My company have a server farm setup like this:
All five servers are VM´s on Hyper-v hosts.
- 2 RDS Gateway servers - windows server 2012 R2
- 3 Terminal Service servers - windows server 2003 R2
Each group of servers have a NLB cluster configured
We have discovered that almost every 30 minutes the one terminal service server with the most user sessions drops the connections so all servers have almost the same amount
and then connection amount to the server that drops it starts to rise again.