Skip to content

Connections to compute nodes


When a job is running on a compute node, the owner of the job can access the node using ssh from a development node. This will connect your terminal to the shell on the compute node where the job is running.

If your job is running on the node amr-070 the command would be ssh amr-070. Once on the compute node you can inspect individual CPU and memory usage with top and htop.

Connecting from your local machine

If you have SSH Tunneling set up, we can extend the ProxyJump framework to connect to compute nodes.

To do this, add the following lines to your .ssh/config:

    Host lac* vim* skl* nvl* amr* nvf* nif* nal* acm*
    User here_you_put_your_net_id
    ProxyJump intel18

The three-letter names correspond to the kinds of nodes available from each cluster. Notice that we are using the development node intel18 to connect to the compute nodes, rather than the gateway.

We can then run a command such as ssh amr-070 from our local terminal (assuming we have a job running on that node). This will connect us directly to that node.


Using OnDemand you can access the compute node that your job is running on in the same way by opening a development node from the Development Nodes menu and then using the ssh command as described above. You can identify the node for interactive sessions from the My Interactive Sessions tab. It is located after the word "Host". Do not include the >_ characters in your ssh command. The image below shows a job running on the node amr-070.

Screenshot of the OnDemand interactive session card with the Host name highlighted.