How would you answer an interview scenario involving Kubernetes resource requests and limits?

A service is frequently CPU throttled and another is OOMKilled. Explain how you would investigate and tune requests and limits.