How would you answer an interview scenario involving Docker logging and observability?

A container restarts intermittently and local logs disappear. Explain how you would design logging and metrics so the next incident is diagnosable.