Version 1 (modified by valva, 11 years ago) (diff)


Required Software

WRF4G need the following software preinstalled:

  • Python, version >= 2.4 and < 3.0.
  • MySQLdb: Ubuntu: sudo apt-get install MySQL-python. Centos, RHEL: yum install MySQL-python

Platform Notes

WRF4G has been run only in 64bits platforms. Currently has been tested under the following O.S.:

  • Ubuntu 10.04 and 11.04: No known issues.
  • Centos 5: No known issues.
  • Debian 6.0: No known issues.


  • Download and unpack the distribution file WRF4G.tar.gz to the installation directory, for example your $HOME directory.
    cd $install_path (i.e cd $HOME)
    tar xzvf WRF4G.tar.gz
  • Set up the WRF4G user environment. (To avoid typing the export command everytime you want to use WRF4G, it is advisable to copy this line in $HOME/.bashrc)
export WRF4G_LOCATION=$install_path/WRF4G
export GW_LOCATION=$WRF4G_LOCATION/opt/drm4g_gridway-5.7

QuickStart? Guide

  • A video with the main steps can be downloaded from here
  • Start the WRF4G Framework:
wrf4g_framework start

  • Launch an example in the local machine:
cd $WRF4G_LOCATION/experiments/single/testc
  • Run wrf4g_ps to see the jobs' status
valva@sipc18:~/pruebasWRF/WRF4G/experiments/single/testc$ wrf4g_ps 
Realization Status   Chunks     Comp.Res   WN         Run.Sta       ext   %
testc       Done     3/3        mycomputer sipc18     Finished       0 100.00
  • wrf4g_ps returns the following information:
    • Realization: Realization name. It is taken from the field experiment_name in experiment.wrf4g.
    • Status: It can be take the following values: Prepared, Submitted, Running, Failed and Done).
    • Chunks [Chunk running/Total Chunks]: A realization is split into chunks. Each chunk is sent as a job.
    • Computer resource: Resource (cluster) where the job is running.
    • WN: Computing node where the job is running.
    • Run.Sta: Job status in the WN (Downloading data, running ungrib, real, wrf, ...)
    • ext: Exit Code. If exit code is different from 0, there has been an error. Error codes are explained in $WRF4G_LOCATION/lib/bash/
    • % : percentage of simulation finished.
  • Check the simulation output and log files in $WRF4G_LOCATION/repository/output

To run a different experiment, you only need to modify the following files according to your needs:

  • experiment.wrf4g: Experiment configuration: type of experment, simulation dates, restarts, physics,...
  • resources.wrf4g: With this file the resources used by a WRF experiment are configured: data repository, number of processes for MPI, version of WRF to run,.. By default there is a file that stores a configuration that affects to every experiment ($WRF4G_LOCATION/etc/resources.wrf4g). If a file with this name (resources.wrf4g) is found in the folder where experiment.wrf4g is stored, this file will be used instead of the application one ($WRF4G_LOCATION/etc/resources.wrf4g)
  • $WRF4G_LOCATION/etc/framework4g.conf: Database and computing resources configuration. After modifying this file, wrf4g_framework restart must be run.

For example, to add a new PBS cluster to run jobs:

  • First modify the file $WRF4G_LOCATION/etc/framework4g.conf to add more resources (More information about how to add resources):
    mycomputer         local://localhost?LRMS_TYPE=fork;NODECOUNT=1
    ui01              local://localhost?LRMS_TYPE=pbs;QUEUE_NAME=estadistica
  • Change the following fields in $WRF4G_LOCATION/etc/resources.wrf4g or copy $WRF4G_LOCATION/etc/resources.wrf4g where experiment.wrf4g is located:
    • The repository data if the cluster do not access them directly.
    • NP to the number of MPI processess you want to use
    • REQUIREMENTS if you want to choose the resource that will run the experiment. REQUIREMENTS ='HOSTNAME = "*ui01*";'
  • Restart the wrf4g_framework: wrf4g_framework restart
  • Submit the experiment with a different name (change the name in experiment.wrf4g)
    cd $WRF4G_LOCATION/experiments/single/snd50c

Veryfing installation

  • List the computing resources available for running jobs.
$ ~/wrf4g$ gwhost 
HID PRIO OS              ARCH  MHZ  %CPU    MEM(F/T)     DISK(F/T)      N(U/F/T) LRMS               HOSTNAME                      
0   1    GNU/Linux2.6.32 unkno 1000    0   1822/3270     2139/8662         0/1/1 FORK               mycomputer 


If you find any problem, please submit us a ticket!!