vRDM SQL Server and Veeam

[ChrisGundry]

Hi all,

After some info please...

We are deploying a new SQL server cluster. We use Nimble storage and Nimble recommend that high performance SQL servers that we use vRDM disks. So we have configured the servers with vRDM and it is indeed flying along very nicely.

So the setup is:
C: is on a VMFS6 volume (no need for it to be vRDM)
D:/ (DATA), L:/ (LOGS, T: (TEMP) are all on vRDM disks.

The vRDM disks are mapped through VMware to the guest. RDM are obviously in virtual mode and set to dependent as we wanted them to be captured by VM snapshots. vRDM pointers are stored in the VMFS6 volume that the C: drive uses so it is all self contained.

We have Veeam configured for integration with Nimble and also have Nimble configured to trigger it's own snapshots at the 'volume collection' level every 15 minutes.

Now I have two questions:
1. We have Nimble snapshots configured to snapshot all the disks of the server, every 15 minutes, which is working fine and I can see all the snapshots on the Nimble side. However, if I locate the server within the 'storage snapshots' view within the backup infrastructure and attempt to do a restore of it, it only sees the VM's C: drive... I can't see the storage snapshots of the other additional volumes without the server OS/VMX file on them. Is this correct? Is it not possible to have it see the other D:,L: and T: disks? At the moment I can't use Veeam to restore the whole VM from storage snapshots, even though there are snapshots for all of the required storage volumes...
My objective is to be able to restore the whole of the server, with all the disks, from storage snapshot via Veeam. Ideally want to be able to use IR as well as full VM restore...

2. We are planning to only do daily Veeam 'backups' of the VM as it is a replicated SQL cluster and we have the storage snapshots in place as well. I expected no issues here as I thought vRDM was fully supported? When attempting a backup I get:
24/02/2020 09:39:51 :: Disk xxxxx_2.vmdk has been skipped due to an unsupported type (raw device mapping disk in virtual compatibility mode)
24/02/2020 09:39:51 :: Disk xxxxx_3.vmdk has been skipped due to an unsupported type (raw device mapping disk in virtual compatibility mode)
24/02/2020 09:39:51 :: Disk xxxxx_4.vmdk has been skipped due to an unsupported type (raw device mapping disk in virtual compatibility mode)

Any help is appreciated.

Thanks

[ChrisGundry]

Sorry, forgot to add something else.

Is it that Veeam just can't do vRDM backups from storage snapshots? If I turn off the snapshot integration for the job then it seems to work OK...

Obviously the other option we can use is to do this as a Veeam Agent backup job, but we were hoping to avoid that if we could.

Thanks

[foggy]

Hi Chris, this is mentioned in the user guide as a limitation - vRDM disks are not supported and are skipped from processing. Thanks!

[ChrisGundry]

Hi Foggy

Thanks for the reply. I had worked that out in the time between posting and your reply. It is unfortunate that is the case

IR is something that we use a lot as it is really powerful. But we are unable to do that with vRDM hosted servers because of this issue. Given that the snapshots exist and have the data on them, it should be possible to bring each of the snapshots on-line and re-connect them to a clone of the VM no? Is this something Veeam can look into adding to the product? Or is it considered not viable because vRDM usage is low?

Thanks

[foggy]

Yes, it doesn't worth the effort. Consider switching to VMDK and gain all the integration benefits.

[ChrisGundry]

Hi Foggy

Nimble recommend vRDM of in guest iSCSI for best performance when using SQL server, you can assign performance policies to get the best from the VM/storage. Their initial recommendation is in guest iSCSI, but that would require us to install and update the Nimble connection manager within the VM as well as use/maintain the Veeam agent. We wanted to use vRDM so that Veeam could at least back up the VM without us needing to use NCM or the Veeam Agent backup method.

With VMFS all the data is treated the same at Nimbles side and has the same performance policy, plus it is another layer so performance is not as good as vRDM. We did a lot of testing and we are seeing lower and more consistent latency through vRDM than through VMFS. So it is hard for us to throw that performance away to go with VMFS for Veeam integration.

It is a shame that IR can't be used with vRDM VMs when the storage snapshots and all the data is there to be used. I guess I can still bring the snapshots online manually, restore the VMFS C: drive/VMX and manually re-add the vRDM disks to the VM. But that is obviously all manual and time consuming vs IR which automates it for you.