I am having some real world problems with a couple of intermittent satcom links and I want to know what I can do to prefer a route.
Currently, I have the most stable satcom preferred using a simple delay metric on the GRE tunnels; however, when that stable satcom link does go down, it causes a reconvergence over the backup link and that causes about a 10-15 second delay in data transmission until it converges on the secondary side…
Now, because I have that delay injected preferring the primary side, when the primary side does eventually come back up, it causes that same 10-15 second loss in connectivity while it re-converges over the primary side…
What technology can I use to make it to where, IF primary goes down, A) it converges over the secondary link a little faster and B) if primary comes back up, is there a way to make it to where it sends an ICMP 10 times over 10 seconds before it prefers the primary again? Just to ensure that the primary is stable before re-converging over the primary side?
The hold timers for EIGRP is 5 seconds, with hellos sent every 1 second.
The keepalive timers for tunnel are 10 1.
Is there anything that I can do to make convergence more efficient and cause less down time? or am I doing everything that I can?
There are several things you can do to speed up EIGRP convergence. Which one you choose will depend on your equipment and your configs.
The first thing that comes to mind is that EIGRP should have a feasible successor for the alternate path. So at the point when your primarily link goes down, it should almost instantaneously install the secondary route without the need for reconvergence. You can check the feasibility condition to see if/why that is not being installed, and if necessary, you can make some tweaks to make sure it gets in there. More info on this can b found at the Introduction to EIGRP lesson. This of course will not resolve the slow convergence when the primary link comes back up.
Another option would be to adjust the EIGRP timers so that they allow for a faster reconvergence. You may need to experiment with various values because satellite links can be somewhat tricky because of their higher latency.
Thirdly, you can try to use bidirectional forwarding detection (BFD) which is an added feature that can further decrease the convergence time of various routing protocols including EIGRP. This will depend on your equipment as well as you topology.
Which you will choose depends upon what you want to achieve and upon the details of your topology and your equipment. Take a look at these and if you have any further questions, feel free to let us know!