Installation

Payu is currently only supported for users on the NCI computing systems, such as Raijin. If you wish to use payu on other systems, see the notes at the end of this document.

NCI Users

Payu is available for all users on Raijin.

To load payu, load the environment module:

module load payu

Local installation

If you want to use the latest version of payu, then you can install it locally from the codebase:

git clone https://github.com/marshallward/payu
cd payu
python setup.py install --user

Payu depends on the following modules:

If you have a recent version of setuptools installed, then these will be installed automatically. But older installations, such as based on distutils, will require a manual installation.

General Use

Payu is not supported for general use, and it would be a tremendous surprise if it even worked on other machines. In particular, the following services are presumed to be available:

  • Environment Modules: Not only do we assume support for environment modules, but we also assume the existence of certain modules, such as an OpenMPI module and particular versions of Python.
  • PBS Scheduler: Payu relies on executables that are provided with most PBS implementations, such as Torque or PBSPro. Most of the argument flags are currently based around PBSPro conventions.
  • MPI: Jobs are submitted via mpirun and most of the argument flags are based on the OpenMPI implementation. We also rely on Raijin’s internal preprocessing scripts for a few tasks.

There are also some additional assumptions based on the architecture of Raijin.

Despite these rather strict requirements, there is opportunity for generalising payu for other platforms, such as through new drivers for alternative schedulers and parallelisation platforms. Please contact the mailing list if you are interested in porting payu to your machine.