I've set up an Apache HTTP server + load balancer (with the mod_proxy_balancer)
After I execute load testing the balancer don't pass the majority of the load towards the after sales servers(simple Java Application over Tomcat)
I am unsure when the apache httpd can't handle the burden, or perhaps is it the burden balancer configuration..
This is the way my conf file appears like:
<Proxy balancer://mycluster> BalancerMember http://xxxxx.amazonaws.com connectiontimeout=200000 max=50000 timeout=200000 ping=200000 BalancerMember http://yyyy.amazonaws.com connectiontimeout=200000 max=50000 timeout=200000 ping=200000 </Proxy> ProxyPass /a balancer://mycluster timeout=200000
Unless of course you actually need some features that Apache provides I'd be enticed to make use of Varnish or HAProxy rather.
In a guess you are drained of threads with no more connections are now being passed towards the origin servers until a thread becomes free.
Your timeout / ping values appear extremely high - what kind of response occasions are the looking to see in the origin servers?
Perhaps you have load examined the person origin servers to determine what degree of load the can deal with, and what is the strain on them like throughout the burden-balanced load test?