Strange behavior with mutual redistribution between EIGRP and EBGP

CodeBloxCodeBlox Member Posts: 1,363 ■■■■□□□□□□
The router performing the mutual redistribution should be preferring the the eBGP routes over the external EIGRP routes. What I am finding is that initially, that is true however, if I break BGP so that the router learns alternative routes to these prefixes and then bring BGP back up, the router still keeps the external EIGRP routes instead of updating with the eBGP routes... There are two points of mutual redistribution in this topology. What's more is the routes learned through EIGRP show up in the BGP table but are showing up in the routing table as D EX routes... It's odd since these routes have a higher AD yet are preferred. I can post configs and topology if needed. Please advise.
Currently reading: Network Warrior, Unix Network Programming by Richard Stevens


  • networker050184networker050184 Mod Posts: 11,962 Mod
    This is a very common thing people see with their MPLS scenarios when doing mutual redistribution. What happens is when the eBGP goes down the EIGRP route is now in the routing table. Now that it is there it is redistributed into BGP. So this is now in your BGP table with a default weight of 32768 for a locally generated route. All good so far and your EIGRP route stays in the routing table. Now, when your BGP session comes back up you get another entry in the BGP table for the prefix now with a default weight of 0 for an eBGP learned route. So basically your BGP table looks like this: locally generated weight of 32768 learned via eBGP weight of 0

    So what happens when your BGP path selection takes place? The route with the highest weight is chosen which is the redistributed route that is already in the routing table and remains there.

    A common fix is to apply a weight to the neighbor higher than 32768 to let the eBGP route win when it comes back into the BGP table and then therefore entered into the routing table.
    An expert is a man who has made all the mistakes which can be made.
  • CodeBloxCodeBlox Member Posts: 1,363 ■■■■□□□□□□
    Nicely said Networker... It is working now by upping the weight as you stated! I was gripping along the lines of something like that being the fix but wasn't sure about what I'd do to implement it. This job as SysAdmin is really awesome and is forcing me to learn some things I hadn't planned to learn until later down the road. I am actually working on CCNP now so all of this BGP knowledge will help me greatly on the exam. I created a GNS3 lab of our production network and am making some changes that will accomodate IPSLA and this was an issue that I noticed in the lab that probably exists in the live environment. What's also odd is that it shows up as an EIGRP external route when I'd expect at the least it show up as a BGP route in the routing table.
    Currently reading: Network Warrior, Unix Network Programming by Richard Stevens
  • networker050184networker050184 Mod Posts: 11,962 Mod
    Definitely a head scratcher I've seen plenty of senior engineers not be able to figure out.
    An expert is a man who has made all the mistakes which can be made.
  • Dieg0MDieg0M Member Posts: 861
    Good one. I would see this in CCNP or a Cisco sub-forum though since this is a problem cause by the Weight attribute.
    Follow my CCDE journey at
Sign In or Register to comment.