Supress flapping BGP neighbors

[lionell01]

Hello!

We have about 4 eBGP neighborships with our customer devices and the links are often highly unstable (international submarine cable issues and the like).

This causes the BGP neighborships to flap. 

- We're using BFD 

- All neighbors advertise the same routes to us and we to them

- Local preference is used to set an order for outbound routing

- AS path prepend is used to influence inbound routing

- If a preferred neighbor goes down, the traffic fails overs to the next but then falls back to the preferred neighbor once it's back up

- This causes issues especially during multiple link flaps and we also have voice traffic on this

Is there a way to force the flapping BGP neighborship to stay down until the link stabilizes (maybe as measured by ICMP)?

Route-dampening won't work here I think as all neighbors are advertising the same routes.

Had read about EEM but that needs manual intervention to get the neighbor backup. Looking for something more automated.

[paul driver]

Hello

---

@lionell01 wrote:

All 4 of the neighbors are in the same AS and advertise the same routes with the exact same attributes.

Would dampening still be a solution?

---

How aggressive do you have BFD, BGP timers, Do you have fast-external failover active? 

[lionell01]

Thanks Paul, Georg, MHM Cisco World for all your responses. The client has now mandated BGP dampening.

@paul,

Our BFD is at 600 mix_rx 600 and multiplier of 3

I hope to test out the below config suggested by MHM Cisco World tomorrow -

route-map DAMP

match next-hop

set dampening <value depend on your monitor>

Will the match

next-hop 

have any specific parameters? Why specifically is this needed? We now have additional links on our router so more eBGP neighbors

[lionell01]

It seems that the match next-hop option is not available in my GNS3 setup. Though it is available on my ASR router.

Please confirm if this will work?

route-map DAMP

match next-hop

set dampening <value depend on your monitor>

[MHM Cisco World]

don't worry I will make GNS3 LAB for you and share the result tonight.

[MHM Cisco World]

this simple network, 
R3 config  AS100  
R2 & R1 config with AS 200 
both R2 & R3 devertsie same Lo which is 10.10.10.10
in

show ip bgp 10.10.10.10

there is before damping 
and there is after damping 

you see that R3 not ignore the prefix it ignore only the path of this prefix to Peer.
Hope this help you to solve your issue

[lionell01]

Hello,

I'm so very thankful for this. Since there's no

next-hop 

key word, I've managed to replicate this in my GNS3 setup as well. What I was doing wrong was flapping the neighbors i.e. shutting and un-shutting f0/0 on R1 in your example which caused BGP to flap but no dampening. Then I shut the loopback (in my case, deleted the static route that was being redistributed into BGP) on R1 and enabled it. This caused dampening to kick in.

This does not help my use case because the we have un-stable international links which cause BFD flap which causes BGP flap. 

Apparently, No penalty is applied to a BGP peer reset when route dampening is enabled. Although the reset withdraws the route, no penalty is applied in this instance, even if route flap dampening is enabled.

[MHM Cisco World · Accepted solution]

in your case not the route is flap the neighbor is flapping 
I do small lab for using IP SLA for add remove the neighbor from BGP config, 
take look.