Using Docker

Available Linux distributions for Dockerfiles are:

• Debian11 (Bullseye): referred as Dockerfile_Debian_Bullseye

• Debian Unstable (Sid): referred as Dockerfile_Debian_Sid

• Ubuntu 22.10 (Kinetic Kudu): referred as Dockerfile_Ubuntu_Kinetic

• Ubuntu 22.04 LTS (Jammy Jellyfish): referred as Dockerfile_Ubuntu_Jammy

• Ubuntu 20.04 LTS (Focal Fossa): referred as Dockerfile_Ubuntu_Focal

Build an image

We take the example of the Dockerfile_Debian_Bullseye Dockerfile. To build a different Dockerfile, you just need to substitute the Dockerfile name as referred above.

To build an image called gym-dssat:debian-bullseye form the Dockerfile named Dockerfile_Debian_Bullseye, simply run:

docker build https://gitlab.inria.fr/rgautron/gym_dssat_pdi.git\#dev:docker_recipes -t "gym-dssat:debian-bullseye" -f Dockerfile_Debian_Bullseye

Test a container

To check the gym-dssat:debian-bullseye image previously built, you can run the default example just with:

docker run gym-dssat:debian-bullseye

Or you can interactively run the Docker image with:

docker run -it gym-dssat:debian-bullseye bash

Docker images with Spack installation

Additionaly, these Dockerfiles to build docker images using Spack are provided. This recipes do not use pre-compiled packages, so gym-DSSAT and all its dependencies are built from sources (see info about gym-DSSAT Spack packages in Spack installation):

• Debian11 (Bullseye): referred as `Dockerfile_Debian_Bullseye_Spack`

• Ubuntu 22.04 LTS (Jammy Jellyfish): referred as `Dockerfile_Ubuntu_Jammy_Spack`

• Ubuntu 20.04 LTS (Focal Fossa): referred as `Dockerfile_Ubuntu_Focal_Spack`

• CUDA on Ubuntu 20.04 LTS (Focal Fossa): referred as `Dockerfile_CUDA_Spack`

• CentOS (latest): referred as `Dockerfile_CentOS_Spack`

These images can be built (it takes a long time, as every dependency will be compiled) and tested the same way than the other gym-DSSAT docker images.