Strange unicast problem when configuring mVPN

I'm going to cross-post this post that I made to the Cisco Learning Network
last night since I haven't been getting any hits on it there. Here is a
link to that discussion if you want to see the topology or get the GNS3
project file with configs.

I'm working on an mVPN lab in GNS3 and am running into a really bizarre
problem. I've attached my topology. The gist of it is that R13 and R14 are
customer routers that can ping each other just fine when I only have the
basic L3VPN configuration in place, but things get weird quickly. For
background, I have OSPF running in my customer areas and BGP is my PE-CE
protocol. R1 and R7 are my vpnv4 peers.

So, to configure mVPN, I started out by turning on PIM-SM on my P routers
and made R3 the RP via BSR. Next, I configured PIM-SM in my customer areas.
So far, no problem.

Next I enabled PIM on the customer-facing interfaces on my PE routers.
Still no problem. Then I configured the mdt default address in the vrf
config and BLAMMO...broken unicast connectivity between 13 and 14. In the
IOS image I'm running at the moment, my PE routers will immediately begin
exchanging MDT information as soon as I configure the address in the vrf.
They use the regular vpnv4 AF for this, unlike newer releases that use the
ipv4 mdt AF.

What in the world could cause something like this? I'm completely at a
loss. I'm not even sure how to troubleshoot it since it's so bizarre. But
once those MDT tunnels come up, things go bad fast.

This is about the best it looks when the MDT tunnels are up:

R13#ping 14.14.14.14 rep 50 time 5

Type escape sequence to abort.
Sending 50, 100-byte ICMP Echos to 14.14.14.14, timeout is 5 seconds:
!!!!!!!!!!!!!!!...!!!!!.!!..!!!!!!!!!!!!!..!!..!.!

Sometimes I get nearly no responses at all.

Even stranger is that it seems like removing the mdt config and bouncing
the PE BGP peers doesn't seem to resolve it reliably. So far, I've found
nothing that fixes my unicast connectivity once it starts breaking. I'm
going to try saving my configs and topology and then restart all of the
routers.

Have you ever seen anything like this? I'm totally stumped.

Thanks,
John

Blogs and organic groups at http://www.ccie.net
Received on Thu Jan 10 2013 - 10:08:44 ART