We have a small cluster of linux machines (~15) with up to 20 users. Over the last few years we have written a few basic R scripts for doing basic statistical analysis. The problem I have had as a newbie system administrator is package management. Currently if someone writes a new script which requires a new package this package needs to be installed on all machines for all users. This seems like a rather caveman strategy (we also do this for python and perl modules). I am currently upgrading our operating systems from rhel5 to CentOS6 and would to do things properly.
Here are my initial thoughts?
In R on the first machine margay (which has CentOs6 running)
.libPaths()
points to "/usr/lib64/R/library" "/usr/share/R/library"
I was planning on creating a default .Rprofile for all users which contains the line .libPaths("/usr/share/R/library" ). I think this will then default install software packages to the aforementioned directory.
Currently the write permissions to usr/share/R/library/ is drwxr-xr-x. The first question would be should I all users to write to this directory or install R packages as root user?
On subsequent machines which I will install CentOS6, I was then thinking of using nfs to automount /usr/share/R/ on all other systems to point to margay:/usr/share/R/. SO I update the OS on machine 2 (lynx) and install R.How do I automount /usr/share/R/ on lynxto point to the equivalent directory on margay.
Does this seem like a sensible solution? If I install one new package on margay, would I be able to automatically run any R scripts which use this package on other machines? What are the pitfalls? Are the same versions of R necessary on all machines
Should I mount /usr/share/R/library on all other CentOS6 machines to point to /usr/share/R/library. Then if I install a new package as root on margay all other machines should see this new package? How do I do this if this is sensible
