My current setup is: 1 workstation (1x 8 threads) which is running the swarm coordinator and my UE Project; and 6 Clients (6x 32 threads)
when i hit Build on the workstation everything looks fine in the swarm agent i can see all pcs working for my workstation.
but after a while (10 min depends on project) all clients stoped working and my workstation with the fewest number of threads has to do the rest (for hours).
i also tryied the avoid local execution function and the default process priority in the swarm agent settings without success
Sometimes there’s tasks lightmass cannot split up across multiple cores, like if you are for example baking 200 objects with 128x128 light maps, and one object with a 4096x4096 lightmap, that 4096x4096 lightmap is going to only run on one core and take the longest to complete, and is your bottleneck, not the number of cores.
Each object in the scene gets sent to a single thread, so at some point it can end up that all the other objects are rendered and one object is still rendering because it uses a large lightmap and can only use one thread for that object. It’s a limitation in Lightmass that it can’t use multiple threads to render an object at the same time.