Scratch that previous bit about qmaster being easy to set up. Woke up to find all the renders had stalled, wasted a bunch of time getting the whole thing working again. Some notes:
- Traffic appears to be routed through the box that submits the jobs, so the submission machine needs to be left on the network.
- If you delete a controller, you have to reassign its slave nodes to the pool before creating a new one, else they will quietly (and invisibly) wait for the original to return.
- Compressor stores locations as relative to its current location, i.e. if you submit your batch from the box that shares the network drive none of the other nodes will be able to find the files.
- Each node needs a different serial, else you will get Quicktime Error: -50.
- The simplest setup seems to be a controller that runs Compressor (submitting the jobs), linked to another box that maintains the network volume/is a slave node. From there you can add additional slave nodes that see the network volume.