MOOSE Configuration

Configure and test your own "MOOSE" framework

Basic Steps

1. mkdir projects
2. cd projects
3. module load moose/26-jul-21
4. git clone https://github.com/idaholab/moose.git
5. cd moose
6. git checkout master
7. export VTKLIB_DIR=${HPC_VTK_LIB} VTKINCLUDE_DIR=${HPC_VTK_INC}
8. Run "./scripts/update_and_rebuild_libmesh.sh --enable-vtk-required --with-vtk-lib=${HPC_VTK_LIB} --with-vtk-include=${HPC_VTK_INC}" and allow it to complete. It should finish without errors.
9. cd test; make -j 4; run_tests -j 4
10. Build and test the "phase field" module:
    1. cd moose/modules/phase_field
    2. make -j 4
    3. ./run_tests -j 4
11. Build and test the "combined" module:
    1. cd moose/modules/combined
    2. make -j 4
    3. ./run_tests -j 4

Additional Information

See Bugzilla Request #43807

More Detailed Instructions

Load a dev Session

Build libmesh and the MOOSE executables in the dev session, but run the tests on the login node. This is an alias you can add to your ~/.bash_profile:

vim ~/.bash_profile
alias devsess =’srundev --mem=32gb --ntasks=4 --cpus-per-task=4 --time=12:00:00’

Load the MOOSE Module

There are multiple versions. Each version was built with different versions of the required libraries. In general, the latest version of the libraries will be used in the latest version of the module. The module listed below is good in general, but you may have specific reasons to use a different version.

module load moose/26-jul-21

Check the module list, you should see something like:

module list

Currently Loaded Modules:
  1) gcc/9.3.0
  2) mkl/2020.0.166 
  3) openmpi/4.1.1 
  4) petsc/3.15.1 
  5) qt/5.12.9 
  6) vtk/8.2.0 
  7) moose/26-jul-21

Clone MOOSE

mkdir ~/projects
cd ~/projects
git clone https://github.com/idaholab/moose.git
cd moose
git checkout master

Export VTK paths

Make sure you export the VTK paths after the MOOSE module is loaded, otherwise the path may be empty.

export VTKLIB_DIR=${HPC_VTK_LIB} VTKINCLUDE_DIR=${HPC_VTK_INC}

Peacock

To use the peacock input file syntax front-end, add the path to the 'python' directory in your MOOSE install as peacock requires the mooseutils python module. E.g.

export PYTHONPATH=/my/moose/dir/python:$PYTHONPATH

Update and Build libmesh

Run the update and build libmesh script. This step takes a while but should finish without errors (though you will see some warnings).

cd ~/projects/moose
./scripts/update_and_rebuild_libmesh.sh --enable-vtk-required --with-vtk-lib=${HPC_VTK_LIB} --with-vtk-include=${HPC_VTK_INC}

Compile MOOSE Tests

Build the MOOSE test executables.

cd ~/projects/moose/test
make -j 4

Run the Test Suites

Now run the tests. The test suite in moose/test is important to check if all libraries are correct. You need to exit the dev session to run the tests. If the build was successful, I would recommend building the module you are interested in and then exit and run both test scripts. Don’t forget that you may need to reload the moose module on the login node.

cd ~/projects/moose/test
./run_tests -j 4 (login node)

Compile and Test Your Own Module

Build the executable and run the specific tests.

cd ~/projects/moose/modules/combined
make -j16 (dev session)
./run_tests -j 4(login node)

General suggestions

If you are not getting a fresh clone of MOOSE, it’s recommended to do a clean-up before you recompile your executables.

cd ~/projects/moose/modules/combined
make clobberall
make -j 4

MOOSE-based Applications

Perform all the previous steps related to the MOOSE installation. Then build your application and test it.

cd ~/projects/your_app
make clobberall
make -j 4
./run_tests -j 4 (login node)

SLURM Job Script

This is an example of a slurm job script:

#!/bin/sh
#SBATCH --job-name=moose                 #Job name
#SBATCH --nodes=1                        #Number of nodes (servers, 32 proc/node)
#SBATCH --ntasks=16                      #Number of tasks/MPI RankS
#SBATCH --ntasks-per-node=16             #Tasks per node
#SBATCH --ntasks-per-socket=8            #Tasks per socket
#SBATCH --cpus-per-task=1                #Number of CPU per task
#SBATCH --mem-per-cpu=3600mb             #Memory (120 gig/server)
#SBATCH --distribution=cyclic:cyclic     #Distribute tasks cyclically 
#SBATCH --time=12:00:00                  #Walltime days-hh:mm:ss
#SBATCH --output=moose-%j.out            #Output and error log
#SBATCH --mail-type=END,FAIL             #When to email user
#SBATCH --mail-user=your-email@ufl.edu   #Email address to send mail to
#SBATCH --account=michael.tonks          #Allocation group name, add -b for burst job

srun --mpi=pmix_v3 ~/projects/moose/modules/combined-opt -i moose_input_file.i