Tutorial#
Launching Jupyter Lab on a Remote Host’s Login Node#
For instance, here is how to start a jupyter lab server running on a login node of a remote host:
❯ jupyter-forward username@supersystem.univ.edu
Launching Jupyter Lab on a Remote Host’s Compute Node#
To launch jupyter lab
on a remote host’s compute node, the user needs to specify the --launch-command
option. The launch command is meant to submit a job on the remote host’s queueing system. Once the job is up and running:
jupyter lab
is launched on the compute nodethe session is port-forwarded to the user’s local machine.
and a link to the jupyter lab session is opened in the local machine’s browser.
Here is an example showing how to launch jupyter lab on Cheyenne’s compute node.
Note
Cheyenne uses the PBS job scheduler
In the following command, we tell the remote host to use the qsub
command to submit a batch job to PBS. In addition, we specify the resources, and set job attributes:
The queue via
-q regular
The project account via
-A AABD1115
The resources: 1 node, 36 cpus via
-l select=1:ncpus=36,walltime=00:05:00
❯ jupyter-forward username@supersystem.univ.edu \
--launch-command "qsub -q regular -l select=1:ncpus=36,walltime=00:05:00 -A AABD1115"
Here’s an example showing how to launch jupyter lab on a remote system that uses Slurm job scheduler,
❯ jupyter-forward username@supersystem.univ.edu \
--launch-command "salloc -A AABD1115 -t 00:05:00 srun"
Launching Jupyter Lab on a Remote Host without port-forwarding#
If and/or when the remote host has nodes that can be accessed via a public IP address, jupyter-forward
provides a --no-port-forwarding
option which disables SSH tunneling. When the --no-port-forwarding
option is active, the Jupyter Lab session is accessible at https:\\<public-ip-address:port>
instead of https:\\<localhost:port>
in the local machine’s browser.
❯ jupyter-forward username@supersystem.univ.edu --no-port-forwarding