I do some strange stuff sometimes for absolutely no reason but to fix things I need to get working the way I want them to, and my latest hack is no exception.
I’ve been wanting to use Blender to render a set of short (<30s) placeholder animations for internal videos, and the issue of rendering them quickly and efficiently has been on the back of my mind for a good while. I’ve looked at a number of render farm management tools out there, but none of them was as easy to use as Loki Render - you basically drop it into a machine with Blender installed, kickstart it with a single command line, it and you get automatic master/grunt detection and connection setup.
You then give the master your
.blender file, and it does all the hard work of re-packaging and distributing the assets to all the grunts, assigning work and collecting the results on its hard disk - no need to mess around with shared filesystems, fetching the results from the remote machines, nothing. It’s a brilliant piece of work, and it works fine on both Linux and Mac - I was able to get it running within 15 seconds across both kinds of machines.
- manually specifying which master to connect to instead of locating it through multicast
- wait around until the master shows up if it’s down, retrying every 30s
It’s a bit picky about the command-line arguments right now, but I have little incentive to improve it since it now does what I need for deploying it in large scale (I’m thinking about 15 nodes with 4-8 cores to start with - tests show I can render a full PAL resolution frame in under a couple of seconds that way, depending on model and effects).
If you told me that would be happening a year ago, I’d have said you were nuts. As it is, and despite having virtually no time to do more interesting stuff, I’m counting my blessings already.