same ASN number at CE router caused network outage

[lerner cisco]

All,

recently faced issue with one of my customer  where both CE router used same ASN number , would be great your inputs on the below scenario

Existing router & under production  with T1 link

 Router A -  ASN 655xx working  with T1 link - connected to PE  X router  

 Router A - source subnet & loopback address : 10.200.20.0 & 10.200.21.0  connected Giga bit interface Gi 0/1  & Gi 0/2 , 10.1.1.1 , 10.20.1.1  lo0 & lo10  adverting out to the PE

Int Gi 0/1 

ip address 10.200.20.1 255.255.255.0

no shut

Int Gi 0/2

ip address 10.200.21.1 255.255.255.0

no shut

int lo0

ip address 10.1.1.1 255.255.255.255 ( management customer )

no shut

int lo 10

ip address 10.20.1.1 255.255.255.255 ( management vendor)

no shut

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

Router B -  ASN 655xx  connected  with  4T1 link - connected to PE  X router  

 Router B - source subnet & loopback address : 10.200.20.0 & 10.200.21.0  only configured ( not connected to LAN )on the Giga bit interface Gi 0/1  & Gi 0/2 , 10.1.1.1 , 10.20.1.2  lo0 & lo10 .

Int Gi 0/1 

ip address 10.200.20.1 255.255.255.0

no shut

Int Gi 0/2

ip address 10.200.21.1 255.255.255.0

no shut

int lo0

ip address 10.1.1.1 255.255.255.255 ( management customer )

no shut

int lo 10

ip address 10.20.1.2 255.255.255.255 ( management vendor)

no shut

New router installed same site parallel with four T1 links , plan was to bring up the local LAN interfaces once all 4 T1ckt tested and working condition  during this process customer faced  connectivity issue.

here the router LAN gi 0/1 & 0/2 configured with same address , the interfaces are not made admin down however cables not connected - so it should not advertise source subnets to cloud

loopback0 have same ip address using on the second router - 10.1.1.1 -  it will advertise however impact only accessing this subnet the router ID will be selected by the new router is highest loop back address loopback10 (10.20.1.2) by default

loopback10 have different  ip address on the second router - 10.20.1.2 -  It will advertise

would be great your inputs what caused the issue -  is same ASN  number on the CE side  reaches to the other CE   again blocks everything or PE router creates the loop issue - please suggest

[Akash Agrawal]

Hi,

What issue exactly you faced? Can you please draw some topology? What i understand that two CE are connected to same PE with T1 link. Same lan ip address assigned on the new CE router's gig interfaces but cables were not connected. Were Gig links in down down state? If they were in up up state and networks will get advertised and it may affect the routing. When issue occurred how did you resolve the issue? Check route for lan subnet on PE and if they recently got refreshed when you disconnected new CE with PE. If that is the case, may be new CE was advertising lan subnets to PE and attracted traffic.

Regards,

Akash

[lerner cisco]

Akash,

issue faced - customer not accessing application , when issue happen new CE router T1 links removed

 router A --------------- PE router ----------- router B

same Asn #                                                   same Asn # 

Also I missed one more parameter here , there are some subnets  advertising out to BGP cloud through statically on the old router pointing 10.200.20.2 LAN device interface IP .  since the LAN interface IP is active the subnets adverting out

On the new router these static subnets also configured pointing at the IP address 10.200.20.2, however at the new router no cables connected, the Gi 0/1 & Gi 0/2  subnets should not advertise .

thinking on the other subnets which are statically advertised from the old router and received at new router  , the ip address 10.200.20.2 also receives on the new router routing table , I have a static below mentioned routes on the new router pointing to 10.200.20.2  which is reachable from the new router . 

     hence on the new router these source networks ( 10.200.30,31,32/24) installed on the routing table and which intern advertise back to PE and routing loop ?? .

ip route 10.200.30.0 255.255.255.0 10.200.20.2

ip route 10.200.31.0  255.255.255.0 10.200.20.2

ip route 10.200.32.0 255.255.255.0 10.200.20.2

[Akash Agrawal]

Hi ,

router B would not accept prefixes coming from router A due to BGP inbuilt loop prevention mechanism.

http://networktalk.wordpress.com/bgp-tutorial-multihoming-practices-cisco/chapter-a-bgp-general-notions-techniques-and-practices-within-an-isp/a-5-bgp-loop-preventions-mechanisms/a-5-a-ebgp-loop-prevention-mechanism-based-on-as-path-attribute/

So in BGP looping is not possible unless untill you change the default behavior. 

My suspect is router B was advertising the lan subnet and attracted traffic. You have to check router b syslogs if Gig interfaces came up during that time or check route age on PE from lan subnet 

-Akash

[lerner cisco]

the  route age is more than 6 weeks , even on router A   no logs Gigi flap , customer not moved cable to new router .

[Akash Agrawal]

In that case, I guess next-hop for BGP routes would be loopback 0 (10.1.1.1) which is same configured on router B. If by any chance to reach 10.1.1.1 PE was preferring router B link then it will also cause outage.