OSPF problem in a GSR12000 with MPLS interfaces

[simhon_doctori]

Hi,

I am using an MPLS interface connected to other vendor routers.

OSPF is set on the MPLS interface.

The OSPF did not synched with the other vendor OSPF due to the following :

00:45:55: OSPF: Rcv DBD from 10.10.55.113 on GigabitEthernet4/0 seq 0x178476E opt 0x42 flag 0x7 len 32 mtu 1500 state EXSTART

00:45:55: OSPF: First DBD and we are not SLAVE

00:46:00: OSPF: Retransmitting DBD to 10.10.55.113 on GigabitEthernet4/0

00:46:00: OSPF: Up DBD Retransmit cnt to 1 for 10.10.55.113 on GigabitEthernet4/0

00:46:00: OSPF: Send DBD to 10.10.55.113 on GigabitEthernet4/0 seq 0x745 opt 0x52 flag 0x7 len 32

00:46:00: OSPF: Rcv DBD from 10.10.55.113 on GigabitEthernet4/0 seq 0x178476E opt 0x42 flag 0x7 len 32 mtu 1500 state EXSTART

00:46:00: OSPF: First DBD and we are not SLAVE

After diagnosing the problem we found out that if the loopback router-id address value is bigger then the neighbores IP address then you will get this problem.

If you will change the router-id ip address for 1.1.1.1 and the MTU is set ok (to be 1500 on the MPLs i/fs) then it will work well !

Regards,

Simhon Doctori.

[Harold Ritter]

The issue seems like it is MTU related. What is the MTU size used by the other side? I don't see anything wrong in the log you provided. Bear in mind that only the router with the lower MTU will complain about its peer having a higher MTU.

Hope this helps,

[simhon_doctori]

Hi,

I am aware of the MTU problem since I saw it happening once, the MTU I use between the interfaces is 1500 and seems to be OK.

The problem is in the OSPF. The RFC explains that the two peers need to select Master and a Slave based on the IP address of both. I think the GSR assumes he is the Master among the two although he should be the Slave.

When I changed the router-id, or loopback address to be the minimum possible, i.e. 1.1.1.1 the problem fixed and the OSPF between the two peers established well.

Regards,

Simhon Doctori

[Harold Ritter]

What version of IOS are you running on the GSR? Who is the other vendor? Does the adjacency stay in EXSTART. What are the RID of the two routers when you have the issue?

Thanks,

[simhon_doctori]

Oh, and another thing -

The RID of the GSR was 12.12.12.12 = 202116108 (Dec)

and the other vendor was 10.10.55.113 = 168441713 (Dec),

In this case the RID of the GSR was the greater, when I changed it to be 5.5.5.5 it had the smaller value, then the OSPF between the two synched.

Thanks,

Simhon Doctori

[Harold Ritter]

The GSR is bahaving properly (ie: declares itself as master when its RID is higher and slave when its RID is lower). As I already mentioned, changing the RID on the GSR is a workaround in this case.

It definitely points toward an issue with the other vendor implementation.

Hope this helps,

[Harold Ritter]

By the way, I had overlooked one detail. It is completely normal for the GSR to declare itself the Master as you stated that it had the higher RID.

In my view there seem to be something wrong on the other box. Setting the RID to a lower value on the GSR is just a workaround.

Any way you could provide the equivalent of a "show ip ospf adj" from the other router.

Thanks,

[simhon_doctori]

Hi Again,

I am using Version 12.0(26)S1 of cisco 12008/GRP.

The other vendor box was toggling between INIT to EXSTART.

I will check if the problem may be on the other vendor equip.

Thanks for your help so far.

Simhon Doctori

[dknov]

Hi,

Sorry for barging in on this. Can you please tell me how this Master/Slave relationship comes into play in conjunction with DR. DR is a focal point for distributing LSD and propagating all the updates on broadcast or NBMA networks.

a. What if DR has a RID of 1.1.1.1 which is lower

than any other RID on a segment?

b. Will it be chosen as Master anyhow?

c. What if a topology change occurs on a Slave

router, it is allowed to update the Master?

Is there the same Master/Slave notion in other routing protocols, specifically ISIS, EIGRP and BGP?

Thank you very much.

David