vDS and LACP for NFS Storage

[fort_sphere]

Hello,

I am looking for some help/advice before I implement LACP on a vDS for NFS Storage for the first time. Our virtual infrastructure was recently upgraded from vSphere 5.5 to 6.0 and the licensing has been upgraded to Enterprise Plus to enable the use of vDS. The existing environment is using vSS for all of the server connectivity (VMs/vMotion/Mgmt/NFS), but the datacenter network team wants to enable LACP on the storage switches/ports and I need to configure the vCenter/vSphere hosts for it.

I have always used multiple uplinks for storage in an active/active configuration on vSS and vDS, but the implementation of LACP on the switches and at the vCenter vDS level is new to me. I have been reviewing documentation and videos on how to implement it, but I find it a bit confusing. From my understanding I will need to complete the following:

1. Create a new vDS in vCenter for NFS Storage (enable jumbo frames)

2. Create new virtual port group for NFS Storage (set VLAN)

3. Create a LAG for NFS Storage under the LACP properties of the new NFS Storage vDS

• 2 x # of ports
• Active mode (so it negotiates with physical switches)
• I am unsure what Load balancing mode makes sense? I believe it needs to match the switch?

4. Migrate VMs away from PoC host before connecting to vDS just in case

5. Migrate network traffic to LAG Wizard - Set LAG as standby uplink

• Manage distributed port group
• Select Teaming and failover
• Select NFS Storage port group
• Set LAG to standby uplinks (leave LB, failure detection, notify switches and failback at default?)
• Accept warning
• Apply

6. Migrate network traffic to LAG Wizard - Reassign physical network adapters of the hosts to the LAG ports

• Add PoC host to NFS Storage vDS
• Manage physical adapters and vmkernel adapters
• Add both NFS vmnics to LAG ports
• Migrate NFS vmkernel adapters to from standard port group to new virtual port group for NFS
• Apply

7. Migrate network traffic to LAG Wizard - Set LAG as only active uplink

• Manage distributed port group
• Select Teaming and failover
• Select NFS Storage port group
• Set LAG to active uplinks (leave LB, failure detection, notify switches and failback at default?) and move everything else to unused
• Apply

At this point I believe that is all the work I need to do to configure LACP for the vDS and vPG for NFS at the vCenter level, correct? At what point in this process do I get the datacenter network team to create the LACP group on the physical switch ports?

Is there anything here I am missing in the procedure? My plan is to leave this for a couple of days before migrating the other hosts in the cluster to the NFS Storage vDS/LAG. When I do that I should just follow the standard vDS migration wizard to add the host and migrate the physical and vmkernel adapters?

Any help would be greatly appreciated.

Thanks and take care!

[daphnissov]

but the datacenter network team wants to enable LACP on the storage switches/ports and I need to configure the vCenter/vSphere hosts for it.

Ok, so let's start here. Why *exactly* do they want (or think they want) to enabled LACP on the vSphere environment? This comes up multiple times a week and each time is usually the same reason. Now, when you say you want to do it with your storage traffic, it becomes doubly important. So I ask what is the reason to want LACP in the first place?

[fort_sphere]

Thanks for the reply.

The reason being as this is a shared customer environment and they want to follow their standards and this one environment is not compliant and is causing a problem for them. And the reason they want to do this on the storage network is that the standard network and storage networks are on different switches and it is only the storage network that is not configured for LACP.

[daphnissov]

The reason I ask is because, especially on NFS networks, the LACP is almost always useless. It doesn't, as one example, give you combined throughput on two disparate links to a single NFS extent when using version 3. In exchange, you greatly complicate you vSphere for unnecessary purposes. When you already have a license that entitles you to use vDS, it actually has more capabilities than LACP yet most people don't know this (or the networking teams who are mandating LACP don't know this).

[fort_sphere]

I agree! This is over complicating an environment that is already working well, but these standards have been put into place and they are forcing this change to happen whether I like it or not.

That being said, from your experience, does the procedure above look correct or am I missing something in this procedure?

Thanks again!