Hi all,
Most of the documentation/blog posts/etc I have found discuss pre-6.0 NFSv3 implementations and design. Would someone mind commenting on my design proposal to see if it makes sense and follows best practices?
This is a (3) node deployment. Each host has (4) 1Gbps interface. My intention is to create two virtual distributed switches:
* dvSwitch1 (2x1Gbps) - management, FT, vMotion, NFS v4.1 for datastores (each on dedicated VLANs)
* dvSwitch2 (2x1Gbps) - VM traffic
As I understand it, NFSv4.1 introduces 'session trunking' support which will allow me to specify multiple IP's for my NFS server(s). This enables multi-pathing and won't require me to use multiple subnets/VLANs to distribute NFS traffic across the physical interfaces that was previously required with NFSv3.
My questions:
* Should I be concerned with how I handle IP addressing on my NFS server? If I simply create multiple virtual interfaces on the NFS server, is that enough to take advantage of session trunking? Do I need to worry about using IP's with unique least significant bits [1]?
* I'm still a bit confused on how to handle physical link aggregation. Would it be best if I used LACP on each pair of physical links (one LACP for dvSwitch1, one for dvSwitch2)? I would prefer to use LACP over LBT due to ease of switch configuration. However, would using LACP over LBT affect the ability of session trunking to distribute load across both physical interfaces in the LACP bundle?
Feel free to point me to any updated white paper that covers NFSv4.1 and vSphere 6.5 if you know of one!
[1] NFS on vSphere Part 1 - A Few Misconceptions - Wahl Network
Thank you!
