Kissel-B
asked on
Server 2016 Hyper-V Issue Guest VM's @ Please wait or have connection issues
I have a two node Hyper-V cluster I am upgrading the hardware to newer gen hardware and upgrading the node os to Windows server 2016.
I am having a issue were any VM's live migrated or started on the new 2016 Hyper-V node sit at Please wait for ages. If I get a login in screen after logging in I get a applying settings screen which also takes for ever. Doing some research it looked like a network connection issue I have checked the switches, settings tried different Nics and even removed the virtual network connection to the guest still issues continue. If I move it back to the 2012 R2 node comes right up logins in fine and have not connection issues. Even when I get a guest to start there are connection issue for instance a DAG member will not see the other on the other node even know it can be pinged. I am at a loss. The newer servers are HP DL580 G7 the older node is DL580 G5. All the guests are 2012 R2 Exchange are 2012. Nics are Built in 4 port 1 GB and I have some 10GB Mellanox connectx-2 EN.
I am having a issue were any VM's live migrated or started on the new 2016 Hyper-V node sit at Please wait for ages. If I get a login in screen after logging in I get a applying settings screen which also takes for ever. Doing some research it looked like a network connection issue I have checked the switches, settings tried different Nics and even removed the virtual network connection to the guest still issues continue. If I move it back to the 2012 R2 node comes right up logins in fine and have not connection issues. Even when I get a guest to start there are connection issue for instance a DAG member will not see the other on the other node even know it can be pinged. I am at a loss. The newer servers are HP DL580 G7 the older node is DL580 G5. All the guests are 2012 R2 Exchange are 2012. Nics are Built in 4 port 1 GB and I have some 10GB Mellanox connectx-2 EN.
ASKER
No I am replacing the hardware so I am just adding a node and removing the old hardware. Disabling the VMQ had no effect.
ASKER
Sorry I did not see the first part the storage is on another server which is connected to the nodes via iscsi. Both are running the iscsi on an arista 7148 10Gb
FYI: There is a difference between disabling VMQ in the operating system versus in the driver itself. Broadcom Gigabit NICs are notorious for causing all sorts of havoc as a result of not abiding by the VMQ spec calling for VMQ on Gigabit adapter ports to be DISABLED.
Disabling in the OS doesn't help.
Since new hardware is being added I suggest doing a Cluster Rolling Upgrade. NOTE: For failover compatibility until the cluster level is brought up to 2016 the CPU Compatibility mode needs to be enabled. This requires the VMs to be shut down.
Disabling in the OS doesn't help.
Since new hardware is being added I suggest doing a Cluster Rolling Upgrade. NOTE: For failover compatibility until the cluster level is brought up to 2016 the CPU Compatibility mode needs to be enabled. This requires the VMs to be shut down.
ASKER
I am experiencing the please wait, applying users settings and applying group policy registry policy on the other node now
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
No I am the only one that can edit it and I have not touched it in years but I agree it seems that the servers are having issues reading it I thought is was confined to the new hardware but the old box is doing it as well
ASKER
Yes you were right something was choking the Group policy delegation. Upon searching the group policy error more in depth it appears that jumbo frames is the issue on switch did not have it enabled once i removed the jumbo frames from the dc's and member servers communications across the switches and nodes came back up.
That sounds more like an RSS (Receive Side Scaling) problem than a Jumbo Frames problem. But, I'm glad that you got it sorted out.
Make sure the Broadcom based Gigabit NIC ports have Virtual Machine Queues disabled at the _driver_ level for ALL physical ports in each node.
Is this a Cluster Rolling Upgrade?