On our cluster I have installed R to a staged location that other users share for use, but I remain the user that manages the staged installed binaries. The cluster doesn't share the user's home location (they share a mounted path where the data is placed). We want to know if we are able to have the users specify where their bioconductor installs go so they don't have to ask for my help every time they want to use packages. I suspect the fact that they have access to the shared R script library location might interfere with the usual mode (since it wasn't installed using root as I am not an admin nor do I have access to become one).

What should I tell the other users to do? Do I have to be the one to install these packages or are there ways to manually choose the library location for a user-specific install?

Make the installed library .libPaths()[1] or other staging libraries read-only. A user trying to install a package will then be prompted to create and install new packages in their home directory, e.g., ~/R/x86_64-unknown-linux-gnu-library/3.1. One hazard here is that biocLite() / install.packages() will complain that installed directory not writeable and similar when a package in .libPaths()[1] becomes out-of-date; this causes user confusion, and potentially incompatible staged and user package versions.

A system administrator might also arrange to specify R_LIBS_USER in a system-wide way. See ?R_LIBS_USER for further information.

The problem is the paths associated with ~/* are in the indivdual system's paths, they are not in the shared fibre-link drive that the cluster machines all have access to, so he wouldn't be able to submit jobs and have them see those packages. It NEEDS to go somewhere that isn't located as a subdirectory of ~