Two ESXi hosts and LACP

[Tcpuniverse]

Hi;

I'm studying vmware vSphere and want to establish LACP between two ESXi hosts and a single physical switch. I created a dvSwitch on vCenter and added both ESXi hosts to that dvSwitch. I also configured LAG on the dvSwitch and added one port on each ESXi host to the LAG. On the other hand, I configured LACP on switch and put it to active mode. There are no any other commands on the switch side with regards to the spanning-tree or vlans.

I started with one port on each ESXi host (only vmnic2 on both hosts) , so the LAG will be consists of 2 physical ports totally. The problem is that one of the ports inside the LAG is active and the other one is suspended (for example, vmnic2 on esxi 1 is active, but vmnic2 on esxi 2 is in suspended state). When I disable the active port and enable the suspended one, this time the other one gets suspended. I read some chapters on the books, but got confusing comments.

[dbalcaraz]

Hi,

Well, you don't need to set up LACP if you want to achieve failover or concurrent links capabilities. It seems that you have a vDS so you could use Load Based Teaming (LBT) instead of setting a LAG.

Anyway, if you want to continue, I would check this article which is quite nice: LACP Configuration in vSphere 6 – Virtual Reality

[Tcpuniverse]

Actually I also referred to that article beside other ones and followed the steps outlined there which are common. But at the end, I have a LAG with only one port active inside it. The Cisco switch displays the 2nd port as "suspended". I turned on debugs to see any possible errors, but got nothing regarding the reason behind this suspension.

[daphnissov]

Because you're trying to do LACP incorrectly. You cannot form a LAG between different hosts' uplinks. An LACP must be created for each host, and the given vmnics for each host connected to those ports. As the previous user mentioned, however, you really do not need LACP in vSphere if you're licensed for vDS. There are almost no valuable use cases for this.

[Tcpuniverse]

At the first, I designed it the way you said, which is a separate LAG per ESXi host. But, while I tried to edit the dvPort Group to use the LAG as uplink, it allowed me to only select "one" LAG as uplink.

For example, I have devices in vlan 100 on both ESXi hosts in the topology. Based on what you said, then I should create LAG-1 on ESXi-host-1 and LAG-2 on ESXi-host-2. But at this point, while editing dvPort group for vlan 100 on dvSwitch, which uplink LAG I should choose? (because vCenter allows me to choose only one LAG as uplink). If I add LAG-1 as uplink for dvPort-Group-vlan100 (which was created on ESXi-host-1), how devices on ESXi-host-2 can communicate with the outside?!!

[Tcpuniverse]

It's disappointing that I got nothing in reply for such a simple/common question in a community as big as this!

[daphnissov]

Again, LACP isn't something you probably need to be messing with, and when you don't get many responses it's because people also don't mess with it. There are better, simpler ways to achieve everything LACP can accomplish as I've stated.