I'm working on setting up scaling policies for an ASG application.
I'm wondering if it is possible (or advisable) to combine Predictive and Dynamic scaling policies for a single ASG.
The peaks for the application load will be relatively cyclical, which should work well for Predictive Scaling. However there could be situations where the load is unusually high during particular events.
What I would like to do, is have the group scale based on forecasts, but also have the ability to scale in response to dynamic unexpected load.
What I want to avoid though, is a situation where the scaling policies are conflicting with each other. For example, the predictive scaling scale-down during heavy load times, because the forecast predicted low load.
It seems it would be easy enough to prevent problematic conflicts. For example, it could scale up if either of the policies, dynamic or predictive, indicated that additional capacity is needed, but only scale-down when both policies show that it's safe to do so.
Can anyone tell me if the scaling policies work in this way when combined. The documentation is fairly clear about how they work in issolation, but doesn't seem to mention what happens if they're combined.
