For example, a Redis sentinel instance monitors a Redis shard made up of several Redis server instances (one master and several slaves per shard). I deployed a server instance as a container within a Kubernetes pod. IP address of the pod may change during some occasions such as after rolling update. This results in failover-abort-no-good-slave because the sentinel instance has obsolete server IP addresses.

How should the sentinel instance picks up the IP address change?

Setting up a hostname for the master and slaves does not work - the sentinel instance will convert the hostname (which is given during initiation) to IP address when overwriting sentinel.conf file.