The thumbnail image is created by chatGPT-4o

This content explains how to use Jupyter notebook at a compute node on HPC environment.



1. Enabling ssh between Polaris Compute Nodes

(https://docs.alcf.anl.gov/polaris/known-issues/)

  1. cd /home/[username]/.ssh
  2. ssh-keygen -t rsa
  3. create authorized_keys file and copy id_rsa.pub into authorized_keys
  4. Try setting up your .ssh/config file like this (swapping out the username placeholder with your ALCF username)
polaris-login-03:~> cat .ssh/config
Host *
User <YOUR_USERNAME>
Compression yes
Protocol 2
ControlMaster auto
ControlPath ~/.ssh/master-%r@%h:%p
ForwardAgent yes
StrictHostKeyChecking no
UserKnownHostsFile=/dev/null


2. Your /home/ directory permissions should be set to 700

chmod 700 /home/<username>


3. Confirm the following files exist in your .ssh directory and the permissions are set to the following:

  1. -rw——- (600) authorized_keys 2. -rw-r–r– (644) config 3. -rw——- (600) id_rsa 4. -rw-r–r– (644) id_rsa.pub 
chmod 600 ~/.ssh/authorized_keys
chmod 644 ~/.ssh/config
chmod 600 ~/.ssh/id_rsa
chmod 644 ~/.ssh/id_rsa.pub


4. Allocate debug node


5. Setting port forwarding using SSH to construct connection between local node and remote node.

From the login node, excute the below command:
ssh -L 8888:localhost:8888 [your_username]@[compute node]


6. Execute jupyter notebook at the compute node

jupyter notebook --no-browser --port=8888


7. Open in local browser


8. Or if you want to use jupyter notebook on Visual code,

  1. Select Kernel
  2. Select Another Kernel
  3. Exisiting Jupyter Server
  4. Copy the URL