BGP and OSPF errors

[transnexaec]

Good morning!

I have a Router Core working with OSPF (5 routers) and BGP (2 routers). Sometimes (4 times at week), the Core have a problem, it loses the BGP connection due an OSPF Error (I am not sure this diagnostic). These are the errors in the folowing order:

%BGP-5-ADJCHANGE: neighbor X.X.X.X Down BGP Notification sent

%BGP-3-NOTIFICATION: sent to neighbor Y.Y.Y.Y 4/0 (hold time expired) 0 bytes

And then:

%OSPF-4-NONEIGHBOR: Received database description from unknown neighbor Z.Z.Z.Z

Is this normal? Can OSPF problems affect BGP? Than you for your help.

[mheusinger]

Hello,

BGP uses TCP to connect to the neighbor IP. So any problem in OSPF can also lead to a loss of a BGP neighbor. This is just because you loose IP connectivity (or it is degraded with too many errors).

The interesting part is the OSPF message. Are there any further messages about loosing adjacency? Is the IP address Z.Z.Z.Z a valid OSPF router in your network (check with sh ip ospf neighbor and/or with show ip protocol)?

If not: consider to set an area password and also configure the OSPF router-id for every process.

Regards, Martin

[transnexaec]

Thank you for your Help Martin. About your questions:

1.- Are there any further messages about loosing adjacency? I coul not see them becasuse this router is an Access Server and the most part of log messages are about the Async Interfase. Can the continuous Up/Down state of these interfases affect the OSPF process? Last week, this OSPF problem have produced from another router P.P.P.P

2.- Is the IP address Z.Z.Z.Z a valid OSPF router in your network? Yes, this IP address and ots loopback are part of my network.

Note: All the routers have not more tha a 60% of CPU utilization. All the links between router are OK. I have attached a diagram.

[ilya.varlashkin]

Properly functioning (speaking of implementation, not configuration) OSPF router will send database description packet only after it heard hello from a router listing that (Z.Z.Z.Z) router as neighbor. So looks like Z.Z.Z.Z is valid neighbor at least for DR on that segment, but the router on which you observed the message somehow didn't hear hello from that router. This could be due to link errors or congestions or overloaded router.

BGP went down either because Z.Z.Z.Z was next-hop to reach that X.X.X.X or because the router on which you observed mesages has same link, congestion or CPU overload as above.

So, both are true: OSPF can cause BGP problems, and also it could be indication that your router or link has problems.

[transnexaec]

Thank you for your help Ilya. All the routers have not more than 60% of CPU utilization. All the links between router are OK. I think is something about de OSPF proccess.

[ilya.varlashkin]

Looking at your diagram I could only assume the router where you observed the problem is the one that has two ethernet and one ATM interfaces (lower right corner of your BGP-speaking area). If typical CPU load of this router is 60%, then it's well possible that it gets many BGP or OSPF updates and performing re-calculations. During this time load can easily go upto 99% and then it starts melting down - not replying to keepalives, loosing packets in general. Try setting running continious ping (for a day or so) towards this router. You could do this by setting RTR/IP SLA probes on router Z.Z.Z.Z and X.X.X.X.

Since it's this router that missed keepalive from BGP peer, and at the same time looks like this router missed hello from an OSPF neighbor , while neighbor actually heard hello (though not necessary from this router if this router isn't DR or BDR); then all speaks for watching carefully for CPU load on this router.

OSPF implementation on Cisco routers is mature and behaves according to requirements, unless you have some rogue router (unlikely from your description), I'd think it's CPU utilisation problem on the router where you observed this message.