nanog mailing list archives
Re: BGP and aggregation
From: Stephen Griffin <stephen.griffin () rcn com>
Date: Sun, 12 May 2002 17:10:35 -0400 (EDT)
In the referenced message, Andy Walden said:
Conditional Router Advertisement: http://www.american.com/warp/public/459/cond_adv.pdf andy
As it sounds like he's using a single AS, the above may not be a fix, since a partitioned AS is still a failure condition. One other solution, which could possibly be used along with the above, is to build a GRE tunnel from the interface facing transitA to the interface facing transitB, and exchanging routing information with a really high metric. This would keep the AS non-partitioned during loss of the inter-city link. General rules regarding collapsing a tunnel by routing the tunnel endpoint via the tunnel, do apply. This, however, is relatively easy to mitigate. The benefit to a tunnel is that it adds no additional state in the DFZ, and also is unlikely to run into the standard problems with deaggregation (that people will undoubtedly eventually filter you). One disadvantage is that you may have some amount of u-turn traffic, but you also won't have traffic following a default route to unreachable entities as you would likely have with any other solution.
Current thread:
- BGP and aggregation Ralph Doncaster (May 11)
- Re: BGP and aggregation Andy Walden (May 11)
- Re: BGP and aggregation Ralph Doncaster (May 11)
- Re: BGP and aggregation Stephen Griffin (May 12)
- Re: BGP and aggregation Andy Walden (May 12)
- Re: BGP and aggregation Stephen Griffin (May 12)
- Re: BGP and aggregation Andy Walden (May 12)
- Re: BGP and aggregation Ralph Doncaster (May 12)
- Re: BGP and aggregation Andy Walden (May 12)
- Re: BGP and aggregation E.B. Dreger (May 12)
- Re: BGP and aggregation Stephen Griffin (May 13)
- Re: BGP and aggregation Ralph Doncaster (May 13)
- Re: BGP and aggregation Scott Granados (May 13)
- Re: BGP and aggregation Andy Walden (May 11)