I've come to the conclusion lately that what is documented isn't
always how it really should be - especially in cases involving MPLS.

Ho hum. Good to have these experiences before the lab than during.

On Thu, Jul 30, 2009 at 12:41 PM, Antonio Soares<[email protected]> wrote:
> Hello Con,
>
> I found that the example shown in the Implementing Cisco MPLS Student Guide 
> is using the network command.
>
> Funny, isn't it ?
>
>
>
> Regards,
>
> Antonio Soares, CCIE #18473 (R&S)
> [email protected]
>
> -----Original Message-----
> From: Con Spathas [mailto:[email protected]]
> Sent: quarta-feira, 29 de Julho de 2009 11:27
> To: Antonio Soares
> Cc: [email protected]
> Subject: Re: [OSL | CCIE_SP] VOL2 - Section 1 - Task 8.2
>
> Gday Antonio,
>
> Using the network statement under BGP as opposed to redistribute connected 
> did indeed result in the same issue you described.
>
> I can see how it would be expected that either method should work however 
> given your experience and a somewhat similar one I had as
> I explained in an earlier email has caused me to think of it like this.
>
> When we use the network statement in BGP and then redist BGP into OSPF, the 
> network advertised is a candidate for redistribution.
>
> If we use the redist connected, then we have explicitly defined this and 
> broken any implied redistribution that could occur.
>
> For example if we have say BGP running (like we do) and redist connected 
> loopback into BGP. Then we decide to redist BGP into OSPF -
> in order to get this loopback into OSPF, we need to explicitly redistribute 
> it into OSPF as well, otherwise it won't appear. This is
> my understanding anyhow. I think it is this functionality that protects the 
> sham link end points under normal conditions.
>
> Anyhow I'm glad we were able to re-create the problem nonetheless.
>
> Cheers,
> Con.
>
> On Tue, Jul 28, 2009 at 10:58 PM, Antonio Soares<[email protected]> wrote:
>> Hello Con,
>>
>> I finally know what it is. When i made this lab i used the network
>> command instead of the redistribute connected under BGP in order to
>> announce the sham-link. This causes a problem because both routers
>> will inject the sham-link end point into the OSPF DB. When i noticed the 
>> problem, i changed from network command to redistribute
> connected and the problem was still there. Now i know this is because the 
> duplicate entries remained in R3 and R9 OSPF databases and
> this information was sent again to R1 and R8.
>>
>> The documentation is not very clear about the way the sham-link should be 
>> advertised:
>>
>> http://www.cisco.com/en/US/partner/docs/ios/iproute/configuration/guid
>> e/irp_ospf_sham_link.html
>>
>> I always thought that using the network command or the redistribute
>> command would give us the same results. It seems this is not true. This can 
>> be a corner case since we have an Inter-AS scenario
> with RR's.
>>
>> Well, at least i learned that the domain-tag command can be very helpful in 
>> a situation like this.
>>
>> I'm sorry for the confusion i may have created here.
>>
>> It would be nice to see the problem replicated in your side :)
>>
>>
>> Regards,
>>
>> Antonio Soares, CCIE #18473 (R&S)
>> [email protected]
>>
>> -----Original Message-----
>> From: Con Spathas [mailto:[email protected]]
>> Sent: terça-feira, 28 de Julho de 2009 21:38
>> To: Antonio Soares
>> Cc: [email protected]
>> Subject: Re: [OSL | CCIE_SP] VOL2 - Section 1 - Task 8.2
>>
>> Gday Antonio.
>>
>> This is truly strange.
>>
>> I copied your configs verbatim into my routers and I still can't manage to 
>> re-create the problem.
>>
>> R1#sh ip ospf neighbor 8.8.8.8 | inc Neigh|up  Neighbor 8.8.8.8,
>> interface address 8.8.8.8
>>    Neighbor priority is 0, State is FULL, 6 state changes
>>    Neighbor is up for 00:07:51
>>
>> The Sham Link is stable.
>>
>> I'm keen to get to the bottom of it in case it happens in a lab!
>>
>> Let's take this offline and report back to the group if we figure it out?
>>
>> Con.
>>
>> On Tue, Jul 28, 2009 at 7:02 PM, Antonio Soares<[email protected]> wrote:
>>> Hello Con,
>>>
>>> I was able to reproduce the problem:
>>>
>>> ++++++++++++++++++++++++++++
>>> R1#sh ip route vrf VPNB 8.8.8.8
>>> Routing entry for 8.8.8.8/32
>>>  Known via "bgp 125", distance 200, metric 0
>>>  Tag 678, type internal
>>>  Redistributing via ospf 39
>>>  Advertised by ospf 39 subnets
>>>  Last update from 6.7.8.8 00:00:23 ago
>>>  Routing Descriptor Blocks:
>>>  * 6.7.8.8 (Default-IP-Routing-Table), from 125.125.125.2, 00:00:23
>>> ago
>>>      Route metric is 0, traffic share count is 1
>>>      AS Hops 1
>>>      Route tag 678
>>>      MPLS Required
>>>
>>> R1#
>>> 01:04:03: %OSPF-5-ADJCHG: Process 39, Nbr 8.8.8.8 on OSPF_SL0 from
>>> FULL to DOWN, Neighbor Down: Interface down or detached R1# R1#sh ip
>>> route vrf VPNB 8.8.8.8 Routing entry for 8.8.8.8/32
>>>  Known via "ospf 39", distance 110, metric 1
>>>  Tag Complete, Path Length == 1, AS 678, , type extern 2, forward
>>> metric 1
>>>  Redistributing via bgp 125
>>>  Last update from 6.7.8.8 00:00:05 ago
>>>  Routing Descriptor Blocks:
>>>  * 6.7.8.8 (Default-IP-Routing-Table), from 8.8.8.8, 00:00:05 ago
>>>      Route metric is 1, traffic share count is 1
>>>      Route tag 3489661606
>>>
>>> R1#
>>> R1#
>>> 01:04:16: %OSPF-5-ADJCHG: Process 39, Nbr 8.8.8.8 on OSPF_SL0 from
>>> LOADING to FULL, Loading Done R1# R1#sh ip route vrf VPNB 8.8.8.8
>>> Routing entry for 8.8.8.8/32
>>>  Known via "bgp 125", distance 200, metric 0
>>>  Tag 678, type internal
>>>  Redistributing via ospf 39
>>>  Advertised by ospf 39 subnets
>>>  Last update from 6.7.8.8 00:00:11 ago
>>>  Routing Descriptor Blocks:
>>>  * 6.7.8.8 (Default-IP-Routing-Table), from 125.125.125.2, 00:00:11
>>> ago
>>>      Route metric is 0, traffic share count is 1
>>>      AS Hops 1
>>>      Route tag 678
>>>      MPLS Required
>>>
>>> R1#
>>> ++++++++++++++++++++++++++++
>>>
>>> The link between R3 and R9 is shutdown.
>>>
>>> I am using 12.2(25)S13 in all routers in this lab.
>>>
>>> The relevant configs:
>>>
>>> R1:
>>> !
>>> interface Loopback1000
>>>  ip vrf forwarding VPNB
>>>  ip address 1.1.1.1 255.255.255.255
>>>  no clns route-cache
>>> !
>>> router ospf 39 vrf VPNB
>>>  log-adjacency-changes
>>>  area 0 sham-link 1.1.1.1 8.8.8.8
>>>  redistribute bgp 125 subnets
>>>  network 31.3.1.1 0.0.0.0 area 0
>>> !
>>> router bgp 125
>>>  bgp router-id 125.125.125.1
>>>  bgp always-compare-med
>>>  no bgp default ipv4-unicast
>>>  bgp log-neighbor-changes
>>>  neighbor 125.8.1.8 remote-as 678
>>>  neighbor 125.125.125.2 remote-as 125
>>>  neighbor 125.125.125.2 update-source Loopback0
>>>  neighbor 125.125.125.5 remote-as 125
>>>  neighbor 125.125.125.5 update-source Loopback0
>>>  !
>>>  address-family ipv4 vrf VPNB
>>>  redistribute connected
>>>  redistribute ospf 39 vrf VPNB match internal external 1 external 2
>>> route-map ospf2bgp
>>>  no auto-summary
>>>  no synchronization
>>>  exit-address-family
>>> !
>>> route-map ospf2bgp permit 10
>>>  set origin egp 1
>>> !
>>>
>>> R8:
>>> !
>>> interface Loopback1000
>>>  ip vrf forwarding VPNB
>>>  ip address 8.8.8.8 255.255.255.255
>>>  no clns route-cache
>>> !
>>> !
>>> router ospf 39 vrf VPNB
>>>  log-adjacency-changes
>>>  area 0 sham-link 8.8.8.8 1.1.1.1
>>>  redistribute bgp 678 subnets
>>>  network 98.9.8.8 0.0.0.0 area 0
>>> !
>>> router bgp 678
>>>  bgp router-id 6.7.8.8
>>>  bgp always-compare-med
>>>  no bgp default ipv4-unicast
>>>  bgp log-neighbor-changes
>>>  neighbor as678 peer-group
>>>  neighbor as678 remote-as 678
>>>  neighbor as678 password ipexpert
>>>  neighbor as678 update-source Loopback0
>>>  neighbor 6.7.8.6 peer-group as678
>>>  neighbor 6.7.8.7 peer-group as678
>>>  neighbor 125.8.1.1 remote-as 125
>>> !
>>>  address-family ipv4 vrf VPNB
>>>  redistribute connected
>>>  redistribute ospf 39 vrf VPNB match internal external 1 external 2
>>>  no auto-summary
>>>  no synchronization
>>>  exit-address-family
>>> !
>>>
>>> I don't remember very well but i think i also loaded R1 and R8 with 12.0S 
>>> and 12.4 and got the same results.
>>>
>>> And i'm using the same process id in R1 and R8. But even with different 
>>> process ids, the problem still happens.
>>>
>>>
>>> Regards,
>>>
>>> Antonio Soares, CCIE #18473 (R&S)
>>> [email protected]
>>>
>>> -----Original Message-----
>>> From: Con Spathas [mailto:[email protected]]
>>> Sent: domingo, 26 de Julho de 2009 9:29
>>> To: Antonio Soares; [email protected]
>>> Subject: Re: [OSL | CCIE_SP] VOL2 - Section 1 - Task 8.2
>>>
>>> Gday Antonio,
>>>
>>> I've tried to re-create the issue you described below with no luck:
>>>
>>> 10.10.18.1/32 is the R1 loopback interface for the Sham
>>> 10.10.18.8/32 is the R8 loopback interface for the Sham
>>>
>>> +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>> R1#sh ip ospf neighbor 98.9.8.8 | inc Neighbor|interface
>>>  Neighbor 98.9.8.8, interface address 10.10.18.8
>>>    In the area 0 via interface OSPF_SL0
>>>    Neighbor priority is 0, State is FULL, 6 state changes
>>>    Neighbor is up for 00:21:00
>>> +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>> R1#sh ip ospf database | beg Type-5
>>>                Type-5 AS External Link States
>>>
>>> Link ID         ADV Router      Age         Seq#       Checksum Tag
>>> 4.0.0.4         31.3.1.1        1325        0x80000001 0x009E9A
>>> 3489661053
>>> 10.10.18.1      98.9.8.8        1282        0x80000001 0x004D4A
>>> 3489661606
>>> 10.10.18.8      31.3.1.1        1279        0x80000001 0x00E82A
>>> 3489661053
>>> 11.0.0.11       31.3.1.1        1279        0x80000001 0x00FC2E
>>> 3489661053 54.5.4.0        31.3.1.1        1325        0x80000001
>>> 0x00D130 3489661053
>>> +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>> R1#sh ip route vrf VPNB 10.10.18.8
>>> Routing entry for 10.10.18.8/32
>>>  Known via "bgp 125", distance 200, metric 0
>>>  Tag 678, type internal
>>>  Redistributing via ospf 125
>>>  Advertised by ospf 125 subnets
>>>  Last update from 6.7.8.8 00:21:34 ago
>>>  Routing Descriptor Blocks:
>>>  * 6.7.8.8 (Default-IP-Routing-Table), from 125.125.125.2, 00:21:34
>>> ago
>>>      Route metric is 0, traffic share count is 1
>>>      AS Hops 1
>>>      Route tag 678
>>>
>>> +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>> R8#
>>> R8#sh ip ospf neighbor 31.3.1.1 | inc Neigh|inter
>>>  Neighbor 31.3.1.1, interface address 10.10.18.1
>>>    In the area 0 via interface OSPF_SL0
>>>    Neighbor priority is 0, State is FULL, 6 state changes
>>>    Neighbor is up for 00:22:02
>>> +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>> R8#sh ip ospf data | beg Type-5
>>>                Type-5 AS External Link States
>>>
>>> Link ID         ADV Router      Age         Seq#       Checksum Tag
>>> 4.0.0.4         31.3.1.1        1386        0x80000001 0x009E9A
>>> 3489661053
>>> 10.10.18.1      98.9.8.8        1338        0x80000001 0x004D4A
>>> 3489661606
>>> 10.10.18.8      31.3.1.1        1341        0x80000001 0x00E82A
>>> 3489661053
>>> 11.0.0.11       31.3.1.1        1     (DNA) 0x80000001 0x00FC2E
>>> 3489661053 54.5.4.0        31.3.1.1        1386        0x80000001
>>> 0x00D130 3489661053
>>> +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>> R8#sh ip route vrf VPNB 10.10.18.1
>>> Routing entry for 10.10.18.1/32
>>>  Known via "bgp 678", distance 200, metric 0
>>>  Tag 125, type internal
>>>  Redistributing via ospf 678
>>>  Advertised by ospf 678 subnets
>>>  Last update from 125.125.125.1 00:22:34 ago
>>>  Routing Descriptor Blocks:
>>>  * 125.125.125.1 (Default-IP-Routing-Table), from 6.7.8.7, 00:22:34
>>> ago
>>>      Route metric is 0, traffic share count is 1
>>>      AS Hops 1
>>>      Route tag 125
>>> +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>> As you can see the Sham-Link is stable, however I have not
>>> manipulated the Domain-Tag as shown in the OSPF database.
>>>
>>> Tag 3489661606 = 11010000000000000000001010100110 1010100110 = AS 678
>>>
>>> Tag 3489661053 = 11010000000000000000000001111101
>>> 1111101 = AS 125
>>>
>>> I tried this with 12.0S and 12.4Mainline code.
>>>
>>> Cheers,
>>> Con.
>>>
>>>
>>>>I found the solution to this problem !
>>>>
>>>>When we have a sham-link inside the same AS, there's no issues with
>>>>routing loops with the sham-links because the external LSA will have
>>>>the same route-tag.
>>>>
>>>>When we have a sham-link between two AS's, and the sham-links are
>>>>advertised by eBGP, there's no problem because the eBGP AD is lower
>>>>than OSPF.
>>>>
>>>>But when we have a sham-link between two AS's and the sham-links are
>>>>advertised by iBGP, there's a routing loop. The routers will prefer
>>>>the OSPF learned route instead of the iBGP because of lower AD.
>>>>
>>>>So the solution is to use the same "domain-tag" under the OSPF
>>>>process in both PE's.
>>>>
>>>>
>>>>
>>>>
>>>>Regards,
>>>>
>>>>Antonio Soares, CCIE #18473 (R&S)
>>>>[email protected]
>>>>
>>>>-----Original Message-----
>>>>From: [email protected]
>>>>[mailto:[email protected]] On Behalf Of Antonio
>>>>Soares
>>>>Sent: segunda-feira, 8 de Junho de 2009 19:08
>>>>To: [email protected]
>>>>Subject: [OSL | CCIE_SP] VOL2 - Section 1 - Task 8.2
>>>>
>>>>My sham-link is flaping. As soon as the sham-link comes up, R1 and R8
>>>>start prefering the OSPF route instead of the iBGP route. I never saw
>>>>this problem in regular MPLS VPNs inside one AS. In this because we
>>>>have an Inter-AS MPLS VPN scenario ?
>>>>
>>>>+++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>>>R1#
>>>>00:03:31: %OSPF-5-ADJCHG: Process 39, Nbr 8.8.8.8 on OSPF_SL0 from
>>>>LOADING to FULL, Loading Done R1# R1#sh ip route vrf VPNB | inc
>>>>8.8.8.8 B       8.8.8.8 [200/0] via 6.7.8.8, 00:00:14 R1# R1#
>>>>00:03:47: %OSPF-5-ADJCHG: Process 39, Nbr 8.8.8.8 on OSPF_SL0 from
>>>>FULL to DOWN, Neighbor Down: Interface down or detached R1# R1#sh ip
>>>>route vrf VPNB | inc 8.8.8.8 O E2    8.8.8.8 [110/1] via 6.7.8.8,
>>>>00:00:00 R1#
>>>>+++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>>>R8#
>>>>00:06:28: %OSPF-5-ADJCHG: Process 39, Nbr 1.1.1.1 on OSPF_SL0 from
>>>>LOADING to FULL, Loading Done R8# R8#sh ip route vrf VPNB | inc
>>>>1.1.1.1 B       1.1.1.1 [200/0] via 125.125.125.1, 00:00:13 R8# R8#
>>>>00:06:44: %OSPF-5-ADJCHG: Process 39, Nbr 1.1.1.1 on OSPF_SL0 from
>>>>FULL to DOWN, Neighbor Down: Interface down or detached R8# R8#sh ip
>>>>route vrf VPNB | inc 1.1.1.1 O E2    1.1.1.1 [110/1] via
>>>>125.125.125.1, 00:00:02 R8#
>>>>+++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>>>
>>>>After a few hours trying to understand why this was happening, i was
>>>>able to make it work tweaking the OSPF AD for the External routes in
>>>>R1 and R8.
>>>>
>>>>Anyone saw this problem in this lab ?
>>>>
>>>>And why in the PG we don't see the sham-link interfaces in R3 and R9 ?
>>>>
>>>>
>>>>Regards,
>>>>
>>>>Antonio Soares, CCIE #18473 (R&S)
>>>>[email protected]
>>>>
>>>
>>>
>>>
>>
>>
>>
>
>
>
_______________________________________________
For more information regarding industry leading CCIE Lab training, please visit 
www.ipexpert.com

Reply via email to