Windows 10 Guest freezes every minute - vif Guest Rx stalled

[Ulrich Thomas]

Hello,

I'm using XenServer 7.1 CU2 with latetest patches installed.

It's running on HPE DL380 Gen9 with Nvidia Grid K2 Card and serves Windows 10 VDI.

 

I few week ago, one of my users complained that the vm is unrespondable for a few seconds every minute.

 

in the kern.log I see the following entries at the time where this freezes happend.

Mar  3 14:10:07 xenserver01 kernel: [56967.225339] vif vif-9-0 vif9.0: Guest Rx ready
Mar  3 14:11:07 xenserver01 kernel: [57027.257711] vif vif-9-0 vif9.0: Guest Rx stalled
Mar  3 14:11:17 xenserver01 kernel: [57037.041781] vif vif-9-0 vif9.0: Guest Rx ready
Mar  3 14:12:18 xenserver01 kernel: [57098.174168] vif vif-9-0 vif9.0: Guest Rx stalled
Mar  3 14:12:26 xenserver01 kernel: [57105.882232] vif vif-9-0 vif9.0: Guest Rx ready
Mar  3 14:13:27 xenserver01 kernel: [57167.174643] vif vif-9-0 vif9.0: Guest Rx stalled
Mar  3 14:13:36 xenserver01 kernel: [57175.854705] vif vif-9-0 vif9.0: Guest Rx ready
Mar  3 14:14:36 xenserver01 kernel: [57236.127123] vif vif-9-0 vif9.0: Guest Rx stalled
Mar  3 14:14:44 xenserver01 kernel: [57243.859153] vif vif-9-0 vif9.0: Guest Rx ready

The vm isn't pingable at this thimes.

If I deactivate/activate the vm network in xencenter this ping pong is over.

 

I'll very pleased if someone can point me to the underlying issue.

 

regards,

Ulrich
 

[Tobias Kreidl]

Make sure there wasn't a duplicate IP address assigned to another instance on the network that conflicts. Also, is NTP properly synchronized on all servers in the pool?

 

-=Tobias

[Alan Lantz]

What changed a few weeks ago? Sounds like congestion to me. There are reports online of bugs in the software causing this sort of thing as well. If you think you are good on resources upgrading to a supported release would probably the the right direction to go.

 

--Alan--

 

 

[Ulrich Thomas]

There aren't any ip conflicts and the time is in sync. 

If you restart the vm or cycle the vm network adapter in xencenter the error is gone for the time but isn't solved.

 

This setup was running over the last 5 years without any issues. The only changes in the past were the hypervisor patches. 

 

problem is related to the xenserver network stack (openvswitch). from inside xen the vm isn't pingable when this "stalled" entry is logged, so every minute there are up to 10 pings that are lost.

Via XenCenter console the vm is responsible and not stucked at all

 

-= Ulrich =-