In Remote Desktop environments where high availability and continuity is important multiple Remote Desktop Session Host Servers will most likely be combined in a server farm. Using a RDS farm users will connect to a farm address and will be load balanced across the RD Session Host servers in the farm. A way of load balancing these sessions could be Microsoft’s Network Load Balancing (NLB). If you ever have to do maintenance on one of the RD Session Host Servers that is currently in production you can use a feature on the RD Session Host that is called drain mode. If a RD Session Host server is running in drain mode existing sessions will continue to run, but new sessions to the server will be denied. The server will eventually run out of sessions and will be available for maintenance. So far, so good.

However, if you use NLB (or any other load balancing software or hardware) that loadbalancing part will in most cases not pick up the fact that an RD Session Host server is in drain mode and will continue to route new sessions to it. As a result, users who are redirected to this server in drain mode will get an error that they can’t connect. Furthermore, as more users will log off from the server in drain mode, more and more users will be send to that server since it is the server with the least load.

Overcoming this issue is rather easy, inside NLB you have the option to set the RD Session Host in drain mode as well. Therefore, with the combination (drain mode on the RD Session Host and drain mode on NLB) you are full proof and can take a RD Session Host out of production without end users receiving connection errors.