[c-nsp] Uneven LACP load-balancing

[Benjamin Lovell]

Not sure if the 3K platforms use hash ID to randomize hash result but  
I think the problem would be that it's the same ASIC doing both hashes  
so that if the hash criteria are the same result will be the same in  
polarizing like effect.

-Ben


On Nov 12, 2010, at 1:14 PM, Brandon Ewing wrote:

> I would still classify this as a bug, or request additional features  
> to
> allow one to alter the layer 3 hash ID and layer 2 hash ID  
> seperately to
> avoid issues such as this where src-dst-ip hashing is required at  
> layer 2.
> My testing appears to indicate that since both the layer 3 hash and  
> layer 2
> hash are the same, links that choose one L3 interface will always  
> choose the
> same L2 interface inside the individual bundle.  Being able to seed  
> them
> independently should resolve this issue.