We help IT Professionals succeed at work.

Packet loss on Windows 2012 R2

2,446 Views
Last Modified: 2014-12-22
Hi,

I have a weird problem that I am trying to solve.

We have a hyper-v host running Windows Server 2012 R2 with 3 VMs. The host has 2 network cards Intel 82574L. The first card is configured as the management card for the host and the second is dedicated to the hyper-v external virtual switch.

What happens is this. As soon as I reboot the host, I get a lot of packet loss (about 50%) on the Host but not the VMs. If I change the static IP address, it works just fine but only until I reboot.

I already changed the cards drivers and also tried disabling IPv6 but no success. I also configured the Host on the second card to see if it was a hardware issue but the problem remained.

Thanks
Comment
Watch Question

CERTIFIED EXPERT
Distinguished Expert 2018

Commented:
Disable VMQ on the NIC.

Author

Commented:
I don't have to option in my network card. I also ran the PS command Get-NetAdapterVmqQueue and none are listed.
Rob GMicrosoft Systems Engineer

Commented:
Do you have a gateway on both nics?

Do not turn off IPV6 on 2k12 server.. Server 2k12+ uses IPV6 to communicate with other windows 7+ machines on the network, it is helpful if you leave that enabled, even if your LAN Switches and Routers don't support it. Otherwise you can make things worse..

Author

Commented:
Hi,

The management NIC has a gateway and the VM NIC is set to use DHCP so that receives one from that. As for the IPv6, it is reactived, it was only a test I did.
Rob GMicrosoft Systems Engineer

Commented:
You can't technically use two gateway's in windows.
It will cause this issue, by default windows will automatically choose the best route to use, (Shortest distance) from point A to point B. It sounds like you are having packet loss because you are running two gateways. You can try to fix this by forcing the primary NIC on the system to the one you want to use for your traffic, but generally, windows will still choose shortest route, which in VLAN's etc can cause the system to send out the traffic, but then not be able to get there.

Here are two tests you can do.
Both are going to cause some blip in network productivity.

1. Go to your network connections list (Control panel>Network and Internet>Network connections)
Press the ALT key on the keyboard, choose advanced, then choose advanced settings
Here you will see adapters and bindings,
move the primary NIC to the top of the list using the arrow keys on the right.
Click OK (May cause a blip)

Try this and see if you notice the issue again..
Chances are it won't fix the issue..

2. Remove the gateway from the management NIC and see if that fixes your issue..
I am 99% positive it will..

(Best description.. NIC 1 sends a packet to Desktop A. Desktop A sends a reply, but uses NIC2 as it is a shorter route, Nic 2 has no idea what Desktop A is talking about, NIC2 drops packet from Desktop A. After some time Desktop A might send replies back to NIC 1 but at that point NIC1 doesn't remember what Desktop A is talking about. This causes Latency and Packet loss)

Author

Commented:
Thanks for your comment.

I had already tried point number 1.

I removes the gateway and still get the same issue unfortunately. One thing to note, the hyper-v was originally using the same network card as the Hyper-V switch and I was having the same issue.

Also, when I reboot, the pings are fine for maybe 15-20 seconds before I start losing packets.

The packet loss are fairly regular, 4 go through and then 4 are lost.
Philip ElderTechnical Architect - HA/Compute/Storage
CERTIFIED EXPERT

Commented:
I suggest the following:

Remove the current vSwitch.
Team the two physical NICs using the Windows Server Team Wizard.
Create a new vSwitch shared with the host OS.
Bind the new vSwitch to the existing VMs.

Please make sure the most recent version of Intel PROSet is installed on the server prior to the above and also that your backups are good!

Also, I have an EE Article that might be of assistance: Some Hyper-V Hardware and Software Best Practices.

A virtualization platform should eliminate as many single points of failure as possible. A bad cable or switch port the vSwitch was bound to would put a full-stop to the VMs if I understand the description rightly.

Author

Commented:
Hi and thanks for you suggestion but we would very much like to keep the two NICs separate. Also, we have several other servesr with this configuration and we don't seem to experience this issue.

Do any of you think that it would be a good idea to try with a different NIC? We have a Intel Gigabit Desktop that we could try.
This one is on us!
(Get your first solution completely free - no credit card required)
UNLOCK SOLUTION

Author

Commented:
It fiex the issue as soon as it was unplugged
Unlock the solution to this question.
Join our community and discover your potential

Experts Exchange is the only place where you can interact directly with leading experts in the technology field. Become a member today and access the collective knowledge of thousands of technology experts.

*This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

OR

Please enter a first name

Please enter a last name

8+ characters (letters, numbers, and a symbol)

By clicking, you agree to the Terms of Use and Privacy Policy.