We had 4 pcs connected to the swarm and they were all running when we built the lighting. Then we split everything down for level streaming and now the swarm nodes dont get utilised during the build despite having a large number of lightmaps to process, so the build is horribly slow. Is this something to do with how swarm only gets used if the required processing power exceeds that of the local machine? Can you force it to use the swarm to speed up the light build? I can ping them from the coordinator and they all show in the coordinator as available, unassigned so they are all active.
I think once a slave computer finishes processing the lightmaps assigned to it, it’s not able to pick a new job and continue to work further… I think I’ve read that somewhere. Would like a confirmation, if it’s true i’m not gonna keep my slave pc just for a tiny difference in render time!
Post an image of the Swarm agent visualizer progress bars.
What happens a lot of time is that there aren’t enough tasks (separate components) to go around. If you have 4 machines to distribute to, and each has 8 cores, the scene needs to have approximately 4 * 8 * 4 components to actually use all those cores. Only separate components can be sent to different machines.
That’s not the case, agents will continue requesting tasks until there are no more to distribute.