none
Hyper-v replica and load balance

    Question

  • I'm after a bit of advice regarding replication for Hyper-v

    I currently have a full domain setup using a SOFS server hosting 10 servers. It works well. I have 3.6tb of total data using storage pools and the server is a good spec. I have purchased and am waiting on delivery of a secondary server that is exactly the same as this one. Same make, memory, storage etc.

    The intention of the secondary server is to provide complete fail-over for the site. I want to set it up right the first time :)

    I have a good solid guide to setup replication. My questions are around balancing the VMs out.

    I intend split the VMs 5 on one server and 5 on the other. Can I do this?

    I plan to have a dc on each then balance out the other 4 depending on use. Can I setup both servers as replica servers then replicate back the VMs not hosted on the server? so SOFS 1 would host servers 1-5 and SOFS 2 would host servers 6-10. SOFS 1 would replica server 1-5 to SOFS 2 and SOFS 2 would replica server 6-10 to SOFS 1. In the event of a failover of either server hosts would jump from one SOFS to the other.

    My aim is no pause in service for the host move would need to be seamless and not drop connectivity for the site.

    Any help or advice would be greatly appreciated.

    Tuesday, February 28, 2017 1:17 PM

Answers

  • Hi Sir,

    >>I intend split the VMs 5 on one server and 5 on the other. Can I do this?

    >>My aim is no pause in service for the host move would need to be seamless and not drop connectivity for the site.

    >>Can I setup both servers as replica servers then replicate back the VMs not hosted on the server? 

    Based on this , building hyper-v cluster may cover your needs . And , "planned failover" will migrate the service running on HA virtual machine from one node to another seamlessly .

    Because , hyper-v replica is not a "real-time" replica . There will be some delay during data replication .

    Hope it is helpful to you .

    Best Regards,

    Elton


    Please remember to mark the replies as answers if they help.
    If you have feedback for TechNet Subscriber Support, contact tnmff@microsoft.com.


    Wednesday, March 01, 2017 12:41 PM
    Moderator
  • Hi Jason,

    Yes, as Tim and Elton said you will need to implement cluster to make that VM HA. Replica is only DR solution in case of VM failure so that you can run/restore it from replica site.

    This may help 

    http://www.altaro.com/hyper-v/failover-cluster-manager/

    -----------------------------------------------------------------------------------------------------------------
    If you found this post helpful, please give it a "Helpful" vote. 
    Please remember to mark the replies as answers if they help.


    • Edited by Nedim Mehic Thursday, March 02, 2017 8:41 AM
    • Marked as answer by Jason2234uk Thursday, March 02, 2017 10:36 AM
    Thursday, March 02, 2017 8:40 AM

All replies

  • Hi Sir,

    >>I intend split the VMs 5 on one server and 5 on the other. Can I do this?

    >>My aim is no pause in service for the host move would need to be seamless and not drop connectivity for the site.

    >>Can I setup both servers as replica servers then replicate back the VMs not hosted on the server? 

    Based on this , building hyper-v cluster may cover your needs . And , "planned failover" will migrate the service running on HA virtual machine from one node to another seamlessly .

    Because , hyper-v replica is not a "real-time" replica . There will be some delay during data replication .

    Hope it is helpful to you .

    Best Regards,

    Elton


    Please remember to mark the replies as answers if they help.
    If you have feedback for TechNet Subscriber Support, contact tnmff@microsoft.com.


    Wednesday, March 01, 2017 12:41 PM
    Moderator
  • As Elton states, if you want no "drop [in] connectivity", you would need failover clustering.  Hyper-V Replica is a disaster recovery technology and not a high availability technology.  And, there is always some drop in connectivity, though it may be a short period of time, when a VM moves from one node to another.  When a VM is live migrated from one node to another, there is a very short period of time while the IP connection is moved from one host to another that the VM loses connectivity to its clients.  Depending upon what the VM is running, this drop in connectivity may or may not be noticed, but it is generally within the acceptable time limit that IP will recover.  However, if a node of the cluster fails, causing the VM to restart on the surviving node, there will be a drop in connectivity while the VM is rebooting.

    "plan to have a dc on each"

    If you do decide to use Replica between nodes instead of failover clustering, there is no need to replicate the DCs.  They already have built in replication and that should be used.  There is no benefit, and possible issues, if you try to use Hyper-V Replica on domain controllers.


    . : | : . : | : . tim

    Wednesday, March 01, 2017 1:42 PM
  • I have miss understood the difference between replica and clustering :)

    Do you have a good guide on failover clustering? My confusion has come from both systems having storage pools and not shared storage making me think I needed to replicate the data.

    So VM 1 on server 1 would automatically be copied to server 2 in the event of a failure on server 1 once clustering is setup?

    Thursday, March 02, 2017 8:19 AM
  • Ok I've done some research and can see my short comings and why I can't understand this :)

    I don't have a SAN so can't do cluster failover as I understand it unless I used a third party to create a Virtual SAN.

    Therefore rather than HA can I aim for DR which will still give me what I'm ultimately after.

    I manage a school network. HA isn't essential we just need failover to some degree. If I replicate half the VMs on to each Hyper-V server I will still get server recovery but not automatically?

    So for example I replica VMs 1-5 to server 2 from 1 and replica VMs 6-10 from server 1 to 2. If server 1 or 2 goes down for a hardware fault I can switch on the downed VMs on the working server.

    Once my server is repaired I can power back up shutdown the VMs from the working servers that didn't have the issues and start up again on the newly repaired server. We're not afraid of a bit of downtime we just don't want days of it while a server is repaired.

    Thursday, March 02, 2017 1:43 PM
  • "So for example I replica VMs 1-5 to server 2 from 1 and replica VMs 6-10 from server 1 to 2. If server 1 or 2 goes down for a hardware fault I can switch on the downed VMs on the working server."

    Just remember to NOT replicate your DCs.  Just leave them running on both hosts.  DCs have built-in replication that is aware of the active directory application.  This is far superior to what Hyper-V Replica can offer.


    . : | : . : | : . tim


    Thursday, March 02, 2017 1:59 PM
  • Thanks for the heads up Tim. I'm just reading an article about this as well as I was curious about another article I read pushing the fact you need a dc on each host and not to set replica but it didn't go into detail as to why
    Thursday, March 02, 2017 2:08 PM
  • In general, if an application (e.g. SQL Server, Exchange, Active Directory) has built-in replication capabilities, it is far superior to use those capabilities than to rely on a generic replication methodology.  The application knows when and how to replicate to ensure application continuity.  A generic replication technology does not have that application awareness.  Can generic replication technology work? Yes. Can it sometimes cause issues with the application? Yes.


    . : | : . : | : . tim

    Thursday, March 02, 2017 2:54 PM
  • Thanks for all the help I have a good clear plan now and a clear understanding of both replica and clustering.
    Thursday, March 02, 2017 3:59 PM