none
Site Resiliency Solution for Hyper-V

    Question

  • Hi all,
           I hope you are all fine.  I have come across a situation where I need to design a solution for the customer keeping site resiliency in mind.   A failover cluster on the site provides fault tolerance and high availability on the SITE, but if something goes wrong in the SITE and we want the services available then how would I achieve this?

    Thanks in advance.

    Wednesday, March 10, 2010 1:26 PM

All replies

  • Lets see....(creativity is your limit)

    Distance clusters of Hyper-V hosts.

    SAN replication of the VHDs to a hot site where there are pre-configured VMs with the replicated VHDs ready to boot.
    (I personally think this is a rather simple yet effective solution)

    Distance clusters of VMs (the VMs are clustered between datacenters lust like you would do with a physical server)

    In any of these, don't forget your network layer - we are only discussing keeping an OS or application running.

    Brian Ehlert (hopefully you have found this useful)
    Wednesday, March 10, 2010 4:01 PM
  • Thanks for the reply.  You have always been a great help.

    I will clarify option one by one.

    1.  Distance Clusters of Hyper-V Hosts
         Does not this require a third-party solution such as GeoCluster or can it be achieved through just windows clustering services.

    2.  SAN replication of the VHDs to a hot site where there are pre-configured VMs with the replicated VHDs ready to boot.
         What exactly are the requirements for that.  I am really interested in this option as I can afford a little downtime.  Does it require anything on the part of SAN like synchronization thing or specific SANs.  Can you guide me to any link where I can study it in detail.  Thanks again.

    3.  Distance clusters of VMs (the VMs are clustered between datacenters just like you would do with a physical server)
         Again, is not this the firt option that you describe, a geographically dispersed cluster.

    I can afford a little lag during the failover and I believe the 2nd option that you have proposed is a good one.  Please, elaborate a little on this.  Thanks again.  God bless you.

    Thursday, March 11, 2010 5:57 AM
  • Hello.  Anyone.
    Thursday, March 11, 2010 12:03 PM
  • MSFT Clustering (active / Active and Active/passive) can span datacenters.

    GeoClustering is taking this to the nth distance (and there are folks doing this - but there is generally some specialized hardware involved when you move to great distances, such as North America with the UK)

    If you are interested in this, then you need to head over to the Clustering information sets.
    Most of the information is application specific however all the same rules apply.
    http://www.microsoft.com/windowsserver2008/en/us/failover-clustering-r2.aspx
    http://www.microsoft.com/windowsserver2008/en/us/failover-clustering-multisite.aspx
    Something old, but with good descriptions of the concepts:
    http://technet.microsoft.com/en-us/library/cc917704.aspx


    SAN replication gets talked about randomly in this formum - Like I mention it is relatively straightforward and simple to administer.
    Yes, SAN replication reqires that the SAN itself is the one that synchronizes data changes between the live SAN and a SAN at the DR location.  So it generally involves licensing, a new feature, a second SAN, yadda, yadda.
    And, it all depends on your SAN vendor (most enterprise SAN solution vendors offer this)
    It is primarily for data replication - it can be used for a hot site or a cold site - it all depends on how you implement it.

    http://www.bing.com/search?q=san+replication


    Brian Ehlert (hopefully you have found this useful)
    Thursday, March 11, 2010 3:49 PM
  • Thanks for the reply.  Thanks for the links, I will definately look into this.

    Following is what I have in mind.  If I am thinking right, kindly, approve or guidem me further.


    A two-node cluster at Production Site connected with shared storage SAN1 holding all the production virtual machines.

    A hyper-V server at DR Site connected with a storage SAN2.

    At the time of creation of virtual machines in the production site, the VHDs will be transferred to SAN2 (that would be present at Production Site and then moved to DR Site)

    any incremental changes in the VHDs will be replicated to the SAN2 in DR Site.

    In case of a Site Failure, We will create the virtual machines in the DR Site using the VHDs present in the SAN2.


    Correct me if I am wrong.


    Thanks and waiting anxiously for the reply.

    Thursday, March 11, 2010 7:03 PM
  • All of that thinking is correct.

    What you need to watch for is having too many VMs on a single LUN.  Not from a number of VHDs angle but from a Disk IO angle.

    Each VM has its own IO and that IO compounds on top of each other when access the physical disks that actually hold the VHDs.

    This is just how you carve up the SAN.
    Brian Ehlert (hopefully you have found this useful)
    Friday, March 12, 2010 5:12 PM
  • Thanks for the reply.  Let me further clarify, few things from you.


    1.  A two-node cluster at Production Site connected with shared storage SAN1 holding all the production virtual machines.

    Clean as a whistle

    2.  A hyper-V server at DR Site connected with a storage SAN2.

    Clean as a whistle

    3.  At the time of creation of virtual machines in the production site, the VHDs will be transferred to SAN2 (that would be present at Production Site and then moved to DR Site)

    Few questions here, how would one transfer the baseline VHDs from SAN1 to SAN2 - would this be the duty of SAN Vendor or Administrator?

    Does the SAN1 and SAN2 need to have same LUNs for each Virtual Machine? Like if VM1 is on LUN1 on SAN1 so it needs to be on LUN1 on SAN2?

    4. any incremental changes in the VHDs will be replicated to the SAN2 in DR Site.

    How would this happen.  Is this also the functionality of SAN and again handled by SAN Administrator or Vendor?


    5.  In case of a Site Failure, We will create the virtual machines in the DR Site using the VHDs present in the SAN2.

    In case of a failure, the replication would stop no matter how it is being performed and who is managing it.  Now, on the DR Site, when we would create new virtual machines and assign the VHDs in SAN2, would not the VHDs be corrupted or what if they could not run?  when replication would be done from san1 to san2, how it would be handled by SAN2 and VHD like it would create an additional file or merge it with existing VHD.


    Thanks for the time and waiting anxiously for the reply.  





    Tuesday, March 16, 2010 8:51 AM
  • Step 5 can be done at any time - as the SAN will be replicating the VHDs over.

    Step 4 is handled by the SAN replication software (see your SAN vendor)

    Step 3 - that is all a personal preference and a limitation of the replications feature of the SAN (brand specific)
    Most SAN replication software creates a mirror - so what you configure at your primary site is mirrored at your secondary site.
    Does it have to be equal - no, it does not.
    Does it have to be able to be configurable and meet Hyper-V rules?  Yes.

    This means that if you have VMs at site A and they all are HA and they are all configured with individuals LUNs (one LUN per VM), then at site B you desire to have just one big LUN with all VHDs - then this big LUN at site B needs to be a CSV LUN (that is the confuguration for many VM VHDs on a single LUN)

    It is all design - by far the easiest to deal with is a full mirror.  As the configuration you are used to looking at every day is also the configuration you will see in a disaster.

    1 and 2 are all personal preference.

    There have been a lot of suggentions along this thread - in the end you have the desiction of how you are going to implement.  Your personal implementation is based on what you need and what you know.
    If there are grey areas - please, ask.

    You also need to understand your SAN and its capabilities - that is not something we can answer here - as this might have a significant impact in defining your implementation.

    BTW - you might want to watch the Citrix Essentials for Hyper-V product for the new "Site Recovery" feature.  It is very similar to what I described but through a different mechanism.


    Brian Ehlert (hopefully you have found this useful)
    Wednesday, March 17, 2010 2:45 PM
  • To add to the answers below, if you are looking for an asynchronous replication product which is storage agnostic, there is a new feature in Windows Server 8 Beta called Hyper-V Replica. This guide http://www.microsoft.com/download/en/details.aspx?id=29016 contains more details on the feature

    Praveen

    Friday, April 06, 2012 2:27 AM
  • Replica is a very specific solution and it is designed for Disaster Recovery not any other consideration.

    The solution is designed to be easy to flip to the replica in the case of an event but not as a way to provide redundancy such as clustering does.  Do not consider Replica as a replacement for high availability, only consider it as a solution for a disaster recovery scenario, such as a hot site.


    Brian Ehlert
    http://ITProctology.blogspot.com
    Learn. Apply. Repeat.
    Disclaimer: Attempting change is of your own free will.

    Friday, April 06, 2012 3:42 AM
  • We use doubletake (no I don't work for them) to do this.  They of course are not the only ones.  We do have SANS in both locations but we made some internal decisions into why we didn't want to use SAN replication.  Doubletake does it at the hypervisor host level (failover and failback).  Our sites are across states and replica via WAN.  Some others that I think are similar are:

    • Neverfail
    • Steeleye
    • I think CA  has one called HA Arc serve or something.

    These solutions basically handle the dirty work for you, work regardless of storage (SAN etc) and generally allow HA failover across WAN and failback.  Perfect for those 'there's a hurricane coming' or site) a burned to the ground so boot everything at site b) 7 states away.

    Friday, April 06, 2012 2:48 PM
  • Hello,

    In addition to the solutions noted for Windows Server 2008 R2 SP1, there is a new solution coming for Hyper-V in Windows Server 2012, Hyper-V replica.  It is used to replicate virtual machines to a DR site for recovery.  You can test it out with these procedures:

    http://blog.concurrency.com/infrastructure/configure-hyper-v-replica-in-windows-8/

    http://blog.concurrency.com/infrastructure/monitoring-hyper-v-replica-in-windows-8/

    More information from the Windows Virtualization Blog:

    http://blogs.technet.com/b/virtualization/archive/2012/03/27/why-is-the-quot-hyper-v-replica-broker-quot-required.aspx

    Nathan Lasnoski


    http://blog.concurrency.com/author/nlasnoski/

    Friday, April 27, 2012 7:22 PM
  • Hello,

    Did this answer your question?

    Nathan


    http://blog.concurrency.com/author/nlasnoski/

    Thursday, May 10, 2012 3:56 AM