I've encounterd a problem on one of our application servers where the server process would lock up and refuse anymore http sessions. We have a setup where load-balancing occurs between two servers. We create two content rules per server, one for http (80), and one for ssh (22) so that we can access the servers through the CSS. We were originally using http keepalives, but this had the effect of shutting down all content rules when the http service was stopped, making the server inaccessible until the keepalive was changed to icmp or an active service port.
I have two questions. The first one is, can I make it so that if a http keepalive fails, we can still access the server via its other content rule through ssh, even while the CSS takes it out of the load-balancing loop (since it is technically "down").
My second question is this. When a server locks up, port 80 is actually still open, though not responding. If we are using http keepalives, how does it react to this? Will it consider the server still up, or will it try to look for some kind of response from the server via port 80, and mark it as down if it doesn't get it?
--A.Hsu
