Today I had a conference call with the people from Emulex responsible for the network adapters, firmware and drivers. As many of you know we’ve had a long lasting issue with HP/Emulex 554FLB CNAs in HP BL460c Gen 8 blade servers in c7000 blade enclosures. After we had replaced Windows Server 2012 with Hyper-V by Windows Server 2012 R2 with Hyper-V on the same hardware, we started to notice virtual machines losing connection. We have multiple guest clusters on top of the Hyper-V clusters and sometimes during Live Migration of one of the cluster nodes, we would see that connectivity was lost, even to the point that a cluster node would be forced to leave the cluster and come back later when the network connection was re-established. In fact that was our single best test to reproduce the problem.
Marc van Eijk en Peter Noorderijk wrote blogs about it and together they got over 200 comments from customers all over the world seeing the same problem, the majority of them having Emulex through an OEM like HP, Dell, IBM and Cisco. In fact we also read similar cases with HP rack servers, other NICs and different hardware combinations which led me to believe there could also be an issue in the networking/teaming stack in Windows Server 2012 R2.
During the months of November and December 2013 we collected a list of 10 registered support cases with either HP, Microsoft or both. At that time my primary contact was a senior escalation engineer at Microsoft who was able to collect several customer cases mostly from European customers. Unfortunately we could not register a support case with HP ourselves because we did not have a support contract with them. But the amount of customers that discovered they had the same problem, grew and grew. There must have been hundreds of open support calls with both HP and Microsoft and the storm of discontent was growing.
According to HP they were dependent on Emulex and HP did not get any feedback either. Also Microsoft was left in the dark for a long time.