Load Sharing with the Loopback Address as a BGP Neighbor - BGP session down

[muhammad shahrir osman]

Hello,

I have one scenario. One end is using Cisco router and the other end using Juniper router have been configured eBGP Multihoming using loopback IP. At each router, two interface have been configured and both static route has been pointing to the eBGP loopback IP. The problem is, for example if the

1st interface/link down, the BGP also will down eventhough the 2nd interface/link up. By right for Multihoming BGP at least one physical link should be up to keep to BGP up. Any one have experienced this sort of problem before? Need some help. Thanks in advance!

Regards,

arel

[lgijssel]

Having a look at some configuration would be helpful.

Cisco side should contain this:

neighbor 2.2.2.2 remote-as xxxx

neighbor 2.2.2.2 update-source loopback 0

neighbor 2.2.2.2 ebgp-multihop [ttl]

Are you sure both paths are working?

Verify both paths carefully, using static routes may not be the best way to create failover redundancy.

regards,

Leo

[muhammad shahrir osman]

Hi Leo,

Thanks a lot for your feedback. Both path are working and all configuration you mention above already included. Please look example config below. If one link down the BGP will drop. Any idea what cause this problem? Thank you

----------------------------

CISCO BGP config

----------------------------

neighbor 2.2.2.2 remote-as xx

neighbor 2.2.2.2 description "SINGA-STM4x2"

neighbor 2.2.2.2 ebgp-multihop 2

neighbor 2.2.2.2 password 7 xxxxxx

neighbor 2.2.2.2update-source Loopback0

neighbor 2.2.2.2 activate

neighbor 2.2.2.2 remove-private-as

neighbor 2.2.2.2 soft-reconfiguration inbound

neighbor 2.2.2.2 prefix-list in-filter in

neighbor 2.2.2.2 prefix-list xxxxxx out

neighbor 2.2.2.2 route-map xxxxx in

neighbor 2.2.2.2 filter-list 1 out

ip route 2.2.2.2 255.255.255.255 x.x.x.33

ip route 2.2.2.2 255.255.255.255 x.x.x.21

Loopback0            1.1.1.1

Juniper BGP config

==============

set routing-options static route 1.1.1.1/32 next-hop x.x.x.22

set routing-options static route 1.1.1.1/32 next-hop x.x.x.34

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 descriptionn "XXXX with Multihop-EBGP"

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 multihop ttl 2

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 local-address 2.2.2.2

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 import no-default

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 import no-smallprefixes

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 import rfc1918-dsua

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 import XXXXIN

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 import XXX-IN

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 family inet any prefix-limit maximum 1000

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 family inet any prefix-limit teardown 90

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 family inet any prefix-limit teardown idle-timeout forever

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 authentication-key "xxxxxx"

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 export no-default

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 export no-smallprefixes

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 export rfc1918-dsua

set protocols bgp group CUSTOMERS neighbor 1.1.1.1 peer-as XXXX

Regards,

arel

[cadet alain]

Hi,

What type of WAN link are you using?

Regards.

Alain.

[muhammad shahrir osman]

Hi Alain,

Do you mean interface? POS interface. Thanks

Regards,

arel

[cadet alain]

Yes and what encapsulation are you using?

Regards.

Alain.

[muhammad shahrir osman]

encapsulation PPP

Regards,

arel

[cadet alain]

Hi,

Can you post output of sh ip route static and sh ip int br | exc una  when both links are up then when one is down

as well as sh run interface.

Regards.

Alain.

[Peter Paluch]

Gentlemen,

In addition to other requested information, would it be possible to post the (complete) output of the show ip bgp neighbor 2.2.2.2 command on the Cisco side?

Best regards,

Peter

[muhammad shahrir osman]

Hi Guys,

Maybe later will do some failover test and capture debugging for both router to check BGP and TCP transaction. I think there is no wrong with the configuration right or am I missing any configuration? because when both links is up no issue arise...

==================

Cisco interface config

==================

interface POS7/1

description XXXXX

ip address X.X.X.22 255.255.255.252

no ip directed-broadcast

encapsulation ppp

ip route-cache flow sampled input

crc 32

pos framing sdh

pos flag s1s0 2

!

interface POS7/2

description XXXX

ip address X.X.X.34 255.255.255.252

no ip directed-broadcast

encapsulation ppp

ip route-cache flow sampled input

crc 32

pos framing sdh

pos flag s1s0 2

!

=====

Juniper

======

- configuration is the same for other interface

> show configuration interfaces so-6/1/1

clocking external;

encapsulation ppp;

sonet-options {

    fcs 32;

    no-payload-scrambler;

}

unit 0 {

    description "XXXXX"

    family inet {

        rpf-check {

            fail-filter URPF-FILTER;

            mode loose;

        }

        filter {

            input CFLOW;

        }

        address x.x.x.21/30;

    }

}

Regards,

Arel

[muhammad shahrir osman]

i'm just curious..erm could be this related to problem at lower layer? example on transmission issue?

Regards,

arel

[cadet alain]

Hi,.

I think it's hard to answer without further info

But your config is good for Cisco and AFAIK on the Juniper side also.

Were you able to ping 2.2.2.2 from 1.1.1.1 when the failover occured and the BGP peering went down?

Regards.

Alain.

[muhammad shahrir osman]

Hi Alain,

When one interface down,one interface up while BGP was down..ping to BGP neighbor was reach able. Thank you

[gerald.suiza]

i would also do a debug on the Juniper..i have seen things on a Juniper that are not so pretty..

Also do a deug for PPP, you never know it may prove something.

hth

[cadet alain]

Hi,

rpf-check {
            fail-filter URPF-FILTER;
            mode loose;
        }
        filter {
            input CFLOW;

Isn't it an ACL ? along with uRPF could it be the culprit?

Regards.

Alain.