BGP Peers

parrysands · Mon Oct 30, 2006 6:20 pm

I have 2 ISPs connected to a Mikrotik Router which I am using for BGP. In Winbox, how can I force all traffic to go to a specific ISP ? If I simply pull the plug on the one I don't want, everything stops working. So I assume I have something setup wrong....Any help would be appreciated. Please remember, I use Winbox, I am not familiar with the Router OS Command Line.

Thanks,

Parry Sands

changeip · Mon Oct 30, 2006 7:49 pm

You can add an additional 0.0.0.0/0 route with a distance of 2 to the second isps network, or setup ecmp gateways. Use check-gateway=arp if you want to automatically remove that route when its unreachable.

parrysands · Mon Oct 30, 2006 9:15 pm

Example:

ISP1 = 1.1.1.1

IPS2 = 2.2.2.2

ME = 3.3.3.3

I assume then that I should have this:
route 0.0.0.0 1.1.1.1 distance 1
route 0.0.0.0 2.2.2.2 distance 2

This is assuming that I wan't my traffic to go out through ISP #1

Thanks,

changeip · Mon Oct 30, 2006 9:29 pm

yep - perfect

Sam

parrysands · Mon Oct 30, 2006 9:49 pm

Thank You very much....I'll give it a shot later and see what happens.

nikhil · Fri Nov 03, 2006 2:29 am

How do you set this up to have incoming transfer come through one of the isps? I have a similar BGP setup

Please advise .

changeip · Fri Nov 03, 2006 5:41 am

To force more traffic in one ISP than the other you can use AS prepending... its not perfect but helps influence the traffic to one side or the other.

nikhil · Fri Nov 03, 2006 7:06 am

How to do this with MT im running 2.9.30 and routing-test. I want to make sure more incoming comes in via 1 of the isps

changeip · Fri Nov 03, 2006 5:56 pm

Here is one filter that I use to make the path look longer for one incoming route:

add chain=level3-out prefix=204.16.xxx.0/24 prefix-length=24 invert-match=no \
action=accept set-bgp-prepend=2 comment="" disabled=no

This adds 2 extra AS-PATHs to the length which makes it less preferred to the outside world.

Sam

nikhil · Fri Nov 03, 2006 6:05 pm

Which version of mikrotik are you using ?

changeip · Fri Nov 03, 2006 7:08 pm

That above filter was from 2.9.26... and it's been up for a whopping 96d now : )

Sam

nikhil · Sat Nov 04, 2006 4:03 am

and your advetising this tou level3 only OR to the other provider as well ?
This makes level3 appear further away so your other provider will become the preferred for incoming correct?

<quote>
add chain=level3-out prefix=204.16.xxx.0/24 prefix-length=24 invert-match=no \
action=accept set-bgp-prepend=2 comment="" disabled=no

</quote>

changeip · Sat Nov 04, 2006 7:25 pm

Yes, for one of our /24s we annouce out both peers we want most of the traffic to come in Cogent... so the level3 path is prepended to make it look less preferrable.

changeip · Sat Nov 04, 2006 7:26 pm

Yes, for one of our /24s we annouce out both peers we want most of the traffic to come in Cogent... so the level3 path is prepended to make it look less preferrable.

changeip · Sat Nov 04, 2006 7:29 pm

Yes, for one of our /24s we annouce out both peers we want most of the traffic to come in Cogent... so the level3 path is prepended to make it look less preferrable.

nikhil · Sun Nov 05, 2006 5:44 am

I did this it has no effect . Want to do exactly the same that you have done to have certain / all classes incoming via a partciular . What I have done is advertised out those classes out to the one I dont want to have incoming from -- yipes . But has no effect on the incoming so far. Looked through the looking glass i dont see anything. How do i check or do i prepend further away (more than 2) ?

BrianHiggins · Sun Nov 05, 2006 6:48 pm

That above filter was from 2.9.26... and it's been up for a whopping 96d now : )

Sam

Sam, have you tried any other version that have been stable for extended time periods? I'm going to be implementing a MikroTik along side of my existing Cisco for BGP (each running eBGP to upstream providers, and iBGP between each other for true failover and redundancy)

I don't have a copy of the x86 2.9.26 package, and I really don't want to load a version that hasn't yet proved it's stability with BGP.

nikhil · Mon Nov 06, 2006 6:58 am

That above filter was from 2.9.26... and it's been up for a whopping 96d now : )

Sam
Sam, have you tried any other version that have been stable for extended time periods? I'm going to be implementing a MikroTik along side of my existing Cisco for BGP (each running eBGP to upstream providers, and iBGP between each other for true failover and redundancy)

I don't have a copy of the x86 2.9.26 package, and I really don't want to load a version that hasn't yet proved it's stability with BGP.

USed 2.9.30 up 40+ days.

BrianHiggins · Tue Nov 07, 2006 3:50 am

USed 2.9.30 up 40+ days.

sure beats the 16 day average I'm getting with my Cisco 7204/NPE300/256mb betweeen reboots... and I've got a copy of x86-2.9.30

anyone else had any BGP implementations running for more then 30 days?

nikhil · Tue Nov 07, 2006 4:38 am

I didnt know the npe300 gave only 16 days its supposed to be very stable if you have the l8st ios

BGP Peers

BGP Peers

Who is online