EIGRP and LDP neighbor flaps

Unanswered Question
Mar 12th, 2010

Hello everyone.  I have an interesting problem.  I am trying to enable multicast on my WAN to enable Enterprise TV with my DMM and DMPs.  The problem is that once I enabled multicast on the WAN and start sending traffic, I will randomly have EIGRP and LDP neighbor flaps.  The EIGRP neighbors will drop and reconnect in about 0.1 seconds according to the logs.  The LDP neighbors will take a few more seconds to reconnect.  I am using all the defaults for EIGRP, and the WAN links are Metro Ethernet.  We are running MPLS VPN across the WAN as well for some of our vendors.  Multicast is configured as sparse-dense mode, with BSR and RP candidates defined for the 10 groups that we have.

I have included a basic WAN diagram to try and help.  R0 is the one that is losing its neighbors.  It will show Neighbor X.X.X.X (VlanX) is down: Peer goodbye received.  I have done debugs on some of its peers, and they will miss some of the hellos that R0 is sending.  The other neighbors on Provider 1 do not drop their adjacencies with each other, only with R0.  This doesn't happen often, and there is no real service interruption, but it is quite annoying.  All of the devices on the WAN are ME6524 running 12.2(33)SXH2a Adv IP Services, except for one 3825 running 12.4(3i) SP Services.

I can post configs and debugs if needed, but they are somewhat complex and will take some scrubbing.  Thanks!

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (1 ratings)
Laurent Aubert Sun, 03/14/2010 - 19:03


It looks to me there are some packet lost in provider 1 backbone maybe due to bursty traffic. Can you correlate the lost of adj with a peak rate of traffic ?



jcrussell Sun, 03/14/2010 - 19:14

I have not yet tried to coordinate the drops with a burst of data, that is a good idea.  I can say there are no packet drops on my equipment's interfaces, though.  When I get in to work tomorrow, I will turn on a stream, wait for a drop, and check the traffic rates at that time.  We have Solarwinds with the Netflow module, so that should be of some assistance.

jcrussell Tue, 03/16/2010 - 11:49

It turns out our provider rate-limits ingress multicasts to 1% of the link bandwidth.  Since we have a 1 Gbps pipe, and a 10Mbps video stream, that's what was causing it.  I called the provider, and they said they will lift the restriction only if a customer calls to complain.  So they lifted it.  Thanks for looking, everyone!

Giuseppe Larosa Wed, 03/17/2010 - 09:30

Hello Jcrussell,

you have been kind in providing a feedback on this issue.

It makes this a complete story

Best Regards



This Discussion