When multiple versions of software are installed, some method is needed to use the version you want to use. The
Modules package is the overwhelming choice of HPC centers for this.
Modules is supplied on the system to manipulate the users's environment variables to run a choice of the needed programs and versions. The most important of these variables are
$PATH, telling the system where to find executable files such as
$LD_LIBRARY_PATH, telling the system where to find shared libraries that an executable calls and requires. You can manipulate the environment variables yourself instead of calling modules, but generally there is no advantage in doing so.
The four critical module commands are
list (currently loaded),
avail (all available),
purge. The combination purge/load clears all loaded modules and sets new ones:
$ module list Currently Loaded Modulefiles: 1) intel/14.0.3 2) impi/5.0.0 $ module avail -------- /share/apps/modules/Modules/3.2.10/modulefiles ------------------------------------------------ dot gcc/4.9.1 impi/5.1.1 mkl/14.0.3 module-info mvapich2/2.1 openmpi/1.8.6 perl/5 gcc/4.7.2 impi/5.0.0 intel/14.0.3 module-git modules null openmpi/1.8.8 use.own $ module purge $ module load intel/14.0.3 impi/5.1.1 $ module list Currently Loaded Modulefiles: 1) intel/14.0.3 2) impi/5.1.1 $ module purge $ module load intel/14.0.3 mkl/14.0.3 mvapich2/2.1 $ module list Currently Loaded Modulefiles: 1) intel/14.0.3 2) mkl/14.0.3. 2) mvapich2/2.1
module avail can be slow to load. A quicker way to see the top-level names only is (on any cluster, this is razor)
$ ls /share/apps/modulefiles abinit cdhit flash idb mkl netcdf pindel reapr tophat abyss changa gadget ifort module-cvs null platform_mpi samtools transdecoder allpaths cmake gamess impi module-info nwchem pqs scipy trilinos augustus cplex gcc infernal modules open64 proj.4 sickle trinity bbmap crb-blast gdal intel mono opencv python siesta usearch bib cuda GenomeVISTA java moose openfoam qiime soapdenovo2 use.own blast cufflinks glpk last mothur openmpi qlogicmpi spades velvet blat dDocent gotoblas2 LIGGGHTS mpiblast orca qmcpack sra-tools ViennaRNA boost deepTools grass maker mpt os quast sunstudio visit bowtie dot gromacs matlab muscle parallel R swat wise2 bowtie2 emboss gurobi mcl mvapich pear randfold swig busco fastqc hdf5 migrate-n mvapich2 perl raxml symphony bwa fftw hmmer miRDeep ncl PGI rDock tcltk
ls -R /share/apps/modulefiles | more gives a full list of every file.
A script file in each user's home directory called
.bashrc also sets environment variables such as
$LD_LIBRARY_PATH at each login. Since the file name begins with a period, the file is “hidden” so it does not register by
ls but does register by
If you use a different shell than
bash such as
tcsh, there will be an analogous file such as
.cshrc which does the same functions in
The default .bashrc loaded with a new account includes three
module loads. You can and usually should edit this file to get the effect you want.
$ cat ~/.bashrc . /etc/profile.d/env-modules.sh ulimit -s unlimited 2>/dev/null ulimit -l unlimited 2>/dev/null #if using goto/mkl blas with mpi these should be set to 1 unless you want hybrid mpi/openmp export GOTO_NUM_THREADS=1 export OMP_NUM_THREADS=1 export MKL_NUM_THREADS=1 #enter your modules here module load intel module load openmpi module load mkl [ -z "$PS1" ] && return PS1='`/bin/hostname -s`:`whoami`:`echo $PWD | sed "s=$HOME=="`$ ' alias ls='ls --color=auto' module list
.bashrc is sourced at the beginning of each interactive job. There is a similar file
.bash_profile sourced at the beginning of each non-interactive job. In our setup, we source
.bash_profile so that the files are effectively the same, thus reducing the maintenance effort, but you can separate them if necessary. Interactive or batch is determined in
[ -z “$PS1” ] && return which drops out of the loop on batch runs, so commands following that are for interactive sessions only, like setting the value of the prompt
$PS1. Commands towards the top of ``.bashrc`` are for both interactive and batch. You can add commands to
.bash_profile for batch jobs only.
For csh-users, module commands should operate identically under
tcsh but they are generally untested.
On multiple-node parallel jobs,
.bashrc can do some unexpected things as the system spawns a shell on each node. Here are some recommended module/.bashrc setups for different cases:
You run only one program, or all the programs you run use the same modules, or each uses different modules that don't conflict
module load … in
[ -z…. The same environment will be loaded from
.bashrc for every interactive session, batch job, and MPI program if any. Many modules, for instance
python, can be assumed to not conflict, though most of the very many combinations have not been tested. Modules that definitely do conflict are MPI modules, only one may be safely used at a time, and multiple versions of the same program, like
gcc/4.9.1. Multiple compilers, such as
intel usually don't conflict, but for MVAPICH2 and OpenMPI, the compiler module is used to load the MPI module, so multiple compiler modules should be loaded in order (1) compiler module you want to use with MPI (2) MPI module (3) additional compiler module. In some cases, gnu MPI programs using MKL libraries can want a library only available from the Intel compiler, so a combination like
module load gcc/4.7.2 mkl/14.0.3 openmpi/1.8.8;module load intel/14.0.3 may be necessary. The last
intel/14.0.3 may show a harmless warning message.
You use different (conflicting) modules for different programs, but only run single-node batch jobs
If you put
module load … in your
.bashrc, every interactive session and batch and MPI thread will load that, which won't be good if it's a slave compute nodes loading the wrong MPI version. Delete or comment out (add leading “#” to) the
module load … in
.bashrc. Put the relevant
module load … in your batch scripts after the
#PBS statements, for example:
#PBS ... #PBS -l node=1:ppn=12 module purge module load R openmpi cd $PBS_O_WORKDIR R < myprog.R --nosave --vanilla
In this case, for interactive sessions such as compiling, type the relevant
module load … in your session before work.
You use different (conflicting) modules for different programs, and also run multi-node MPI jobs
This is a more difficult case. The first two solutions won't always work. If a module is set in a batch script using multiple nodes, the module definitely applies to the MPI processes running in the first or “master” compute node (usually the first and lowest numbered assigned node in our batch configuration) but does not necessarily apply to the “slave” compute nodes, depending how different MPI types issue remote threads. Multiple nodes imply MPI is being used, and the solution varies by MPI type. A certain form of the
mpirun statement for each MPI type is required. See the MPI article for more details.