Cluster rendering for blender

❗❗❗This Project is NOT PRODUCTION READY❗❗❗

Since blender 2.78 there was no avilable cluster rendering option for individuals, that I was aware of. So I decided to write such a solution myself. Right now is more of a proof of concept, rather than a complete product, but hey, it works.

How does it work?

Cluster controller is expected to export a network folder. Once a zipped .blend project gets dropped into it, this file get analyzed and frames to render get split between all configured render devices. Every render device copies the project to it's own disk space and than renders assigned frames. Once rendering is done every frame is sent back to the network exported folder.

Compatibility

Feature	Compatible?
CYCLES	✔️
EEVEE	❌
GPU RENDER	✔️
Image Sequence Render	✔️
Any Other Output Format	❌

This project was developed to run on Linux and MacOS hosts, however there's nothing preventing it from running on Windows platform.

Setup

Setup is divided into two main parts:

Cluster Controller Setup
Device Setup

Cluster Controller Setup

1. Set up network exported folder
2. Download and unpack latest cluster_controller.zip from release tab
3. Edit configs/cluster_main.json
listen_path is where you want this program to listen for files to render
temp_path is folder used for temporary file. Best to leave as /tmp on linux and macOS
port is the port used for both web dashboard and device communication
eg.

{
  "listen_path" : "/exports/listen_folder",
  "temp_path": "/tmp",
  "port": 2452
}

4. Edit device files in devices directory. Filenames don't matter as every file in this directory is parsed.
Evey device needs seperate file in this directory. Set up as following:
ip - is devices up address
hwid - is a unique identifier, make sure that corresponding device has same hwid set
performance - is determining what part of the render this device will recieve (higher is more frames to render)
port - leave default if you didn't change it in the device

{
  "ip": "172.16.0.22",
  "hwid": "whatever",
  "performance": 1,
  "port": 2540
}

5. Run cluster_main.py

Device Setup

1. Mount cluster exported file
2. Unpack cluster_device from cluster_controller.zip on your device
3. Install blender
4. edit config/config.json
port - port on whitch to operate
blender_command - command that runs blender, can be full blender path
hwid - unique id, make sure you set the same on cluster side configuration for this device
tmp_path - directory for temporary files
network_mount - path to mounted network folder
server_address - servers ip address
server_port - cluster controllers port, set to same as in cluster_main.json

{
  "port": 2540,
  "blender_command": "blender",
  "hwid": "whatever",
  "tmp_path": "/tmp",
  "network_mount": "/exported/by/controller",
  "server_address": "172.16.0.21",
  "server_port": 2452
}

5. run cluster_device.py

Note that if you run both cluster controller and device on one system you need to set up 2 separate tmp folders

Usage

1. Mount network exported folder on your workstation
2. Copy zipped blender project to mounted folder

Roadmap

1. Add EEVEE support
2. Make everyhong secure (cuz now it's not)
3. Add support for various output formats
4. Docker support