My answer was too abridged?

The "client" partition (i.e. not VIO Server) (initially) has one virtual ethernet adapter. This adapter has one or more VIO servers that are providing the SEA function. In other words, the client has no conception of etherchannel.

In one or more VIO servers there is/are physical ethernet adapters that get bridged to a virtual adapter with the same PVID as a client (assumption is that all ethernet communication  is VLAN unaware). If there are multiple VIO servers, each of the virtual adapters that may function as "trunking device" (check box can access external network) also needs to set a priority - 1 (one) being the highest priority. Additionally, the vio partitions share a seperate vlan that they use to send control information. The mkvdev command needed will look something like:

$ mkdev -sea ent0 -vadapter ent2 -defaultid 1 -attr ha_mode=auto ctl_chan=ent3

If you want to make an etherchannel, then you should do that on the VIOS with the physical ports. Assume eth0 and eth1 are dual, single port cards, eth2 is the virtual adapter that is to be the trunking device, eth3 is the control channel.

The commands would be as follows:

$ mkvdev -lnagg ent0 ent1         # creates ent4
$ mkdev -sea ent4 -vadapter ent2 -defaultid 1 -attr ha_mode=auto ctl_chan=ent3 # creates ent5

Create an IP address on ent5 as needed. It does not have to have an IP address to function as a bridge.

I dont know if etherchannel is supported with two virtual adapters. Generally, with dual VIO servers there is one virtual ethernet adapter in the client partition. This virtual adapter is in the same LAN as two shared ethernet adapters (one in each VIO) - with one VIO as primary SEA, and the other as backup SEA.

If there is an etherchannel - this is usually defined in the VIO and the etherchannel entX device is the device that gets bridged by the SEA.

Does this answer you question sufficiently? Or does it raise more?