A dockerized SheepIt Renderfarm client with Nvidia CUDA support

in order to update the sheepit client itself, you only need to restart the container, it will download the latest version on startup

Instructions (CPU)

Start the image with:

docker run -it \
 --name "Sheepit" \
 -e user_name="<username>" \
 -e user_password="<password_or_public_key>" \
 -e gpu=none \
zocker160/sheepit-client <additional_sheepit_arguments>

Instructions (GPU)

In order to make this image work, you need

• Docker >= 19.03
• Nvidia GPU driver
• nvidia-docker2

An official guide by Nvidia can be found here⁠.

Start the image with:

docker run -it \
 --name "Sheepit" \
 --gpus all \
 -e user_name="<username>" \
 -e user_password="<password_or_public_key>" \
 -e gpu=OPTIX_<0/1/2/...> \
zocker160/sheepit-client <additional_sheepit_arguments>

Usage

important: if you don't specify -e gpu, a list of all supported GPUs will be shown

• -e user_name / -e user_password specify usename and password of your sheepit account
• -e cpu(optional) to override cpu core autodetection; e.g. -e cpu=4
• --gpus all this enables the passthrough to the GPU(s)
• -e gpu specify the Nvidia GPU to use for rendering (only one at a time is possible / lanuch multiple clients for more GPUs)
• -h(optional) set hostname
• -e ui=<text/oneLine>(optional) change the CLI mode; default: text

important: you can also pass any additional argument directly to the sheepit-client by replacing <additional_sheepit_arguments> (e.g. -memory 5G)

hint: if you are using a DockerGUI like Portainer, you put those arguments into the CMD or Command field

Get into the server terminal

docker attach <Container_name>

press Ctrl + c in order to exit after finishing the current frame

press Ctrl + p and then Ctrl + q in order to detach from the terminal

Instructions Vast.ai

• create new template
• image path / tag: zocker160/sheepit-client:vast
• version tag: vast
• launch mode: SSH
• on-start script:

screen -dmS sheep /startapp.sh -compute-method GPU -gpu OPTIX_0 -login zocker_160 -password 2nuZxTC1bxmkeFETiqK0RDpqKqYBcjb9EFAOH2CH

or you can use this template link⁠.

Parameters

• -gpu OPTIX_<0..n>: selects GPU0 as the GPU to use for rendering
• -compute-method <CPU/GPU/CPU_GPU>: CPU: CPU only, GPU: GPU only, CPU_GPU: both
• -login: your sheepit username
• -password: your sheepit render token (no I did not leak my password, it is a render token)

Disclaimer

I am not affiliated with vast, nor do I get any benefit, but I personally just think, that it is a great option for sheepit, since they offer very good prices for the hardware you are getting.

Extra bits

Docker Hub page
GitHub page⁠SheepIt client repo⁠
SheepIt website⁠