[c-nsp] OSPF inside VRF - Cisco Juniper Interoperability

[Junaid]

Hi,

Just want to share my findings:

The problem as suspected was a bug on Cisco side - CSCsg42488 as
pointed out by Oliver Boehmer. The work around employed was to use the
knob: "route-type-community vendor" for the OSPF instance inside the
VRF on Juniper PE.

Thanks once again Oliver for the solution. Now CE1 is also getting
Type 3 LSAs from CE2.


Regards,
Junaid

On Wed, Aug 27, 2008 at 3:12 PM, Junaid <junaid.x86 at gmail.com> wrote:
> Hi,
>
> I am caught up in what seems to be a Juniper Cisco interoperability
> issue. I am running OSPF with customer inside VRF. Topology is
> something like the following:
>
> CE1 ---[Area 0]--- PE1 ---- P1 --- P2 --- PE2 ---[Area 6]--- CE2
>
> The two P routers are acting as route reflectors.
>
> CE1, CE2 and PE1 are Cisco devices while rest are Juniper M-series
> routers. The problem I am facing is that CE1 routes received at CE2
> are Inter-area which is what is required (no redistribution into OSPF
> is done on CE1 and CE2). However, CE2 routes received by CE1 are Type
> 5 (E1). The documentation states that inorder to preserve the route
> types, domain IDs should be same on both PE routers. I have set domain
> ID to be 1.1.1.1:512, this was done on cisco via the command:
> "domain-id type 0105 value 010101010200" and on juniper as: "domain-id
> 1.1.1.1:512" in the OSPF configuration inside the VRF. Also on Juniper
> the domain-id was added into the ospf routes when redistributing them
> into MBGP.
>
> The problem seems to be with the Cisco PE1 router that can't seem to
> interpret the route-type attribute generated by Juniper (seen in the
> output as 0x306:0:393472):
>
> PE1#sh ip bgp vpnv4 all 10.254.20.254
> BGP routing table entry for 1:103:10.254.20.254/32, version 550
> Paths: (1 available, best #1, table VPN_OSPF)
>  Not advertised to any peer
>  Local
>   <PE2_Loopback_IP> (metric 4) from <P1_Loopback_IP> (<P1_Loopback_IP>)
>     Origin IGP, metric 2, localpref 100, valid, internal, best
>     Extended Community: RT:1:103 OSPF DOMAIN
> ID:0x0105:0x010101010200 0x306:0:393472
>
> 10.254.20.254/32 is advertised by CE2 (assigned on one of its loopback
> interfaces). Now the domain ID is fine but it seems that Cisco is
> unable to interpret the route-type attribute. 393472 translates to
> 60100 where 6 is the area ID, 01 says that it is type 1 LSA and and
> last two bytes are options are not used in this case. Upon receiving
> this route via MBPG, PE1 injects a type 5 LSA towards CE1 (confirmed
> on CE1 by enabling debugging) where it should inject have injected
> type 3:
>
> OSPF: Ack Type 5, LSID 10.254.20.254, Adv rtr 10.254.1.1, age 5, seq 0x80000001
>
>
> If I replace the Juniper PE2 with a Cisco then on PE1 seems to
> interpret the route-type attribute correctly (OSPF RT:0.0.0.6:2:0) and
> inject type 3 LSA
> towards CE1 and CE1 receive the routes as inter-area:
>
> PE1#sh ip bgp vpnv4 all 10.254.20.254
> BGP routing table entry for 1:103:10.254.20.254/32, version 676
> Paths: (1 available, best #1, table VPN_OSPF)
>  Not advertised to any peer
>  Local
>   <PE2_Loopback_IP> (metric 2) from <P1_Loopback_IP> (<P1_Loopback_IP>)
>     Origin incomplete, metric 11112, localpref 100, valid, internal, best
>     Extended Community: RT:1:103 OSPF DOMAIN
> ID:0x0005:0x010101010200 OSPF RT:0.0.0.6:2:0 OSPF ROUTER
> ID:10.254.2.1:512
>
>
> Debug output:
>
> OSPF: Ack Type 3, LSID 10.254.20.254, Adv rtr 10.254.1.1, age 1, seq 0x80000001
>
> Any idea what is causing this behavior? Any solution? Will appreciate any help.
>
>
> Regards,
> Junaid
>