I have a clustered installation of Hyperion Planning version 126.96.36.199. When one of the nodes fails over and the user is redirected to another node, he has to re-login to Planning. I guess we could fix that by enabling sticky sessions but I am not sure how.
I guess it should be done in the OHS installed on the load balancer ?
Any advice is appreciated.
This depends on how you're load balancing and yes it would be either OHS or the load balancer. The term is normally called "Cookie Session Persistence" where the load balancer detects a cookie based session with the cookie name JSESSIONID (I think) and it persist the connection to the server for the duration of that JSESSIONID. I know there are documents out there specifiying specs for configuring an F5 load balancer. If I find the link I will post it.
What you are talking about is not going to be solved by sticky/persistent sessions. This just ensures that once the user comes with a specific ip the load balancer does not flip that user to the other server. In fact, if you did not have sticky sessions you wouldn't even be able to log into planning because the session would not persist log enough for you to login, it would redirect you back to the login screen.
In the case of an actual fail over - the session is left on the failed server and the new server has no idea about that session that was just lost. While I think this problem is solvable (during failover, the new server picks up all the sessions on the lost node) - most clients are ok with having clients re-login in case of a failed server to avoid complexity.
Edited by: Nicholas King on Oct 14, 2012 12:45 PM