[nsp] CEF routing bug?

Ilker Temir itemir at cisco.com
Thu Jan 30 18:05:18 EST 2003


I was talking about the Ed's spesific case .

Of course a cef entry is created for any kind of route in the routing table.

However a directly connected entry (those which needs l3-l2 mapping like
ethernet) creates multiple /32 adjacencies for the hosts on that network.

Ilker

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Ilker:
>
> Really, I thought the adacency was created when there is an IGP route
> in the routing table and the NHR is available.
> Is that not where the cached adjacency comes from when we do a sh ip
> cef detail.
> Could you clarify?
>
> ==DMT>
>
> > -----Original Message-----
> > From: Ilker Temir [mailto:itemir at cisco.com]
> > Sent: Thursday, January 30, 2003 11:24 AM
> > To: Roy
> > Cc: cisco-nsp at puck.nether.net
> > Subject: Re: [nsp] CEF routing bug?
> >
> >
> > > It looks like CEF creates the /32 adjacency based on the
> > ARP irrespective
> > of
> > > whether or not its the correct route.
> >
> > Wrong interpretation, CEF creates a /32 adjacency only if the entry
> > is directly connected. Adjacency is created when there is no IGP
> > route in the
> > routing table.
> >
> > Ilker
> >
> > > The basic bug is that the route used is different based on
> > whether or not
> > CEF is
> > > on and that seems to be a bug.
> > >
> > > Ilker Temir wrote:
> > >
> > > > If you lose the route from your IGP, the network will be
> > > > directly
> > connected.
> > > > And CEF will create /32 adjacency based on ARP, as long
> > as this spesific
> > > > adjacency exists it will be used even if the route is
> > relearned from IGP
> > > > (since adjacency created from IGP will be less spesific).
> > > >
> > > > If arp times out, corresponding /32 CEF adjacency will be
> > removed and
> > there
> > > > will be no problem. However arp will never time out in
> > this case (unless
> > > > remote goes down). And to me, it is logical. Why would you want
> > > > to
> > remove an
> > > > arp entry when the directly connected destination is
> > still reachable ?
> > > >
> > > > Better change your design.
> > > >
> > > > Ilker
> > > >
> > > > ----- Original Message -----
> > > > From: "Ed Ravin" <eravin at panix.com>
> > > > To: <cisco-nsp at puck.nether.net>
> > > > Sent: Thursday, January 30, 2003 12:33 AM
> > > > Subject: [nsp] CEF routing bug?
> > > >
> > > > > Roy writes:
> > > > > > I have two routers ("A" and "B") and both are
> > directly connected to
> > LAN
> > > > > > "C" ( a /26 subnet) [...]
> > > > > > For various reasons, I prefer the traffic on router
> > "A" to go to LAN
> > "C"
> > > > > > via router "B".  I do this my entering more specific
> > routes for LAN
> > "C"
> > > > > > pointing to router "B" (two route of /27).  This
> > works fine until I
> > turn
> > > > > > on CEF.  When I do that, the direct connection is always
> > > > > > used
> > instead of
> > > > > > the indirect routes.
> > > > >
> > > > > I'm seeing similiar problems on two different routers -
> > a 7206 with
> > > > > 12.2(7c) and a 7513 with 12.0(21)S1.  On one router,
> > routes received
> > > > > via OSPF or RIP do not override CEF when the CEF route is on
> > > > > a FastEthernet interface.  On the other router, the
> > dynamic routes are
> > > > > used when the router is restarted, but if the dynamic routing
> > > > > protocol drops the route and CEF takes over, we can't
> > lose the CEF
> > > > > route without rebooting the router again.
> > > > >
> > > > > _______________________________________________
> > > > >
> > >
> >
> > _______________________________________________
> > cisco-nsp mailing list  cisco-nsp at puck.nether.net
> > http://puck.nether.net/mailman/listinfo/cisco-nsp
> > archive at http://puck.nether.net/pipermail/cisco-nsp/
>
> -----BEGIN PGP SIGNATURE-----
> Version: PGP 7.0
>
> iQA/AwUBPjlTkQgiZycqTvq3EQKsdQCgwtcUoBZjKLVyImgBfmtLKU284x4An0Ul
> So17gXy0Ndzo40zbQewY54JK
> =ybCQ
> -----END PGP SIGNATURE-----
>



More information about the cisco-nsp mailing list