jpalpant/fah-client repository overview

⁠Folding@Home Docker Image

A thin wrapper around FAHClient using dumb-init⁠. Easy to use, exposes the full FAHClient interface, and dumb-init provides all the benefits of a full init system.

Since v0.1.2, it's built on nvidia/cuda:10.1-runtime-ubuntu18.04 modified to include OpenCL, so GPU support is possible if you configure nvidia-container-runtime⁠ or if you are running on a platform that supports GPU-allocation, like GKE. This container works equally well without a GPU, however.

This image is published to quay.io as quay.io/jpalpant/fah-client⁠ and also to DockerHub as jpalpant/fah-client⁠ (since v0.1.4).

⁠FAHClient Version

⁠Usage

⁠Help

docker run -it quay.io/jpalpant/fah-client --help

⁠Minimal (CPU-only, anonymous)

Pass --fold-anon to FAHClient to start folding immediately, without waiting for any other configuration.

docker run -it quay.io/jpalpant/fah-client --fold-anon

⁠Minimal GPU (anonymous)

You'll need to configure nvidia-container-runtime⁠ or nvidia-docker⁠ on your system to provide GPU access to your image, or otherwise make GPUs available to container images.

For nvidia-container-runtime: docker run -it -e NVIDIA_VISIBLE_DEVICES=all quay.io/jpalpant/fah-client --fold-anon --gpu=true For nvidia-docker: docker run -it --gpus all quay.io/jpalpant/fah-client --fold-anon --gpu=true

⁠Run as a user

Use FAHClient arguments to specify the user or team you'd like to use:

docker run -it quay.io/jpalpant/fah-client --user=myname --team=12345

⁠Persistent the data directory to save changes to config.xml as well as work-in-progress

The working directory of the FAHClient is /var/opt/folding by default.

With volumes: docker run -v persistent_volume:/var/opt/folding quay.io/jpalpant/fah-client --fold-anon

By path: docker run -it -v /path/to/folder:/var/opt/folding quay.io/jpalpant/fah-client --fold-anon

You can remove --fold-anon if you have configured config.xml manually. You can do this using FAHClient, or via the FAH web UI.

docker run -it -v /path/to/folder:/var/opt/folding quay.io/jpalpant/fah-client --configure

⁠Exposing the web UI

FAHClient comes with a built-in web UI on port 7396. Bind to 0.0.0.0 and expose the port to access it:

docker run -it -p 7396:7396 quay.io/jpalpant/fah-client --fold-anon --web-allow=0/0 --allow=0/0

Then visit http://localhost:7396⁠

⁠Using FAHViewer

This image does not include FAHViewer, a package for viewing the models your host is simulating in real-time. However, if you have this program installed, you can connect to this container on port 36330:

Start FAHClient: docker run -it -p 36330:36330 quay.io/jpalpant/fah-client --fold-anon --command-enable=true --command-address=0.0.0.0

Connect using FAHViewer: FAHViewer --connect=127.0.0.1:36330

⁠Image

⁠Dockerfile

gitlab.palpant.us/justin/folding/docker/Dockerfile⁠

⁠Tags

Currently I'm making :latest (every commit to master) and periodic tagged versions like v0.1. Happy to consider a system that follows FAHClient's versions more closely, or another system such as making smaller CPU-only images along with GPU-supporting images.

⁠Contact

Email: [email protected]⁠ Twitter: @justin_palpant⁠