The entire purpose of a network is to get data from point A to point B. If for some reason we aren't able to get data where we need it, it can sometimes be a pain to pinpoint exactly where the issue manifests itself. But through the process of elimination, pinpointing where routing issues manifest themselves shouldn't be too difficult.
Whenever I experience issues with a node being unable to communicate to a specific server or network, I like to work my way from their workstation back to the switch stack until I find the issue. To start, I check the obvious things, such as what the IP address is (or if the machine even has one) and then I also check the routing table. If the problem is intermittent, you would likely want to test the cable. For some reason, I've come across quite a few instances where a problem resulted from a bad cable. I don't know why, but it seems that other administrators I know, don't have this luck. But it never hurts to run a cable tester on...