I’m trying to improve scheduling of our render nodes and I’m hoping I can mine some data from deadline about what each slave is doing throughout the day. Essentially creating a timesheet like human employees would fill out to determine billable hours, etc. Node1 = 8 hours on job 2, 8 hours on job 4 and 14 hours on job 5 for example.
I can’t quite figure out what information I need to be pulling out of the repository to figure this out. To complicate things a little, often single tasks (frames) are started on one machine and then resumed on another (using Maxwell render and resuming the mxi files) so I can’t just add up each task’s time, since I believe it won’t capture the fact that it spent 4 hours on node1 and then 5 hours on node2. Instead it would assign all 9 hours to node2, is that correct?
If I can’t get it straight from the repository information, I can always just set up a script that runs every 10 minutes or so and checks what each slave is doing and saving that out, but scanning the repository seems a little more elegant.
Any advice?