Prerequisites

Before you begin, make sure you have Jupyter installed on your personal computer. Here’s how to do it:

Installation via Python

Create a virtual environment:

$ python3 -m venv jupyter_test

Activate the virtual environment:

$ source jupyter_test/bin/activate

Install Jupyter Lab:

$ pip install jupyterlab

Note: If you are using Anaconda (version 2024.06), Jupyter is already included.

Accessing Jupyter on the Frontend ($HOME)

Part I: Starting a Job with a Jupyter Notebook Server on the vm-projet

Step 1: Log in to your account

Log in to your explor account using SSH:

$ ssh -p <port> login@193.54.9.82
login@193.54.9.82's password:

Step 2: Load the necessary modules

Load the necessary modules to run Jupyter on the vm-projet:

[login@vm-projet ~]$ module load anaconda3/2024.06
[login@vm-projet ~]$ source $HOME_ANACONDA/anaconda.rc
(base) [login@vm-projet ~]$

Step 3: Launch Jupyter Notebook

Launch Jupyter Notebook with the following command:

(base) [login@vm-projet ~]$ jupyter notebook --no-browser --port=8888 --ip=0.0.0.0

Example output:

[I 2025-03-06 10:52:44.630 ServerApp] Jupyter Server 2.14.1 is running at:
[I 2025-03-06 10:52:44.630 ServerApp] http://vm-projet:8888/tree?token=856bf559f7b34d11a5c87e4e755b7221d4757ef001bf94b6

Note the provided URL, as you will need it to connect to the server.

Part II: Opening a Notebook through an SSH Tunnel

Step 1: Open the SSH Tunnel

Open an SSH tunnel in a shell session on your workstation:

$ ssh -fNL 8888:localhost:8888 -p <port> login@193.54.9.82
login@193.54.9.82's password:

The first port (8888) corresponds to your local machine. You can adjust it if necessary.
The second port (8888) is that of the server, as specified when launching Jupyter Notebook on the vm-projet.

Step 2: Launch the Notebook

Activate your virtual environment and launch Jupyter:

$ source jupyter_test/bin/activate
(jupyter_test) user@work-station:~$ jupyter notebook

Step 3: Access the Notebook

In your browser, open the URL provided by the notebook server, replacing the vm-projet name with localhost:

Replace:

http://vm-projet:8888/tree?token=856bf559f7b34d11a5c87e4e755b7221d4757ef001bf94b6

With:

http://localhost:8888/tree?token=856bf559f7b34d11a5c87e4e755b7221d4757ef001bf94b6

You should now be connected to your Jupyter notebook running on the vm-projet.

Accessing Jupyter on Compute Nodes

Part I: Starting a Job with a Jupyter Notebook Server on a Compute Node

Step 1: Log in to your account

Log in to your explor account using SSH:

$ ssh -p <port> login@193.54.9.82
login@193.54.9.82's password:

Step 2: Request a Reservation

Request a reservation for an interactive connection to a node, for example for a GPU node:

[login@vm-projet ~]$ salloc -A projet -N1 -p gpu -w noeud -t 1:00:00 srun --pty bash

Example output:

salloc: Pending job allocation 1344365
salloc: job 1344365 has been allocated resources
salloc: Granted job allocation 1344365
salloc: Waiting for resource configuration
salloc: Nodes gpf01 are ready for job

Step 3: Load the necessary modules

Load the necessary modules to run Jupyter on the node:

[login@node ~]$ module load anaconda3/2024.06
[login@node ~]$ source $HOME_ANACONDA/anaconda.rc
(base) [login@node ~]$

Step 4: Launch Jupyter Notebook

Launch Jupyter Notebook with the following command:

(base) [login@node ~]$ jupyter notebook --no-browser --port=8888 --ip=0.0.0.0

Example output:

[I 2024-11-22 18:50:20.263 ServerApp] Jupyter Server 2.14.1 is running at:
[I 2024-11-22 18:50:20.263 ServerApp] http://node:8888/tree?token=0cd965475476ce5b0a5235c4f6fc5f0c82123eb0e8c5b354

Note the provided URL, as you will need it to connect to the server, as well as the name of the compute node.

Part II: Opening a Notebook through an SSH Tunnel

Step 1: Open the SSH Tunnel

Open an SSH tunnel in a shell session on your workstation:

$ ssh -fNL 8888:node:8888 -p <port> login@193.54.9.82
login@193.54.9.82's password:

The first port (8888) corresponds to your local machine. You can adjust it if necessary.
The second port (8888) is that of the server, as specified when launching Jupyter Notebook.

Step 2: Launch the Notebook

Activate your virtual environment and launch Jupyter Lab:

$ source jupyter_test/bin/activate
(jupyter_test) user@work-station:~$ jupyter notebook

Step 3: Access the Notebook

In your browser, open the URL provided by the notebook server, replacing the node name with localhost:

Replace:

http://node:8888/tree?token=0cd965475476ce5b0a5235c4f6fc5f0c82123eb0e8c5b354

With:

http://localhost:8888/tree?token=0cd965475476ce5b0a5235c4f6fc5f0c82123eb0e8c5b354

You should now be connected to your Jupyter notebook running on the compute node.