I'm connected to a vendor using a Cisco VPN client that has local LAN access disabled.
So when I can connect to my company's website, let's say www.xyz.com, the traffic goes through the vendor's network.
Our website's DNS points to a load balancer, and we have several servers behind it. Connections going through the VPN seem to bypass the load balancer entirely, and go directly to a specific server. This is weird. Usually it's not a problem, but if we take that server out of the load balancer, connections from my machine start dropping.
It seems like the Cisco VPN client is creating a TCP connection with the destination server the first time it connects through DNS, and then subsequent requests go directly through that connection instead of going through the load balancer. Am I correct, and is there anything I can do to get around this?
Best Answer
No, odds are your load ballancer has some sort of stickyness enabled so that you always hit the same server on the backend. As everyone on that VPN has the same public IP, you all hit the same server.
About all you can do it disable the stickness and do a true round robin on the load ballancer.