Forum Discussion
Hi
-
An existing TCP session will still work to same server, but a new TCP session will be load-balanced, which means you then have a 50/50 chance of hitting the same server. Also remember the 50mins means 50mins after the F5 last saw traffic from that source ip to that VIP. My recommendation would be don't use source IP - use an http and a oneconnect and a "cookie insert" profile.
-
If all you are after is balancing each TCP connection, then just don't configure source IP persistence on the application layer VIP. Each TCP connection will be individually load-balanced. If you want to individually load-balance HTTP requests, then enable the http and oneconnect profiles. I assume that the application VIP is stateless and so doesn't require persistence from the client to an application server?
- pgsmith_120398Dec 04, 2013NimbostratusHey IheartF5, Thanks for the info. I was under the impression after reading the white papers that persistence was started from the beginning of the session and did not extend based off idle time. Is this not the case? All the traffic between the web server and application server is TCP only, its not HTTP so we do not have the option to use Cookie based persistence. We are using cookie based persistence on the web VIP but not the application VIP. If I don't configure some sort of persistence on the application VIP doesn't that mean that, according to your helpful information above, if a connection goes idle for 5 minutes the user will be directed to a new web server? The application layer has a 2 hour idle timeout limit that I need to at least match. The web to application tier is stateful TCP traffic between the web and application tier. However users cannot directly access the application servers, I think that's what you're asking. Basically I have TCP traffic from the web tier to the application tier. Those sessions can sit idle for 2 hours before the application kills them. I need to at least match that. If what you stated above about the persistence being based on idle time not the initial connection time then it would make sense to me to set it to just two hours. I was wondering if there's a way I can load balance each unique TCP connection that's not IP or cookie based.