learned a lot just from that master class. for example... cluster autoscaling not able to change a machine type (CPU/memory) as opposed to just adding/removing nodes based on pod requirements. My original plan seems more sound now.