Hi! We're using Round Robin to distribute traffic between members. During pool member failures we see that the traffic is not balanced out evenly across the members even after a few hours. What could be the reason?
There could be many reasons behind this behavior. If using persistence clients would keep returning to the same member unless the persistence timeout has been reached for the connection/session in question. So for very active clients that might take a very long time.
Another reason is if you use long lived connections such as web sockets. If the client connection never dies the LB would not re-balance it. If you want to ensure the good distribution of clients I'd recommend the predictive load balancing method in most cases and to stay away from Round Robin.
If you're really particular and wants to load balance per request you might want to look into a OneConnect. Just make sure you also enable, and log X-Forwarded-For though or you'll get application logs where you can't trust the source IP address.