ExtremeSwitching (EXOS)

Creating an MLAG between two 440-G2's. The ISC will be an LACP of ports 15 and 16 and if I understand it, this LACP link must have a static MAC assigned on both sides of the ISC (on each peer). Is there a recommendation in terms of the MAC to use (do you just make one up)?

Also, mostly out of curiosity - there's a note about downstream LACP switches (one port to each peer) that discusses ensuring that the MLAG ports (ISC LAG) be configured as LACP too (which I'm doing). It reads as though this is a required config (I can't use a single 10Gb ISC connection). What's the reason? Is it because of the shared MAC issue too? Thanks!

Hi Eric,

Use the command

configure mlag peer lacp-mac

Thanks Brad. So to confirm, the shared mac on each mlag lacp/isc shared port will be the mac of one of the two core 440's. Same on both sides of the link. I'm assuming l3/l4 both there and on the downstream 440's (lacp with one port to each mlag core) will provide the most balanced lacp use?

Okay, so after a little more digging I have another question about this. The KB shows that this command is in support of MLAG ports (ie: the downstream connections to the MLAG pair) using LACP (like a switch with an LACP split between MLAG peers). In one of the KB docs, the command simply says "configure mlag peer mac

You also need to enable LACP on the MLAG ports in SW1 and SW2
MLAG Port SW1
en mlag port 1:1 peer "core-sw2” id 1
en sharing 1:1 group 1:1 algorithm address-based L3_L4 lacp

MLAG Port SW2
en mlag port 1:1 peer "core-sw1" id 1
en sharing 1:1 group 1:1 algorithm address-based L3_L4 lacp

I guess it all boils down to the fact the the ISC is really nothing more than a normal LACP trunk to which the ISC VLAN is connected and across which the peer is defined, along with other VLAN's that are needed to support the "cross-core" LACP connections? Does that pretty much sum it up?

Thanks Ron. Missed that. So even though each switch is only singly connected you still need LACP to make the LAG work. I was thinking that somehow the MLAG port definition inherently did that, but then I guess it wouldn't work for different (non LACP) LAG types. Cool and - thanks!

That is a correct conclusion.