I tried to replace Spring Security 2.0.2's default SessionRegistry with a custom SessionRegistry implementation which uses either EHCache or a JGroups' ReplicatedHashMap to store the principals and sessionIds. I tried both approaches, but neither of them makes the concurrent login filter work as expected (ie. allow only one concurrent login for each user regardless their used cluster node).
I can post my modified SessionRegistry code but I doubt the problem is there (it's a plain replacement of Maps to Caches or ReplicatedHashMaps, nothing else). Any clue what could be wrong with such a setup?
Generally, is it a naive theory to handle concurrent logins in a cluster this way?



