BTW round-robin is totally bad. I was too tired last night to think about it, but if one of your boxes goes down you might hit the right box for the app to respond and then all the JS fails to load because that request has gone off to the dead machine?
No solution is completely foolproof I suppose. I've been looking at using HAProxy for load balancing just because it handles all the SSL termination stuff as well but then you're back to the issue of a single point of failure
Yeah, if you want SSL with cluster you'd have to have nginx/haproxy running on each app server... Having one or two LBs is less firewall to manage as well, because the app servers can be on a private network with no public network ports open other than maybe SSH
Cluster is definitely a cool idea though, so will keep an eye on it :+1: