14.7 PE-CE Routing with OSPF

I have seen numerous posts from different people inquiring about why we have different domain-ids on PEs R5 & R6 in this example, but I haven't seen any concrete explanation and I am trying find clarity, so please bare with me....

I understand that when OSPF routes from one CE are sent to it's PE, the PE redistributes the OSPF routes into MP-BGP and when it gets to the Remote PE, it is redistributed back into OSPF. Once this is done, the down-bit is set and the routes are sent to the CE as LSA Type 3 routes Because the down bit is set, these routes wont get advertised back into BGP.

In this lab, I understand that we change the domain-id to make to the BGP routes redistributed from the Remote PE to the CE LSA Type 5 External routes (which causes the router to bypass the down-bit check).

I was cool with that logic, until I went to go do 14.8 (Sham-Link) Lab and I noticed that when I made the domain-ids on the PEs the same, removed the VRF-Lite configs off of SW1 to make it a pure CE and configured the OSPF process globally, I noticed that all of the routes that used to be E2 are now IA routes.

My question is, why is it that when SW1 became a pure CE, all the routes learned from the PE became IA routes opposed to when it was running OSPF 100 VRF VPN_A, all the routes were E2? (scratching head)

I have played around with the lab quite a bit, making the domain-ids the same, verifying that the down-bit was set and watching the result, which was all of the OSPF E2 routes disappearing from the table. Now that I am seeing the IA routes in the table (with the same domain-id), i see that down-bit is set, but the routes are still in the table...

I have looked through a few books in regards to this topic, but none of them go too in-depth with it... 

 

Thanks in advance for any help with this!

 

 

 

Comments

  • My question is, why is it that when SW1 became a pure CE, all the routes learned from the PE became IA routes opposed to when it was running OSPF 100 VRF VPN_A, all the routes were E2? (scratching head)

    I'm not going answer this question directly but here goes with what the domain id does.

    When you configure OSPF as you PE-CE protocol on multiple sites - effectively the MPLS cloud is imagined as a super backbone - which is above any area within your connected OPSF areas.

    Simply put if each PE running OPSF has the same process number - then OPSF routes redistributed on another PE will show the OPSF routes as IA.  The reason is the domain-id is taken from the OSPF process-id in the bgp extended community attributes.

    If one of your PE routers uses a different OSPF process id - the routers will be seen as these routes as E2.

    To get arround this you can set the domain-id so that this is used within the BGP extended community attributes.

  • Thank you for your response Welshy.

    In regards to what you said, I understand what the domain-id does and why when the domin-ids are the same, the routes redistributed to the CE become IA routes.

    What I dont understand is why when SW1 was setup with the OSPF 100 vrf VPN_A process and I set the domain-ids the same on the PEs, the routes on SW1 were E2 and not IA.

    Does it have something to do with SW not being a pure CE and using VRF-Lite?

    This is what I am trying to understand...

  • What I dont understand is why when SW1 was setup with the OSPF 100 vrf VPN_A process and I set the domain-ids the same on the PEs, the routes on SW1 were E2 and not IA.

    This does suprise me both the PE and your switch would have to have an interface in the same area - did you look at OSPF on your PE?

     

  • I will re-do this lab today and and see if I can find anything that is configured wrong on the PE.

    Thanks...

Sign In or Register to comment.