You have the following routing design. You discover that Compute Engine instances in Subnet-2 in the asia-southeast1 region cannot communicate with compute resources on-premises. What should you do?