nic teaming with physical switches in a lag.

[gerrydaly5]

A question that I cant necessarily find the answer to. If I have two e/w switches and they are stacked as one with a LAG, am i correct in saying the best NIC teaming setting should be active/active for each distributed port group?

[AjayChanana]

Please check the below link to understand the configuration.

VMware Knowledge Base

[gerrydaly5]

Thanks Ajay,

Sorry besides LACP, (which we dont have configured in vcenter) the physical switches are stacked with a DAC and LAG configured on the physical switches, so my questions is are there predefined/recommended nic teaming settings when certain physical switch configurations are in place?

(apologies if I'm skimming over all the in's/out's of this, vlan trunking/tagging etc) I just cant easily find the answer.

[rphoon]

You should be matching what your physical switch configuration is.

If you have a LAG on the physical, then you should run a LAG on the uplinks as well.

Make sure that the load balancing algorithm (eg. IP-SRC-DST) matches on both the physical switches and the uplink LAG.

Have a look at the VMware KB below for sample configurations and supported algorithms.

VMware Knowledge Base

[daphnissov]

What license are you using for ESXi in your case?

[gerrydaly5]

We have vCloud Suite 6 Enterprise.

[daphnissov]

If you have a vCloud Suite Enterprise license, you have vSphere Enterprise Plus, which means you're entitled to use a vDS. With that being the case, there is really no need to use LAGs of any kind as the vDS contains a far better option called load-based teaming. This is highly recommended over any LAG technology at the physical layer.

[gerrydaly5]

Thanks for all the replies guys, so I had vmware networking team review the configuration and our switch vendor do the same and both found no underlying configurations. (thankfully).

The issue I was facing was with random and intermittent networking issues from the 2 uplinks per host in my cluster.

Long story short I we updated the NIC driver from one that is on the HCL to another lower number but newer driver and initial testing appears this has resolved our issue!.