RDM Mapping Files when using NFS and iSCSI data stores

[LFC]

My customer uses NetApp storage and VM's are predominantly stored on NFS data stores.

The customer uses a large number of MS SQL Servers which have SnapDrive and SMSQL on the SQL VM's. The different DB and Logs drives are mounted as iSCSI (physical mode) RDM's into the guest.

When I try to protect the VM's I get errors relating to the RDM drives being 'not-replicated' (all RDM disks are properly SnapMirrored).

The RDM pointer files are stored on a dedicated iSCSI data store, as we cannot store them on NFS data stores. Currently there are 20 VM's with RDM pointer files on the iSCSI data store. 

If I SnapMirror the RDM Mapping LUN, the RDM disks can then be protected in the VM, however the RDM data store is added to the Protection Group along with the shared NFS data store and the actual RDM LUNs.

I could see how this could work in a complete DR event, however if I wanted to failover a subset of VM's the RDM data store (which contains RDM pointers for multiple VM's) would need to be failed over, and this would not be possible as VM's which are not being failed over would still be using this DS.

As I see it, I need either

1. Dedicated RDM iSCSI data store per VM (or group of VMs sharing a Protection Group)

2. Move the OS disks to an iSCSI data store and store the RDM's with the VM files

3. Do not use RDM's anymore and transition my RDM's to VMDK's and integrate SnapDrive / SMSQL with vCenter

This must be a fairly common occurrence as this implementation is pretty standard (vSphere 5.5 / SRM 5.5 on NetApp ONTAP 8.2.1)

Has anyone got experience of either of the above 3 options and which works best?

Thanks

[memaad]

Hi,

All this 3 scenario work fine. To keep it simple for management, I would prefer to store RDM mapping files with VM files.

Regards

Mohammed Emaad

[shivaprasada]

All the 3 options will work well.

[LFC]

In the end I went with Option 2. I created an iSCSI data store and re-mapped the RDM pointer files from a small number of VM's to be stored with the VM itself. Further testing with SRM worked well and we plan to implement this solution across the board.

Thanks

Sean

[bayupw]

Don't forget to double check the NetApp technical report for deploying SRM

Deploying VMware vCenter Site Recovery Manager 5 with NetApp FAS/V-Series S

To be able to protect a VM with SRM and the NetApp SRA, all parts of the VM must exist on only one NetApp controller or MultiStore vFiler unit, in both the protected and recovery site.

Relationships where the data (vmdk or RDM) owned by any individual virtual machine exists on only one array (physical controller or vFiler unit) at each site are supported.

So make sure the vmdk & RDM for individual VM is in the same array.

Relationships in which the data (vmdk or RDM) owned by any individual virtual machine exists on multiple arrays (physical controller or vFiler) are not supported.

Thanks,

Bayu