"gweatherall" wrote: I want to set up a clustered environment with a number of partitions (say A, B and C). We will have two (or more) nodes in each partition (so A-1, A-2, B-1, B-2 etc). I need to have session replication within each partition (for fail over), but not between partitions (as this would limit the total number of useful machines in the cluster). How can I distribute incoming HTTP requests in this case? I understand the basic mod_jk idea of sticky sessions looking at the session id, but that only half solves the problem. Mod_jk does not seem to understand the clusters back end, and therefore it would be only luck that when the primary target failed the secondary target selected as the replacement would be in the same cluster (for example, if the session was stuck to A-1 and it failed, I want to try A-2, A-3 etc before trying B-1, B-2, C-1 etc). I would guess this is a fairly common scenario, but after a lot of google searching I have not come up with a solution.
