talascu
commented
10 years ago Indeed. As odd as it may sound, I never thought about this possibility.
At the moment we are focusing on the following development directions:
- conflict detection. If conflicts are "global" we could issue a warning when a potentially conflicting state is entered.
- conflict avoidance. If conflicts are "local" we could generate the plan, then from it generate a "conflict-graph" and finally find a minimal set of machines/locations that need to be used in order to avoid conflicts.
As for capacity constraints we are starting to think about the possibilities to integrate them, but as for now we don't have much material on this topic.
If the above issues turn out to be too complex, we could, in a first phase, do what you suggest and report violations of the deployment plan.
rdicosmo
It would be extremely useful to read all the specifications of the Aeolus model (including conflicts and capacity constraints), and after finding a plan, verify which constraints are violated, and emit an appropriate warning (maybe different classes of warning for the 3 different violations... provides abused, requirements undersatisfied, and conflicts violated).
This will help the used decide whether the plan is viable, or allow an iterative approach to find a better plan when needed.