hi all,
I am experiencing this strange issue with my TMG 2010 in a vm on hyper-v 2008 R2 running on a Fujitsu RX300 S7 with an HP NC360T and Intel I350:
When I am using a synthetic network adapter in hyper-v as my external nic it looses every about 2-3 days its connection with the default gateway. ARP:
xxx.xxx.227.153 Unreachable Incomplete
I have a second "external" nic (legacy) running, which never lost its connection and arp was always good.
Additionally with the synthetic driver I had a feeling that TMG was unable to route packets correctly through either of the two nics...
An example:
Nic 1: 10.1.1.10-.1.15 / DG: 10.1.1.1 (synthetic, Intel I350)
Nic 2: 192.168.1.10-.1.15 / DG: 192.168.1.1 (legacy, HP NC360T)
I set up a network rule, which says From Computer XY -> External | Outbound IP: 10.1.1.14
While this rule was active with nic1 being a synthetic adapter this computer XY never got any replies back... I found out with wireshark, that the windows tcp ip stack sent all packets through NIC2 but with the source ip of NIC1!?
I then read the article about how server 2008 + manages outbound ip connections and interface selection (longes prefix) and everything was clear and understandable.
But now I changed nic 1 to legacy because of the three-day-GatewayARP-dying problem and suddenly with both nics being legacy my example from above works perfectly correct!
I am now able to set up a static nat by selecting some pcs that should surf through nic (and isp) #1 and the rest surfs through nic #2.
Is this possible or just a coincidence?? Has anyone a good explanation why legacy adapters seem to be better suited in my case?
Thank you!