Changes to ARCHER Python Environment

To improve the consistency and simplify the setup on the Python environment on ARCHER for users we are planning to make a number of changes during the At Risk session on Wednesday 12 August 2015.

A full description of the planned setup can be found in the new Python chapter of the ARCHER User Guide:

• Using Python on ARCHER

The main changes are highlighted below. Please let the ARCHER Helpdesk know if you have any questions or concerns about these changes.

Anaconda will become the default Python environment for users

The current default Python environment for users on login is a custom version of Python 2.7.6 compiled for the ARCHER compute nodes (the python/2.7.6 module).

From Wednesday 12 August 2015 the default Python environment on login will become the Anaconda scientific Python distribution through the anaconda/2.2.0-python2 module. This environment is designed for data analysis on the login and post-processing nodes and includes a large number of commonly used Python modules.

For a full description of the Anaconda Python environment on ARCHER see:

• Python for data analysis: the Anaconda distributions

python modules will become python-compute

The current python modules that are compiled for use on the ARCHER compute nodes will be renamed python-compute to make their purpose clearer.

Additional Python packages that have been compiled for use on the compute nodes (by linking against performance or parallel libraries) will be renamed to be prefixed by pc- to make it easy for users to see which packages are available.

For a full description of the compute node Python environment on ARCHER see:

• Python for HPC: the native distributions

List of renamed modules

The following table lists the Python modules that will be renamed as part of this work.

Modules to be removed

The following modules will be removed from the system as the packages are supported through the Anaconda environment:

• docultils
• ply
• six