M-Star CFD Docker Container¶

Download the dockerfile package: mstar-ubuntu24-cuda12.zip

This docker image can run the following M-Star components:

M-Star CFD Python Pre API
M-Star CFD Solver

Requirements:

M-Star floating license
Docker
Nvidia Driver (525.60.13 or newer)
Nvidia Container Toolkit: https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html

Build the image¶

docker build -t mstar:1 .

If you only need the solver in the container, you can use the Dockerfile.solver instead:

docker build -t mstar-solver:1 -f Dockerfile.solver .

Run tests (optional)¶

The included script do_tests.sh will run a number of tests to make sure the container is working correctly.

Modify the mstar_LICENSE environment variable to be your license server port@host.

docker run --rm \
            -v $PWD:/w \
            -w /w \
            -e mstar_LICENSE=5053@licenseserver \
            --user $(id -u):$(id -g) \
            --gpus all \
            -it mstar:1 ./do_tests.sh

Run solver with an existing case¶

docker run --rm \
            -v $PWD:/w \
            -w /w \
            -e mstar_LICENSE=5053@licenseserver \
            --user $(id -u):$(id -g) \
            --gpus all \
            -it mstar:1 mstar-cfd-mgpu -i input.xml -o out --gpu-auto --force

Run the M-Star Python Pre API¶

This example uses the included “test.py” script to create the agitiated test case example. Note that we are invoking python3.11 which is setup to use the M-Star API.

docker run --rm \
            -v $PWD:/w \
            -w /w \
            -e mstar_LICENSE=5053@licenseserver\
            --user $(id -u):$(id -g) \
            --gpus all \
            -it mstar:1 python3.11 test.py