Vscode-jupyter: Using remote SSH and using Jupyter with Allow Unauthorized Remote Connection

[lgonzalezsa]

Bug: Notebook Editor, Interactive Window, Editor cells

Steps to cause the bug to occur

1. Use remote SSH and connect
2. Enable "Allow Unauthorized Remote Connection" setting in the remote SSH server
3. Connect to JupyterHub via existing URI
4. Open notebook in SSH server
5. Try to run a code cell

Actual behavior

Error: Failed to connect to remote Jupyter notebook.
Check that the Jupyter Server URI setting has a valid running server specified.
https://my-host
TypeError: request to https://my-host?5187757 failed, reason: unable to verify the first certificate

Expected behavior

Be able to connect to JupyterHub as same as I can when I do directly from my laptop.
Need to use compute power of my-host

Your Jupyter and/or Python environment

Please provide as much info as you readily know

• Jupyter server running: Remote
• Extension version: 2020.1.58038
• VS Code version: 1.41.1
• Setting python.jediEnabled: true
• Python and/or Anaconda version: 3.7.3
• OS: Linux (Ubuntu):19.10
• Virtual environment: conda

Developer Tools Console Output

Microsoft Data Science for VS Code Engineering Team: @rchiodo, @IanMatthewHuff, @DavidKutu, @DonJayamanne, @greazer

[IanMatthewHuff]

@lgonzalezsa . Thanks for the report on this. So you are using remoteSSH features of VSCode to connect to the JupyterHub machine? If you are doing that, then our extension would actually be running on the remote JupyterHub machine and you could treat your scenarios as local scenarios without using the remote connection URI.

But if you are using JupyterHub as your big compute machine I would think that the easiest way might just be to do this:
https://blog.jupyter.org/connect-to-a-jupyterhub-from-visual-studio-code-ed7ed3a31bcb
You don't have to use remote-SSH in this scenario, just have a notebook server started up on the JupyterHub machine and connect to it directly. Does this work for your usage?

[lgonzalezsa]

Thanks @IanMatthewHuff for your response.
In my scenario, I am using github and I have all my notebooks in a repo, depending on the type of workload, need to access server resources like filesystem or just the task could be quite heavy and long for few hours. Using @dask as well; then connecting remotely as needed from my laptop is a really nice Jupyter feature, then have similar capability with VSCode will be really useful.
Also, open to other suggestions/recommendations.

[lgonzalezsa]

I was doing additional testing based on @IanMatthewHuff suggestion to treat my scenario as local and is running fine but each time I connect is a fresh kernel meaning need to start my entire notebook again.
I think was doing wrong interpretation of VSCode capability to connect to existing URI from my remote JupyterHub server; trying to mimic the web client with VSCode to connect to existing notebook and get ipython capabilities for lint/debug my code.
Even if using remoteSSH feature I was able to connect to the URL allowing the unauthorized remote connection, the expected result was not to connect to an existing already running kernel if not mistaken.
Also reading https://github.com/microsoft/vscode-python/issues/7014#issuecomment-566698680, seems that is what I need, continue researching, if you can point me in the right direction greatly appreciated!
thanks for VSCode and really nice features!

[IanMatthewHuff]

Hey @lgonzalezsa . So that issue that you linked to was our kernel picking support that we added in our last release. With that support you should be able to connect to a remote server (not using remoteSSH, just using the "Python: Specify local or remote Jupyter server for connections") and when the notebook or interactive window opens using the kernel picker in the top right to select a kernel that is already running and active.

[lgonzalezsa]

Yes!, was so happy to see that, testing it right now. you guys rock! amazing tools/ collaboration