I'm trying to setup 2 remoteapp servers with a bit of fail-over in the design. I keep running into a few dead ends where there is no load balancing... or connections take a ridiculous amount of time to reconnect if the gateway fails... or broken sessions won't resume when reconnecting (starts new session instead)...
Can someone explain how the whole process functions a bit more clearly? Or give me a definitive answer on what works with what and what must be on a separate server from what... etc...
My understanding of the process is as follows... please correct me where I'm wrong:
- Client attempts to connect to FQDN provided using RDP. Firewall is set to reject RDP traffic (port 3389 stuff).
- Client uses public FQDN to make HTTPS connection to 1 of the two gateway services (using 1 public IP NAT'd to private IP that's clustered to both servers using NLB).
- Gateway service looks at the server farm members and bridges connection from HTTPS client to either remoteapp server in round-robin fashion (using private IP & rdp).
- remoteapp server connects to session broker server which redirects the gateway service to the correct server (either reconnects existing session or creates a new session in a load-balanced manner).
Where I'm at now... The gateway servers are working like they should... but it seems that load-balancing isn't happening at all with the connection broker. All connections (from gateway to RDP) end up on the same server... unless that server becomes unavailable (i.e. shutdown/yank network cable/etc...). Anyone have any ideas?