So the setup I built for v8supercars.com.au is two Varnish servers fronting two productions servers with an admin server updating the DB for content.

One of the production servers was always running a much higher CPU useage then the other and upon investigating further, it wasn't just that one was handling the traffic better, it had all the traffic while the other one had minimal traffic being sent to it.

There is a load balancer that handles the traffic to the two Varnish server and that was working fine because basically each Varnish server had very similar loads but then the Varnish servers were sending most of the traffic to one server and not the other.

Server load post Varnish change - FusionReactor reports show activity load on the servers after changing the active server in Varnish

A quick look at the Varnish setup showed that both had the servers in the same order under the 'director' setting, effectively making backend server "V8SCWEB03" the primary, hard working server. So a quick edit to swap the order on one of the Varnish servers straight away fixed the problem.

The graphs above show the activity being received by both server both before and after the change.