Setting up LAGs on multiple ESXi hosts

[MykolaSverdel]

Hi All

I'm not an absolute newbie in vmware, however I have used only standard vSwitches and failover link configuration.

Now I'm trying deploy much more complicated configuration. I have installed 5 ESXi hosts (2 NICs each) and vCenter and configured Distributed Switch. After that i have moved my physical NICs to this one and configured 5 LACP LAG (2 NICs each). So, now I have 5 port groups, each of them includes vmkernel port of one of ESXi host and setted up for using corresponding LAG. My problem is misunderstanding how to setup virtual machines portgroups correctly. Now I can only set one LAG as active (no multiple LAGs supported) for every single portgroup. What will happen if I move VM to different host?

I have searched it via google and VMWare docs, but found nothing apropriate, so I will be grateful for any peace of advices or theoretical info or best practices references about this topic.

[daphnissov]

Is there any particular reason why you want to use a LAG with a vDS here? Are you aware that you do not require a LAG to achieve concurrent link utilization as well as failover capabilities? I see this quite commonly where users are under the (false) impression that a LAG is the only way to achieve this on ESXi because it was a necessity on traditional bare metal. In actuality, if you're licensed to use a vDS (Enterprise Plus or vSAN), then there exists a technology that comes with it called Load-Based Teaming (LBT) which can achieve all of these benefits and is, in fact, more robust than using LACP. So in virtually all instances of users attempting to LAG with a vDS you should be using LBT with a vDS instead.

[MykolaSverdel]

First of all, thank you for your answer.

Yes, I know about Load-Based Teaming, its advantages and how it works. Moreover, I had already set this configuration up and was convinced that it works well. And frankly saying there aren't any particular reasons for using LAG.

However, now my aim is to figure out how to set up my hosts and VMs exactly using LAGs. Unfortunately, I still have not found any proper docs, so two possible ways for me are to conduct an experiment and to ask more experienced people for an advice or link.

particular reason

[daphnissov]

I didn't say there weren't any reasons for using a LAG. I'm asking if you have one in this case or if you just are curious to know how it's done.

[MykolaSverdel]

Actually, I'm just curious, and the lack of documentation spurs my curiosity. So, I don't have urgent need for this knowledge if you meant this.

[daphnissov · Accepted solution]

There are several articles and blogs out there which can give you this idea, for example: LACP Configuration in vSphere 6 – Virtual Reality

[MykolaSverdel]

Many thanks for the link. I'll definetly write if it helped when I read it and test the configuration on physical environment.

[MykolaSverdel]

Link you recommended me was really helpful. My childish mistake was creating 5 separate LACP LAG instead of one that would included physical adapters of all ESXi hosts. The quoting from your link (LACP Configuration in vSphere 6 – Virtual Reality​):

At the Esxi host proxy switch level, only one physical NIC can be connected to one LAG port. However on the vDS, one LAG port can have multiple physical NICs from different hosts connected to it.

And I check it, quoting from VMware Docs (LACP Support on a vSphere Distributed Switch​):

On a host proxy switch, you can connect one physical NIC to only one LAG port. On the distributed switch, one LAG port can have multiple physical NICs from different hosts connected to it.

Well, I was being too stupid for find it by myself. But now I know that skiming of introduction is bad way to understand topic deeply. Now my configuration works perfectly, so thank you so much for your time and patience.