BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
When migrating from a standard switch to a VMware vDS, how do we handle Link Aggregation Control Protocol (LACP) if we are using Virtual Switching System (VSS) uplinks to connect to physical switches?
If you can allow an interruption on the VMs, this will simplify the process. Migrating the VMs to another host won't help because the port group will not exist when you try to bring the VMs back with vMotion.
First, let's clarify that the load-balancing algorithm used in this case is "Route based on IP hash." The image below shows where this is configured on a standard virtual switch.
If VMs can be powered down
When downtime is possible, you don't have to change Link Aggregation Groups (LAG) on your physical switch. Let's say that vmnic0, vmnic1 and vmnic2 are active adapters used in your standard switch and are in an LAG. Remove them from the standard switch and add them to the distributed switch; this is when the VMs will no longer be connected to the standard switch on this host. When the three physical network interface cards (NICs) are connected to the distributed switch, the virtual machines can be connected to the port group on the distributed switch, and they will be on the network again.
If VMs need to stay powered on
If the VMs need to keep running, then I would start by disabling the IP hash algorithm on the vSwitch -- there is no need to change the physical switch -- then remove two NICs from the switch and leave at least one physical NIC attached to the switch. This will affect available bandwidth, so you should perform this procedure during off-peak hours. Add the two NICs to the distributed switch -- which is also in the default load balancing, no LACP -- and connect the VMs from the old port group to the new port group. When they are all moved to the distributed switch, remove the last NIC and add it to the distributed switch and configure LACP and the IP Hash load-balancing algorithm on the distributed switch.