Building a 12-node render farm, end to end

Nobody on the team had built a render farm before, which meant the job was equal parts research, procurement, and improvisation. Twelve nodes, one queue, and a community of 3D artists who needed it to just work.

I picked Flamenco for the front-of-house scheduling because artists could submit jobs without thinking about infrastructure, and paired it with AWS Thinkbox Deadline underneath for the actual render management across nodes.

The part nobody warns you about is thermal and power planning for twelve machines running flat out in one room. Design instincts turned out to transfer here too. The real interface most people interacted with was a status board, and that got the same attention any product screen would.