Backup via NFS vs from storage snapshot

[smithaetg]

I've got a NetApp cluster providing storage for vSphere 7.0.3 environment via a NFS v3 datastore. Normally the VM backup job works fine backing up from the storage snapshot. However, one day I changed the password to the NetApp cluster, without changing it in VBR. The next day I saw that all of the VMs backed up using NFS. I also noticed that some of the hard disks backed up faster via NFS than they did using the primary storage snapshot. So after putting the correct password in VBR and fixing the initial error, that got me thinking about how to manually change the backup job to use NFS. I want to see if that speed bump was a one time thing, or if it would be consistently faster. I've read through and studied all of the documentation I can find on integrating NetApp cluster datastores with VBR and I can find nothing on "forcing" the job to use NFS rather than the storage snapshots.

So my questions for the forum:
Besides "forgetting" the password to the NetApp cluster, is there a way to force the job or proxy to use NFS?
What has the forum members experiences been when backing up via NFS vs from a storage snapshot? Right now backing up from a storage snapshot doesn't put any extra stress on our production environment if backing up a large VM sneaks into production hours. Does backing up a large VM during production hours via NFS cause any production lag? Does anyone else have a similar setup and can offer advice?

[rennerstefan]

Hi

not sure I understand correctly but I guess you mean if there is any difference between backing up via storage snapshot integration vs using direct NFS backup, right?
Because even der backup from storage snapshot will use NFS in your case as NFS is used to preset the volume to the NetApp.
The only way to enforce direct NFS over backup from storage snapshot using NFS is by disabling storage snapshot processing in your job settings as the job will then use the regular transport modes.
In any case it will be NFS in both scenarios regardless of you use BfSS or direct NFS.
The difference could only be that your system has somehow a policy in place that limits the processing rate in NetApp when reading from a snapshot vs the original volume.

Let me know if that helps and what you find out.

Thanks

[smithaetg]

Hi Stefan!
Thank you for taking time to reply.

if there is any difference between backing up via storage snapshot integration vs using direct NFS backup, right?

Yes.

Because even der backup from storage snapshot will use NFS in your case as NFS is used to preset the volume to the NetApp.

That's what I'd assume too.

When backing up from a storage snapshot, I get log events of:
Using backup proxy VMware Backup Proxy for retrieving Hard Disk 1 data from storage snapshot on [name of NetApp cluster].

When backing up via NFS, I get log events of:
Failed to prepare VM for processing from storage snapshot, failing over to using VM snapshot.
Using backup proxy VMware Backup Proxy for disk Hard disk 1 [nfs].

I've also had this job backup VMs via nbd. When this occurs, I get log events of:
Using backup proxy VMware Backup Proxy for disk Hard disk 1 [nbd]

So, when I stop backing up from a storage snapshot, how do I make sure the backup job backs up via nfs rather than nbd? I'm guessing the ndb backup is using NFS, yet it is so much slower because its going through the esxi hosts to perform the backup rather than communicating directly with the datastore. Correct?

The only way to enforce direct NFS over backup from storage snapshot using NFS is by disabling storage snapshot processing in your job settings as the job will then use the regular transport modes.

How?

The difference could only be that your system has somehow a policy in place that limits the processing rate in NetApp when reading from a snapshot vs the original volume.

Interesting theory. I know there is no policy like that when I log into the NetApp's system manager. However, I don't know enough about NetApp's management OS to state for certain that this could NOT be possible. (So, in other words, I guess it could be possible.)

[ronnmartin61]

To disable backup from storage snapshot in a job open it for editing and in the Storage dialog select "Advanced" --> "Integration" and uncheck "Enable backup from storage snapshots"

[smithaetg]

Hi Ronmartin61,
Thank you for that info. Is there a way to be sure that the backup job will use NFS rather than NBD after disabling backup from storage snapshots?

[ronnmartin61]

That's a function of the transport method selection process employed by proxy. From the transport mode section in the proxy properties you can set "Direct Storage Access" and deselect "Failover to network mode..." to force direct access.

[smithaetg]

Thank you Ron. I will do that for tonight's backup. I'll let you know how it works tomorrow.

[smithaetg]

Ron and Stefan,
Thank you very much for your input to my request. After disabling the backup directly from storage snapshots, the backup job used NFS. I let the backup job backup via NFS for a few days. In comparing the VBR job history for this job I can see no real difference in speed between the two types of backup processes. So I've changed the backup job back to backing up via storage snapshots. Thank you again for helping me figure out how to accomplish this test.

[ronnmartin61]

You're very welcome!

[DaStivi]

Hi there,
Just double check what network interfaces are uses while storage snapshot is transfered... Not sure if this is running through mgmt interfaces that might be connected with 1g only... At least on some older machines without only 10g interfaces...

Also make sure you're not ending up having hotadd on some VMs because this scenario with NFS v3 and hotend propaply freeze VMs more sooner then later... As this is some locking issue in VMware since the beginning... Should be fixed with NFS v4 but this on the other hand is more complex to deploy...

[smithaetg]

Hi Stephan,
Thank you for that advice about making sure what network interface cards are being used. I've checked the VBR VM, the ESXi hosts, the NetApp cluster, and the switch they are all connected through. As far as I can tell this process should be using the 10 Gbps network end to end. The only NICs that aren't 10 Gbps are the NetApp clusters' management nics. However these are also going through that 10 Gbps switch at auto-speed.

I don't understand your second paragraph. All of my VMs reside on this NetApp datastore. Each of my VMs in this job state "Preparing for hot backup" as the third action in their backup job history. It takes the job about 3 and a-half minutes to prepare all of the VMs for storage snapshot.