Dell PowerSwitch LAG LACPIn a datacenter spine-leaf model, 2 leaf switches are in the Top of Rack (ToR). Typically, the switches do not stack because they are not designed to stack. By design, each switch operates independently. This is for high availability. When both switches are working, there are 2 paths from the rack to the spine/core layer. In the event one switch goes down, one path remains. Having the switches stacked would result in a single path to the core. If the stack goes down, all paths to the spine goes down; not the desired result.
Lower in the rack, servers hum along with multiple NICs. This could be a single server or a modular server like a Dell VRTX or an Dell FX. Each host in the chassis will have to independently go to each ToR switch.
While the ToR switches are independent, they still need to communicate. This is done through a vPC connection. The vPC connection is broken into 2 separate links; a peer-link and a keep-alive link. The peer-link sends the data over a high-speed connection (40G) for synchronization. Because the link is vital, 2 ports on each switch (say 1/49 and 1/50) should be used and bonded together through a LAG/Team/PortChannel. LAG's show as Po1, Po2 and so on.
The other link, a keep-alive link, is a simple heartbeat link and can be a slower connection (1G) such as the management port; mgmt0. It just asks/responds to keep-alive queries. If it fails, it automatically takes action.
Just as the switchports have a LAG, the NICs on the server should have a LAG. Windows Server calls this a TEAM and Linux calls this a BOND (Intel calls this Advanced Network Services - ANS). I have other articles on creating a TEAM/BOND, so I won't do it here.
The goal is to have 2 NICs in a Team with one NIC going to the top switch and one NIC going to the bottom switch. Horizontally, the Nexus switches talk to each other to share information over the Peer-Link and check their other's status through the keep-alive link. Altogether, graphically, it looks like this:
The TEAM has different modes; STATIC (hard assignment to Po1) or DYNAMIC/LACP (automatically negotiate and assigned).
For LACP to work, configuration must be done on both side; on the OS and on the Switch.
On the OS, enable NIC Teaming, LACP and Dynamic, like so:
For Dell PowerSwitches, the LAG is set with simple numbers; 1 for Po1. Just find the port that is needed (hint: use the ADDRESS-TABLES tab to lookup the MAC) and assign the LAG number. After the LAG number is assigned, click the LACP box to toggle and show an "L". Click APPLY.
Afterwards, find the next switchport and complete the same steps.
Where some get confused is that they create the LAG with all the ports/members and then they try to change to LACP. This will error with a message, "Cannot add a dynamic member to a LAG with static members." You have to remove all the ports/members from the LAG and add them one at a time.
NOTES:
https://networkdirection.net/articles/virtual-port-channels-vpc/virtualportchannels/