Description
Is your feature request related to a problem? Please describe.
While customer wants to control max value of nodepool dynamically for cost management purpose, it is found out that the unneeded time will be reset. The scaling in operations will be affected by this.
A demo set-up is here:
Describe the solution you'd like
Unneeded time will not be reset unless CAS is being turned off.
Describe alternatives you've considered
Allow to set max node allowed to be created per scale-out event.
Additional context
The context of this feature requests is: If we sent in Jobs in, when Pods can be completed within a very short time, as the time that nodepool completes its scaling-out operation, most Pods will no longer need many nodes.
For example: I send Jobs which requires 100 nodes, but during the scaling-out event, some Jobs already completes itself, and when the scaling-out completed, the Jobs only needs 50 nodes now (however at the time there are already 100 nodes, and this cause waste).
- So one solution is that customer should try to dynamically change the max value for nodepool, but this approach is blocked by "unneeded time reset" issue.
- Another solution is allowing customer to set max node number that allowed to be scaled out for each scale-out event.
- Or combing the above two, which is better.