A General Guide on How to Use Pronto

This guide provides a brief overview of understanding how to utilize Pronto.

Prerequisites

  • You should be familiar with the command line and have a basic understanding of how it works.
    • Learning how to write a bash script would be extremely helpful. This can save you valuable time without having to retype a sequence of commands multiple times.
    • There are also plenty of resources online that can help you.
  • Access to Pronto
  • A terminal to type your commands in
    • For Windows, you can use PuTTY, Powershell, or use the Windows Subsystem for Linux.
    • For Mac, you can use the regular terminal (Under Applications > Utilities > Terminal)
    • For Linux, you can either search for Terminal or just right click on the desktop and click "Open Terminal"
  • Understanding of software modules

Overview

Pronto has plenty of different compute nodes available for your needs.

Here's a list of nodes (with hardware specifications) that are available to you.

Each compute node has different specifications and is suitable for different workloads.

Please see this guide for how to pick the best compute node to run your job on. This is important. Please read it.

Logging in and using Pronto

Open up the terminal that you are using and connect to pronto using:

ssh <YourNetIDHere>@pronto.las.iastate.edu

If you are using PuTTY, omit the ssh and NetID portion and type in "pronto.las.iastate.edu" in the hostname field. You will be prompted to login and/or enter your password. Please note that when typing your password, no asterisks, letters or spaces will appear. Just type in your password and press enter. This is done to protect your password from people looking over your shoulder.

pronto_window

Your terminal should look similar to the image above after logging in.

You are on the head node now. You cannot run any software on the head node, so you need to allocate a compute node. This is done through SLURM, a workload manager. You will need to utilize SLURM in order to run the jobs that you wish to run.

Please refer to this guide and use it as a reference when you want to allocate compute nodes and run future jobs on the clusters.

Storage space

We provide 10GB of home storage for every user on pronto, home should generally not be used for storing data that you'll be working on.

For working data you'll want to use the space available in your /work directory. This directory is named after your PI's netid. If your PI's netid is jones, then your work directory is at /work/LAS/jones-lab. If you are unsure, contact your PI or research IT.

Data on /work is not backed up! It is for temporary use only!

Running a Job

In the background

The best way to run your job would be writing a simple bash script and submitting it via sbatch. This will allow your job to run even while you're not at your computer.

To write the script, you can tweak our example scripts, or use the job script generator. Then you can either upload your script (see the Transferring Files section below) or create the bash script on the compute node with an editor like vim, nano, etc.

To submit the job, run: 

sbatch myscript.sh

The job will enter a queue and run in the background.

To check the status of the job, you can run

squeue --me

Interactively

Alternatively, you can use the srun command to allocate a compute node and be placed on it. This will start an interactive session which is useful for debugging or for programs that require user input. Note that these jobs could potentially be lost if you lose internet connection or if the VPN reconnects. It is recommended you use the screen command or tmux.

srun --time=01:00:00 --cpus-per-task=1 --partition=interactive --pty /usr/bin/bash

Advanced Options

Please see the job scheduler documentation if you want to add more configurations such as allocating a GPU, specific compute node, writing a job script etc. You can run jobs up to 31 days maximum.

Available Software

We have a lot of software available as loadable modules. You may encounter a "command not found" or an error along those lines. This means you need to load a few modules before running your software.

To see the full list of available modules:

module avail

To search for a module, or see available versions of a specific module, you can run:

module spider <software name>

For example, to see available versions of python, run:

module spider python

To load a module, run:

module load <module name>

For example:

module load python/3.8.8-ucekvff

Transferring Files

File transfer to and from the cluster is available via Globus.

If you want to use scp or SMB, you will need to use the data transfer node. This node is specifically designed for transferring files across different nodes/clusters quickly and reliably.

The FQDN (IP Address) you will be using will be:

prontodtn.las.iastate.edu

Login with your NetID and password.

To use scp, you will first need to ssh to prontodtn.las.iastate.edu. You may be prompted for your password twice. You should enter it both times to ensure that you get a Kerberos ticket. If you are not prompted for it a second time, run

kinit

After entering your password, you can disconnect from ssh. scp should then work.

Guides

Troubleshooting

Here is a link to a guide that helps you troubleshoot common issues you might encounter:

Troubleshooting common issues

If you wish to run a job using R, please refer to this documentation.

Other issues

Please email researchit@iastate.edu if you still require assistance.