Hello Gurus....
The other day I ran into a redistribution scenario that I was surprised to find out didn't really work as expected.
I have one router R1 which runs 2 x OSPF processes as well as BGP (all BGP neighbours belong to other AS' so eBGP throughout).
The topology looks like this:
R4 ----(OSPF Process 2)----- R1 -------(eBGP)---------- R2
|
(OSPF Process 1)
|
R3
Again, R1 uses OSPF process 2 to talk to R4, BGP to talk to R2 and OSPF process 1 to talk to R3.
The way I wanted redistribution to work in R1 is as below:
- OSPF 1-> BGP
- OSPF 2-> BGP
- BGP -> OSPF 2
- OSPF 1 receives only a default route as it's a stub site.
I expected OSPF 1 to redistribute the routes learned from R3 into BGP, then for BGP in turn to redistribute those routes into OSPF 2 and onwards to R4 and R2.
What I found instead is that R4 only ever received the original BGP routes from R1 and R2. The OSPF routes that were learned from R3 and redistributed to BGP never made it to R4 (OSPF 1->BGP->OSPF 2 redistribution didn't work).
However, I could see the routes that originated from R3 in both R1 and R2's routing table.
In the end, the only way I could make R4 see R2's routes was to enable redistribution between OSPF processes in R1:
What is the mechanism that prevented BGP from redistributing to OSPF process 2 those routes that it learned from OSPF process 1?
regards,
Leo
