After adding a new powerful Hyper-V server based on Server 2012 R2 on a Dell with Broadcom NetXtreme Gigabit adapters I had to struggle with heavy packet loss on the Hyper-V guests, which I just relocated from an older Windows Server 2008 R2 machine.
Played around with driver updates on the host, removing and readding virtual swich and guest network adapters, tried all the well known tricks about disabling offloads and increasing buffer size for send and receive buffer, however, I got no solution. And the most issues with network performance, Broadcom, Hyper-V were pointed to Teaming, which I had not active.
Today I spotted, that another pair of virtual machines on another Dell server with Broadcom network card running Server 2012 developed the same response issues.
Running ipconfig /all, checking the physical address entry and validating the found information with the arp cache of the pinging machine did now show the fact, that both hosts assigned the same physical addresses to the virtual network card of their guests!
C:\> arp -a
192.168.10.100 00-15-5d-99-65-01
192.168.10.200 00-15-5d-99-65-01
No wonder, why network connectivity was so flaky, since the switch could not decide, whom to serve. Manually adding a unique network address entry to the network card properties of the guest machines on one host left those without connectivity, but the issue of the other machines was immediately resolved. So I had to visit the Virtual Switch Manager, change the MAC address range there and remove and add the adapters from/to the guests again in the Hyper-V console.
While this issue should have not happened at all, I just post this information here, since it might helpful for others with multiple Hyper-V hosts based on Windows Server 2012.
Best greetings from Germany
Olaf