Writing notebook server cookie secret to /home/wlaw/.local/share/jupyter/runtime/notebook_cookie_secret The log file will output something like this: Check the log output to find out the IP we need to use to create an SSH tunnel: Now we need to send this as a job to Slurm.Ĥ) Now, you can check that your job is running:ĥ) Wait 10 minutes after your job starts for Jupyter to start. This tells Slurm to launch a Jupyter Lab server on the node with your requested resources, over port 8888. Note: You can activate a custom Jupyter conda environment if you want to use custom packages by changing the path for the “source activate /path/here” section of the above code. Or, provide an alternate path for your log output.

Replace the part of your #SBATCH -output=home//jupyter.log with your SUNetID. Source activate /share/sw/open/anaconda/3 Paste the following text into your sbatch script, and save the file. A tunnel must be created as you cannot directly SSH to Slurm nodes on Nero. Once running, we are going to connect to the jupyterlab instance with SSH port forwarding from our local laptop. To summarize: We are creating a slurm job that runs jupyterlab on a Slurm node, for up to 2 days (max is 7). That way, your jupyter lab server gets placed on a Slurm node with enough resources to host it. With Nero On-Prem, you’ll need to submit this as a Slurm job. This documentation specifically covers this process. This enables you to run your Jupyter server on a Slurm node with GPUs. Several scenarios come up where you might want to run your Jupyter server via Slurm.

Jupyter notebook ssh tunnel