redistribute EIGRP into BGP based on best path / reported distance (RD) or some other value

[will]

I have two main datacenters (DC's), which each could obtain an EIGRP route. This learned EIGRP route is from a VPN tunnel RRI (reverse route injection), which either terminates in one DC or the other. So the DC's will not have the same origin sending them the route at the same time. they can share this route with each other via EIGRP however, via a DC P2P link.

Next, each DC can also redistribute the learned route into BGP for sharing with other eBGP peers. In practice, if I don't do anything specific, both DC's will redistribute the EIGRP route into BGP with the same path length/cost (just different AS's for each respective DC). I want the DC which learned the EIGRP route to be the one that redistributes this into BGP.

what I can envision is the EIGRP RD will change as the learned route hops to the far-side DC. Based on this value, I would like to either distribute into BGP or not distribute into BGP. then I can let the best DC do the eBGP redistribution.

Is this possible? is there another method for achieving this?

thx in advance!

[Philip D'Ath]

How about you redistribute this RRI route into eigrp and bgp at both sites - but DON'T redistribute it from EIGRP into BGP?

[will]

hey phil, thx for quick reply. if I redistribute the RRI route into BGP at both sites, it will be come an equal hop path out on MPLS network from both DC's. I want to prefer the DC that originally got the RRI.

And I cannot redistribute into EIGRP at both sites, but by doing the redis. at DC1, the EIGRP relationship with DC2 will carry the route across.

After some more research today, I think I can do this as follows, but its going to require OSPF, as the RRI device is ASA and ASA doesnt support 'tags' in EIGRP.

1. receive route on ASA

2. on ASA, tag it with route tag 65001/65002 for each DC respectively. these are just tags representing the AS. not actually used as AS. DC1 tags with 65001; DC2 with 65002.

3. on ASA, RRI route into IGP, which has to change to OSPF. :(

4. OSPF carries the route everywhere (both DC's).

5. redis. the OSPF route into BGP at both DC's, but use a route-map to match the tag.

6. the site without the incorrect tag will just drop the route; the site with correct tag does the redis.

I'm still playing around with running BGP solely and removing the IGP (EIGRP/OSPF) entirely, but not quite sure I want to do that yet. may lab it up tomorrow to the extent I can.

[Philip D'Ath]

You wont be able to distribute RRI into BGP at both sites - because it will only be up at one site at a time.  Hence only one of your two sites will advertise it via BGP.

[will]

hey phil, once I distribute the RRI into either EIGRP or OSPF, I can set the tag. this tag should be carried through the IGP, and be usable for manipulating the BGP redistribution. the P2P DC link runs the IGP. the MPLS is a second network in/out of each DC. so both DC's do have awareness of the RRI route.

Im still testing this solution, but I think I can set the origin somehow with BGP based on the EIGRP or OSPF route tag. prepend AS based on a route tag is not supported within BGP - at least that's what the error message implies.

If I cannot add the origin in some way as a path prepended, I think I can do this with communities and filtering the MPLS inbound BGP RRI route (from the other DC), based on a preset community.

Here are come cool links on BGP origin tagging from IGP:

http://packetlife.net/blog/2009/jan/19/bgp-route-auto-tagging/

https://ccieblog.co.uk/bgp/bgp-automatic-tag-as-path-tag

[milan.kulik]

Hi,

1) I guess there shoud be different next-hop when the EIGRP prefix would be learnt from the other DC?

So possibly you could use ip next-hop match within the redistribution route-map?

2) Or you could tag the RRI prefix within EIGRP using a different tag in each DC and match the tag values?

See http://www.cisco.com/c/en/us/support/docs/ip/border-gateway-protocol-bgp/49111-route-map-bestp.html#cmdsredist

for details.

Best regards,

Milan