[c-nsp] multipath BGP not balancing equally.

[Rodney Dunn]

That's usually caused by routes not being the same on the paths.
This is a hard problem to solve.

Is there any way we could prove the prefix distribution is the exact 
same over the paths?

I don't know of a way other than dumping the output for every route in 
the RIB looking for the next hop.

Rodney



bas wrote:
> Hi,
> 
> I have an issue with unequal multipath BGP loadbalancing
> It is a 6500 / SUP720-3BXL running 12.2.18SXF16
> 
> There are four eBGP sessions to a transit carriers ASN, all with full table
> 
> However one out of four interfaces sends about 2Gbps less than the other three.
> 
> RTR-HV7#sh int ten 2/2 | i output rate
>   1 minute output rate 6357052000 bits/sec, 546295 packets/sec
> RTR-HV7#sh int ten 3/1 | i output rate
>   1 minute output rate 8509719000 bits/sec, 729490 packets/sec
> RTR-HV7#sh int ten 3/3 | i output rate
>   1 minute output rate 8721235000 bits/sec, 746980 packets/sec
> RTR-HV7#sh int ten 4/4 | i output rate
>   1 minute output rate 8592400000 bits/sec, 734864 packets/sec
> 
> All four sessions have the same settings (in the same peer-group)
> Through netflow I've tried to deduct if there are specific ASN's not
> chosen through the nexthop that has less traffic, but that does not
> seem to be the case.
> 
> I've looked at "ip cef load-sharing algorithm universal" however that
> seems to already be the default algorithm in current IOS versions.
> 
> With any prefix I test through "sh ip cef x.x.x.x detail" it seems all
> four paths are used.
> 
> Thanks in advance,
> 
> Bas
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/