BGP Route Issues

Blake Dunlap ikiris at gmail.com
Mon Aug 19 02:42:24 UTC 2013


Local Pref (which is common by the way to be set so customers > peers >
transit). AS Path doesn't beat it.
 You can only request people follow the routes you want ingress, there's
nothing you can do to force them to take a path to you short of
deaggregation, and that only works until they notice it, and it irratates
the rest of the world as well by using additional route slots.

Poor routing is purely a viewpoint problem, not necessarily in agreement
between all parties.

-Blake


On Sun, Aug 18, 2013 at 8:33 PM, Christopher Karel <chris.karel at gmail.com>wrote:

>   Good evening,
>
>     I'm hoping you guys might be able to offer some advice or insight into
> a BGP problem I've got.  I've noticed some strange routes between our
> network, AS27270, and AS22943.  It looks like both our networks are dual
> homed.  One ISP as the primary, and the other used as a backup, with path
> prepending to prevent it from actually being used except in an outage.
>  However, our route to 22943 appears to be using their backup link.  (27270
> 4323 7018 22943 22943 22943 22943 22943 22943 22943 22943 22943 22943
> 22943)  Which is strange, because we can reach their primary ISP without
> any such rigmarole.
>
>     Playing around with Looking Glass servers indicates that Cogent
> (AS174) has a similar backup route to our network.  (174 22402 27270 27270
> 27270 27270 27270 27270)  Everywhere else I check seems perfectly sane.
>  But since Cogent is essentially in-between the two networks I'm
> troubleshooting, I would assume that the other network has a similar route.
>  But Cogent won't talk with me about this, since I'm not a customer.
>
>     So as far as advice goes, is there a common issue that might result in
> such poor routes in both directions?  Any further troubleshooting that I
> should be doing?  Or any ideas on how to help remedy things that appear to
> be outside our network/ISP?  Or are we doing something so wrong that this
> is all my fault?
>
> I'd really appreciate any input on this.
>
>


More information about the NANOG mailing list