So the last couple of days I have tried to wrap my head around Direct NFS support which is coming in Veeam v9. The cool thing about this feature is that Veeam has a custom built NFS agent, which will go directly to the NFS share (only needs READ access) and export the snapshot data when doing a backup.
Now important that Veeam is configured against a vCenter server ( I tried many times against an ESX directly and then NFS Direct didn’t really work.
When setting up a Direct NFS backup solution, we need to first setup a Veeam Backup Proxy as we would in other scenarioes. We need to include the Veeam Backup Proxy in the virtual vSwitch that Nutanix provisions within ESX (Note: Do not change the vSwitch, just add the VM to the vSwitch network)
Then define an IP address to the Veeam Backup Proxy within the vSwitch so it can communicate with the Controller VM.
Note that since the vSwitch is an internal only switch, we should setup a Backup proxy per node to maximize the performance. Even thou in this scenario it will work to do NFS direct on this node against other node as well, but then we will be pushing the traffic across the Controller VM network. So when setting up backup jobs try to make it so it uses the local proxy on the host which the virtual machines recides on, this will give the best troughput.
We also need to whitelist the IP address of the proxy so that it can allow access ot the NFS share (Which in the case of Nutanix will be the Storage Container which virtual machines resides on) This can be done on a container level or at a cluster level.
Next we need to “force” Veeam to use the storage network on the proxies to do backup traffic. Which can be done in the central management pane within Veeam.
Lastly we need to rescan the storage attached to the infrastructure which will allow Veeam to see the new NFS datastores and see that they can access it using NFS direct. This can be done here.
We can see from the statistics of this job that it is using NFS in the first screenshot
We can also see in the backup job log file for the VM
and that we are using regular hotadd in the second one.