BGP Dampening refusing to work.. What have I missed?

[graham smart]

Hi Guys,

Im trying to minimise the problems caused when a neighbor flaps.

Im trying to use dampening.

I have the below config to dampen any route..

ip prefix-list NETWORK_FLAPPING seq 5 permit 0.0.0.0/0 le 32

route-map NETWORK_DAMPENING permit 10

match ip address prefix-list NETWORK_FLAPPING

set dampening 15 750 3000 60

router bgp (ourAS)

bgp dampening route-map NETWORK_DAMPENING

#sh ip bgp dampening parameters

dampening 15 750 3000 60 (route-map NETWORK_DAMPENING 10)

  Half-life time      : 15 mins       Decay Time       : 1800 secs

  Max suppress penalty: 12000         Max suppress time: 60 mins

  Suppress penalty    :  3000         Reuse penalty    : 750

Right.. Now i flap the neighbour remotely..

Jun 13 12:21:19.822 BST: %BGP-5-ADJCHANGE: neighbor xx.xx.xx.xx Down Peer closed the session

Jun 13 12:21:30.642 BST: %BGP-5-ADJCHANGE: neighbor xx.xx.xx.xx Up

Jun 13 12:21:36.262 BST: %BGP-5-ADJCHANGE: neighbor xx.xx.xx.xx Down Peer closed the session

Jun 13 12:21:40.010 BST: %BGP-5-ADJCHANGE: neighbor xx.xx.xx.xx Up

Jun 13 12:21:46.695 BST: %BGP-5-ADJCHANGE: neighbor xx.xx.xx.xx Down Peer closed the session

Jun 13 12:21:50.511 BST: %BGP-5-ADJCHANGE: neighbor xx.xx.xx.xx Up

host#sh ip bgp dampening flap-statistics

host#

Nothing appears in flaps? nothing appears in dampened pathes either..

Does anyone have any idea what im missing here? ( Fyi, Theres plenty of routes being advertised by the bgp peer, So something should be dampened... surely? )

Thanks

-G

[Giuseppe Larosa]

Hello Graham,

dampening looks for change of state of specific route that is a cycle of advertisements and withdrawns.

you are making the whole BGP session to flap that is a different thing.

In your case the peer sends a BGP Notify message to close the session it is not sending BGP updates with the withdrawn section filled with specific IP prefixes.

A withdrawn is an update that advertises that a specific prefix is not available anymore so it should be removed from BGP table.

This is the reason why you don't see BGP dampening in action in you tests

You should for example shut and unshut a loopback interface on BGP peer to see the associated IP prefix dampened on local node.

Hope to help

Giuseppe

[graham smart]

Thanks for the information.

That makes more sense.

Could you think of a better way to protect the router from flapping peers?

Sometimes we can have a lot flapping and it spikes the CPU something silly. Would be nice to have a way to hold a peer down for a bit if it flaps too much.

But thanks for the information.

[matarasov]

Hi Graham,

Were you able to figure out how to protect from flapping peers?

We got similar scenario, where our neighbor flapped few times.

Thanks,

[cadet alain]

Hi,

if these are eBGP peers peering on directly connected links then disabling bgp fast-external-fallover would make the neighbour declared down after 180 sec by default.

interface x/x ---> going to flapping peer

ip bgp fast-external-fallover deny

Regards.

Alain

Don't forget to rate helpful posts.

[AJAZ NAWAZ]

I know this was a long time ago, but BFD dampening seems like it would be a better fit?. Perhaps that feature was not there at that time, but even now its only available in IOS-XE anyways

hth