R Guide

R is a GNU project for statistical computing and graphics. The R language is widely used among statisticians for developing statistical software and data analysis. See Wikipedia.

There are several versions of R installed on the HPC Cluster. Users can install their own packages in their home directories.

Rstudio cannot be used on HPC

We apologize for the inconvenience that has brought to you. Please write and debug your R code on your own computer and copy it to cluster to run. Thank you for your cooperation.

Loading R module

To list available versions of R, type

  module avail r

At the time of writing, the most up-to-date version installed on the cluster is 4.1.2. To load it, run

  module load r/4.2.1

To make R 4.2.1 autoload on login

  module initadd r/4.2.1

Interactive R use with slurm

To run an interactive R session with 24 cores using the "general" partition, you will want to do the following:

fisbatch --ntasks=24 --nodes=1 --exclusive

Once you are in an interactive session, you can load one of the R modules and start working with it interactively.

module purge
module load r/4.2.1
R
... # here will be the interactive commands with R
exit

Install an R package

Local package install

If we want to install an R package, for instance, data.table, we advise first creating a directory to store the locally installed packages.

The command below will create the directory rlibs in your home directory.

mkdir ~/rlibs

Next, assuming an R module has already been loaded, we start an R session and inform R to look for packages installed at ~/rlibs/ too.

R
.libPaths("~/rlibs")
install.packages("data.table", lib = "~/rlibs", repo = "https://cloud.r-project.org/")

Note that:

1. We need to specify the repo from which the packages will be downloaded from. For a list of options, take a look at https://cran.r-project.org/mirrors.html.

2. We need to set lib when installing the package to tell R where to install it.

Now, whenever you start a new R session or use Rscript to run something, you will need to tell R that your packages are stored in the ~/rlibs directory. There are two ways to do it. In the first one, is to add

.libPaths("~/rlibs")

to the beginning of all of your scripts (and execute it once when working on an interactive session).

The next option is to create a .Rprofile file. This can be achieved by running

echo '.libPaths("~/rlibs")' > .Rprofile

Some packages depend on other libraries and are harder to be installed locally. For example, sf is a package to deal with spatial (GIS) data. It depends on geos, gdal, and proj. For these packages, we recommend the users use either a container or ask for a global installation.

Global package install

Please submit a ticket with the packages you would like installed and the R version, and the administrators will install it for you.

Submitting jobs

Serial

Assume that you have a script called helloworld.R with these contents:

cat('Hello world!')

Submit to Slurm scheduler using sbatch

 sbatch -n 1 R CMD BATCH helloworld.R

Submit to Slurm scheduler with multi-threading:

 sbatch -n 1 -c 20 --exclusive R CMD BATCH helloworld.R 
 # use "-c 20" to setup multi-threading for R

When the job completes output will be written to helloworld.Rout

MPI

The Rmpi package has to be installed to work with MPI in R. In addition, you have to either install locally or load the module of a specific MPI implementation.

An example of how to install Rmpi using the module openmpi/gcc/64/1.10.7 can be found below. Note that, the package snow has to be installed as well.

module load r/4.2.1
module load openmpi/4.1.4
R
.libPaths("~/rlibs") # assuming you are installing your 
                     # packages at the ~/rlibs folder
install.packages("Rmpi", lib = "~/rlibs", repo = "https://cloud.r-project.org/",
                 configure.args = "--with-mpi=/gpfs/sharedfs1/admin/hpc2.0/apps/openmpi/4.1.4/")
install.packages("snow", lib = "~/rlibs", repo = "https://cloud.r-project.org/")

To submit a MPI slurm job, we created the submit-mpi.slurm file (see code below). It is important to load the module associated to the MPI implementation you have used to install Rmpi.

#!/bin/bash
#SBATCH -p general
#SBATCH -n 30

source /etc/profile.d/modules.sh
module purge
module load r/4.2.1 openmpi/4.1.4

# If MPI tells you that forking is bad uncomment the line below 
# export OMPI_MCA_mpi_warn_on_fork=0

Rscript mpi.R

Now create the mpi.R script:

library(parallel)

.libPaths("~/rlibs")

hello_world <- function() {
    ## Print the hostname and MPI worker rank.
    paste(Sys.info()["nodename"],Rmpi::mpi.comm.rank(), sep = ":")
}

cl <- makeCluster(Sys.getenv()["SLURM_NTASKS"], type = "MPI")
clusterCall(cl, hello_world)
stopCluster(cl)

Run the script with:

sbatch submit-mpi.slurm

In your slurm output you will see a message from each of the MPI workers.

Read R's built-in "parallel" package documentation for tips on parallel programming in R: https://stat.ethz.ch/R-manual/R-devel/library/parallel/doc/parallel.pdf