Using DPF Through Docker

You can run DPF within a container on any OS using docker.

There are several situations in which it is advantageous to run DPF in a containerized environment (e.g. Docker or singularity):

  • Run in a consistent environment regardless of the host OS.

  • Portability and ease of install.

  • Large scale cluster deployment using Kubernetes

  • Genuine application isolation through containerization.

Installing the DPF Image

There is a docker image hosted on the DPF-Core GitHub repository that you can download using your GitHub credentials.

Assuming you have docker installed, you can get started by authorizing docker to access this repository using a personal access token. Create a GH personal access token with packages read permissions according to Creating a personal access token.

Save that token to a file with:

echo XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX > GH_TOKEN.txt

This lets you send the token to docker without leaving the token value in your history. Next, authorize docker to access this repository with:

GH_USERNAME=<my-github-username>
cat GH_TOKEN.txt | docker login docker.pkg.github.com -u $GH_USERNAME --password-stdin

You can now launch DPF directly from docker with a short script or directly from the command line.

docker run -it --rm -v `pwd`:/dpf -p 50054:50054 docker.pkg.github.com/pyansys/dpf-core/dpf:v2021.1

Note that this command shares the current directory to the /dpf directory contained within the image. This is necessary as the DPF binary within the image needs to access the files within the image itself. Any files you wish to have DPF read will have to be placed in the pwd. You can map other directories as needed, but these directories must be mapped to the /dpf directory for the server to see the files you wish it to read.

Using the DPF Container from Python

Normally ansys.dpf.core attempts to start the DPF server at the first usage of a DPF class. If you do not have ANSYS installed and simply wish to use the docker image, you can override this behavior by connecting to the DPF server on the port you mapped with:

from ansys.dpf import core as dpf_core

# uses 127.0.0.1 and port 50054 by default
dpf_core.connect_to_server()

If you wish to avoid having to run connect_to_server at the start of every script, you can tell ansys.dpf.core to always attempt to connect to DPF running within the docker image by setting the following environment variables:

export DPF_START_SERVER=False
export DPF_PORT=50054

Or on windows:

set DPF_START_SERVER=False
set DPF_PORT=50054

Where DPF_PORT environment variable is the port exposed from the DPF container and should match the first value within the -p 50054:50054 pair.

And DPF_START_SERVER tells ansys.dpf.core not to start an instance and rather look for the service running at DPF_IP and DPF_PORT. If those environment variables are undefined, they default to 127.0.0.1 and 50054 for DPF_IP and DPF_PORT respectively.