I am configuring NFS in a UCS environment where there are 2 x 10Gb physical uplinks configured in a LAG.
Should I configure my NFS virtual port group on my DVS as active / active or active / standby?
Same as per my other reply, always use active/active where possible. This will give you the maximum available bandwidth.
There are some gotchas with NFS, though, which you may want to be aware of. NFS doesn't support native multipathing. So if you want to make use of both 10Gb uplinks, you need to make sure you have more than one source/destination pair of IPs between the ESXi host and the storage array. You can achieve that by splitting NFS volumes between two storage controllers. Or if all volumes are on one controller you will need to assign a second IP address to the controller.
Same as per my other reply, always use active/active where possible. This will give you the maximum available bandwidth.
There are some gotchas with NFS, though, which you may want to be aware of. NFS doesn't support native multipathing. So if you want to make use of both 10Gb uplinks, you need to make sure you have more than one source/destination pair of IPs between the ESXi host and the storage array. You can achieve that by splitting NFS volumes between two storage controllers. Or if all volumes are on one controller you will need to assign a second IP address to the controller.
In vSphere 6.0 they have support for NFS 3 and NFS 4.1. NFS 4.1 does add support for session trunking multi-pathing in vSphere, but at the trade off of not supporting other features like SIOC, Storage DRS, SRM, etc. There is a good comparison chart here that lays out what is supported with each protocol.
I'm a block storage guy so I haven't set it up yet.
Totally agree to go active / active if possible, especially if you have VDS and NIOC. With the standard switch a lot of the times you get stuck having to use active / standby to separate traffic since you don't have NIOC. When we had a bunch of gig interfaces we could do physical separation, but with fewer 10G we end up having to manipulate active / standby.
ok great info - thanks again