The PETSc and FEniCS Dockerfiles and linked images here might be of interest to anyone running at an HPC facility that supports NERSC's Shifter. https://github.com/jedbrown/jedbrown-dockerfiles IMO, Shifter is far superior to any other HPC deployment mechanism and really should be supported everywhere. Using SLURM, it boils down to running (once) shifterimg pull docker:jedbrown/petsc:latest then sbatch --image docker:jedbrown/petsc:latest ... with srun -n 64 shifter ./your-petsc-app No need to futz with HPC build environments, modules, etc. You can build in the Shifter environment on a login node (see README). If you must use Python or other shared libraries, performance at scale is better (https://docs.nersc.gov/development/shifter/overview/) than any other low-hassle Python deployment technique. Comments welcome. I think we should consider automating the publishing of nightly updates for 'maint' and 'master', perhaps with a few common build configurations (int64, etc.). You'll notice that the Dockerfiles have some manual repairs for overzealous RPATH. We should have a configure option to avoid RPATH, and perhaps default to not using RPATH for any paths that don't need it (i.e., those that are already managed by the system ldconfig).