Hyper-V VM's randomly losing connectivity

[d0pefish]

OK so this is our setup, and until yesterday has worked fine.

 

3 HyperV hosts set as a failover cluster. Each host has about 5 or 6 virtual servers on.

 

Now as of yesterday, the servers running on Hyperv-1 failed to respond to ping requests randomly (some clients could ping a server, others timed out). If you left the ping running and live migrated the server, soon as it started on another HyperV host, all pings worked.

Ping from Hyperv-1 > VM and VM > Hyperv-1 is fine, as are all pings to Hyperv-1 itself.

This is only happening on this host, the other two are fine and migrating servers to them solves it. So to me, an issue with Hyperv-1 host somewhere, but where is the problem.

Have tried looking at the VM adapter settings on the host, all seems fine. As I said, until yesterday this setup had worked fine, but is now causing issues.

 

Any ideas would be welcomed as it's getting to the point of not knowing what else to try!!

Just as an addition, the ping test was merely a test (it's not all that was tested), the servers appear to refuse all connections, so no network shares, no RDP, nothing. But you still have some clients that CAN get on their shares and RDP to them, so its not a complete cut-out

[StephenHardy]

Sounds a bit like DNS, but could be the NIC order... Might be worth checking

[StephenHardy]

Your failover cluster host to host Ping May be going through the failover NIC

[StephenHardy]

Also check on the failover NICs and any SAN connections, everything apart from IP4/6 (6 if you use it) is disabled

[d0pefish]

of the two NICs teamed up for VM traffic, disabled one and it seems to have smoothed out, so first impressions are now it was the VM traffic team having a faulty NIC. now going to see if the error lies with cabling or card.

cheers for the pointers

[SYNACK]

Have you considered the NIC Drivers, I had a server where the Intel Drivers were as flakey as anything till upgrading to the latest ones that allowed segregation to VMs and they were still slow. Are you able to look at upgrading your hosts to Server 2012 R2 as this supports link aggregation without involving dodgey vendor drivers. I know I had a similar issue back on server 2008 with drivers and hyper-v that was only supported by taking the vendor supplied cack out of the loop - that was the latest HP teaming software that reportedly had issues unless installed in the correct order, it had issues even when using the order they suggested.