[c-nsp] PoP's, Routing and GRE

[Bruce Pinsky]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Mark Tohill wrote:
> We have the following scenario:
> 
>  
> 
> A newly installed DSL PoP.
> 
>  
> 
> | H1------S1----R1----ISP1
> 
> |    \    /\     /
> 
> |     \  /  \   /
> 
> |      \/    \ /
> 
> |      /\     \
> 
> |     /  \   / \
> 
> |    /    \ /   \
> 
> | H3----- S2----R2----ISP1
> 
>  
> 
> H1/2 are dual-homed hosts.
> 
> S1/2 are WS-2970G-24's.
> 
> R1/2 are 7204VXR's with NPE-G1's connecting to redundant links in
> providers PoP.
> 
>  
> 
>  
> 
> We have a problem whereby we have 2 x /24's of fixed IP /32 users and
> another /25 of same who were being announced via BGP (3 x Class C's) to
> the old PoP. These users need to get transferred. We could announce the
> two class C's from the new PoP but what about the /25? Whose going to
> route a /25!!!??!! Keep in mind the second half of the /25 must remain
> in original PoP.
> 
>  
> 
> We thought GRE tunneling might get this static traffic across to new PoP
> without getting it bounced back to us via BGP. We set up tunnels on the
> old PoP edge and on new PoP edge 7200's.
> 
> We then pointed the 2 /24's and /25 across tunnel. The tunnels stayed no
> more that half a minute and then dropped.
> 
>  
> 
> I know it's not pretty but that's the way assignments have happened.
> Have we any other options here to sort out these customers?
> 
>  
> 
> Any thoughts welcome.
> 
>  

Were the tunnel endpoints assigned out of the 2 /24's and the /25?  If so,
then you ended up with a recursive route through the tunnel and it would
come down since the route to the tunnel endpoint would be through the
tunnel itself.

- --
=========
bep

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (MingW32)

iD8DBQFDmIc9E1XcgMgrtyYRAhrcAKD299gZIte6xPvEnHcRTDPs54R1wwCfetda
gox5hxwpceBA7I8bdMplbi0=
=QyCb
-----END PGP SIGNATURE-----