Scale up up to minSize

[haugenj]

Fixes #274

[imuqtadir]

Should we add additional check as a defense-in-depth against negative targetCapacity?

[imuqtadir]

My assumption is we can provide minSize to be negative (?)

[haugenj]

minSize should never be negative because of config.jelly

but maybe this doesn't protect against configuration as code?

Perhaps we should have a more explicit check that minsize >= 0 when initializing the cloud and throw an error at that point?

[imuqtadir]

Let's not throw error, maybe just log that as an error.

[imuqtadir]

Do we need to introduce updatedState. Once the fleet gets updated at this point, the currentState obj remains stale. Is there any way you can avoid this?

[haugenj]

not sure I get what you mean, can you elaborate?

[imuqtadir]

I'm trying to understand why introduce updatedState instead of currentState.

Sample snippet:

currentState = new FleetStateStats(currentState, targetCapacity);
instead of
final FleetStateStats updatedState = new FleetStateStats(currentState, targetCapacity);

[haugenj]

ah I see. I added it because I thought it made it clearer that the state was being changed.

and I changed updateByState to return FleetStateStats because we were modifying the state inside of that method but it wasn't getting captured

Related, do you know what benefit we get out of constructing a new object instead of having a setter for targetCapacity? Since currentState is a local var it seems like it'd be fine?
i.e. why do we need
currentState = new FleetStateStats(currentState, targetCapacity);
instead of
currentState.setTargetCapacity(targetCapcity);

[imuqtadir]

I guess we are updating number of active instances too so probably creating new object instead of just updating targetCapacity. We should probably just re-use currentState object to point to the newer state object and let updateByState return FleetStateStats.

[imuqtadir]

lgtm