Charles Sprickman <spork@inch.com> writes:
> We have made an agreement with another provider in the co-lo space to back
> each other up... I'm wondering what my options are here; it's strictly
> for backup purposes. I've considered announcing routes to both upstreams,
> padding heavily to the backup, and doing a floating static default.
The big pain in such a situation is that padding generally won't do
what you want. An ISP will _generally_ localpref its customer routes
up so that they will beat out stuff that it hears from its peers,
regardless of how hard you pad the path.
Sometimes, it will be possible to put a community on that route so
that the upstream will localpref it as if heard from a peer, but still
re-announce it if that route is the best one heard from downstream.
Or you may be able to get a custom one-off configuration to do the
right thing.
The problem is, having this work out is predicated heavily upon having
a bgp-clueful person working for your mutual-defense partner, and a
hook to a clueful person at the upstream ISP who isn't kept from
working with you by configuration management software, policies, etc.
> Is
> there really any point in taking two full views for this simple
> configuration?
>
> We connect to both upstreams via FE, and both hit a switch before the
> router. In a situation where an upstream's router is down, but their
> switch is up, would there be any way to detect that in my simplistic
> floating static scenario?
You could always send default to each other via BGP; that way, if the
session dies the default goes away (better than static).
---Rob
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:13:29 EDT