03-23-2012 07:35 AM - edited 03-07-2019 05:44 AM
In the past, I've had two routers, with 2 GRE tunnel interfaces each. When one tunnel would go down, EIGRP would instantly converge, and begin passing traffic through the secondary tunnel. I've since moved the secondary tunnel to a second router at each location, and convergence now takes roughly 15 seconds. Is this due to the fact that multiple neighbor relationships existed between the two routers, and because of this the primary tunnel was able to send a "Goodbye" packet via the secondary tunnel, enabling rapid convergence? As it stands it appears that the 15 second delay is a result of hello / hold timers being exceeded. If this is the case, I'll look into using something like Bidirectional Forwarding Detection; I just wanted to make sure the behavior I'm seeing is expected. I'd like to avoid putting band-aids on something if there is something significantly wrong with the underlying configuration.
Solved! Go to Solution.
03-26-2012 01:52 PM
Yep. If layer 2 doesn't tell EIGRP that the link is gone, you're at the mercy of the hold timer if you don't use BFD.
03-23-2012 05:53 PM
Comments inline with DS
In the past, I've had two routers, with 2 GRE tunnel interfaces each.
DS Where do these tunnels terminate? Just between these two routers?
When one tunnel would go down, EIGRP would instantly converge, and begin passing traffic through the secondary tunnel. I've since moved the secondary tunnel to a second router at each location, and convergence now takes roughly 15 seconds.
DS Does this mean you now have a box of four routers?
A----
B
. |
A'----
B'
Is this due to the fact that multiple neighbor relationships existed between the two routers, and because of this the primary tunnel was able to send a "Goodbye" packet via the secondary tunnel, enabling rapid convergence?
DS. No. EIGRP treats two links between two Routers as independent peers and can't use one link to send info pertaining to the other peer relationship.
As it stands it appears that the 15 second delay is a result of hello / hold timers being exceeded. If this is the case, I'll look into using something like Bidirectional Forwarding Detection; I just wanted to make sure the behavior I'm seeing is expected. I'd like to avoid putting band-aids on something if there is something significantly wrong with the underlying configuration.
DS Apparently the link down is no longer happening or being reported to EIGRP, so you're relying on hold timer expiration. Thats what you need to explore.
Sent from Cisco Technical Support iPad App
03-26-2012 01:29 PM
In an attempt to keep things simple, I think I may have done more harm than good. The connectivity I'm describing is taking place in a hub / spoke VPN topology, where there are two hub sites. Because of this, losing connectivity to one hub will not cause the multipoint tunnel interface to go down. If EIGRP is waiting for an interface state change to trigger convergence, then I think this is my problem. I'm not sure how it was working so well before, but I think I have an idea of why it isn't working now. I'm still thinking that BFD is my best bet.
03-26-2012 01:52 PM
Yep. If layer 2 doesn't tell EIGRP that the link is gone, you're at the mercy of the hold timer if you don't use BFD.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide