Sign in
 
HomeWindows Server 2012Windows Server 2008 R2Windows Server 2003LibraryForums
Windows Server TechCenter >
> 
> 
Hyper-V Cluster Failover Scenarios

Answered Hyper-V Cluster Failover Scenarios

  • Wednesday, February 27, 2013 6:32 PM
     
     
    Sign In to Vote
    0

    Hi All,

    My question is basically identical to this post: http://social.technet.microsoft.com/Forums/en-US/windowsserver2008r2virtualization/thread/3b126efb-5c69-45a3-8f2d-3dd08bf66a74

    There was no answer to that post, so I'll phrase mine a little differently....  Since in the Hyper-V cluster, the guest only networks are not cluster resources, a failure of one of these does not trigger a failover event.  Does anyone have any suggestions as to the best way to address this, or is the answer just "guest cluster"?

    Thanks!

     

All Replies

  • Wednesday, February 27, 2013 8:56 PM
     
     Answered
    Sign In to Vote
    2

    Hi!

    Fail-over clustering is a HA solution to ensure VM availability in the aspect that it is indeed running. This is, as you probably know, merely a way to assure VM functionality in case of a Hyper-V host outage.

    Since you are allowed to construct private as well as external networks and generally install anything you like in a virtual machine, there is no solid way to perform online availability checks from a host cluster point of view, without customization, deviating from best-practice guidelines or maybe even cause damage or instability to the cluster itself.

    Configuring cluster resources and dependecies are sometimes necessary, but only to enforce that virtual machines are able to fail-over properly.

    There are ways to achieve what you wish to accomplish, and you were actually touching the subject yourself, constructing a guest cluster is one solution, but there ought to be several more. The problem these days, I think, is that most people forget to treat their virtual servers as servers. Hyper-V should help consolidating and facilitate management of servers, but is expected to handle practically everything, thus adding a layer of unforseen complexity in regards of administration.

    My point with all of this, is that it's probably not at the Hyper-V or cluster levels you should apply these modifications, but rather monitor guests and trigger failovers at desired events from an onlooking system.

    With Hyper-V 2008 and 2008 R2, Microsoft gave us high availibility. With 2012 and the system center 2012 products, they have taken the step further with continous availibility.

     I haven't tried any of this or read up much on the subject, but from what I understand the Orchestrator in cojunction with SCOM, you should be able to design flow-charts and tailor these kinds of things exactly as you wish to see them happen. Monitoring servers SNMP, ping, services, etc. can be done with a lot of softwares such as SCOM, nagios, OP5, but with Orchestrator I'm under the impression that you can also automize taking actions, for example if a virtual machine stops responding to ping, it should immediately failover.

    Powershell is another thing, that came out strong and capable with 2012. Powershell is probably what you should script your actions with, if you wan't to knit something together yourself.

    All in all, there are far more variables at guest level to take into consideration when looking for this type of solution, than at host level. If a Hyper-V host was allowed to see everything in a guest, totally transparently, to decide on what actions to take, then it would probably not be considered secure enough to recommend putting to use in production environments. The host/guest isolation is overall a good thing.