I have two sites with a stretched DAG between them. By "sites" I mean two AD sites that correspond with different physical and network circuits.
When I take an individual mailbox database and move it to a second site, I want to make sure all the clients connect to that site directly.
What is the purpose of the RpcClientAccessServer setting on a database that has replicas in a different site?
When should the RpcClientAccessServer be updated? (e.g. when making the other site active?)
The RpcClientAccessServer is the CAS address the client wants to use to access their mailbox. I normally see this assigned to a DNS record that resolves to a LB VIP in a single-site DAG.
Since it doesn't sound like you're load balancing at all, but have two active sites with copies in the opposite site for DR, I would recommend creating a couple of DNS records such as cas1.mydomain.local using the IP of the CAS server in site1, and cas2.mydomain.local for the CAS server in site2. You would assign the RpcClientAccessServer value for mounted copies in site1 to cas1.mydomain.local, and cas2.mydomain.local for the mounted copies in site2. Give the records a low TTL, like five minutes. This will keep users with mailboxes in their respective site connected to their local CAS.
It is important that these names cannot be resolved from the internet, or your initial connections using Outlook Anywhere will be very, very slow because the clients will attempt an RPC connection first and wait for it to fail before trying HTTPS.
In a site1 failure scenario, all copies would be mounted in site2, so you would point cas1.mydomain.local to the same IP that cas2.mydomain.local points to. The DNS name you use for your external URLs would probably have to be updated to bring external access over to site 2. You wouldn't regularly be updating the value on the mailbox databases.
There are some unknowns in your description, so the above makes some assumptions, but I hope it helps you understand what the value is for.
