I have VMware Fusion 7.0.0 (2103067) with a centos 6.x guest. Everything works great unless I push the network a little too hard (not very hard as it turns out), and then DNS lookups become intermittent. Because the VM is using NAT then the DNS lookups are to the VMware name server. By writing a simple script which uses dig to lookup the DNS A record for a hostname every 0.1 seconds, I can instantly see when DNS failures start and more importantly when the DNS failures stop. When the failures start, if I immediately stop all network activity except for the digs every 0.1 seconds, then the failures will always stop 60 seconds later.
My suspicion is that the VMware Fusion NAT is causing this problem because of two pieces of evidence:
Evidence item #1: I read somewhere that the NAT timeout for a particular state is 60 seconds.
Evidence item #2: If I change the NIC to bridged then the failures do not happen.
I'm guessing that the NAT state table is not big enough / has too long timeouts, and that once all the states get used then the DNS problems start happening and will last for about 60 seconds until an existing state times out. I have Googled and found similar problems solved for VMware ESX but tuning the configuration parameters of the state table; e.g. number of states / packets. However, I am unable to find information regarding a similar tuning fix for VMware Fusion.
Do such NAT configuration parameters exist for VMware Fusion? Or, am I barking up the wrong tree with my suspicions? If so, what else could be to blame?
Why don't I just switch to bridged networking? Because the NAT is definitely needed for this particular setup. Why don't I switch to a multi-NIC NAT and bridged setup. This works and is my fall-back plan. However, I would prefer just having NAT for security reasons.