Re: Multi-homed BCP38

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 1/9/14, 7:09 AM, John R Levine wrote:
> I was at a meeting talking to ops people from some large ISPs, who tell
> me that when they tell their large customers about BCP 38, the customers
> say forget it, because they're multihomed.  I gather the situation is
> typically that the customer has multiple address ranges, say from
> providers A and B.  Normally traffic from range A goes out through
> provider A, and vice-versa, except sometimes when it doesn't.  Sometimes
> it's failover, or it may be deliberate asymmetic routing.  The customers
> may not be running BGP, or if they do, they don't want to announce range
> A to provider B for business reasons I don't entirely understand but
> that are not going away.
> 
> The ISPs tell me that the customers are often ISPs themselves, so there
> are lots of address ranges, far more than anyone could track manually
> even if they wanted to.
> 
> I see BCP 84, which is now ten years old.  The ISPs are aware of it, but
> it doesn't seem to have done the trick.  I can think of some hacks to
> pseudo-announce ranges for filtering purposes, but surely I am not the
> only person to have noticed this problem.  What have people done to
> address this issue?*  I figure the first thing to do is to understand
> what's failed before.

Peers the do URPF strict, and also having regionally distinct prefix
range advertisements (e.g. we peer in multiple regions) for peers  are a
problem e.g. due to internal issues traffic that was flowing through a
port where the prefix was announced is now flowing towards me through a
port which it is not, likewise in the other direction, in a case where
the out of region prefix would due to local pref be reached via a
transit provider, if the transit route goes away  may well blackhole at
the other peer if it instead crosses your backbone to another exit. We
use loose to prevent this in the inboud direction with those peers  we
also announce covering agregates to particular URPF strict peer.


> Regards,
> John Levine, johnl@xxxxxxxxx, Taughannock Networks, Trumansburg NY
> Please consider the environment before reading this e-mail.
> 
> * - other than calling the customers stupid, which they are not, and is
> not helpful
> 


Attachment: signature.asc
Description: OpenPGP digital signature


[Index of Archives]     [IETF Annoucements]     [IETF]     [IP Storage]     [Yosemite News]     [Linux SCTP]     [Linux Newbies]     [Fedora Users]