Test scenario as below
Server1 : Mongo and Rabbitmq instance running
Server2: Nuve and basic app running
Server3: portal, cluster-manager(master), session-agent, audio-agent, video-agent,webrtc-agent
Server4: portal, cluster-manager(master), session-agent, audio-agent, video-agent,webrtc-agent
Question is When server3 went down how the app knows to connect server4, also from app user1 logged in to server3 using socket running on server3 and user2 logged in to server4 using socket running on server4. does these 2 users will be in same conference? how token is servered.
what is best configuration for making Intel CS as HA
Please refer the Conference Server Guide to get all components' HA policy. For your question, cluster manager hold all room resource allocation and one room holds one session controller node. If one server is down and some room components are running on it, then according to the components' different HA policy, system will either auto recover or notify impact user to relogin.