Clearing Previous Kernels from JupyterLab History?

JupyterLab
1

I’ve been running into an issue while authoring an extension where Jupyterlab will attempt to connect to kernels which don’t exist anymore, which has been slowing down startup times. If its helpful, I get errors like this:

[W 2026-06-03 23:16:35.510 ServerApp] wrote error: 'Kernel does not exist: 363dd351-cb08-4564-b83b-f55aa025e09e'
    Traceback (most recent call last):
      File "/Users/.../.virtualenvs/nod-ryhi/lib/python3.14/site-packages/tornado/web.py", line 1880, in _execute
        result = await result
                 ^^^^^^^^^^^^
      File "/Users/.../.virtualenvs/nod-ryhi/lib/python3.14/site-packages/jupyter_server/auth/decorator.py", line 73, in inner
        return await out
               ^^^^^^^^^
      File "/Users/.../.virtualenvs/nod-ryhi/lib/python3.14/site-packages/jupyter_server/services/kernels/handlers.py", line 75, in get
        model = await ensure_async(km.kernel_model(kernel_id))
                                   ~~~~~~~~~~~~~~~^^^^^^^^^^^
      File "/Users/.../.virtualenvs/nod-ryhi/lib/python3.14/site-packages/jupyter_server/services/kernels/kernelmanager.py", line 513, in kernel_model
        self._check_kernel_id(kernel_id)
        ~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^
      File "/Users/.../.virtualenvs/nod-ryhi/lib/python3.14/site-packages/jupyter_server/services/kernels/kernelmanager.py", line 544, in _check_kernel_id
        raise web.HTTPError(404, "Kernel does not exist: %s" % kernel_id)
    tornado.web.HTTPError: HTTP 404: Not Found (Kernel does not exist: 363dd351-cb08-4564-b83b-f55aa025e09e)

It seems like when JupyterLab restarts, it looks for previous connection files and tries to start them up again. How can I prevent this from happening? Is there a way to disable this within JupyterLab? It seems like this is originating from JupyterLab (or at least the client) because the Jupyter server reports the associated 404 error:

[W 2026-06-03 23:23:33.145 ServerApp] 404 GET /api/kernels/25994376-79eb-4b7e-a89b-a45a841e6521?1780554213141 (``596974b570b24182a540183b64f4a51a@127.0.0.1``) 0.84ms referer=http://localhost:8888/lab/workspaces/auto-t/tree/.nod/connection/moduleC.ipynb.

Any pointers would be really helpful! Happy to provide more context about my setup, but I figured this was the crucial info. Thank you so much!

2

I actually figured out what was going wrong – jupyter was not keeping a history at all, but I was generating and then quickly deleting notebook files that jupyter could see and was trying to load kernel sessions for. I’ll leave this in case its useful to anyone else.

3

Hi Erawn, could you please explain a little bit more your solution? I had same issue, my job running automatically, so I put extra steps (restore and reactivate customized kernel) before trigger the job. Could you please share your example? Thank you!

4

So I think I finally figured out the issue, which is that other open notebook tabs from previous sessions were still making requests to kernels which no longer existed. I wasn’t able to figure out how to automatically close those tabs, but I set the kernel to not automatically restart if it died (which was appropriate for my use case anyways) and that fixed the problem for me. Hope that helps!

5

I encounter the same problem.
Old JupyterLab tabs from previous sessions were still open in my browser and kept sending requests to kernels that no longer existed after a Windows 11 update. This corrupted the JupyterLab workspace.

In my case, the side effects are very visible:

  • JupyterLab reopens my notebook from the wrong working directory (~/notebooks instead of ~/mon_projet/notebooks),
  • the active kernel does not appear in “Change Kernel”,
  • notebook titles show “(2)” or “(3)”,
  • and all relative paths break: I must use ../../mon_projet/... instead of the correct ../mon_projet/....

Closing all JupyterLab tabs, quitting the browser completely, killing Jupyter, and restarting from the correct project directory did not fully fix the issue — the wrong workspace keeps coming back.

Any idea how to force JupyterLab to forget old tabs and rebuild a clean workspace?

6

In my very basic understanding Jupyterlab will Autosave your documents unless you untick via Settings-AutoSave Documents.

I suspect there could be other solitons Z2JK level. However, no idea how it would be possible right now.