Hey everyone,
There's an interesting discussion going on about manually scaling revisions. The current 'workaround' people are doing is to manipulate the underlying PodAutoscaler annotations.
This works because we're not currently reconciling the scale annotations/labels from the Revision to the PodAutoscaler. But given the PodAutoscaler is an internal type I'm debating whether we should support this. The PR [1] below would allow mutations of the revision's annotations to be the way to adjust the min/max scale. This seems like a good approach since Revision's are meant to be user facing.
I'm trying to get a sense if we introduce such a change if this would be disruptive to anyone. If so please reply with your comments (ideally on the pr).
cheers,
- dave protasowski
serving lead/steering