Remotely Using Onyx#

If you want to use Jupyter notebooks on Onyx nodes, there are a few extra steps you need to do. There is a video at the end of this page walking through the process.

Install Anaconda#

Connect to the Onyx head node. From the command line, run:

$ ssh

On Windows, you can use MobaXterm.

Follow the instructions in Software Installation to install the software. I recommend accepting all defaults, which will set up your shell to automatically activate Conda next time you log in. Due to network file system performance, I recommend using Miniconda instead of the full Anaconda platform, so we only install the packages we need.

Log out of Onyx.

Preparing Off-Campus Access#

Directly connecting to — called the ‘head node’ — is fine for transferring files, installing the software, and other light work, but not for substantial computing. For actually doing work, we need to connect to one of the ‘Onyx nodes’, individual computers in the Boise State CS department computing system. These computers are numbered with 2–3 digit numbers and have names like

Only the Onyx head node accepts connections from outside the Boise State network. To connect to one of the normal Onyx nodes, you need to be on the Boise State network. There are a few ways to do this from the outside.

OpenSSH ProxyJump#

Set up SSH to use a ‘jump host’ (sometimes called a ‘bastion host’). OpenSSH has direct support for jump hosts, and this is probably the easiest way to set things up on macOS and Linux. Recent versions of Windows (20H2 and later) also include OpenSSH with support for jump hosts as an optional package.

If the directory ~/.ssh (a directory named .ssh in your home directory) does not yet exist, create it:

mkdir ~/.ssh

To set up jump host connections, modify the file ~/.ssh/config (the file config in the .ssh subdirectory of your home directory) to contain the following lines:

Host onyxnode*
    User username

    User username

If the file does not yet exist, create it. The User lines make SSH automatically fill in your username when you connect to Onyx in the future, so it’s more convenient to connect.


The MobaXterm SSH client for Windows includes built-in support for jump hosts. Unfortunately, you will need to create a new session profile for each Onyx node you try to connect to.

To use a jump host, in the MobaXterm “Session” window, under “Network Settings”, click “SSH gateway”. In the window that opens, enter your Onyx connection information: the host, and your Boise State username.

Then, when you connect with this profile, it will use as your jump host and allow you to directly connect to an Onyx node.

Boise State VPN#

You can also use the Boise State VPN. For this, you will need to set up Duo Security for the two-factor authentication. I have requested VPN access for all students in the class.

Connecting to Nodes#

On Linux and macOS, once you on the university network (either physically or by VPN) or have set up your jump host, you can connect to a node from a terminal:

$ ssh

If you are using ProxyJump, SSH will prompt you for your password twice — once to set up the proxy, and again to log in.

On Windows with MobaXterm, enter the node name as the host to connect to when creating a new SSH session.

Once you are logged in to a node, you can run uptime to see how heavily-loaded it is, and w to see if anyone else is logged in. An idle node will have a load average of close to 0 when you run uptime.

If the node is in use, log out and try a different node. Consult the beginning-of-the-year e-mail from Ben for a list of available nodes.

All nodes (and the head node) share the same home directory, so your files will be available no matter which node you use.

Using Jupyter#

Once you have logged into the Onyx node you want to use, change into the directory where you’re keeping your class files or the specific project you want to work on and start the notebook server:

$ cd CS533/project-dir
$ jupyter notebook --no-browser

The --no-browser option is to keep Jupyter from trying to start a browser; if you’re logged in through MobaXterm, it usually sets up X11 forwarding, and without this option Jupyter will start Firefox on the Onyx node and try to display it on your computer. It will be much more efficient to use Jupyter’s network interface than to tunnel an entire browser over remote display.

The notebook server will print out some log messages; at the bottom you will see something like

    Or copy and paste one of these URLs:

You now need to set up an SSH tunnel to connect to the server. Note the port number from the URL (the numbers after localhost:, in this case 8888). Create an SSH tunnel that tunnels that port from your local computer to the port on your node. From the command line, you can do this by opening another terminal and running:

ssh -L 8888:localhost:8888

This tunnels port 8888 from your local computer to the same port on localhost on the remote computer (the Onyx node). You need to change 8888 and onyxnode03 to match your port number and Onyx node.

The forwarded port (in ssh -L, the first port number) does not have to match the remote port. If you have something else running on your computer using port 888, you can forward any port you want.

Once you have the port forward or tunnel set up, go to the URL indicated in the Jupyter Notebook output in your web browser. The tunnel makes localhost:8888 on your computer route to localhost:8888 on the Onyx node, and everything works.

Demo Video#