09-10-2009 11:23 AM
If following condition is met.. what will be end result:
PE1 - VRFA - prefix:1.1.1.0 - rd 1:1
PE2 - VRFB - prefix:1.1.1.0 - rd 1:1
if these routes are shared with a third PE and if exchanged bw these PEs..
09-10-2009 12:56 PM
Hello Mukarram,
the RT still plays a role here so let's suppose it is the same also the RT.
A third PE and a route reflector server will think that this is a multihomed VRF site connected to PE1 and PE2.
if this is not true you have 50% probability to hit the right PE / VRF site.
PE1 and PE2 should prefer connected route (if this is connected) or route learned by CE, to the iBGP route that comes from the other one.
(this was not true some years ago we had to use neigh weight to prefer eBGP CE before using site of origin)
if the RT is different these are IP prefixes in different VPNs and no problem can occur.
Hope to help
Giuseppe
09-10-2009 06:34 PM
Hi,
Say PE1 is configured as follows-
ip vrf CUST1
rd 1:1
route-target both 1:1
and
PE2 is configured as-
ip vrf CUST1
rd 2:2
route-target both 2:2
If 1.1.1.0 prefix is advertised to PE3 which is configured to accept (import) both RTs 1:1 and 2:2, then I think PE3 will install 1.1.1.0 from both PE1 and PE2 with different next-hop IP addresses.
Although, if you think a step further, what will happen when it advertises 1.1.1.0 prefix (of course using redistribution) to its connected CE router?
09-10-2009 10:57 PM
hi amit,
good question..
RD is used to make IPv4 routes unique within the MPLS-VPN domain, but when they are redistributed (imported) from MP-BGP to VRF therez no way to differ between two similar routes...
and i guess this results due to poor design parameters...
09-10-2009 11:02 PM
hi Guislar,
"if the RT is different these are IP prefixes in different VPNs and no problem can occur"
the point is that if the routes are imported from two VRFs using same prefixes... when route is to be installed in VRF , RT is left behind...
09-10-2009 11:26 PM
Hi
You are right there will be a huge problem because the ambiguity will of routes will occur. This is the reason, the routes will be unique during extranet. One more solution for this is that use vrf nat.
regards
shivlu jain
09-10-2009 11:22 PM
Hi
If the same route is advertised with same RD and RT a problem will be there because everytime the best route is selected by RR. Now RR will see the same route from two different PEs and it will select the best route according to the PE metric.
This can be checked by using the command where metric will play signofocant role
show ip bgp vpnv4 vrf xyz
regards
shivlu jain
09-10-2009 11:28 PM
Hi shivlu,
same RT is okk..
kindly consider following scenario:
PE1 is configured as follows-
ip vrf CUST1
rd 1:1
route-target both 1:1
PE2 is configured as-
ip vrf CUST1
rd 2:2
route-target both 2:2
If 1.1.1.0 prefix is advertised by PE1 & PE2 (with different RDs) to PE3 which is configured to accept (import) both RTs 1:1 and 2:2,
Uptil now the two routes were distinguished by an RD.
But when the routes are redistributed from MP-BGP into VRF, what is left are two identical IPv4 prefixes from different PEs.... and the routing results would be unexpected.. ????????
09-10-2009 11:32 PM
Hi
You are right there will be a huge problem because the ambiguity will of routes will occur. This is the reason, the routes will be unique during extranet. One more solution for this is that use vrf nat.
regards
shivlu jain
09-11-2009 05:29 AM
Mukarram,
As Giuseppe indicated, this is far from ideal as the RR or the 3rd PE will select only one VPNv4 best path (RD + IPv4 prefixe), so only one of these two advertisements will be chosen regardless whether the RT is the same or not. Using the same RD for different VPN should really be avoided.
Regards
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