I would like to find out if it is possible to auto-suspend a queue, based on the state of multiple other queue's.
The situation is as follows:
Say we have a cluster with 30 slots and have 3 departments working on it with an equal share. What we want to create is 1 high priority queue per department with 10 slots (the official share) and 1 lowprio queue with 20 slots (so they can use everything is no one else is using the cluster).
I have already figured out that you can make a queue a subordinate, but in this case this does not suffice. Because you can only seem to link it to 1 queue. In this case, I want a lowprio queue of a department to be automatically suspended when the highprio queues of the other departments are completely full.
So the question is, is this possible with subordinates? Or is it possible to work with some kind of logic of complexes? So that for example if all slots are taken from the other highprio queue's it is suspended. But only on the condition that both the other queues are full.
Off course, I do not know if auto-suspend is really necessary if I work with a slots consumable on the host and work with urgencies and nice values. I guess the highprio will that way always get precedence over the other queues?
It is just that an auto suspend would be easier to follow up on.