restore

[Lyra-Brown]

Description

The "Connect to New Server" option disappears from both the Kernel Picker and Command Palette after some time or after VS Code updates. Uninstalling and reinstalling the extension via the VS Code UI does not fix the issue. The only way to restore the functionality is to manually delete the extension folder from the terminal.

Environment

VS Code version: 1.108.0
Colab extension version: v0.1.6
Jupyter extension version: v2025.9.1
OS (Linux | Mac | Windows): Linux (Ubuntu24.04)

Repro Steps

1. Install the Google Colab extension on Linux.
2. Use it successfully to connect to a runtime.
3. Close VS Code and wait for some time (or after an extension auto-update occurs).
4. Open a .ipynb file.
5. Try to search for Google Colab: Connect to New Server in the Command Palette or Kernel Picker.
6. The option is missing.

Expected Behaviour

The "Connect to New Server" option should always be available when a .ipynb file is open.

Actual Behaviour

The option disappears completely. Even reinstalling the extension via the VS Code GUI does not bring it back. It seems like old extension files or cache are corrupted and not being cleaned up properly on Linux.

Additional context

Workaround found: I found that manually deleting the extension folder fixes the issue temporarily:
Bash

rm -rf ~/.vscode/extensions/google.colab*

After running this command and reinstalling the extension, the menu reappears.

Logs

Logs from Colab in the Output panel:

(No logs available as the command does not appear)

P.S.

This is my first time submitting an issue on GitHub, so please let me know if there are any mistakes or missing information. Also, I am using a translation tool, so I apologize for any unnatural English.

[Lyra-Brown]

Screencast.from.2026-01-12.12-33-39.webm

[kevineger]

Thanks for the report @Lyra-Brown ! Let's try and figure out what's going on here. I've not seen other reports of this issue and can't repro it myself. I'm hoping you can help me collect a bit more info!

Could you enable trace logging (screenshot below) in the VS Code settings for Colab. There's a similar setting for the Jupyter extension.

Once enabled, repro the issue and paste the logs back here. Thanks a bunch!

[Lyra-Brown]

Hi @kevineger,

Apologies for the delay in getting back to you.

I enabled trace logging and reproduced the issue as requested.
Just to clarify, these logs capture the sequence from the very beginning of VS Code startup, with the Colab extension already installed and active.
Here are the logs from the output channel when the connection fails (it seems to return an empty list [])

[2026-01-17T13:34:06.394Z] [Info] Visual Studio Code: 1.108.0
[2026-01-17T13:34:06.394Z] [Info] Remote: N/A
[2026-01-17T13:34:06.394Z] [Info] App Host: desktop
[2026-01-17T13:34:06.394Z] [Info] Jupyter extension version: 2025.9.1
[2026-01-17T13:34:06.401Z] [Trace] Kr.provideJupyterServers called with
                                   {
                                     "a": false,
                                     "b": null
                                   }
[2026-01-17T13:34:06.401Z] [Trace] Kr.provideJupyterServers returned (sync)
                                   []
[2026-01-17T13:34:08.133Z] [Trace] Kr.provideJupyterServers called with
                                   {
                                     "a": false,
                                     "b": null
                                   }
[2026-01-17T13:34:08.133Z] [Trace] Kr.provideJupyterServers returned a Promise
[2026-01-17T13:34:08.152Z] [Trace] Kr.provideJupyterServers Promise resolved, args and error:
                                   {
                                     "a": false,
                                     "b": null
                                   }
                                   []

Let me know if you need any further information. Thanks!

Screencast.from.2026-01-17.22-34-22.webm

[kevineger]

Sorry, I have no idea what the problem is. Everything about your logs is normal (when you have no servers, provideJupyterServers returns []).

Next places to check:

• Do you see anything in Extension Host or any of the other output channels?
• It's as if the Jupyter extension is failing to invoke us. Could you enable trace logging for Jupyter and pass that along?