[c-nsp] MPLS affecting normal IP cache flows

[Oliver Boehmer (oboehmer)]

Andy Saykao <> wrote on Wednesday, August 06, 2008 5:19 AM:

> Hi All,
> 
> I've deployed MPLS across parts of our core network and everything
> appears to be working fine. I've also got MPLS VPN's going which is
> the main reason for us rolling out MPLs in the first place.
> 
> However, I've run into a problem with netflow on one of the PE routers
> that affects normal IP flows when mpls is enabled on the interface.
> The PE router having this problems is a cisco 7206VXR (NPE-G1) running
IOS
> Version 12.3(22). Other PE routers are not showing this problem but
> they are 7301's running a different IOS.
> 
> What I'm finding is that when I enable "tag-switching ip" on interface
> Gi0/2 which forms part of our MPLS core (as seen below), the netflows
> for normal IP traffic isn't as it should be. Doing a "show ip cache
> flow" on the PE router only shows a few flows going through for normal
> IP traffic and we'd expect more IP cache flows to be going through
> because lots of customers hang off this PE router. When we remove the
> "tag-switching ip" from the interface, flows are back to normal.
[..]
> Any ideas as to why enabling mpls would be affecting normal IP cache
> flows? I can only suspect that it's some IOS bug with the IOS we're
> running .

do you filter any LDP advertisements? If you just enable LDP on the
interface (or TDP for that matter), "regular" IPv4 traffic will also be
label-switched as LDP advertises labels for all non-BGP IPv4 prefixes in
your RIB. Can you do a "show mpls forwarding-table <prefix>" for a
prefix you would expect to see in the cache? Do you see a incoming label
(which is != Pop label)?

	oli