As of OpenShift Container Platform 4.4, follow the procedure to restore to a previous cluster state in order to recover from lost master hosts.

If you have a majority of your masters still available and have an etcd quorum, then follow the procedure to replace a single unhealthy etcd member.