This is an old revision of the document!
Table of Contents
Accessing Available Software
The cluster uses environment modules to control what software is loaded in your environment. Certain software will be loaded by default, but there are other versions available to keep analyses consistent across a study or additional software that you may wish to load. Modules work by adding specific environmental variables needed for each package and/or adding them to your default search path.
What's Loaded?
- “module list” will show you what is currently loaded
[cmp12@blade08 ~]$ module list Currently Loaded Modulefiles: 1) fsl/6.0.3 2) freesurfer/7.1.1 3) afni/21.1.04 4) ants/2.3.4 5) mrtrix3/3.0.2 6) singularity/3.6.3 7) matlab/R2019b
What's Available?
- “module avail” will show you what additional software or alternative are available
[cmp12@blade08 ~]$ module avail ------------------------------------------ /etc/modulefiles ----------------------------------------------- afni/18.2.15 dtk/0.6.4.1 fsl/5.0.10 laynii/2.0.0 miniconda singularity/2.5.2 afni/21.1.04 ffmpeg/4.1 fsl/5.0.6 mango/4.0.1 mpi/mpich-3.0-x86_64 singularity/3.6.3 ansysEM/19.0 fidl/2.66 fsl/5.0.9 mango/4.1 mpi/mpich-x86_64 slicer/4.6.2 ansysEM/19.5 freesurfer/0_dev fsl/6.0.0 matlab/R2014a mpi/openmpi-x86_64(default) trackvis/0.6.1 ants/2.3.4 freesurfer/5.3.0 fsl/6.0.3 matlab/R2017a mricron weka/3.8.4 c3d/1.1.0 freesurfer/6.0.0 hcp_pipelines/3.27.0 matlab/R2018a mrtrix3/3.0.2(default) camino/2.1 freesurfer/6.0.1 hcp_workbench/1.4.0 matlab/R2019b(default) mrtrix3/mact dcm2niix freesurfer/7.1.1 hcp_workbench/1.4.2 matlab/R2021a mrtrix3/tissue dramms/1.5.1 fscan hcp_workbench/1.5.0 medinria/2.2.3(default) rar/5.70 dtiprep/1.2.4 fsl/5.0.1 laynii/1.5.6 medinria/4.0.0 simnibs/2.0.1
Get more info about a module
To get more info about a specific module use the “module display” or “module whatis” command.
[cmp12@blade01 ~]$ module display fsl ------------------------------------------------------------------- /etc/modulefiles/fsl/6.0.3: module-whatis loads the fsl v6.0.3 environment setenv FSLDIR /usr/local/packages/fsl-6.0.3 setenv FSLSUBMIT 0 setenv FSL_DIR /usr/local/packages/fsl-6.0.3 setenv FSL_BIN /usr/local/packages/fsl-6.0.3/bin setenv FSLTCLSH /usr/local/packages/fsl-6.0.3/bin/fsltclsh setenv FSLWISH /usr/local/packages/fsl-6.0.3/bin/fslwish setenv FSLOUTPUTTYPE NIFTI_GZ prepend-path PATH /usr/local/packages/fsl-6.0.3/bin ------------------------------------------------------------------- [cmp12@blade01 ~]$ module whatis fsl fsl : loads the fsl v6.0.3 environment
How to modify the environment
- to load additional software use “module load modulename(s)“
- just calling the modulename will load the highest version number of an available software, unless there is a “default” designation in the listing
[cmp12@blade08 ~]$ module load laynii dcm2niix fscan [cmp12@blade08 ~]$ module list Currently Loaded Modulefiles: 1) fsl/6.0.3 3) afni/21.1.04 5) mrtrix3/3.0.2 7) matlab/R2019b 9) dcm2niix 2) freesurfer/7.1.1 4) ants/2.3.4 6) singularity/3.6.3 8) laynii/2.0.0 10) fscan
If there is a package already loaded you should use the switch command to change versions
- “module switch modulename/version”
[cmp12@blade08 ~]$ which feat /usr/local/packages/fsl-6.0.3/bin/feat [cmp12@blade08 ~]$ module switch fsl/5.0.10 [cmp12@blade08 ~]$ module list Currently Loaded Modulefiles: 1) **fsl/5.0.10** 3) afni/21.1.04 5) mrtrix3/3.0.2 7) matlab/R2019b 9) dcm2niix 2) freesurfer/7.1.1 4) ants/2.3.4 6) singularity/3.6.3 8) laynii/2.0.0 10) fscan [cmp12@blade08 ~]$ which feat /usr/local/packages/fsl-5.0.10/bin/feat
- to remove a loaded package “module rm modulename” - to unload everything “module purge”
[cmp12@blade08 ~]$ module rm fsl [cmp12@blade08 ~]$ module list Currently Loaded Modulefiles: 1) freesurfer/7.1.1 3) ants/2.3.4 5) singularity/3.6.3 7) laynii/2.0.0 9) fscan 2) afni/21.1.04 4) mrtrix3/3.0.2 6) matlab/R2019b 8) dcm2niix [cmp12@blade08 ~]$ module purge [cmp12@blade08 ~]$ module list No Modulefiles Currently Loaded.
Make it consistent
You can add module calls to your submission scripts to keep the environment consistent for your jobs or you can add module calls to your ~/.bashrc initialization file to have a specific set of modules across all of your cluster jobs or interactive sessions.
[cmp12@blade01 ~]$ tail -n 10 ~/.bashrc module purge module load fsl/6.0.0 freesurfer/6.0.1 matlab/R2018a [cmp12@blade01 ~]$ module list Currently Loaded Modulefiles: 1) fsl/6.0.0 2) freesurfer/6.0.1 3) matlab/R2018a
If there are certain packages that are not available you can always email biac-help@duke.edu to request them. Additionally, standard linux software will not show up in the module list ( ie: cmake3, gcc ). Usually the which or locate command is an easy way to see of they are available.
[cmp12@blade01 ~]$ which python3 /usr/bin/python3 [cmp12@blade01 ~]$ which cmake3 ~/bin/cmake3
Setting Up Experiment Specific Conda Environments
Conda is available via modules to setup and maintain your own specific environments if you want to use specific versions of python ( or R ) modules that are not tied to the default cluster installation.
First, you want to unload the default python3 and load miniconda via modules to begin:
[cmp12@blade29 ~]$ module unload python3 [cmp12@blade29 ~]$ module load miniconda [cmp12@blade29 ~]$ conda --version conda 22.11.1 #set the default location to download packages to TMPDIR so you do not fill up your home directory [cmp12@blade29 ~]$ export CONDA_PKGS_DIRS=${TMPDIR}/pkgs
Now, create your environment in a mounted location. For example and Lab directory for general access from your other lab members, or in an experiment so it can be used throughout all the analysis of an entire experiment.
conda create -p /mnt/munin/Song/Lab/Chris/myenv python #activate it conda activate /mnt/munin/Song/Lab/Chris/myenv #install some of the python modules you may need (/mnt/munin/Song/Lab/Chris/myenv) [cmp12@blade29 ~]$ which pip3 /mnt/munin/Song/Lab/Chris/myenv/bin/pip3 (/mnt/munin/Song/Lab/Chris/myenv) [cmp12@blade29 ~]$ pip3 install numpy scipy nibabel ipython #verify (/mnt/munin/Song/Lab/Chris/myenv) [cmp12@blade29 ~]$ which python /mnt/munin/Song/Lab/Chris/myenv/bin/python (/mnt/munin/Song/Lab/Chris/myenv) [cmp12@blade29 ~]$ pip3 freeze ipython==8.11.0 jedi==0.18.2 matplotlib-inline==0.1.6 nibabel==5.0.1 numpy==1.24.2 .......
Now you have a consistent environment that you can use/maintain throughout the experiment.
To deactivate just run “conda deactivate” which will remove the environment from your path.
To make it consistent, you can modify your ~/.bash_profile ( or ~/.bashrc ) to automatically add miniconda and activate the environment:
module remove python3 module add miniconda conda activate /mnt/munin/Song/Lab/Chris/myenv
Alternatively, you can activate within submission scripts and/or make sure you are explicitly calling your own version of python within scripts:
[cmp12@blade29 ~]$ //mnt/munin/Song/Lab/Chris/myenv/bin/python Python 3.10.4 (main, Mar 31 2022, 08:41:55) [GCC 7.5.0] on linux Type "help", "copyright", "credits" or "license" for more information. >>> import numpy >>> numpy.__version__ '1.24.2' >>> numpy.__file__ '//mnt/munin/Song/Lab/Chris/myenv/lib/python3.10/site-packages/numpy/__init__.py'
Add something else to your environment ( like R ):
(/mnt/munin/Song/Lab/Chris/myenv) [cmp12@blade29 ~]$ conda search r-base #if you don't see the version you want you can add additional channels conda config --add channels conda-forge conda config --set channel_priority strict #then search again conda search r-base #install the version you would like to use (/mnt/munin/Song/Lab/Chris/myenv) [cmp12@blade29 ~]$ conda install r-base=4.2.2 (/mnt/munin/Song/Lab/Chris/myenv) [cmp12@blade29 ~]$ which R /mnt/munin/Song/Lab/Chris/myenv/bin/R (/mnt/munin/Song/Lab/Chris/myenv) [cmp12@blade29 ~]$ R --version R version 4.2.2 (2022-10-31) -- "Innocent and Trusting" Copyright (C) 2022 The R Foundation for Statistical Computing Platform: x86_64-conda-linux-gnu (64-bit)
Other Software
Matlab Libraries : /usr/local/packages/MATLAB
Premade Singularity Images : /usr/local/packages/singularity/images/