Running Jupyter Notebook on Remote Server with Slurm (Gypsum)

There are two ways to connect to gypsum server at gypsum.cs.umass.edu.

With no jump host

Log into gypsum on one terminal:
ssh username@gypsum.cs.umass.edu
Then start an interactive job (requesting a node with cpu/gpu):
srun -p m40-short -t 0-01:00 --pty bash
Or to add arguments to request additional resources, such as using
srun -p m40-short -t 0-01:00 --gres=gpu:1 --pty bash to request a gpu. Note: after the node has been request, you should be able to see the node name on your prompt username@node001, or you could use ifconfig to find the ip address to use in place of the node name as well. With gpu requested, the node name will not be started with node001, but maybe node012, be sure to look for the node name before tunneling.
Last step, launch a jupyter notebook session with no browser:
jupyter notebook --no-browser --port 9099
To run the jupyter notebook session on the background, tmux or screen could be used.

Start a ssh tunneling session to tunnel through login node and the interactive job node:
ssh -t -t username@gypsum.cs.umass.edu -L 7077:localhost:8088 ssh node001 -L 8088:localhost:9099
Note: this command forwards the port 9099 on node001 to the port 8088 on the gypsum login node, and then forward it to the port 7077 on the local host.
Finally, open a web browser, enter the address, and have fun:
localhost:7077

[Different]Log into gypsum on one terminal:
ssh -J username@gypsum-gateway.cs.umass.edu username@gypsum.cs.umass.edu
Then start an interactive job (requesting a node with cpu/gpu):
srun -p m40-short -t 0-01:00 --pty bash
Or to add arguments to request additional resources, such as using
srun -p m40-short -t 0-01:00 --gres=gpu:1 --pty bash to request a gpu. Note: after the node has been request, you should be able to see the node name on your prompt username@node001, or you could use ifconfig to find the ip address to use in place of the node name as well. With gpu requested, the node name will not be started with node001, but maybe node012, be sure to look for the node name before tunneling.
Last step, launch a jupyter notebook session with no browser:
jupyter notebook --no-browser --port 9099
To run the jupyter notebook session on the background, tmux or screen could be used.

[Different] Start a ssh tunneling session to tunnel through login node and the interactive job node:
ssh -t -t -J username@gypsum-gateway.cs.umass.edu username@gypsum.cs.umass.edu -L 7077:localhost:8088 ssh node001 -L 8088:localhost:9099
Note: this command forwards the port 9099 on node001 to the port 8088 on the gypsum login node, and then forward it to the port 7077 on the local host.
Finally, open a web browser, enter the address, and have fun:
localhost:7077