BGP Flap on a 100M Metro Ethernet Circuit

routeinet · ‎02-23-2009

Hello All,

We've 100M Metro Ethernet circuit connecting MPLS CER-PER Links.

We're seeing random BGP flap between CER & PER from the CER side.

Each BGP flap is causing anywhere from 30 seconds to 3 minutes of downtime.

Question #1: Could CER-PER bgp timer (timers bgp 15 45) mistach cause the BGP to flap ?

Questions #2: What are Cisco's recommdenation to monitor Metro Ethernet circuit(Since it's rare that we'll see the GE or FE go down on a Metro ethernet circuit) ?

Feb 23 04:12:30: %BGP-5-ADJCHANGE: neighbor 6x.1xx.1xx.121 Down BGP Notification sent

Feb 23 04:12:30: %BGP-3-NOTIFICATION: sent to neighbor 6x.1xx.1xx.121 4/0 (hold time expired) 0 bytes

Feb 23 04:13:09: %BGP-5-ADJCHANGE: neighbor 6x.1xx.1xx.121 Up

Feb 23 05:15:08: %BGP-5-ADJCHANGE: neighbor 6x.1xx.1xx.121 Down BGP Notification sent

Feb 23 05:15:08: %BGP-3-NOTIFICATION: sent to neighbor 6x.1xx.1xx.121 4/0 (hold time expired) 0 bytes

Feb 23 05:15:52: %BGP-5-ADJCHANGE: neighbor 6x.1xx.1xx.121 Up

Feb 23 10:37:13: %BGP-5-ADJCHANGE: neighbor 6x.1xx.1xx.121 Down BGP Notification sent

Feb 23 10:37:13: %BGP-3-NOTIFICATION: sent to neighbor 6x.1xx.1xx.121 4/0 (hold time expired) 0 bytes

Feb 23 10:37:51: %BGP-5-ADJCHANGE: neighbor 6x.1xx.1xx.121 Up

Feb 23 11:13:05: %BGP-5-ADJCHANGE: neighbor 6x.1xx.1xx.121 Down BGP Notification sent

Feb 23 11:13:05: %BGP-3-NOTIFICATION: sent to neighbor 6x.1xx.1xx.121 4/0 (hold time expired) 0 bytes

Feb 23 11:16:13: %BGP-5-ADJCHANGE: neighbor 6x.1xx.1xx.121 Up

royalblues · ‎02-24-2009

The timers wont cause the BGP flaps as the lowest timer would be negotiated and be used between the BGP neighbors

Seeing the timing betwen the flaps, it more seems to be a case of a cirucit nmisbehaving. It could also be due to congestion in the circuits or high CPU utilization on the devices

Narayan

routeinet · ‎02-24-2009

Thank you very much for your response.

I can rule out congestion or high CPU utilization because Load on the circuit and the CPU utilization is minimal,

txload 3/255, rxload 3/255 and less than 1% CPU utilization at the time of BGP flap.

Kerem Gursu · ‎02-24-2009

Hi James,

Having a debug bgp output will be very useful. Do you have a chance to use this debug and share the output?

Regards,

Kerem