LACP for NSX hosts?

[ESCjkoch]

We have an existing five-host cluster in our primary datacenter using LACP. We're building our secondary datacenter and would like to continue to use LACP.

However, we're adding NSX to both environments, but there is "supported but highly discouraged" language for LACP in one of the older NSX documents I was reading.

Have many people deployed NSX with underlay links being LACP?

[Carlos_E]

Hi,

It depends on your network topology and Cluster design.

In my deployment I ended up with two clusters, one Edge Cluster and one Compute Cluster, all of the existing Distributed Port Groups belong to the Edge Cluster and the Compute Cluster communicates via NSX Bridges.

In this design I couldn´t use LACP on the Edge Cluster but I´m using LACP with 4 ethernet´s on the Compute Cluster :

http://www.vmware.com/content/dam/digitalmarketing/vmware/en/pdf/products/nsx/vmw-nsx-network-virtua...

I did this months ago and it seems like years ago! NSX is wonderful but it´s quite the rabbit hole, I´ve struggling with so many other issues that I barely remember all of the reasons behind those design decisions I made.

I do remember I wanted to use LBT and ended up having to roll that back when I came across that above matrix.

Good luck.

[tspires]

From my understanding, the reason you don't want to go with LACP in NSX is because of issues with dynamic routing over LACP links from your ESGs.

Cisco generally doesn't support dynamic routing when using LACP in a virtual port channel configuration which is a really common way to connect to your physical network from an ESG. If you can guarantee that your uplink with LACP supports dynamic routing than there's not problem using LACP on your edge, but you have to take note that you could have to revert in the future if you ever switch to a vendor who doesn't support dynamic routing via that implementation.

What I tend to do is use LACP on my compute clusters if it's already in place, and source port based balancing on my edge cluster.

[SaurabhRV]

Done that,

Make sure you have Host prepared with NSX it creates DvPortgroup for SWVTEP i.e. vmk1, you have correct nic teaming policy and sequence set to LACP rather than just uplinks.

-Saurabh jain

[ferdi_bulbul]

VMware Reference Design guide recommends not to use LACP on Edge and Compute Clusters as below;

• The recommended teaming mode for the ESXi hosts in edge clusters is “route based on originating port” while avoiding the LACP or static EtherChannel options.
• The recommendation for edge clusters is to select the SRCID/SRC-MAC hash as teaming options for VXLAN traffic. An architect can also extend this recommended approach for the compute cluster to maintain the configuration consistency, automation ease, and operational troubleshooting experience for all hosts’ uplink connectivity.
• AS a summary the “route based on originating port” is the recommended teaming mode for VXLAN traffic for both compute and edge cluster.