Solved: EIGRP-BGP Mutual Redistribution loop prevention

kfarrington · ‎09-09-2005

Guys,

I have a very interesting question.

I have multiple CE routers that are mutually redistributing between EIGRP and BGP. (multiple redist points in the network)

Now the loop prevention route-maps I have does the following

A prefix originates from BGP, and get redistributed into EIGRP via the route map and sets an EIGRP tag of 11111.

So now the prefix is in EIGRP and we are saying in EIGRP - redistribute EIGRP routes into BGP via a route map and if the tag is matched, donot redist

So the EIGRP prefix is not sent back to BGP!

This is protection from BGP into EIGRP and back into BGP only, ie if we receive an external BGP prefix from the carrier network.

This is fine and works dandy between mutual redist points.

BUT WHAT ACTUALLY HAPPENS ON THE LOCAL ROUTER?

I am wondering what will happen from EIGRP into BGP and here are my thoughts (ie, a local campus route that needs to go out to the carrier network).

A prefix originates in EIGRP (say AD of 90), and we tell EIGRP to redist into BGP via route-map. The prefix is now in BGP as it has NO tag of 1111.

Now the route is in BGP, we tell the prefix to be redist into EIGRP via route-map and it goes back into eigrp now with a tag of 11111.

Now the prefix in in EIGRP, we tell it to go back into BGP and it gets denies becuase it now has a tag of 11111.

Would this be the way it works, or is there an internal EIGRP or router mechanism, that would stop a route going from EIGRP into BGP and then back into EIGRP?

I would just like to clarify this point as it may not be the most efficient of routing policies I have and may have to look at setting a route-map to stop this from happening.

!

route-map bgp-to-eigrp permit 10

set tag 11111

!

route-map eigrp-to-bgp deny 10

match tag 11111

!

route-map eigrp-to-bgp permit 20

!

filtering...

router eigrp 1

redistribute bgp 11111 metric 10000 100 255 1 1500 route-map bgp-to-eigrp

network 155.195.0.0

no auto-summary

!

router bgp 11111

bgp log-neighbor-changes

bgp deterministic-med

neighbor External peer-group

neighbor External remote-as 22222

neighbor External timers 2 6

neighbor Internal peer-group

neighbor Internal remote-as 11111

neighbor Internal update-source Loopback0

neighbor x.x.x.x peer-group External

neighbor x.x.x.x peer-group Internal

!

address-family ipv4

redistribute eigrp 1 route-map eigrp-to-bgp

ruwhite · ‎09-09-2005

You're generally okay with this setup as far as just breaking the routing loop, but it is possible for the redistributing router to choose the BGP route over the EIGRP route (administrative distance) or the EIGRP route over the BGP route (order of operation between BGP and an IGP). What I would try and do is to ask the SP to pass communities for me, and use the communities like a tag, so you can set a community on the EIGRP->BGP side, and then block on it on the BGP->EIGRP side. That way, even if the BGP route wins, it won't be redistributed back into EIGRP.

Note that this isn't always optimal, though--if you have a split site, and the split off part has a backdoor link, you could end up not being able to route to destinations you actually have a physical path to. It doesn't sound like you have a backdoor in this situation, so it probably doesn't matter, really.

:-)

Russ.W

View solution in original post

ruwhite · ‎09-09-2005

You're generally okay with this setup as far as just breaking the routing loop, but it is possible for the redistributing router to choose the BGP route over the EIGRP route (administrative distance) or the EIGRP route over the BGP route (order of operation between BGP and an IGP). What I would try and do is to ask the SP to pass communities for me, and use the communities like a tag, so you can set a community on the EIGRP->BGP side, and then block on it on the BGP->EIGRP side. That way, even if the BGP route wins, it won't be redistributed back into EIGRP.

Note that this isn't always optimal, though--if you have a split site, and the split off part has a backdoor link, you could end up not being able to route to destinations you actually have a physical path to. It doesn't sound like you have a backdoor in this situation, so it probably doesn't matter, really.

:-)

Russ.W

kfarrington · ‎09-10-2005

brill. thx fella.

We have a standard set of communities for LP and other BGP descision attributes. will create another community for redistribute.

ta fella