Linux Workstations

Contributors: Nathan TM Huneke, Harry Fagan, Nick Hedger, Yukai Zou
Maintainers: Nathan TM Huneke

Why Linux?

Linux is a lightweight (small size) platform which consumes not as much memory as a Windows OS. This feature is particularly useful when it comes to computing-intensive environment, where a lot of memory and storage spaces are in need. For neuroimaging analysis, it is almost essential that you use a UNIX - based operating system (Mac or Linux), since popular software packages such as FSL, AFNI and FreeSurfer do not run on a Windows operating system.

A number of workstations at the University of Southampton run a flavour of Linux operating system known as Red Hat Enterprise Linux (RHEL) 7. Some of these have been configured for neuroimaging analysis. Current machines that can be used for this purpose include:

  • uos-15263

  • uos-210028

  • uos-211997

Note

For more information on the University’s RHEL7 machines, see http://linuxdesktops.soton.ac.uk/

Setting Up the Machine

To use the above machines effectively, you need to go through a number of initial set up steps.

0. Grant Access

First of all, your access to the specific University linux workstation needs to be granted. This is typically done by the admin (i.e. you would need to find out who that person is) of that workstation, who will raise an iSolution ticket on behalf of you.

Note

Please provide your UoS username to the admin.

1. Log in

These machines can be accessed remotely via remote desktop apps or SSH.

Remote Desktop Apps

For MacOS, you can use “Microsoft Remote Desktop”. If you already have it installed for connecting to “Win 10 Student” at the University, then this would come handy. For Windows, you can use “Remote Desktop Connection”, which is automatically installed on University machines running Windows. You should be able to download and install either app on your personal machines.

Note

If you are off site when logging in then you will need to connect to the University’s VPN first.

  1. On your machine, click on Start and search for Remote Desktop Connection

  2. Type in the computer name as above followed by .clients.soton.ac.uk (e.g. uos-210028.clients.soton.ac.uk) and click connect

  3. Input your University of Southampton username and password when prompted

Alternatively, if you would like to connect via remote desktop app on Linux, you can use Remmina by following an instruction here.

SSH

To connect remotely via SSH, launch a Terminal window, and enter the following command:

Note

If you are off site when logging in then you will need to connect to the University’s VPN first.

ssh username@uos-#.clients.soton.ac.uk - Replace username with your University of Southampton username, and # with the specific workstation ID.

Please enter your University of Southampton password when prompted, and then hit Return.

To display application GUI’s from the remote machine on your computer (e.g. the GUI for FSL or Matlab) you will need to connect with X-forwarding enabled by using the -X flag (or on Mac, -Y):

ssh -X username@uos-#.clients.soton.ac.uk

This is particularly useful when using Matlab remotely. Click here for more information on common problems with X and how to solve them.

Tip

What if I am using a Windows machine?

You can install a terminal emulator and use the commands as above. For example, MobaXterm has all the features needed to achieve X-forwarding on a Windows machine. Alternatively, you can also use Git BASH, a BASH emulation used to run Git from the command line, which is already installed on all University Windows machines.

Tip

I can’t connect via SSH and I got a warning message: “REMOTE HOST IDENTIFICATION HAS CHANGED!” What should I do?

The issue here is most probably due to a new (randomised) host ID has been generated, which is usually a result of system upgrade.

Why is it causing a warning?

ssh is warning you that this host ID is different to what it previously saw when you connected to the computer, since it has no way to know whether (as in this case) it is a legitimate change made by the machine owner, or whether someone has maliciously tampered and may be trying to steal information by getting you to connect to a different computer entirely.

How to solve?

There are several ways to work around:

  1. If this is the only computer that you use ‘ssh’ or related commands (e.g. rsync, scp, sftp) to connect to, then you can safely just delete the file where it keeps its log of host key fingerprints:

    $ rm /h/.ssh/known_hosts

then try the ssh command again, confirm you wish to store the host ID, and proceed as usual.

  1. Alternatively, if you wish to preserve the old file, you can rename it to a different filename:

    $ mv /h/.ssh/known_hosts rm /h/.ssh/known_hosts_old

then try the ssh command again. In this case, a new known_hosts file will be generated.

  1. Alternatively, if you use ssh for other purposes as well (e.g. connect to other computers) and want to keep their host key fingerprints intact, then you will need to delete just the relevant lines from the file.

You could edit the file with vi or nano if you are comfortable with those editors, or, you could use Wordpad or Notepad++ (but not notepad since it is likely not to ‘understand’ the line endings) in Windows, finding the file at h:\.ssh\known_hosts

2. Install Conda

Most of the software you will need has been pre-installed on the machine. However, Conda (a distribution of Python) needs to be installed on a per-user basis.

To install conda, first navigate to your home directory:

$ cd ~

Then download the installer:

$ wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

Then run the installer:

$ bash Miniconda3-latest-Linux-x86_64.sh

Follow the prompts on the installer screens. Use q to skip to the end of the license agreement and accept. When asked, save conda in a hidden directory in your home directory to prevent other software interfering with it. We suggest:

$ ~/.conda/Miniconda3

Once the installer has finished, close and reopen the terminal window. Then type conda --version to check it has installed successfully.

3. Create a conda environment for your software and analyses

What is a conda environment?

An environment is a self-contained collection of conda packages. If you change one environment, your other environments and other installed software are not affected. Some software, particularly DataLad, seems to interfere with other software on the Linux machines. It is therefore safest to use this software in a self-contained environment.

Conda environments can be easily activated and deactivated as needed.

Create your environment

We have created a ready-to-use conda environment that you can clone.

To create your environment, open a terminal. If conda is installed, you should see that you are currently in the base environment, which is signified like so:

(base) [nh6g15@uos-211997 ~] $

Next, install mamba:

$ conda install -c conda-forge mamba

Next, clone the environment description into your home directory with:

$ git clone https://github.com/nhuneke/imagingenv ~/imaging-env

And now create the environment:

$ cd ~/imaging-env
$ mamba env create --file imaging-env.yml

Activating and deactivating your environment

To activate your environment use the following command:

$ conda activate imaging-env

You should now see this environment is active in the terminal, like so:

(imaging-env) [nh6g15@uos-211997 ~] $

You will now be able to use all the software present in this environment.

To deactivate your environment, use the following command:

$ conda deactivate

4. Set up FSL

If this is your first time logging in on the Linux machine then you will need to set up your shell environment to use FSL. Each user’s shell setup is stored in a file called .bash_profile. Open this file in a text editor:

$ gedit ~/.bash_profile

At the end of the file copy and paste the following lines:

FSLDIR=/usr/local/fsl    # NOTE: This is default; modify this line to match your local path
. ${FSLDIR}/etc/fslconf/fsl.sh
PATH=${FSLDIR}/bin:${PATH}
export FSLDIR PATH

Then logout and log back in. FSL will now be ready for use.

Using Space Effectively

The Linux machines as set up by the University are made up of three broad partitions:

  • The root partition /

  • The home partition /home/, or ~ for your personal home directory

  • The local partition /local/

The root and home partitions are very small. You will not have access to the root partition anyway as changing things here can be dangerous. Only administrators can access this and make changes. But your home directory is yours. However, this partition is only 50GB in size in total, for all users!.

For this reason, you should not store neuroimaging data here, as a single dataset can easily reach more than 50GB in size. Instead, you should use the scratch space which is in the local partition. Each user has their own scratch space at:

/local/scratch/<user id>

Replace user id with your username.

Danger

The scratch space is not backed up!

The scratch space is for current work, like a drafting space. It is not backed up. You should regularly make backups of your work here and once the project is finished store the data in a permanent store, such as the university research filestore. DataLad can help with this (add ref).