none
Hyper-v, Clustering, and Virtual Switch Failover

    Question

  • Hello,

    We have a failover cluster with the following: (FYI this system has been up and running for years with the problem below and was set up by my predecessor so I don't know how some things were done)

    • Dell 2950 2ea with 8 NIC ports
    • Dell MD3000i SAN using iSCSI
    • Windows 2008 R2 Enterprise
    • 2 Cisco 3750 POE, Layer3 switches with Stackwise Cable connecting them

    Each server has the following NIC connections:

    • 1 for Mgmt (not usable by Hyper-v)
    • 1 for heartbeat (direct connection not through a switch),
    • 2 for switch 1 (not usable by host system, Teamed using Intel "Virtual Machines Load Balancing)
    • 2 for switch 2 (not usable by host system, Teamed using Intel "Virtual Machines Load Balancing)
    • 2 connections to the SAN with redundant controllers.

    Hyper-v has 2 different virtual switches set up, "DATA" and "Services", initially they were set up without VLAN support but I had to add a VLAN to our network and put some VMs on that subnet so I changed the "Services" virtual switch for trunking to the Cisco 3750s, the "DATA" virtual switch is still using access ports on the switch with no VLAN tagging.

    What is happening is when I live migrate or through a failover the VM's forget which virtual switch they are supposed to use and the default to the "Services" virtual switch even if they were assigned to the "DATA" virtual switch.  The ones assigned to the "Services" virtual switch of course stay and that virtual switch but forget what VLAN they are supposed to be on, when a failover occurs everyone comes running, "Is Outlook out?"  I would like the VMs to stay on the correct virtual switches.

    In Server Manager under Features - Failover Cluster Manager - Networks I have the following:

    • Heartbeat (self explanitory)
    • Host_Management (self explanitory)
    • MD_RAID_M0_P0 (iSCSI to SAN)
    • MD_RAID_M0_P1 (iSCSI to SAN)
    • MD_RAID_M1_P0 (iSCSI to SAN)
    • MD_RAID_M1_P1 (iSCSI to SAN)

    The networks for VM traffic are not part of the cluster networks.  This is where I show my inexperience, are they supposed to be there?  Is this why I have my problem during Live Migration and failover, the cluster is unaware of the networks?  (FYI the virtual switches are named EXACTLY the same on both cluster nodes; I did read that in some documentation).

    Wednesday, February 29, 2012 10:20 PM

Answers

  • Hi.

    Did you configure the Network Settings (not the virtual network itself) of the VM via Hyper-V Manager or via Failover Cluster Manager?

    It is very important that storage and network settings are configured through the Failover Cluster Manager so that the cluster is aware of that.

    If you did it via the Hyper-V Manager then you have to do a "Refresh Virtual Machine Configuration" in the Failover Cluster Manager.

    Olaf

    • Marked as answer by Meseptiamus Friday, March 02, 2012 4:13 PM
    Thursday, March 01, 2012 7:13 AM

All replies

  • Hi.

    Did you configure the Network Settings (not the virtual network itself) of the VM via Hyper-V Manager or via Failover Cluster Manager?

    It is very important that storage and network settings are configured through the Failover Cluster Manager so that the cluster is aware of that.

    If you did it via the Hyper-V Manager then you have to do a "Refresh Virtual Machine Configuration" in the Failover Cluster Manager.

    Olaf

    • Marked as answer by Meseptiamus Friday, March 02, 2012 4:13 PM
    Thursday, March 01, 2012 7:13 AM
  • HI,

    AS per expert, NIC teaming will impact how you can place virtual machine on different VLANs. Hence you should check your hardware manufactures for the latest drivers and setup instructions.

    The network for VM traffic shouldn’t be part of cluster. 

    Please check whether you have added same VLANs on both the switch port so that the moved VM can start communicate through the new NIC on other node.

    Thursday, March 01, 2012 7:33 AM
  • I found the answer for this about 2 hours after I posted the question.  Yes you have to change the network settings through the Failover Cluster Manager to make the cluster aware of the settings/changes.  As soon as I did that all was well in Neverland.

    I also found that if you try and push too many VMs to a cluster node dangerously short of memory the BSODs start to invade Neverland as well.

    Anyway thanks for the responses and the help.

    Friday, March 02, 2012 4:17 PM
  • Hello -

    When I manage the virtual machine through the failover cluster manager, it brings up the Hyper-V manager. When I make the NIC setting change in the Hyper-V manager it does not keep the setting through the failover, thus keeping the original NIC setting from when the virtual machine was created.

    WOuld you be able to point me in the right direction of managing this virtual machine from the failover cluster manager?

    Thank you,

    Thursday, May 03, 2012 2:48 PM
  • Hi.

    Where did you right-klick?

    On the left side and choosing "Manage Virtual Machine"?

    Then you are right ad this "only" brings up the Hyper-V Manager. If you go this way you have todo a "refresh virtual machine configuration" afterwards so that the changes get written into the cluster properties.

    I would recommend to only select the virtual machine on the left side and the do a right click on the virtual machine in the middle windows and choose "settings".
    This will show the common Virtual Machine Settings window, BUT when you did an OK klick after your changes the cluster properties will be modified afterwards.

    Hope that helps

    Regards

    Olaf

    Thursday, May 03, 2012 3:07 PM
  • Olaf,

    Thanks! That is what I was looking for.

    Thursday, May 03, 2012 4:11 PM