VMware View configuration in disaster recovery scenario

[VDIWannabe007]

Hi I am trying to design a DR solution for my virtual desktops with the following infrastructure:

3 x ESX hosts at datacenter 1 (live site)

3 x ESX hosts at datacenter 2 (DR site)

SNapmirror of VMFS volumes (Netapp) between datacenter 1 and datacenter 2.

I am using VMware view 3 as the virtual desktop infrastructure and each user is mapped to their own specific desktop in VMware View connection manager ( 1 to 1 mapping, no connection pools used).

In the event that datacenter 1 fails we would then switch over to datacenter two where the VMFS volumes with the virtual desktop configuration and system files have been replicated to. The virtual desktops would be regsitered to virtual center at Datacenter 2 through a script and would be brought online.

The issue I face is that the VMware View 3 connection manager would still think that the virtual desktops are located in datacenter 1 as the VMware View connection manager has 1 to 1 mappings for the virtual desktops and the datacenter that the desktops are located at are also defined. The datacenter settings in the connection manager would still be pointing to datacenter 1. How would I change this in a DR situation?

Any help would be gratly appreciated.

[mjsvirt]

I would hope at some point that SRM supports View. This would help the challenge that you have.

[mreferre]

SRM would obviously be the preferred method.

However, IF the NetApp can provide transparent IP failover between the two replicated storages and IF a stretched cluster configuration can be implemented (dependending on the distance between the two sites).... this might be an alternative. In case a site fail you would loose half of your systems but VMware HA should be able to restart the VMs on the servers in the surviving datacenter (assuming the storage has failed over transparently w/o requiring manual intervention).

Massimo.

[KFM]

Hmm...tricky one this one...

Without the advent of some application that will manipulate the View ADAM database and configuration to update the whereabouts of the replicated desktop VMs you may have to manually re-populate the View Connection server with the Individual desktop assignments.

Whilst SRM can automatically re-inventory datastores, re-map networks etc, it can't (at this point in time and in my humble knowledge) re-write the View ADAM database/configuration file purely because now we're delving into the application layer in the VM, i.e we're no longer working at the VMware resource layer - if that makes sense? Having said that, it would be cool if a future release of SRM could do this

In light of the above, I see a possible pseudo-solution (pseudo because it's not fully automated nor tested by me, but I'm just throwing up suggestions).

1. In addition to the Standard server in the live site, create a View Replica server in the DR site.

2. Add the DR VC server into View - you now have two VCs registered in View. This will now allow you to see the desktop VMs that you have re-inventoried when you fail over to DR.

3. During DR, manually add the desktop VMs into View (you will be able to choose which VC server your desktops are registered to - choose the DR one obviously).

4. Disable the desktop (or un-entitle) in the (now offline) live site and enable/entitle the DR desktops.

5. Do DR stuff, get live site online etc etc

6. When the live site comes back online, disable the DR desktop (or un-entitle) and enable/entitle the live site desktops and with luck the change should get replicated to the live site Connection server.

If you can find a script that will automate the creation of pools and Individual Desktops that would ease things a lot....esp if you have a lot of desktop VMs to add back into View!

KFM

[andreleibovici]

Hi KFM,

Have you been able to acomplish or achieve the desired results? I have a similar scenario to solve; 2 sites using NetApp NFS and SnapMirror replication. View4 connection server in the primary site and replica at DR site. vCenter 4U1 clustered with MS Cluster.

SRM is an option, however between sites I've got fiber and can run workload from diferent sites & storages.

Cheers,

Andre