Hello,
I have configured load-balancing for terminal services running on windows servers. All works fine, but sometimes (ones per 1-3 days) is session to terminal server dropped. tcpdump shows reason - communication to client is coming not from VIP address, but from real server address (in other words: session is running 2, or 5 hours and suddenly is session dropped, because tcp issue and load-balanced to other server). Real server is operational during this issue.
here is a configuration:
server side: 10.1.3.0/24
client side: 10.1.0.0/24
(the same config for other services)
service trkat-tsapp7
port 3389
protocol tcp
keepalive type tcp
keepalive frequency 30
ip address 10.1.3.7
active
content trkat-tsapp
vip address 10.1.0.19
protocol tcp
port 3389
sticky-inact-timeout 660
advanced-balance sticky-srcip
add service trkat-tsapp7
add service trkat-tsapp6
add service trkat-tsapp5
add service trkat-tsapp4
add service trkat-tsapp3
active
clients are remote office workers, therefore sticky time 660min. tcp session is handshakes only ones by connecting client to terminal server and sticky timer is initiated. customer requires connecting to the same server after network connectivity problem.
There are around 20-100 users per real server.
I think this configuration is ok. Or it's well know problem with terminal services load balancing?
martin
