Scaling Render Power with Flamenco Orchestra

When working on an open movie at Blender Studio, the final weeks of production are always intense. Shots are locked, lighting is finalized, and suddenly we need to render thousands of frames at full quality. Our in-house render farm handles daily work well, but during these crunch periods, we often need to scale up quickly.

Cloud Rendering with Flamenco

Rather than investing in hardware that sits idle most of the year, we turn to cloud computing. We spin up a standalone Flamenco instance in the cloud (complete with its own Manager and Workers) render the frames there, and sync the results back to our local storage.

The cloud-based farm operates independently, so there's no complex integration with our local infrastructure. When rendering is done, it can be dismantled.

Introducing Flamenco Orchestra

To achieve this, we built a simple set of scripts using Python and OpenTofu, which can be used to deploy and manage a render farm with less than 30 minutes. It's called Flamenco Orchestra.

This setup allows you to deploy a farm on either Hetzner, DigitalOcean or Google Cloud Platform with just a few commands. The configuration includes:

• Flamenco Manager with attached NFS storage for project files and render output
• Flamenco Workers that automatically connect to the Manager on boot
• Automated provisioning of Blender and all required dependencies
• Firewall rules configured for secure operation

In our case we primarily use GPU droplets for OptiX rendering, and Spot instances for CPU-based Cycles rendering. Both setups support scaling the number of workers up or down as needed.

Get Started

In order to use this system, basic knowledge of the command line is required. This repository includes detailed instructions for deploying to either cloud provider:

https://projects.blender.org/studio/flamenco-orchestra

You'll find step-by-step setup guides, troubleshooting tips, and helper scripts for syncing rendered frames back to your local machine.

Contributions Welcome

This configuration works for our needs, and we welcome contributions that:

• Improve usability or documentation
• Add support for additional cloud providers (AWS, Azure, etc.)
• Optimize cost or performance
• Enhance security configurations

If you've built something similar or have ideas for improvements, let us know!