In looking at the deployment guide for CUPS7 there are several deployment models for balancing users within subclusters and references to the use of SRV records to facilitate seemless failover between failed nodes for subscribe and message requests. While this makes sense for server to server communications there seems to be a gap in the end user experience where the subcluster name would not be used for a similar load balanced or active standby login process via the CUPC client.
The initial testing with the SRV records indicate a correct list of system however the CUPC client doesnt seem to utilize them for initial login.
Anyone have any insight on the CUPC interaction on a multinode deployment and the underlying DNS framework to support it.
1) There's no logon traffic from CUPS to CCM. All logon request was handled by CUPS alone. (if LDAP authentication was used, it's CUPS+LDAP).
2) After 1st-time logon, CUPC retrieves a list of servers from CUPS and store it locally on the PC. On next time logon, if the "home server" was unavailable, CUPC will look at its local file and try the next one.
More details can be found in http://www.lulu.com/content/5552336