FEniCS TACC Docker image

Here we build a Docker Image with a working FEniCS installation based upon the TACC tacc-ubuntu18-mvapich2.3-ib and tacc-ubuntu18-impi19.0.7-common images.

Supported systems: Frontera, Maverick2.

Images are available on DockerHub: mvapich compiler, impi compiler

For more information regarding containers at TACC see here.

MVAPICH Compiler

Build the image (locally)

To build the image:

cd tacc-mvapich2.3-ib
docker build -t fenics-2019.1-tacc-mvapich2.3-ib .

To upload the image on DockerHub

docker login
docker tag fenics-2019.1-tacc-mvapich2.3-ib uvilla/fenics-2019.1-tacc-mvapich2.3-ib:latest
docker push uvilla/fenics-2019.1-tacc-mvapich2.3-ib:latest

Run the image on your laptop using Docker

To run the container locally using docker, use the command

docker run -e MV2_SMP_USE_CMA=0 -e MV2_ENABLE_AFFINITY=0 -ti -v $(pwd):/home1/ uvilla/fenics-2019.1-tacc-mvapich2.3-ib:latest /bin/bash 

Your current directory will be mapped to /home1 on the container. The enviromental variables MV2_SMP_USE_CMA=0 and MV2_ENABLE_AFFINITY=0 are needed when running local to avoid some nasty MPI errors.

Run the image on TACC resources via Apptainer

Download the Docker image from DockerHub and save it in $SCRATCH. Note: we must request a node from the interactive queue via idev to use apptainer.

idev -N 1
cd $SCRATCH
module load mvapich2-x/2.3
module load tacc-apptainer/1.1.8
apptainer pull docker://uvilla/fenics-2019.1-tacc-mvapich2.3-ib:latest

From a compute node, you can test the container with

MV2_SMP_USE_CMA=0 ibrun apptainer run $SCRATCH/fenics-2019.1-tacc-mvapich2.3-ib_latest.sif hellow

First time run of hIPPYlib

Before being able to use hIPPYlib in parallel, the FEniCS jit compiler needs to create new folders in your $HOME/.cache directory. To do so, you need to run python -c "import hippylib" in serial (again from a compute node).

git clone https://github.com/hippylib/hippylib.git
cd hippylib
MV2_SMP_USE_CMA=0 ibrun -n 1 apptainer run $SCRATCH/fenics-2019.1-tacc-mvapich2.3-ib_latest.sif python3 -c "import hippylib"

Running hIPPYlib in parallel from the interactive queue

Now you can run the subsurface Poisson example in parallel

cd applications/poisson
MV2_SMP_USE_CMA=0 ibrun apptainer run $SCRATCH/fenics-2019.1-tacc-mvapich2.3-ib_latest.sif python3 model_subsurf.py

Slurm submission to the normal queue

For an example of (non-interactive) job submission see hippylib-ex.slurm. The script assumes that the apptainer contanaier is in $SCRATCH/fenics-2019.1-tacc-mvapich2.3-ib_latest.sif and hIPPYlib in $SCRATCH/hippylib.

sbatch hippylib-ex.slurm

Intel MPI compiler

Run the image on your laptop using Docker

To run the container locally using docker, use the command

docker run -e FI_PROVIDER=sockets -ti -v $(pwd):/home1/ uvilla/fenics-2019.1-tacc-impi19.0.7-common:latest /bin/bash 

Your current directory will be mapped to /home1 on the container. The enviromental variables FI_PROVIDER=sockets is needed when running local to avoid some nasty MPI errors.

Run the image on TACC resources via Apptainer

Download the Docker image from DockerHub and save it in $SCRATCH. Note: we must request a node from the interactive queue via idev to use apptainer.

idev -N 1
cd $SCRATCH
module load tacc-apptainer/1.1.3
apptainer pull docker://uvilla/fenics-2019.1-tacc-impi19.0.7-common:latest

From a compute node, you can test the container with

ibrun apptainer run $SCRATCH/fenics-2019.1-tacc-impi19.0.7-common_latest.sif hellow

First time run of hIPPYlib

Before being able to use hIPPYlib in parallel, the FEniCS jit compiler needs to create new folders in your $HOME/.cache directory. To do so, you need to run python -c "import hippylib" in serial (again from a compute node).

git clone https://github.com/hippylib/hippylib.git
cd hippylib
ibrun -n 1 apptainer run $SCRATCH/fenics-2019.1-tacc-impi19.0.7-common_latest.sif python3 -c "import hippylib"

Running hIPPYlib in parallel from the interactive queue

Now you can run the subsurface Poisson example in parallel

cd applications/poisson
ibrun apptainer run $SCRATCH/fenics-2019.1-tacc-impi19.0.7-common_latest.sif python3 model_subsurf.py

MVAPICH Compiler with Python3.12

The Docker image can be found on DockerHub: mvapich compiler.

We select package versions compatible with Python 3.12. Please find the updated list below:

fenics-dijitso            2019.2.0.dev0
fenics-dolfin             2019.2.0.dev0
fenics-ffc                2019.2.0.dev0
fenics-fiat               2019.2.0.dev0
fenics-ufl-legacy         2022.3.0
jupyterlab                4.1.2
matplotlib                3.8.3
mpi4py                    3.1.5
mshr                      2019.1.0
notebook                  7.1.0
numpy                     1.26.4
petsc4py                  3.20.2
pybind11                  2.11.1
scipy                     1.12.0
slepc4py                  3.20.1