I have a (2) Node Windows 2008 R2 Hyper-V cluster.
I have several highly available VM's, one of which is a file server with a 1TB data volume.
Currently, this is configured with the following attributes:
- The file server DATA VOLUME is an NTFS volume on an Equallogic SAN
- The volume is configured as a volume in Microsoft Cluster Services
- The volume is configured as a PASS-THRU volume and attached to the file server VM
In Cluster Services, this allows the ownership of the Data Volume object to be independent of the ownership of the VM object, and when live migrating the VM, the volume object does not necessarily follow the VM and has to be migrated separately. While this does complicate the resource configuration, being able to snapshot the large filesystem as a filesystem and not as a VHD (at the SAN level) does have benefits.
This is causing some issues lately, and I need to connect the resources a bit more cohesively. What would be the best path?
1. Convert the filesystem to a VHD and attach the VHD to the File Server
2. Leave the filesystem as it is as a volume on the SAN, and add (2) new NIC's to the physical host machines on the storage segment so that the FileServer can connect to the volume via iSCSI from WITHIN the VM
2a. This means that the VM would negotiate connection to the iSCSI volume from within the VM, and would re-negotiate connections to the iSCSI volume after the VM was migrated to another VM-Host
Does anyone have an opinion on whether I should do option 2 at all? I know that this is the way that Microsoft indicates that SQL / Exchange should be installed in a VM environment, (although they recommend against the VM's being highly available in those installs). Does anyone see any issue with allowing a Hyper-V VM to make direct iSCSI calls to a volume with additional NIC's in the storage segment?