====== How do I get an interactive shell on a compute node? ======

Sometimes you may want to run some test code on a compute node, or compile something interactively on a compute node to ensure that your code works as intended before running the main job.

You can request a short-term interactive shell session by running the following from a login node:

<code lang=shell-session>
$ srun --pty --partition=bigmem /bin/bash
</code>

This will queue a request for a BASH session on a node in the //bigmem// partition and the command will pause until the request is granted. Usually you will get a session opened within a few seconds, but this may take longer in periods of intense activity. Once the session is started you can run normal commands on that node, just as you would on a login node, including loading software modules. To exit the session, simply use the Control-D key combination, or type 'exit', as you would with any other shell. You will then be returned to the login node from where you requested that session.

If you need a session on a partition which is only accessible to projects with funds, then remember to include your account / project code in the request:

<code lang=shell-session>
$ srun --pty --account=myhpc_group --partition=GPU-L /bin/bash
</code>

This requests a session on the //GPU-L// partition, with resource use billed against the //myhpc_group// HPC Project account.

For further info on **srun**, check out our [[started:slurm_basics#srun|Introduction to Slurm]] pages.

----

[[:faq:index|Back to FAQ]]