Skip to content

Setting up HARP on ATOS

Jump to bottom
Samuel Viana edited this page Jun 17, 2024 · 19 revisions

Instructions for starting to use HARP on ATOS.

These notes were created by adapting to ATOS installation steps described here:

https://harphub.github.io/harp_tutorial/

  • First, connect to hpc-login with display capabilities (-X), load the proj, netcdf & R modules, and optionally also rstudio. Suggestion: add these modules to your .bashrc. Then launch R or rstudio through command line:

ssh -X hpc-login

module load proj netcdf4 R rstudio

R or rstudio &

  • In the rstudio console or R console, follow next steps:
  • Install the "remotes" package and use to install the last tagged version of harp from the github repository:

install.packages("remotes")

You will be asked to use a personal library path for the installation; answer "yes" and accept the path suggestion. Then select a CRAN mirror close to your location.

remotes::install_github("harphub/harp", ref = "v0.2.2")

If you get an error related to "GITHUB_PAT", you're probably using a github key stored in your .Renviron file. You need to unset it with Sys.unsetenv("GITHUB_PAT") to make the previous command work.

You will probably be asked to update other packages with dependencies. Whenever asked, select "update All".

  • Install extra packages for grib & FA reading (only in case that you need to interpolate model data to station locations)

(But first, make sure you have the ecmwf-toolbox module loaded in your terminal (it loads eccodes, which is needed for the Rgrib2 package)):

remotes::install_github("harphub/Rgrib2")

remotes::install_github("harphub/Rfa")

  • Extra packages useful for different tasks, i.e. deploying shiny web apps in shinyserver.io :

install.packages("rsconnect")

install.packages("tidyverse")

install.packages("here")

install.packages("argparse")

  • After the installation, exit R or rstudio and launch it again. Check that the harp package can be loaded:

library(harp)

  • ############ The content below is optional, follow it only if you need to get to know the harp system in depth ##########

  • If you want to follow the tutorial from the url above, you will need to download the test data:

remotes::install_github("harphub/harpData")

  • And then create a new project for the tutorial:

Click on:   File > New Project

Select:   - New Directory   - New Project

Choose a directory name (e.g. harp_tutorial)

Browse to a directory under which your project should reside

Click on:     Create Project

  • Linking the test data to the project directory:

library(here)

set_here()

Now you will create a data directory and link the contents of the harpData package to the data directory.

dir.create(here("data"))

for (data_dir in c("grib", "netcdf", "vfld", "vobs")) { file.symlink(from = system.file(data_dir, package = "harpData"), to = here("data")) }

If you were successful with these steps, you're ready to start using HARP i.e. with one of the examples in the HARP tutorial:

https://harphub.github.io/harp_tutorial/interpolating-model-outputs.html

Link to useful HARP information

Clone this wiki locally