enabling group sticky

To meet zero downtime requirements, we have 4 Apache proxy servers with two clusters (with session replication within cluster turned on) of backend application. We would like to configure (in front of Apache servers) load balancer groups in LB and enabling group sticky. In this configuration, one group of Apache servers (say, A1 & A2, running on two different physical servers) will be pointing to Cluster 1 and the second group (say, A3 & A4) will be pointing to Cluster 2.

 

 

On a sunny day, the load balancer will open the new sessions to all the 4 Apache servers (A1, A2, A3 and A4) in round-robin fashion with Cookie based session persistence. When one of the server fails (say, A1), it will forward all the existing connection on that server to another server in the same group (in this case, A2) as long as it is available. Since both these servers are pointing to same cluster and we have session replication enabled, it should not cause any problem. If both the servers in a group fail (it can only happen if both the physical servers crash, which is a case of full outage), then it will forward the requests to the servers in another group, which will result in some errors and the CSRs has to re-login.

 

 

Please let us know how we can go about configuring ‘group sticky’ to meet this requirement.