Customizing RQD Rendering Hosts

Build custom RQD container images to deploy as OpenCue rendering hosts

This guide describes how to customize the default RQD container image published on Docker Hub. The default RQD container image doesn’t include any rendering software. This guide explains how to create a custom Dockerfile that builds on the basic opencue/rqd image to install rendering software. You can adapt the basic ideas in this guide for many other types of software, including commercial rendering packages, such as Maya.

Before you begin

This guide follows on from the OpenCue quick starts for macOS and Linux. Before you work through the steps in this guide, make sure you have successfully started the OpenCue sandbox environment and run a basic command-line test job. You’ll also need all of the software and source code you used in the quick start.

Sample Dockerfiles

The OpenCue project includes sample Dockerfiles to illustrate how to install additional software for RQD containers. OpenCue currently includes sample Dockerfiles showcasing the following:

Blender installation
Nvidia CUDA installation for GPU accelerated rendering

To view the sample Dockerfiles:

Open a terminal.
Change to the directory you cloned or downloaded the OpenCue repository to. For example, if the OpenCue directory is in in your home directory, run the following command:
```
cd ~/OpenCue
```
Change to the subdirectory which includes the sample Dockerfiles.
```
 cd samples/rqd
```

The sample Dockerfiles are listed within their respective subdirectories.

Reviewing the sample Blender Dockerfile

The sample Blender Dockerfile showcases the Blender installation process and environment variable setup. You can update the sandbox environment to build and run the sample Dockerfile so that you can submit and run a rendering job using Blender than just the basic command-line tools illustrated in the quick start.

Before you update the sandbox to run the sample Dockerfile, you might find it useful to review the source code for the sample container.

Run the following command to review the sample Dockerfile:

cat blender/Dockerfile

The command outputs the contents of the Dockerfile.

The first section of the file indicates that this Dockerfile builds on the basic opencue/rqd container image hosted on Docker Hub:

# Builds on the latest base image of RQD from Docker Hub
FROM opencue/rqd

The next section installs all of the dependencies required to run Blender 2.79 on the CentOS operating system installed in the opencue/rqd container image:

# Install dependencies to run Blender on the opencue/rqd image
RUN yum -y update
RUN yum -y install \
        bzip2 \
        libfreetype6 \
        libgl1-mesa-dev \
        libXi-devel  \
        mesa-libGLU-devel \
        zlib-devel \
        libXinerama-devel \
        libXrandr-devel

The next section sets up parameters for the Blender installation directory and download source.

# Set Blender install directory
ARG BLENDER_INSTALL_DIR=/usr/local/blender

# Set Blender download source
ARG BLENDER_DOWNLOAD_SRC=https://download.blender.org/release/Blender3.3/blender-3.3.3-linux-x64.tar.xz

The final section downloads and extracts the archive for Blender to the provided installation directory, in this case /usr/local/blender.:

# Download and install Blender
RUN mkdir ${BLENDER_INSTALL_DIR}
RUN curl -SL ${BLENDER_DOWNLOAD_SRC} \
        -o blender.tar.xz

RUN tar -xvf blender.tar.xz \
        -C ${BLENDER_INSTALL_DIR} \
        --strip-components=1

RUN rm blender.tar.xz

The final command verifies the Blender installation.

# Verify Blender installation
RUN ${BLENDER_INSTALL_DIR}/blender --version

If you’d like to learn more about the configuration of the default opencue/rqd container image, view the source code for rqd/Dockerfile in the master branch on GitHub.

Reviewing the sample CUDA Dockerfile

The sample CUDA Dockerfile extends the default RQD container image to support GPU rendering on supported Nvidia Hardware. This requires Nvidia Container Toolkit to be installed on render nodes as a prerequisite.

Run the following command to review the sample Dockerfile:

cat cuda/Dockerfile

The command outputs the contents of the Dockerfile.

Updating the sandbox environment

To build and run the sample Dockerfile in the sandbox environment, you need to update the docker-compose.yml file that defines the deployment. For a production system, you might make a similar change to update the configuration files for a container management platform, such as Kubernetes.

Complete the following steps to configure the sandbox environment to build and run the sample Dockerfile:

Open the sandbox/docker-compose.yml file in your preferred text editor.
Find the following lines:
```
  rqd:
    image: opencue/rqd
```
Replace the lines from the previous step with the following code:
```
  rqd:
    build:
      context: ./
      dockerfile: ./samples/rqd/blender/blender2.79-docker/Dockerfile
```
This change configures Docker Compose to build your local copy of the Dockerfile in the samples directory instead of using the opencue/rqd image on Docker Hub.
Save your changes.
Before you start Docker Compose, delete any existing OpenCue sandbox environment containers:
```
docker-compose --project-directory . -f sandbox/docker-compose.yml rm
```

To re-deploy the sandbox environment, run the following command:

docker-compose --project-directory . -f sandbox/docker-compose.yml up

Submitting a rendering job

To run a sample rendering job, you’ll need a sample .blend Blender file. If you don’t have an existing .blend file, the Blender project publishes a variety of demo resources.

Note {: .callout .callout-info}Make sure you download a demo file that works with version 2.79 of Blender or earlier.>

After you download a suitable .blend Blender file, move it to the /tmp/rqd/shots directory. The sandbox environment is configured so that both your host machine and the RQD container can access the /tmp/rqd/shots directory.

If you’re starting CueSubmit and CueGUI in the OpenCue sandbox, you need to set the values of the following environment variables in the Python venv environment you created in the quick start:

source venv/bin/activate
export OL_CONFIG=pyoutline/etc/outline.cfg
export CUEBOT_HOSTS=localhost

If you want to submit a Blender job type in the sandbox environment, then you must also update the CueSubmit configuration:

Copy the example CueSubmit config file:

cp cuesubmit/cuesubmit_config.example.yaml sandbox/cuesubmit_config.yaml

Open sandbox/cuesubmit_config.yaml in your preferred text editor.
Update the value of BLENDER_RENDER_CMD to match the installation location in the RQD container image:
```
BLENDER_RENDER_CMD : "/usr/local/blender/blender"
```
Set the value of the following environment variable to update the location of your custom CueSubmit configuration file:
```
export CUESUBMIT_CONFIG_FILE=sandbox/cuesubmit_config.yaml
```
Run the following command to start CueSubmit:
```
cuesubmit &
```

To test submitting a Blender job to OpenCue, see Submitting jobs.

After you submit a job to OpenCue, you can monitor progress in CueGUI.

Stopping and deleting the sandbox environment

To delete the resources you created in this guide, run the following commands from a shell:

To stop the sandbox environment, run the following command:

docker-compose --project-directory . -f sandbox/docker-compose.yml stop

To free up storage space, delete the containers:

docker-compose --project-directory . -f sandbox/docker-compose.yml rm

What’s next?

Learn more about OpenCue concepts and terminology.
Install the full OpenCue infrastructure.