2

I am posting here, because reading a lot of documents on MS site and technical forums have caused only more confusion.

The equipment i have: 2 IBM X3650 M3 servers; both have 2 72GB SAS disks configured as RAID1, 4 10/100/1000 ethernet ports and 72GB RAM. Operating system is Windows 2008 R2 Enterprise.

1 DS 3400 SAN with 9 400GB SAS drives configured as RAID 6. Connection to the both servers is done via fibre adapters. At the moment it has 1 array which is configured as 4 logical disks - 2 for each server.

1 48 port gigabit switch.

At the moment both servers have full Windows 2008 R2 OS running; they are also both AD controllers and run Hyper-V service. Both servers host 3 virtual machines at the moment. VHD files are located on the logical disks in the SAN.

Is it possible:

  1. Configure those servers so that when one X3650 server fails (i.e is physically destroyed) then the other X3650 will run those 3 VM-s automatically so that the users and administrators won't notice anything (only errors on the event log), they don't have to urgently do anything (i.e configure something on the server that is running) and of course virtual machines will run on another server like nothing has happened.

  2. If that is possible, do i have to include double disk space for VHD files? I.e if the total amount of disk space for 6 VHD files is 1,5TB then i have purchase additional SAS disks to expand the array.

  3. At the moment i use DS Storage Manager 10 to assign which server has which logical disk on SAN; how is Windows able to grab a logical disk hosted in the SAN that belongs to another server to itself automatically then?

Best regards, Lauri

1 Answer 1

1

Hyper-V failover clustering works using a shared volume, not replication.

You would need to create a "host group" on your DS3400 and assign access to your logical disk in question to the host group instead of a single host - this will allow clustered configurations with a shared disk.

On the Hyper-V side, you would need to enable Cluster Shared Volumes and Failover Clustering and set up your VMs accordingly. Check the hardware requirements named in the linked documentation site - especially in regard to the support status.

All that said, "failover clustering" would not allow for seamlessly continuing a running VM in the event of a hardware failure. Your VMs just will be restarted on the other cluster node - inducing downtime for the time of guest OS startup and of course the loss of non-permanent state. But you will be able to do planned "live migrations" where no downtime is noticeable for maintenance purposes.

3
  • So am i correct that using only the software that is included in MS Windows 2008R2 Enterprise, DS3400 and other software which is available for free - it is not possible to achive the following setup: 1. Everything works, each server hosts 3 VM-s using half the resources they have (memory, cpu etc.) 2. Something bad happens on the first server, then second will operate all the six VMs almost automagically i.e the administrator can start the repairing process of the first server. 3. Second server fails, and again the first one will grab those VMs to itself. Oct 11, 2012 at 10:11
  • @LauriAndres This is not what Hyper-V failover clustering is going to do for you currently. This kind of feature has been announced for Hyper-V (actually already in V1), but is not yet available - allegedly due to stability concerns. VMWare provides this kind of functionality with vSphere calling it "Fault Tolerance", but it is not available with the free editions of vSphere. What you can do with what you have is set up clustering on your guests - but this would require application/service support.
    – the-wabbit
    Oct 11, 2012 at 18:16
  • @LauriAndres A lot of Microsoft services do run in a clustered configuration - these are the Windows-built-in services which are HA-capable, a number of Microsoft server software would run in a clustered configuration as well or come with own HA mechanisms (e.g. Exchange, SQL Server). The benefit of running clustering in the guest is the ability of having an active/active cluster. With hypervisor-based failover clustering, you have a passive node which consumes resources but is not able to answer requests unless activated.
    – the-wabbit
    Oct 11, 2012 at 18:28

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .