Using +NEW freezes Jupyterlab completely

[michaelaye]

Describe the issue:
After pressing +NEW button to start a cluster:

• The launching never finishes, seemingly stalls.
• I cannot CTRL-C the Jupyter lab server in the terminal any more.
• I cannot open existing notebooks anymore.
• I do not get any log related to having tried to launch a dask cluster in the terminal.
• I still can get out of the tmux (the detach hotkey still works, so that i can kill the session that way.

As usual, this "WORKED BEFORE" (TM) on this machine, but I hadn't used it for several weeks. :(

Minimal Complete Verifiable Example:

• create a new env with python 3.11
• mamba install jupyterlab dask_labextension
• run jupyter lab . in a folder with work projects (sub folders with notebooks in them)

Anything else we need to know?:

I saw that a new py311 env automatically gets bokeh > 3, so I mamba-installed bokeh<3 to make dashboards work.

Environment:

• Dask version: 2023.1.0 (labextension 6.0.0)
• Python version: 3.11
• Operating System: Linux CentOS 7.5.1804 (core)
• Install method (conda, pip, source): conda

[michaelaye]

Phew, found the issue. was caused by temp files redirected to non-existing node: dask/distributed#7491

[roxyboy]

I think I'm getting a similar issue with the following errors showing up on Terminal:

[W 2023-02-25 22:10:57.992 ServerApp] wrote error: 'Cluster failed to start: Scheduler failed to start.'
[E 2023-02-25 22:10:57.996 ServerApp] {
      "Host": "localhost:8000",
      "Accept": "*/*",
      "Referer": "http://localhost:8000/lab?",
      "User-Agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/110.0.0.0 Safari/537.36"
    }
[E 2023-02-25 22:10:57.997 ServerApp] 500 PUT /dask/clusters?1677381055482 (127.0.0.1) 1691.83ms referer=http://localhost:8000/lab?

[roxyboy]

The issue may have to do with the new version of Notebook..?