NFS 4.1 multipathing

[scidoner]

Hi guys!

Is there any chance to get Veeam to support NFS 4.1 multipathing when using Direct NFS transport mode?

My ESXi hosts are connected to a NFS 4.1 share across 2x 1GbE links, which is working correctly.

If Veeam would connect to the NFS 4.1 share via the two links as well, I feel like it would seriously improve performance.

Cheers.

[Andreas Neufert]

I can confirm that we do not use NFS 4.1 mutlipathing at DirectNFS or Backup from Storage Snapshot (NFS).

We will treat this as a feature request. Based on policy of Veeam we can not comment about the future implementations and timelines.

[scidoner]

Thanks for the reply + confirmation that it is currently not supported (saves me going mad trying to configure it!).

No problem at all, I hope to see it implemented some day so thanks for acknowledging it as a feature request!

[jameskilbynet]

Yep I have just come across this issue in an environment ( and confirmed in my lab)

It would be useful if this could be implemented

[Kazz]

Has this ever been put in the works?

[Andreas Neufert]

It will not be in v11. We can not comment about the versions or updates after it.

[orb]

Multi pathing and NFS has always a funny subject. VMware support only session trunking mode Synology, VNX, Unity, nfsd on linux support this mode while NetApp support only pNFS and Oracle has his own client dNFS but work with NetApp. NFSv3 seems to have better performances than v4.1. It is supposed to be fixed in 4.2 but VMware doesn't support it yet while NetApp does (still fresh though)

Oli

[soncscy]

Maybe someone can explain to me, but is NFS just a really fragmented protocol?

Oli's comment is...well, frustrating, not because of anything Oli specifically said, but because all these vendors apparently just do their own implementation. I get this is part of the 'charm' of an open protocol, but at the same time, it makes me nervous to invest time in it because I'm not sure if a given hardware vendor will have the "right" implementation.

Is there a good resource to get atop modern NFS implementations? Or is it just something you need to chase down per vendor?

[orb]

@soncscy

I understand your frustration but NFS is like CIFS, it has a multi-client usage. NFSv4.1 exists since 2010, 4.2 came around 2017, but still today most NFS implementation with VMware uses v3. The advantage of NFS against blocks back in the days is you didn't need any block storage knowledge to have it work, it was more flexible and most the guys had *nix background. This document is good to read https://www.netapp.com/media/10720-tr-4067.pdf but as you notice they focus on client usage, not really VMware. VMware documentation is still a good read to reminds you what they can do or not but also VMware functionalities you lose by choosing 4.1 https://docs.vmware.com/en/VMware-vSphe ... C328F.html

You can emulate block like functionalities like multipathing or multi-link with 2 IPs with NFSv3 but you need to do all the work at the network/switch/vswitch level. You could do that when you had only 1 GbE links but the technical complexity it adds isn't worth the hassle when you have 10GbE or higher links. https://wahlnetwork.com/tag/nfs/, most of the articles are from 2012 but still relevant today when it comes with NFS/VMware.

Oli

[orb]

Multi pathing and NFS has always a funny subject. VMware support only session trunking mode Synology, VNX, Unity, nfsd on Linux support this mode while NetApp support only pNFS and Oracle has his own client dNFS but work with NetApp. NFSv3 seems to have better performances than v4.1. It is supposed to be fixed in 4.2 but VMware doesn't support it yet while NetApp does (still fresh though)

Oli

Correction : NetApp still doesn't support NFS 4.2

[Kazz]

I stumbled across an old post here -> vmware-vsphere-f24/nfs-4-1-multipathing-t51493.html
at the time multipathing was not supported, but the last comment suggested it would be available in v11. Did multipathing ever make it in v11 and is it available in v12?

Looking to increase the transfer speeds to a synology nas and looks like multipathing is my only option.

My VBR proxy is a VM with 2 NICs, first nic is data vlan and the second nic is an unrouted nfs vlan.
If I multipathing is possible, would I need 2 vswitches for nfs and 1 nic from each switch added to VBR proxy or multiple vnics from the same vswitch ?

Thanks

[foggy]

Hi Kazz, I can see no comments here stating that it would be in v11. It is still an open feature request that wasn't prioritized enough due to amount of requests so far.

[orb]

Kaz,

I've noticed that your goal is to increase your transfer speed. NFS multipath won't be of much help in this regard. Instead, you should consider transitioning to iSCSI. You can create 2 (minimum of 12 disks) to 4 LUNs (24 disks or more) and stripe them with a block size of 128k/256k on a single logical volume. This approach will allow you to maximize your performance, especially when using SATA drives. The actual results will depend on your specific storage configuration. Alternatively, you can also do a SOBR, assign each repo to a NIC

As a side note:
I've heard that NFS trunking has been removed in Synology DSM 7. To enable it, you may need to use the shell.
There are also rumors that NetApp is working on NFS trunking.

[Kazz]

@orb

Thank for your comment and a suggestion. Is NFS multipathing meant to help with a link failure and not with the speed? We only have 1 available port left on the Synology and can't switch NFS dedicated nic to be a second nic for iscsi yet as we have other devices accessing Synology via NFS.