Hi Manoj
Your assumption is wrong. As Christian pointed out in a previous post, if the single node in the primary site fails, the node in the secondary site will not take over the internal VIP. The internal VIP is only valid for a given site. The secondary site would typically have its own internal VIP (assuming more than one member). This is why I suggested an external load balancer for your case.
And yes, the clustering requirement for multi site does differ to the clustering requirement for multi-master.
Multi-master clustering (i.e. the primary site) is intended for, and best suits, a single data center. If you have several data centers, then you should use multi site clustering between the data centers. Your primary site will still use multi-master clustering between the nodes in the site. Though, in your case that's not relevant, as you only have a single node in the primary site
Note: Above response is for the general case, not specific to AWS.
Regards
Pearse