locked
Windows 2008 R2 HyperV Cluster RRS feed

  • Question

  • Hello there

    A few questions:

    We have an existing Windows 2008 Standard HyperV server non clustered. We want to re-use this and utilise it as one of the nodes of a new R2 HyperV cluster we are planning for. What would be the easiest way to do this. I'm assuming here that it is probably going to be a wipe and reload, or is there another solution that enables me not to have to wipe the box and restore VHDs. If not would I be able to easily restore our existing VMs on the new cluster once that was built using DPM, any issues with this?

    and lastly

    I understand that a HyperV machine should have a minimum of 2 physical NICs, one for management and one dedicated for VMs. In order to make use of more NICs to provide increased throughput and spread the load over multiple NICS what is current best practice? I guess I can team HP NICs and configure the Vswitch to use the team NIC or I could also create additional virtual switches that binds to an additional physical NIC. I cant seem to find much information on this and wondered what most people do. Obviously one NIC is not going to be able to cope with 10-15VMs.

    Your help is much appreciated

    Thanks

    Adam
    Tuesday, March 9, 2010 7:08 PM

Answers

  • Hi,

     

    You can upgrade the existing Windows Server 2008 Standard Hyper-V Server to Windows Server 2008 R2 Hyper-V server.

     

    Note: The failover cluster feature is not available in Windows® Web Server 2008 R2 or Windows Server® 2008 R2 Standard.

     

    Before you upgrade the system, please delete or merge the snapshots of the VMs, shutdown the VMs. At the same time, please ensure that the processor of the existing computer from the same manufacturer of your new computer(all from Intel or AMD, you can’t mix them in the cluster).

     

    As you mentioned you have VM backup from DPM, please note that you can’t restore the VMs to Windows Server 2008 R2 Hyper-V computer if you took an online backup on Windows Server 2008 Hyper-V computer.

     

    Yes, we recommend that you use a dedicate NIC for host management. By the way, as network teaming is only provided by hardware vendors, you may have known, we don’t support Hyper-V network teaming officially. You may contact the manufacturer of the hardware vendor to have more information on compatibility information. You may also wait for other customer's experience sharing on this.

     

    More Information

    -------------------------

    Microsoft Support Policy for NIC Teaming with Hyper-V

    http://support.microsoft.com/kb/968703

     

     

    Best Regards,

    Vincent Hu

     

    • Marked as answer by Vincent Hu Wednesday, March 17, 2010 7:15 AM
    Friday, March 12, 2010 9:28 AM
  • I actually had to perform this exact same action.  If you are planning on using SCVMM (if you weren't, look into it!), get your SCVMM server(s) up and running and then take these steps:

    1) Add your current server as a host in SCVMM
    2) "Store" your current VMs into the SCVMM library
    3) Remove your host server from SCVMM
    4) Once your host server is free of all VMs, reload the two servers with Server 2008 R2
    5) Configure the servers, add Hyper-V role, add the virtual networks, install Cluster services.
    6) Build your cluster
    7) Add your clustered hosts into SCVMM
    8) Create some test VMs, do some migrations, just test your configuration and make sure everything is golden
    9) Deploy the stored VMs onto your cluster

    You may have to configure some of your network settings afterward on the VMs, but this shouldn't be that big of deal.
    • Marked as answer by Adam42 Thursday, March 18, 2010 10:31 PM
    Monday, March 15, 2010 1:34 AM
  • I can't praise SCVMM enough (although it does have its quirks as any MS product).  The Self-Service portal has also made it convenient for me to have my Web and SIS admins manage their servers and create new servers from a template I made them instead of asking me every two weeks to build them a new dev server.

    Also, I might add since I didn't notice it on your first message.  I originally had these servers set up with the NICs teamed, but after some problems with getting too many ARP packets after a live migration.  I am limited to 4 NICs on the servers, so I ended up using two for iSCSI, one for management & cluster communication (I would love to split that up), and one dedicated to VM traffic.  Now it needs to be said, most of my servers are not network intensive at any given time.  Being an K-12 educational instutution, we are usually victims of server sprawl and our servers get very limited usage.  This is why we moved to a virtual environment in the first place, to better allocate and use resources instead of buying a 3-4 thousand dollar server that gets 5-10% utilization and eats up more space than necessary in my datacenter. Given that we only have a 1gb backbone anyhow, and only 1gb going into my blade chassis, so whether I limit the VMs to 1gb at the chassis or the server level, it ultimately doesn't matter.

    • Marked as answer by Adam42 Thursday, March 18, 2010 10:31 PM
    Monday, March 15, 2010 7:52 PM
  • Hello,

    I'd agree with the system of storing a VM in the SCVMM library, or simply exporting and importing it.  This will allow you to bring the server into a cluster without risking the VMs on it.

    SCVMM is an awesome tool.  Here are a quick overview of its capabilities:

    * Centralized management of all VMs and hosts
    *  VM and Template library
    *  Centralized storage of ISO files
    *  Physical to Virtual conversions
    *  Self VM provisioning

    Have a good day!

    Nathan Lasnoski 
    • Marked as answer by Adam42 Thursday, March 18, 2010 10:32 PM
    Tuesday, March 16, 2010 5:11 AM

All replies

  • Hi,

     

    You can upgrade the existing Windows Server 2008 Standard Hyper-V Server to Windows Server 2008 R2 Hyper-V server.

     

    Note: The failover cluster feature is not available in Windows® Web Server 2008 R2 or Windows Server® 2008 R2 Standard.

     

    Before you upgrade the system, please delete or merge the snapshots of the VMs, shutdown the VMs. At the same time, please ensure that the processor of the existing computer from the same manufacturer of your new computer(all from Intel or AMD, you can’t mix them in the cluster).

     

    As you mentioned you have VM backup from DPM, please note that you can’t restore the VMs to Windows Server 2008 R2 Hyper-V computer if you took an online backup on Windows Server 2008 Hyper-V computer.

     

    Yes, we recommend that you use a dedicate NIC for host management. By the way, as network teaming is only provided by hardware vendors, you may have known, we don’t support Hyper-V network teaming officially. You may contact the manufacturer of the hardware vendor to have more information on compatibility information. You may also wait for other customer's experience sharing on this.

     

    More Information

    -------------------------

    Microsoft Support Policy for NIC Teaming with Hyper-V

    http://support.microsoft.com/kb/968703

     

     

    Best Regards,

    Vincent Hu

     

    • Marked as answer by Vincent Hu Wednesday, March 17, 2010 7:15 AM
    Friday, March 12, 2010 9:28 AM
  • I actually had to perform this exact same action.  If you are planning on using SCVMM (if you weren't, look into it!), get your SCVMM server(s) up and running and then take these steps:

    1) Add your current server as a host in SCVMM
    2) "Store" your current VMs into the SCVMM library
    3) Remove your host server from SCVMM
    4) Once your host server is free of all VMs, reload the two servers with Server 2008 R2
    5) Configure the servers, add Hyper-V role, add the virtual networks, install Cluster services.
    6) Build your cluster
    7) Add your clustered hosts into SCVMM
    8) Create some test VMs, do some migrations, just test your configuration and make sure everything is golden
    9) Deploy the stored VMs onto your cluster

    You may have to configure some of your network settings afterward on the VMs, but this shouldn't be that big of deal.
    • Marked as answer by Adam42 Thursday, March 18, 2010 10:31 PM
    Monday, March 15, 2010 1:34 AM
  • Hi Vincent

    thanks for the information.

    If I dedcided not to use NIC teaming, what is the next best approach for spreading the load of multiple VMs over multiple physical NICs?

    Thanks

    Adam
    Monday, March 15, 2010 4:02 PM
  • Great thanks Scott. I was going to be looking at SCVMM at some point, however it was on the back burner, not any more though.

    thanks for your help

    Adam
    Monday, March 15, 2010 4:06 PM
  • I can't praise SCVMM enough (although it does have its quirks as any MS product).  The Self-Service portal has also made it convenient for me to have my Web and SIS admins manage their servers and create new servers from a template I made them instead of asking me every two weeks to build them a new dev server.

    Also, I might add since I didn't notice it on your first message.  I originally had these servers set up with the NICs teamed, but after some problems with getting too many ARP packets after a live migration.  I am limited to 4 NICs on the servers, so I ended up using two for iSCSI, one for management & cluster communication (I would love to split that up), and one dedicated to VM traffic.  Now it needs to be said, most of my servers are not network intensive at any given time.  Being an K-12 educational instutution, we are usually victims of server sprawl and our servers get very limited usage.  This is why we moved to a virtual environment in the first place, to better allocate and use resources instead of buying a 3-4 thousand dollar server that gets 5-10% utilization and eats up more space than necessary in my datacenter. Given that we only have a 1gb backbone anyhow, and only 1gb going into my blade chassis, so whether I limit the VMs to 1gb at the chassis or the server level, it ultimately doesn't matter.

    • Marked as answer by Adam42 Thursday, March 18, 2010 10:31 PM
    Monday, March 15, 2010 7:52 PM
  • Hello,

    I'd agree with the system of storing a VM in the SCVMM library, or simply exporting and importing it.  This will allow you to bring the server into a cluster without risking the VMs on it.

    SCVMM is an awesome tool.  Here are a quick overview of its capabilities:

    * Centralized management of all VMs and hosts
    *  VM and Template library
    *  Centralized storage of ISO files
    *  Physical to Virtual conversions
    *  Self VM provisioning

    Have a good day!

    Nathan Lasnoski 
    • Marked as answer by Adam42 Thursday, March 18, 2010 10:32 PM
    Tuesday, March 16, 2010 5:11 AM
  • Thanks to you all, very helpful indeed and very much appreciated.

    Adam
    Thursday, March 18, 2010 10:31 PM