We have global settings for:
We set up one of our plugins to detect errors for licenses and set a limit group for that license to the max amount that our facility has. Due to hyper threading on some of the newer machines, they are using up more licenses than a normal machine would. The “In Use” area for the limit group still reports that a single license is being used, yet 4 are really being used. Is there an easy way for us to have the plugin reque a task without causing an error because it’s really unpredictable which machine is going to be picking up the job (whether its a machine with 2, 4, 8 or 16 cores). Perhaps there’s a method so that a job with a particular limit group will ignore the global Failure Detection settings?
Thanks!
-Brett Bronson
Hi Brett,
The Limit Group feature was designed with the assumption that a single machine won’t use more than one license. What it sounds like is that in this specific case, you need a way to tell machines how many limit group “stubs” they need before they can render a job. If there aren’t enough available for that machine, then it can’t pick up the job. We think others might find such a feature useful in cases where licenses used is based on proc count, so we’ve added this to the wish list and will look at adding this feature in an upcoming release.
Unfortunately, there really aren’t any workarounds at the moment that will still allow you to fully utilize your farm when using this limit group. The only workaround I can think of is to setup up the limit group to ensure the license count will never overflow. For example, if you have 20 licenses and 10 machines, and 5 machines have 4 cores and 5 machines have 2 cores, then set your limit group limit to 5. This makes the assumption that all 10 machines have 4 cores, so you will never use more than 5x4=20 cores. Of course, if a 2 core machine picks up the job, then essentially you have 2 licenses sitting idle, but at least your jobs won’t be erroring out.
Cheers,