You deploy to App Engine and want instance count to scale by request rate while keeping at least 3 idle instances available. Which scaling setting do you use?

Automatic scaling is the correct choice because it allows App Engine to scale instances based on request rate, which is a key requirement. Specifically, setting minidleinstances to 3 ensures that there are always at least three instances running and ready to serve traffic, fulfilling the requirement for 3 idle instances. Manual scaling would fix the instance count, not allowing scaling by request rate. Basic scaling is designed for applications that can tolerate startup latency and has different parameters for instance management, not directly supporting a minimum number of idle instances in the same way as automatic scaling.