Francisco Villamil

Logo

Carlos III - Juan March Institute
Department of Social Sciences
Universidad Carlos III de Madrid
francisco.villamil@uc3m.es

               

About
Research
Teaching
Data & Code
Other

Setting up R on a Google Cloud VM instance

Steps to set up a Google Cloud instance to run R, including dependencies needed to install packages for spatial statistics (Note: it is much easier to run R on AWS, see Louis Aslett’s website).

For connecting, see Google helps pages:

As of November 2021, instances in the Free Tier include e2-micro in us-west1 (and some other regions). It might change, so check the Google pricing page.

Instance details:

...@instance1:~$ lsb_release -cdir
Distributor ID:	Ubuntu
Description:	Ubuntu 16.04.7 LTS
Release:	16.04
Codename:	xenial

A basic R installation:

sudo apt-get update && sudo apt-get upgrade
sudo apt-get install r-base r-base-dev

However, by default, at least in my case, the only version of R available on the standard software sources is 3.2.x. To get R 4.0:

# Install dependencies to add new repositories via HTTPS
sudo apt install dirmngr gnupg apt-transport-https ca-certificates software-properties-common
# Add CRAN repository, depending on Ubuntu flavor
sudo apt-key adv --keyserver keyserver.ubuntu.com --recv-keys E298A3A825C0D65DFD57CBB651716619E084DAB9
sudo add-apt-repository "deb https://cloud.r-project.org/bin/linux/ubuntu $(lsb_release -cs)-cran40/"

Although in my case, it only work after I changed https to http. Edit the file (sudo nano /etc/apt/sources.list) and change last line to:

deb http://cloud.r-project.org/bin/linux/ubuntu xenial-cran40/

Install R:

sudo apt-get update && sudo apt-get upgrade
sudo apt-get install r-base r-base-dev

Install the basic packages needed to run rgdal:

sudo apt-get install gdal-bin proj-bin libgdal-dev libproj-dev

The spdep package might require additional libraries (error message: Configuration failed because libudunits2.so was not found). To fix this:

sudo apt-get install libudunits2-dev

The sf package requires GDAL 2.0.0, which is not in the standard Ubuntu library. In this case, installing an older version of sf does not seem to work. The solution is to add Ubuntu GIS unstable releases to the system’s software sources.

sudo add-apt-repository ppa:ubuntugis/ubuntugis-unstable
sudo apt-get update
sudo apt-get install libgdal-dev

Alternatively, see this thread on Stackoverflow.

One last thing. I often use pdfcrop to get rid of white margins in graphs (especially maps). It is not installed by default. To get it, install the texlive-extra-utils package:

sudo apt-get install texlive-extra-utils

Additional

To create the SSH key for a new computer:

ssh-keygen -t rsa -f ~/.ssh/[keyfilename] -C [full username]

Restrict access:

chmod 400 ~/.ssh/[keyfilename]

And remember to copy the public key (.pub) to SSH keys in the instance configuration (in Metadata). You can access your public key via e.g. cat ~/.ssh/key.pub.

You should now be able to connect via:

ssh -i ~/[path to private key] [short username]@[external IP]

And using scp into the VM in the same way:

scp -i ~/[path to private key] [localpath] [short username]@[external IP]:[path]

In case the process is too long, use tmux (or some alternative, see this) to keep make running and be able to access the screen again after disconnecting ssh.

tmux
make

And in the next session:

tmux attach

Use ctrl-b d to detach or get back to console.

To get a rough estimate of memory available:

cat /proc/meminfo | grep "Mem"

To run R and print out and save to file both stdout and stderr (Not sure if this works in Ubuntu, perhaps need to add /usr/bin/bash?):

Rscript --no-save --verbose file.R 2>&1 | tee file.Rout

In case Rscript fails to execute, see this.


Sources