BGP route dampening.

[nikhil.kulkarni]

Hello,

I need some clarity with reference to BGP route dampening. Few days ago some audit reports suggested that the Internet facing routers that run BGP with the ISP need to be configured for route dampening.

Now when I am reading some documents I came across a statement saying "Only applied to inbound announcements from eBGP peers".

My doubts:

1) Why would anyone want to configure BGP route dampening on Internet facing routers when it applies only to incomming routes?

2) If I do not configure route dampening on my Internet facing routers, id anyone else other than me affected?

Regards,

Nik

[Nandan Mathure]

Hi Nikhil,

Route dampening is used to avoid route-flaps which would trigger numerous advertisements in the internet. Mostly used between the ISPs to punish misbehaving route.

When you are receiving routes from the ISP and the link is flapping or ISP has some issue then the routes will no longer be stable. If you are further advertising them to your peer (unless you are a transit AS) then it would be a problem.

Route dampening is a good option if you are advertising routes to other AS than the one you learned from i.e traffic is transitting your AS.

Thanks,

Nandan.

[nikhil.kulkarni]

Thanks Nandan,

I need some more clarity, sorry to bother you.

I am not a transit AS. I have 2 internet facing routers connected to each other via iBGP. Further both my routers are connected to different eBGP peers(different ISP's). What I mean is router 1 connected to ISP 1 router. and router 2 connected to ISP 2 router. both router 1 and router 2 run iBGP.

Now, if router 1 gets some flaps from ISP 1 would it advertise these flaps to router 2? If yes, would this cause router 2 to get overloaded doing route calculation?

What if my own LAN link flaps? Will the internet routers need to do any recalculation because of me?

I mean what is the risk I run and other run due to me?

Regards,

Nik

[Nandan Mathure]

Hi Nikhil,

Since you are not a transit AS that would lessen the chances of needing the BGP dampening.

Well but now since you are dual multihomed and you have iBGP running, R1 learned routes would be advertised to R2 (vice versa) and it would certainly have effect of route-flaps. In this case you can use the BGP dampening feature and avoid learning those routes from one ISP and thus prefer them from another.

Also your LAN flaps would not create any issues as I belive you are not advertising it to the ISP. Also even if you are, it will be taken care by the ISP as you will be its eBGP peer and it will be on the receving end. So you need not worry about it.

Please rate if useful.

Thanks,

Nandan Mathure.

[cadet alain]

Hi Nandan,

if we refer to this link  http://www.cisco.com/en/US/docs/ios/12_2/ip/configuration/guide/1cfbgp.html#wp1002395

The routes external to an autonomous system   learned via iBGP are not dampened. This policy prevent the iBGP peers   from having a higher penalty for routes external to the autonomous   system.

So it will have no effect on external flapping routes learned from an iBGP peer.

Regards.

Alain.

Don't forget to rate helpful posts.

[Nandan Mathure]

Hi Alain,

Yes thats right, we need to apply it on router connecting ISP (eBGP router R1 and R2 in this case). As per the statement iBGP learned routes cannot be dampened but iBGP peer will be affected if routes are not dampened on ISP facing routers. So it is necessary to enable route dampening on routers facing ISP.

Thanks,

Nandan Mathure

[nikhil.kulkarni]

Thanks a ton Guys.

Again my original confusion remains the same, should I implement this dampening or should I not? In my opinion this does not really affect me and my routers can live happy without implementing this feature.

1) I am not a transit AS so any flaps do not affect anyone else.

2) Further in case any of my LAN links flaps I will not be able to control what happens to my ISP, my ISP can control what happens to the entire Internet BGP routers.(similar to the above point)

3) Even if a couple of routes flap somewhere on the Internet and are propogated to my routers by my ISP not much of CPU is wasted getting them recalculated.

The above points are just from my understanding prespective.

I have a question on the actual configuration of BGP route dampening.

When you talk about half-life you define this as "The amount of time to wait before decrementing the dampening penalty, ranging from 1 to 45 minutes. The default half-life is 15 minutes."

Now assuming that the route does not flap for 15 mintes half life will reduce the penalty value. By what number would the half-life decrease this value?

Regards,

Nik

[cadet alain]

Hi,

half-life is the time it takes to reduce the penalty to half its value.

take a look here:http://www.cisco.com/en/US/docs/ios/12_2/ip/configuration/guide/1cfbgp.html#wp1002395

Regards.

Alain.

Don't forget to rate helpful posts.

[stmillet]

Nikhil,

BGP route dampening is no longer best practice and not recommended, especially on Internet facing links.

Anectodally no ISP uses this feature anymore, also it breaks more than it actually fixes. The modern routers have ample CPU to support constant BGP updates without affecting performance, especially if you are only holding a subset of the full BGP routing table, so in short don't turn it on.

I hope this helps

Cheers

[sagriffin]

https://tools.ietf.org/html/bcp194 disagrees with this position. BGP route flap dampening is still a Best Common Practice.

[paul driver]

Hello

---

@nikhil.kulkarni wrote:

Thanks a ton Guys.

Again my original confusion remains the same, should I implement this dampening or should I not? In my opinion this does not really affect me and my routers can live happy without implementing this feature.

1) I am not a transit AS so any flaps do not affect anyone else. - you can apply dampening globally which will dampen any received ebgp flapping route(s) in  bgp or specific to a route/es or ASN and then it will only apply to those routes or asn

2) Further in case any of my LAN links flaps I will not be able to control what happens to my ISP, my ISP can control what happens to the entire Internet BGP routers.(similar to the above point) - applying dampening on your rtrs wont effect your ISP

3) Even if a couple of routes flap somewhere on the Internet and are propogated to my routers by my ISP not much of CPU is wasted getting them recalculated.- depends how much recalculation is performed due to the flapping route(s)

The above points are just from my understanding prespective.

I have a question on the actual configuration of BGP route dampening.

When you talk about half-life you define this as "The amount of time to wait before decrementing the dampening penalty, ranging from 1 to 45 minutes. The default half-life is 15 minutes."

Now assuming that the route does not flap for 15 mintes half life will reduce the penalty value. By what number would the half-life decrease this value? Using the default values, the reuse limit =750 sec and each flap = 1000 it will take 3 flaps to dampen a route=3000 so after 15 mins the dampened route penalty would have decreased by half  to 1500 , but that is still above the reuse limit of 750 so once it goes below 750 value it will be usable again unless you manually clear the dampened route/s then it will be usable straight away

---