Installing Software for Your Research Group
It is quite likely that standard Red Hat-provided software will not meet your needs, e.g. Python version, R version, various Perl modules, etc. In such cases, it is probably best to install them in your group directory.
General Notes
Many packages assume you will be installing on your own computer, and therefore the instructions would not be suitable for installation on a shared computing cluster like Picotte.
Most software packages distributed as source code come with some installation instructions. Here are some things to watch out for.
- sudo is for running commands as an administrator. You will not be able to do this. Read on for other options.
- apt-get is used on Ubuntu and other Debian-based Linux distributions. Proteus uses Red Hat, which uses yum. In either case, only administrators are allowed to use those commands, as they install software system-wide.
- For Python modules, see Python#Installing Your Own Python Modules
- For R packages, see R#Third-Party R Packages
- For Perl, use cpan with the
local::lib
module[1]
Spack on Picotte
If you are installing software on Picotte, we suggest using Spack as it deals with a lot of the non-standard details of many scientific software codebases, and can handle installation and use of multiple compilers and versions of compilers. Spack can also create appropriate modulefiles.
PATH environment variable and ~/bin
Linux looks at the environment variable PATH
for a list of directories
(separated by colons) which may contain executables. To add a directory
to path in a persistent way, such that it happens at every login, modify
your ~/.bashrc file:
# .bashrc
# Source global definitions
if [ -f /etc/bashrc ]; then
. /etc/bashrc
fi
# User specific aliases and functions
export PATH=~/bin:${PATH}
Installation Details
We suggest the following layout:
- base install directory:
- Picotte:
/ifs/groups/myresearchGrp/opt
- Picotte:
- which means that standard GNU-like software installation would have
- Picotte:
PREFIX=/ifs/groups/myresearchGrp/opt
- Picotte:
- create your own Module File[2] so that your users may do something like "module load myresearchGrp-python" to load your group's version of Python
Compiling
As mentioned elsewhere,[3] there are many choices to be made when compiling software. If you wish to have code optimized specifically for Intel CPUs, use the Intel Composer XE toolset. For AMD CPUs, use Open64 and the ACML math library. If you don't want to have such specific versions, you should use the GNU compiler (GCC). By default, all users have the Bright-supplied GCC 4.8.1 module loaded.
GNU-style Software
GNU-style software is distributed with a configure script, which queries the system and environment for prerequisites, and appropriate compiler settings. The main one to change is the prefix, which sets the base directory for installation. Following the recommendation above, a typical invocation would be:
[juser@picotte001 somepkg-3.1.2]$ ./configure --prefix=/ifs/groups/myresearchGrp/opt
This should result in the following directories, or some subset:
/ifs/groups/myresearchGrp/opt/bin
/etc
/include
/lib
/share
You will have to modify your environment to add the bin
directory to
PATH
, and lib
to LD_LIBRARY_PATH
. We recommend doing this with a
modulefile.[4]
References
[1] Perl local::lib module CPAN page
[2] Environment Modules#Creating Your Own Module Files
[3] Compiling Quick Start Guide
[4]