Comware

Hello!

let me explain the setup scenario:

I am building a new cluster with several nodes. Each with 4 x 10GBit interfaces. For HA reasons, 2 of the interfaces should point to swich A and the other 2 to switch B.

Switch A and B is a IRF stacked pair.. 

IRF is LACP aware but my question is how to connect the nodes:

- create one BAG with 4 x 10,GBit, 2 of them pointing to swich A and 2 of them to switch B

-  create 2 BAGs, first of the BAG to switch A and the second BAG to switch B

In the past, I used the second method but I am not sure if this is the best setup.

Thans for any suggestions

Regards

Juergen Northe

Thanks for any advice

Regards Juergen

So if you have, at core, a two members IRF Stack (you're calling IRF members 1 and 2 as Switch A and B, that's OK) and various edge Switches equipped with 4 x 10G interfaces...go with your first idea: create one LACP BAG with 4 x 10G members interfaces (this on relevant edge Switch(es) and on the IRF Stack too) and conveniently distribute 2 physical BAG member links to IRF Member 1 and remaining other 2 physical BAG member links to IRF Member 2.

Doing so you will have a single 4 x 10G BAG connected to the IRF Stack which, from the point of view of edge Switch(es), is seen as a single L2/L3 virtual Switch...clearly you should pay attention to throughput between IRF Members - IRF Port(s) - too (I assume you're interconnecting them accordingly): consider to enable, at system level, the lacp traffic-redirect-notification option on the IRF Stack *and* to keep enabled by default the long LACP timeout interval too.

In other words, avoid to create two separate BAGs on the same edge Switch, connecting the first BAG to IRF Member 1 and the second to IRF Member 2.

Hi Parnassus,

thank you very much for your support and usefull hints!

Regards

Juergen