Docker

Instructions on using Docker to run ModelE
Docker container only contains gfortran
ModelE is on the host computer, and accessed via mounts

Dockerfile

This Dockerfile sets up the virtual machine

FROM ubuntu:22.04

LABEL authors="Ken Mankoff"
LABEL maintainer="[email protected]"

# system environment
ENV DEBIAN_FRONTEND noninteractive

RUN apt-get -y update && apt-get install -y --no-install-recommends --no-install-suggests \
      gfortran \
      libopenmpi-dev \
      libnetcdf-dev \
      libnetcdff-dev \
      make \
      wget \
  && apt-get autoremove -y \
  && apt-get clean -y \ 
  && rm -rf /var/lib/apt/lists/*

RUN echo LANG="en_US.UTF-8" > /etc/default/locale

ENV LANGUAGE en_US.UTF-8
ENV LANG C
ENV LC_ALL C
ENV LC_CTYPE C

ENV SHELL /bin/bash

# create a user
RUN useradd --create-home user && chmod a+rwx /home/user
ENV HOME "/home/user"
WORKDIR /home/user

ENV MODELERC /home/user/.modelErc

# # switch the user
USER user

# CMD ["/usr/bin/bash", "--version"]
CMD ["gfortran", "--version"]

Build

docker build -t modele .

Run

 # run it with default CMD
docker run -it modele

# run interactively
docker run -it modele bash

# run with your user, ${HOME} mounted at /user/home, and ../modelE mounted at /modelE
docker run -it --user $(id -u):$(id -g) --mount type=bind,src=${HOME},dst=/home/user --mount type=bind,src=$(pwd)/../modelE,dst=/modelE modele bash
ls /modelE
ls /ModelE_Support
exit

Use locally

Set up ModelERC

You can mount your host modelE source code folder anywhere in the container
- In this example, the source is mounted at /modelE
- In the example above, we mounted the host ../modelE to the container /modelE with --mount type=bind,src=$(pwd)/../modelE,dst=/modelE
You can mount your host ModelE_Support folder anywhere in the container
- In this example, ModelE_Support is mounted at /ModelE_Support and the ${MODELERC} file expects it there
In this example, ${MODELERC} on the host is ${HOME}/.modelErc.docker, and in the container, is found at /home/user/.modelErc.docker
The Dockerfile also sets the following bash environment variables
- MODELERC=/home/user/.modelErc but we override it when we launch docker to point to /home/user/.modelErc.docker
- You can inject (or override) environment variables into docker with --env VAR="value"
Below is an example file is saved as ${HOME}/.modelErc.docker.
- We need to update MODELERC to point to this, knowing that host ~/ will be mounted at /home/user/
- We need to tell docker to mount the source code and ModelE_Support folder as expected below

# This file contains global options for modelE. 
# By default it assumes that the directory structure for modelE runs
# is set under /ModelE_Support .

## Directory structure ##

# DECKS_REPOSITORY - a directory for permanenet storage of run info.
# All rundecks that you create will be copied to this directory. 
DECKS_REPOSITORY=/ModelE_Support/prod_decks

# CMRUNDIR - directory to which all run directories will be linked.
# This directory will be searched by most scripts for locations of 
# specific runs.
CMRUNDIR=/ModelE_Support/prod_runs

# GCMSEARCHPATH - directory to search for gcm input files.
# All necessary input files should be copied or linked to this directory.
GCMSEARCHPATH=/ModelE_Support/prod_input_files

# EXECDIR - path to directory with modelE scripts and with some
# executables. This directory should contain the scripts from modelE/exec.
EXECDIR=/ModelE_Support/exec

# SAVEDISK - a directory where all run directories (which will contain
# all output files such as rsf, acc etc.) will be created. This should
# be big enough to accomodate all model output.
SAVEDISK=/ModelE_Support/huge_space

## External libraries ##

# Some of these options can be provided by environment modules (if you 
# use them). Specify here only what is necessary. Options specified 
# here will overwrite options proviided by environment modules.

# NETCDFHOME - path to location of netcdf installation directory. 
# NETCDFHOME=/opt/netcdf/3.6.3
NETCDFHOME=/usr
NETCDFLIBDIR=/usr/lib/x86_64-linux-gnu

# MPI - set to YES if you want to compile the model for parallel 
# execution on multiple CPU cores. Keep in mind, that functional 
# MPI library should be installed on your computer and its type 
# and location should be specified below.
# This option can be overwritten from the compile line.
MPI=YES

# MPIDISTR - the MPI distribution you are using. Currently supported 
# distributions are: 'intel, 'openmpi', 'mpich2', 'mvapich2', 'SCALI',
# 'mpt' 
MPIDISTR=openmpi

# MPIDIR - path to the MPI installation directory. (Needs to be set
# only if compiler can't find correct MPI library and include files by
# default)
MPIDIR=/usr

# MPILIBDIR - path to the location of MPI library. Set it only if 
# it is different from the default $MPIDIR/lib
MPILIBDIR=/usr/lib/x86_64-linux-gnu/openmpi/lib

# MPIINCLUDEDIR - path to location of MPI include files. Set it only
# if it is different from the default $MPIDIR/include
MPIINCLUDEDIR=/usr/lib/x86_64-linux-gnu/openmpi/include/

# ESMF5_DIR - path to the installation directory of ESMF (version 5)
# library. (Required only for Cubed Sphere simulations)
# ESMF5_DIR=

# ESMF_BOPT - optimization level of ESMF library. (Should only be used
# togeteher with ESMF5_DIR)
# ESMF_BOPT=O

## Architecture and compiler

# ABI - Application Binary Interfaces. This variable specifies the
# architecture you are using. The valid values are '64' and '32'. 
# On most modern systems you should use '64'. Use '32' if your
# hardware or compiler support only 32-bit binaries.
ABI=64

# COMPILER - specifies the Fortran compiler you are using. Currently
# only 'intel' and 'gfortran' are supported. ('nag' has partial
# support on development branch.) If you are using Modules for
# Environment Management, then this variable may already be set in the
# environment. In this case you don't need to set it here.
COMPILER=gfortran

## General User Preferences ##

# MAILTO - email address of the user. When the program ends/crashes
# all notifications will be sent to this address. Leave empty 
# or unset if you don't want to receive these emails
MAILTO=

# UMASK - the value of 'umask' you want to use for model runs. The files
# inside the run directory will have permissions set according to this
# mask.
UMASK=022

# OVERWRITE - can "gmake rundeck" overwrite files already in repository?
# (i.e. in the directory DECKS_REPOSITORY)
OVERWRITE=NO

# OUTPUT_TO_FILES - if set to YES all errors and warnings will be sent
# to files with the names <source_name>.ERR
OUTPUT_TO_FILES=YES

# VERBOSE_OUTPUT - if set to YES gmake will show compilation commands
# and some other information. Otherwise most of the output will be
# suppressed
VERBOSE_OUTPUT=NO

Compile

Note, before compiling ModelE you might need to make some changes for it to work with gfortran, such as

git diff

diff --git a/config/compiler.gfortran.mk b/config/compiler.gfortran.mk
index a991e1a05..41878da23 100644
--- a/config/compiler.gfortran.mk
+++ b/config/compiler.gfortran.mk
@@ -63,10 +63,10 @@ endif
 # flags needed for particular releases
 
 FFLAGS_RELEASE =
-ifneq (,$(filter 10 11 12,$(GFORTRAN_RELEASE)))
+ifeq ($(GFORTRAN_RELEASE),10)
 FFLAGS_RELEASE += -fallow-argument-mismatch
 endif
-ifneq (,$(filter 8 9 10 11 12,$(GFORTRAN_RELEASE)))
+ifneq (,$(filter 8 9 10,$(GFORTRAN_RELEASE)))
 FFLAGS_RELEASE += -fwrapv
 endif

docker run -it \
       --user $(id -u):$(id -g) \
       --env MODELERC="/home/user/.modelErc.docker" \
       --mount type=bind,src=${HOME},dst=/home/user \
       --mount type=bind,src=$(pwd)/../modelE,dst=/modelE \
       --mount type=bind,src=$(pwd)/../ModelE_Support,dst=/ModelE_Support \
       modele \
       bash

cd /modelE/decks/
make rundeck RUNSRC=E4M20 RUN=E4M20_docker_test01
../exec/get_input_data -w E4M20_test01 ../../ModelE_Support/prod_input_files/

make clean RUN=E4M20_docker_test01
make -j setup RUN=E4M20_docker_test01

Run

docker run -it \
       --user $(id -u):$(id -g) \
       --env MODELERC="/home/user/.modelErc.docker" \
       --mount type=bind,src=${HOME},dst=/home/user \
       --mount type=bind,src=$(pwd)/../modelE,dst=/modelE \
       --mount type=bind,src=$(pwd)/../ModelE_Support,dst=/ModelE_Support \
       modele \
       bash

cd /modelE/decks/
# My computer has 4 cores, each dual thread, total of 8 CPU
../exec/runE E4M20_docker_test01 -cold-restart -np 2
../exec/runE E4M20_docker_test01 -np 4 # 18 model-yr/human-day
MPI_FLAGS=--use-hw-cputhreads ../exec/runE E4M20_docker_test01 -np 8 # 15 model-yr/human-day

Deploy for sharing

# docker tag local-image:tagname new-repo:tagname
docker tag modele mankoff/modele:gfortran
docker login -u "user" -p "pass" docker.io
docker push mankoff/modele:gfortran

Use on discover

Can’t run docker on discover
Can run docker images using singularity.
singularity is like docker but for HPC systems
Easily build micro (or full) VMs to run whatever applications you want.

Singularity example

As an example, if you need to run lynx to do some web-browsing from discover,

# Pulling down VMs can take some space, don't do it in ~/
cd ${NOBACKUP} 
export SINGULARITY_CACHEDIR=${NOBACKUP}/.singularity/cache

mkdir -p singularity
cd singularity

module load singularity
singularity pull lynx.sif docker://nbrown/lynx
singularity exec -B ./:${TMPDIR} lynx.sif lynx http://www.giss.nasa.gov

ModelE in Singularity running via SLURM batch jobs

cd ${NOBACKUP} 
export SINGULARITY_CACHEDIR=${NOBACKUP}/.singularity/cache

mkdir -p singularity
cd singularity

module load singularity
singularity pull modele.sif docker://mankoff/modele

# TODO - Build script to submit SLURM job that launches runE in singularity container

Name		Name	Last commit message	Last commit date
Latest commit History 1 Commit
Dockerfile		Dockerfile
README.org		README.org
modelErc.docker		modelErc.docker

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

Docker

Dockerfile

Build

Run

Use locally

Set up ModelERC

Compile

Run

Deploy for sharing

Use on discover

Singularity example

ModelE in Singularity running via SLURM batch jobs

About

Releases

Packages

Languages

NASA-GISS/docker

Folders and files

Latest commit

History

Repository files navigation

Docker

Dockerfile

Build

Run

Use locally

Set up ModelERC

Compile

Run

Deploy for sharing

Use on discover

Singularity example

ModelE in Singularity running via SLURM batch jobs

About

Resources

Stars

Watchers

Forks

Releases

Packages 0

Languages

Packages