This raised lot of questions/confusions due to the hack involved in adding primary dns as the management server itself even if the user is allowed to change its value it never gets reflected. 

Right now when the user edits the following 

DNS1 
DNS2
DNS3 

We add DNS1 to be the management server IP forcibly even if the user provided input is different. Now the DNS provided by user might be legitimate for his cluster and he would want to do that. Rather than we doing a hack in the backend. 

For a sysadmin his provided values not getting reflected is a bad example. 

Now that if forcibly modify the DNS1 value to management server IP. Why do we even ask the user to modify it in first place. 

We just provide the USER information their that

Internally DNS1 -> UI Management server IP -> xxx.xxx.xxx.xxx
Internally DNS2 -> UI DNS1 -> xxx.xxx.xxx.xxx
Internally DNS3 -> UI DNS2 -> xxx.xxx.xxx.xxx

Primary DNS happens to be the private DNS and secondary DNS happens to be the public DNS. This is confusing, and the slave node is not able connect to the master node.

I would like to conclude that, whether we are showing first server as dns server or not, there is no control on name resolve.  For example, first server is referred as mmc.mylan.com by his local DNS server will not be used inside storage network.  As server names may differ from local DNS name and internal storage name, its using own DNS server always.  BTW ejabberd needs a dns server with these names.  If there is a DNS failure, ejabberd will go crazy.

We have a plan to remove ejabberd already.  I think this eliminate running own dns server.

Needs to be tested