Single IP routing problems through Level3

Matthew Kaufman matthew at
Sun Jun 15 22:07:21 UTC 2008

Matt Palmer wrote:
> A multiple-link bundle which is load balanced by source/destination pair
> with an undetected dud link?  I hadn't thought of that, but it does make an
> *awful* lot of sense....

I've also seen interesting OSPF misconfigurations that resulted in a 
router doing path-wise load balancing between the live link and an 
unroutable destination address that went into the bit bucket.

On the Cisco boxes I was using at the time, the hallmark of 
load-balancing into a dead path was that every other IP address worked, 
but then every some number of addresses (12, as I remember, but I'm not 
100% sure) the polarity flipped, so that whichever of odd vs even had 
worked before was now the one that didn't for the next N addresses. 
Artifact of the hash algorithm in use, no doubt.

Matthew Kaufman
matthew at

More information about the NANOG mailing list