Hi,

I am currently trying to run confluence data center nodes on Kubernetes.
Since the underlaying network layer does not support multicast, I am stuck with tcp_ip cluster.type.
As pods get assigned new IP-addresses on rebuild and I need to rebuild the pods in order to change the configuration, there seems to be a circular dependency I cannot get over.

As I see many people struggling with this exact part on K8s, I'd like to suggest the property of cluster.peers to be DNS names, a bit like CONFSERVER-53527, but without the synchrony-part.

The config-property would be extended like this:

<property name="confluence.cluster.peers">192.168.1.1,"confluence-0.confluence.cluster.local"</property> <!-- A comma-separated list of node IP addresses, without spaces -->

The naive behavior-changes of the cluster-node discovery would consist of a simpe DNS-query while parsing the config, then recheck the DNS every time a node fails.

This change would directly enable Confluence to be run on Kubernetes.