We also have Apache web servers installed on the node for serving up our Sorry Service, the load balancer directs traffic here when the whole pool is down. Also hosted by Apache are up-down pages whereby we direct the load balancer to consider a node up or down. Should our monitoring determine a node need a break from normal traffic, I can mark that node down and cause traffic to divert to other nodes before there is a real problem. It can stop receiving other work without having to shut down the node.