From FarmShare

Contents 1 Looking at installed packages 2 Which R are you using? 3 Installing CRAN Packages 4 R Sample Job 5 Another R Sample Job 6 Links 7 building our local R 8 lapack issues

Looking at installed packages

library();

Which R are you using?

Try run

 which R

Try run

 R --version

Installing CRAN Packages

Most CRAN packages can be installed per-user by running install.packages() in an interactive session:

install.packages("package_name", dependencies = TRUE)

R initially attempts to install to /usr/local/lib/R, but will prompt for the creation of a library subdirectory in ~/R (if necessary) and fall back to installation there when the initial attempt fails. If your package requires dependencies available from the standard Ubuntu repositories you can submit a HelpSU ticket requesting installation. We can install from the Debian/Ubuntu package repositories or into the shared FarmShare fs.

You can, of course, install R libraries into any arbitrary path and just add that path to your R env. That will probably break the next time R is upgraded to a new version, since your packages are built with the older version.

NOTE: when you install a package in corn, it will be available to you in Barley.

R Sample Job

Here's an example R file that generates a large array, fills it with some random numbers, then sleeps for 5mins. This happens to use up almost exactly 8GB of RAM.

$ cat 8GB.R 
x <- array(1:1073741824, dim=c(1024,1024,1024)) 
x <- gaussian()
Sys.sleep(300)

Here's an example SGE submit script that runs that R file.

You can submit it with just

 qsub r_test.script

Here are the output files that I get, one from stderr, one from stdout

$ cat r_test.script.e497 
tset: standard error: Function not implemented

Undefined tty
stdin: is not a tty

$ cat r_test.script.o497 
Warning: no access to tty (Bad file descriptor).
Thus no job control in this shell.
R version 2.12.1 (2010-12-16)
Copyright (C) 2010 The R Foundation for Statistical Computing
ISBN 3-900051-07-0
Platform: x86_64-pc-linux-gnu (64-bit)
R is free software and comes with ABSOLUTELY NO WARRANTY.
R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.
Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.


> x <- array(1:1073741824, dim=c(1024,1024,1024)) 
> x <- gaussian()
> Sys.sleep(300)
>

Those errors about tty and job control have something to do with shell startup and terminal settings, and are normal. Or specify the 'sh' shell in your job script

 #get rid of spurious messages about tty/terminal types
#$ -S /bin/sh

In the mail that you get about the ending of the job, the maxvmem number is actually incorrect, it is a known bug in this version of SGE. The R script on this page actually uses 8GB of vmem.

Another R Sample Job

R script, let's call it R-rjags.R

print("Hello World")
library(rjags)
#this just loaded some settings from that library
print("Finished")

Job script, let's call it R-jags.submit.script

#!/bin/bash

# use the current directory
#$ -cwd
# mail this address
#$ -M chekh@stanford.edu
# send mail on begin, end, suspend
#$ -m bes

R --vanilla --no-save < R-jags.R

Submit it to the test queue with a small memory requirement:

 qsub -l mem_free=200M -l testq=1 R-jags.submit.script

Looking at the output files, it errored out because R can't find the package rjags. You have two alternatives:

• include the R library from /mnt/glusterfs/software
• use modules to specify the full R install from /mnt/glusterfs/software

The first way, you would add this line to your R script:

 .libPaths(c("/mnt/glusterfs/software/free/R-2.15.0/lib/R/library", "/usr/lib/R/library"))

The second way, your script will look like this:

$ cat R-jags.submit.script
#!/bin/bash

# use the current directory
#$ -cwd
# mail this address
#$ -M chekh@stanford.edu
# send mail on begin, end, suspend
#$ -m bes

eval `tclsh /mnt/glusterfs/software/free/modules/tcl/modulecmd.tcl sh autoinit`
module load R-2.15.0
R --vanilla --no-save < R-jags.R

Links

Some other departments have some other more detailed examples:

• http://wiki.genomics.upenn.edu/index.php/HPC:ExamplesR
• http://me.eng.uab.edu/wiki/index.php?title=R-userinfo
• https://www.stanford.edu/dept/statistics/cgi-bin/projects/stat-sysadminwiki/index.php/R_Jobs

building our local R

Here's how I usually do it.

• cd /mnt/glusterfs/software/free
• wget http://cran.cnr.berkeley.edu/src/base/R-2/R-2.15.1.tar.gz
• tar zxvf R-2.15.1.tar.gz
• cd R-2.15.1
• ./configure
• make
• don't "make install"
• write new FarmShare module, e.g. /mnt/glusterfs/software/free/modules/tcl/modulefiles/R-2.15.1

lapack issues

If you see messages like:

  unable to load shared object '/usr/lib/R/modules//lapack.so':

most likely you're mixing R versions and libraries.

Use the module load command at the top of the page and figure out if you've set R library path to point to directories with older libraries at some point.

Test script:

$ cat lapack.r 
data(iris)
zz = lm(Sepal.Length ~., data = iris) 
summary(zz)

$ R --no-save < lapack.r