2

I'm trying to set up remote access (with VSCode) to the GCP VM that's setup with Notebooks AI. However, when I ssh into the VM I don't have write permissions for /home/jupyter so cannot edit any of the notebook files.

I have tried both gcloud compute ssh and setting up local aliases with gcloud compute config-ssh.

My best guess is that the users are different. It looks like the terminal on JupyterLab is logged in as jupyter@[instance...] while when I ssh in its myname@[instance...]. Checking permissions of /home/jupyter/, it's owned by user jupyter of group jupyter. I also tried adding users to the jupyter group with sudo usermod -a -G but that didn't do the trick. When I try to ssh in as jupyter@[instance...] from anywhere else I get permission denied (public key).

I can edit files once logged in if I use sudo vim ..., but that won't help for VS code.

EDIT: a partial solution is to open up permissions using sudo chmod 777 /home/jupyter/*. However, that's probably a hackish, unsafe way to do it. Moreover, it only works on existing files -- new files will still only be writable by whichever user created them.

Craig
  • 835
  • 1
  • 7
  • 21

1 Answers1

4

To SSH into the notebook instance as the “jupyter” user, an SSH key should be generated for that user and be added to the notebook VM instance. Also, please make sure that the notebook instance VM has the appropriate firewall rule to allow the SSH connection. The following are the steps that would create an SSH connection to the “jupyter” user which has the write permissions.

  1. Run the following commands on the local machine to generate the required SSH key:

     ssh-keygen -t rsa -f ~/.ssh/jupyter-ssh-key -C jupyter

    “jupyter-ssh-key” → Name of the pair of public and private keys (Public key: jupyter-ssh-key.pub, Private key: jupyter-ssh-key)

    “jupyter” → User in the VM that we are trying to connect to

     chmod 400 ~/.ssh/jupyter-ssh-key

  2. In the Compute Engine console, edit the VM settings to add the contents of the generated SSH public key. Detailed instructions can be found here.

  3. Initiate the SSH connection from the local machine to the notebook VM:

     ssh -i ~/.ssh/jupyter-ssh-key jupyter@<external-ip-of-notebook-vm-instance>

    If the SSH connection succeeds, the same can be followed in VSCode.

  4. In VSCode, select the “Remote-SSH: Connect to Host” option from the command palette. Enter the above ssh -i command to add the notebook VM instance as a recognized host. A new VSCode window will appear where we have been logged in as the “jupyter” user.

Pooja S
  • 550
  • 2
  • 9
  • Great, that worked extremely well. Thanks for helping out. (One nit: I think you forgot to add the link for "detailed instructions" although they were easy to find) – Craig Sep 15 '21 at 03:50
  • This no longer seems to work -- I now have an issue adding the SSH keys to VMs. When adding in the console, I get the error below. The format does not match the help page here (https://cloud.google.com/compute/docs/connect/add-ssh-keys#gcloud_3), but, either way, neither format works. Any ideas? "Invalid key. Required format: " – Craig Jan 02 '22 at 01:04
  • Does the code need to be a certain number of bits? Should their be carriage returns? I've tried just about every possible format for the keyblob without luck – Craig Jan 02 '22 at 01:05
  • Argh, silly mistake, but now fixed -- need to copy the value in ~/.ssh/jupyter-ssh-key.pub, not ~/.ssh/jupyter-ssh-key for it to work. – Craig Jan 02 '22 at 15:28