Software stack

In Cecile software is provided systemwide through a so called software stack, a collection of software, generated by using the Spack package manager.

A software stack contains a number of software necessary for your analysis and much more. If any software you need is missing from the stack you may contact cecile-admins-l at ovgu.de to request it.

Type of stacks

There are two kinds of stacks available:

1. Current stack: A stack that is kept stable for a long period of time. A new stable stack is created periodically from scratch, which means that the new current stack will include only new up-to-date software versions; however the old versions of current stacks are still available on the cluster.
2. Experimental stack: A flexible stack that is going to be frequently updated also upon user's request. A software requested by a user will be added to the experimental stack; as soon as a new current stack is created the new software will be also added to current.

Multiple software versions in experimental stack

Continuous updating in the experimental stack can lead to have multiple versions of the same software (e.g. python ~gcc@11.0 vs python ~gcc@11.1) in the experimental stack, therefore be careful to load the correct version when using the experimental stack. Usually back compatibility between software versions is mantained, but sometimes this might not be the case and some feature might have been changed.

How to use the stacks

Current stack:Experimental stack

• Source the current stack, by running the following command (please, be aware of the space between the dot and the path)

  . /software/current/env.sh
  

• Now load a software using module load <software name>

  module load py-pandas
  

• To see what software is available run the following command

  module avail
  

• Source the experimental stack, by running the following command (please, be aware of the space between the dot and the path)

  . /software/experimental/env.sh
  

• Now load a software using module load <software name> (in case a specific version is needed, check the software version with module avail)

  module load py-pandas
  

• To see what software is available run the following command

  module avail
  

Software names in the stack

Before loading a module it is advisable to check how the software you are interested in is called in the stack. For example all python software is preceded by py-, thus to load pandas you need to type: module load py-pandas and R software is preceded by r-. This naming convention needs to be used when loading the software, in your code software must have their usual names.

New and old versions of current stacks and how to use them

As pointed out, the current stack is updated and rebuilt periodically; this procedure ensures that the cluster always includes up-to-date versions of all software. However, the older versions of current stacks are still available on the cluster. To distinguish them from the up-to-date current stack, the older versions of current stacks get a suffix indicating the creation date, current_<yyyy-mm-dd>.

To use an older current stack you only need to source the correct stack:

. /software/current_<yyyy-mm-dd>/env.sh

For example:

. /software/current_2024-07-28/env.sh

Medusa software stack

In case you need to use the old software stack used within the Medusa cluster, please contact the admin at cecile-admins-l at ovgu.de

Available current stacks and included software

The following tables represent all the current stacks (up-to-date and older) currently available on the cluster and software included in each of them.

currentcurrent_2024-07-28

Table of current stack

Software	Version
bids-validator	2.4.1
bidsonym	0.0.6_2025-10-24
connectome-workbench	2.1.0
eyelink	2.1.1197.0_v26.04
fmriprep	25.2.5
freesurfer	8.2.0-1
fsl	6.0.7.22
gdb	17.1
git-annex	10.20230408
git	2.53.0
gnuplot	6.0.0
hdf5	1.14.6
headcase-pipeline	1.0.0_2024-05-16_cca8bf2
heudiconv	1.4.0
laynii	2.10.0
matlab	r2025b
mricron	1.2.20211006
mriqc	24.0.2
plantus	0.1.1
py-bidscoin	4.6.2
py-bidskit	2025.11.7
py-datalad	1.4.0
py-dcm2bids	3.2.0
py-eye2bids	0.1.2026-03.16
py-flake8	7.3.0
py-glmsingle	main
py-jupyterlab	4.5.8
py-matplotlib	3.10.9
py-memory-profiler	0.61.0
py-mne-bids	0.18.0
py-mne	1.11.0
py-mypy	2.1.0
py-neurokit2	0.2.13
py-neurora	1.1.6.12
py-nibabel	5.4.2
py-nilearn	0.13.1
py-nipype	1.11.0
py-numba	0.65.1
py-numpy	2.3.5
py-palettable	3.3.3
py-pandas	2.3.3
py-pip	26.1
py-pybv	0.7.6
py-rsatoolbox	0.2.0
py-scikit-learn	1.8.0
py-scipy	1.17.1
py-seaborn	0.13.2
py-sphinx	9.1.0
py-statsmodels	0.14.6
py-virtualenv	21.4.1
py-wesanderson	0.0.4
python	3.14.5
r-afex	1.5-1
r-bayesfactor	0.9.12-4.8
r-brms	2.23.0
r-dplyr	1.2.0
r-emmeans	2.0.2
r-ggplot2	4.0.3
r-irkernel	1.3.2
r-rstan	2.32.7
r-tidyverse	2.0.0
r	4.5.3
simnibs	4.6.0

Table for current_2024-07-28 stack

Software	Version
bids-validator	1.14.6
bidsonym	0.0.6
deepprep	25.1.0
environment-modules	5.4.0
eyelink	2.1.1197.0
fmriprep	21.0.2
fmriprep	23.0.2
fmriprep	24.0.0
freesurfer	7.4.1
fsl	6.0.7.4
gdb	14.2
git-annex	10.20230408
git	2.45.2
gnuplot	6.0.0
hdf5	1.14.3
headcase-pipeline	1.0.0_2024-05-16_cca8bf2
heudiconv	1.1.6
laynii	2.7.0
matlab	r2023b
mricron	1.2.20211006
mriqc	23.1.0
py-bidscoin	4.1.1
py-bidskit	2023.9.7
py-datalad-hirni	0.0.8
py-datalad	0.18.4
py-dcm2bids	3.1.0
py-eye2bids	0.1.dev1
py-flake8	6.1.0
py-glmsingle	main
py-jupyterlab	4.0.1
py-matplotlib	3.9.0
py-memory-profiler	0.61.0
py-mne-bids	0.15.0
py-mne	1.7.1
py-mypy	1.8.0
py-neurokit2	0.2.4
py-neurora	1.1.6.10
py-nilearn	0.10.3
py-nipype	1.8.6
py-numba	0.58.1
py-numpy	1.26.4
py-palettable	3.3.3
py-pandas	2.1.4
py-pip	23.1.2
py-pybv	0.7.5
py-rsatoolbox	0.2.0
py-scikit-learn	1.5.1
py-scipy	1.14.0
py-seaborn	0.13.2
py-sphinx	7.4.5
py-statsmodels	0.14.0
py-virtualenv	20.24.5
py-wesanderson	0.0.3
python	3.11.9
r-afex	1.3-0
r-brms	2.19.0
r-emmeans	1.8.5
r-rstan	2.21.8
r-tidyverse	2.0.0
r	4.4.0
rabies	0.5.0
simnibs	4.5.0
texlive	20220321

Matlab software stack

The Matlab installation provided in the stack does include a variaty of toolboxes. In general we differentiate between toolboxes provided by MathWorks and third party toolboxes.

MathWorks toolboxes are toolboxes from https://www.mathworks.com/products.html like e.g. Image Processing Toolbox or Statistics and Machine Learning Toolbox and are included automatically.

Third party toolboxes like SPM or FieldTrip are also available in the central software stack but have to be loaded manually.

Current stackExperimental stack

/software/current/matlab_toolboxes/

e.g load SPM individually

addpath /software/current/matlab_toolboxes/spm12

/software/experimental/matlab_toolboxes/

e.g load SPM individually

addpath /software/experimental/matlab_toolboxes/spm12

Don't use genpath to load all third-party toolboxes at once

Although matlab provides the possibility to load all toolboxes within a path by using genpath, this is not recommended to do because some of the toolboxes don't work well together and might break. For example spm will not start anymore.

Matlab shortcuts fix

While interacting with Matlab, you might notice that shortcuts such as Ctrl-C or Ctrl-V do not work as expected, this behavior is due to a different editor keys-shortcuts binding, in the image below you can see that Paste is indeed bound to the Ctrl-Y shortcut.

If you go to preference, then to shortcuts and select Windows Default Set and press Apply, the regular shortcuts will be repristinated.

To verify that the process was successful, do a right click with your mouse and you will see that now Paste is bound to the familiar Ctrl-V.

JupyterLab

What is JupyterLab?

JupyterLab is a web-based interactive environment for notebooks and code. For in depth information, please visit Jupyterlab page.

How to use JupyterLab on Cecile

JupyterLab works only via Open OnDemand

1. Once you are on Open OnDemand, go to Interactive Apps and select JupyterLab (image below) and a configuration window will be shown.

   

2. Configure your JupyterLab session (see image below):

   • Kernel: It allows to choose the language you want to use inside JupyterLab. The available kernels are python and R, by selecting either one of these kernels you are able to use only the selected language within JupyterLab. By selecting all you have both python and R kernels and all the relative libraries available.

   • Jupyterlab start directory: There are two starting directory options: user home, which allows you to start and use JupyterLab in your home, and /data/projects/ which allows you to start JupyterLab in /data/projects/ and to freely navigate to your project via the JupyterLab GUI. Keep in mind that if you select user home as starting directory, you cannot navigate to /data/projects/ and vice versa.

   • The remaining configuration options are the same as any other interactive session.

   Loading all libraries takes time

   By selecting all in kernels, you have more flexibility inside JupyterLab, but it requires about 1 minute to load all the libraries. Please be patient.

   

3. Launch JupyterLab

   

4. Now you are on JupyterLab and you can work on notebooks or code.

   

5. Once you are done working with JupyterLab, please shut down the interactive session: File --> Shut Down