Our cutting-edge render farm infrastructure relies heavily on Flamenco, a specialized system optimized for equitable task distribution across multiple high-performance workstations. While the architecture is cloud-compatible, it is primarily designed for operation within a local network environment, which can also be configured for secure remote access via a VPN. The infrastructure employs a hybrid approach, serving not only as a powerhouse for routine, large-scale rendering projects, but also offering the flexibility to handle ad hoc, single-task render jobs efficiently.
- Release: a collection of jobs that are run in a related context
- Job: a file that defines a set of tasks to be run
- Task: a single command to be run on a single worker
- Worker: a render worker node instance
Following above concepts, the main volume should look like this:
├── jobs # reserved by Flamenco
├── file-store # reserved by Flamenco
├── quick
├── releases
│ ├── <release-name>
│ │ ├── store
│ │ │ ├── <job-name.blend>
│ │ │ ├── <job-name.blend>
│ │ │ ├── ...
│ │ ├── out
│ │ │ ├── <job-name>
│ │ │ │ ├── ######.png
├── dist
│ ├── <release-name>
│ │ ├── xyz.png
│ │ ├── xyz.png
│ │ ├── ...
├── archives
│ ├── <release-name.tar.gz>
│ ├── <release-name.tar.gz>
│ ├── ...- When certain object has updated after the release is created.
# plan a new release
farm plan release
# update a specific job
farm replace job <job-name>
# finalize the release, rename output files as expected, move to dist folder
farm harvest release <release-name>
# kill job(s) with a specific query
farm kill jobs <query>
# cancel job(s) with a specific query from a specific release
farm kill release.jobs <query>
# push a prior independent job to the farm
farm seed job [file] <job-name>
# archive a release
farm archive release <release-name>
# reset the farm, clean up all the history
farm slash-and-burn