-
Juan Escobar authoredJuan Escobar authored
README_MNH_CONDA 5.56 KiB
#MNH_LIC Copyright 1994-2021 CNRS, Meteo-France and Universite Paul Sabatier
#MNH_LIC This is part of the Meso-NH software governed by the CeCILL-C licence
#MNH_LIC version 1. See LICENSE, CeCILL-C_V1-en.txt and CeCILL-C_V1-fr.txt
#MNH_LIC for details. version 1.
#
# Written by : Juan ESCOBAR , for the Meso-NH Support , 16/07/2021
#
###############################################################################################
# #
# How to install the python package needed for the new graphic script of the version MNH-55X #
# test case with "conda" or "miniconda" #
# #
###############################################################################################
#
# MAP
#
# I) Installation from Scratch , on your PC for example
# a) Installation of mini-conda
# b) Creation of the conda environmenent "mnh_conda_cartopy_offlinedata"
# c) Using the environment with mesonh
# d) Optional: export the environment with "conda pack"
#
# OR
#
# II) Installation from a packed tar.gz file, created with "conda pack"
# for Linux Supercomputer without internet connection
# I) Installation from scratch, on your PC for example
# =======================================================
# I- a) Installation of mini-conda
----------------------------------
# If you don't have already the "conda" command in your PC
# first install "miniconda"
# Creation of the install directory, for example in your $HOME :
export MNH_MINICONDA=${HOME}/MINICONDA
mkdir -p ${MNH_MINICONDA}
cd ${MNH_MINICONDA}
# get/install the last version of miniconda for X64 ( here with python 3.9 )
wget https://repo.anaconda.com/miniconda/Miniconda3-py39_4.9.2-Linux-x86_64.sh
bash Miniconda3-py39_4.9.2-Linux-x86_64.sh
# during the installation
# - accept the licence -> yes
# - give the path of installation > #path -> /home/{your_login}/MINICONDA/miniconda3
# when installed initialization of miniconda
source ${MNH_MINICONDA}/miniconda3/etc/profile.d/conda.sh
# REM : the other solution is via "conda init bash" which changes the "~/.bashrc"
# but I don't like it ...
# I- b) Creation of the conda environment "mnh_conda_cartopy_offlinedata"
---------------------------------------------------------------------------
#
# Now , creation of the conda environment "mnh_conda_cartopy_offlinedata"
# /!\ about 3Go of disk space are needed
#
conda create -n mnh_conda_cartopy_offlinedata
conda activate mnh_conda_cartopy_offlinedata
conda install -c conda-forge netcdf4 cartopy matplotlib
#
# this is the minimum packages needed for the MesoNH script of the test cases
#
#
# for computer without internet connection ( for example compute nodes of super computers )
# you could add this package to download in advance cartopy data files
#
conda install -c conda-forge cartopy_offlinedata
# REM : this files add about 2OO Mo and are located in "$CONDA_PREFIX/share/cartopy"
( CONDA_PREFIX is set when you initialize/ source the (mini)conda command
# I-c) Using the environment with mesonh
-----------------------------------------
#
# Now to use this conda environment with MesoNH , the best way is to put
# these lines at the end of your "profile_mesonh{the_good_one}"
#
export MNH_MINICONDA=${HOME}/MINICONDA
source ${MNH_MINICONDA}/miniconda3/etc/profile.d/conda.sh
conda activate mnh_conda_cartopy_offlinedata
# and then use mesonh as usual by sourcing this file
cd .../MNH.../conf
. profile_mesonh{the_good_one}
# I-d) Optional: export the environment with "conda pack"
-----------------------------------------------------------
#
# After this, if you were to export your binary installation in a "tar.gz" file to another 'compatible' computer
# first add the 'conda-pack' package
conda install -c conda-forge conda-pack
# And now export the environment with
conda pack -n mnh_conda_cartopy_offlinedata -o mnh_conda_cartopy_offlinedata.tar.gz
# II) Installation from a packed tar.gz file , created with "conda pack"
# for Linux Supercomputer without internet connection
=========================================================
#
# On Supercomputer without internet connection , like TGCC/CEA Irene or CINES/Occigen
# you could not install package with the conda command.
# As a work-around you could download and use a "tar.gz" as prepared on my Linux OpenSuse PC .
# ( as explained in the previous sessions )
#
# REM : you don't need to load any 'module files' like python to do this work-around
# <-> no conda/python needed in your PATH
# First, on your PC computer download the "tar.gz" I prepared and uploaded on the MesoNH web page
wget http://mesonh.aero.obs-mip.fr/mesonh/tmp/MINICONDA/mnh_conda_cartopy_offlinedata.tar.gz
# Then, to use it for example on Irene or Occigen computer
# Copy the "tar.gz" for you PC to Super computer in $CCCWORKDIR (or $WORKDIR ) :
# From your PC
scp mnh_conda_cartopy_offlinedata.tar.gz irene:.
# On Irene
cd $CCCWORKDIR
mv ~/mnh_conda_cartopy_offlinedata.tar.gz $CCCWORKDIR/.
mkdir mnh_conda_cartopy_offlinedata
tar xvf mnh_conda_cartopy_offlinedata.tar.gz -C mnh_conda_cartopy_offlinedata
# Activation of the package, for example at the end of your "profile_mesonh{the_good_one}
source $CCCWORKDIR/mnh_conda_cartopy_offlinedata/bin/activate
# That's all, you can now use the packages included in the "tar.gz"
#/!\
#/!\ Warning: you don't have "conda" so you cannot modify this installation/package .
#/!\
Good tests .