Q on OSPF: re-advertisement towards the same router.

[SandevChopra07800]

Hi, I am working on a design situation where I want to advertise the OSPF default-route towards the same router from where it was learnt but with a higher metric. The use case is if the site looses its site-local internet then it uses the cross site`s

default-route

This is only possible if the WAN router sends the

default-route 

back to the local Core SW. I havent lab`d this yet, but does OSPF allow to advertise the DF back on the same interface from where it was learnt using a

route-map

with a higher metric?

[MHM Cisco World]

R1-R2 

R1 learn defualt route from other router inject it to R2 

Here R2 learn defualt route from R1 and from where else?

[SandevChopra07800]

Refer attached sample topology. So in normal scenario this is how the

 default-route

propagation happens in Site1 (Just explaining Site1 in here, Site2 will be same):

Site1EdgeFW -> Site1R2 -> Site1R1. Site1R1 and Site2R1

(the WAN rtrs) also exchange the

default-route

over iBGP vpnv4 address family but it is not installed in the routing table since each site prefers its local OSPF learnt

default-route. 

Use case is if the

Site1EdgeFW 

goes down or its local connectivity to internet goes down, everything in Site1 can dynamically route over Site2`s internet. This will only be possible if Site1R1 sends a default-route back again to Site1R2 with a higher metric using the same layer3 interface. Does OSPF allow this?

[MHM Cisco World]

But ibgp have AD=200 and opsf have AD=110/90 so sure ospf is more prefer than ibgp'

Site1 will use defualt route via opsf but in same time it bgp table contain defualt route with r failure because it have better path via opsf 

If ospf is loss the route will dynmaic inject into rib from bgp table.

I will be sure by run lab and test failed conditions.

[MHM Cisco World]

PART1
as I mention before the router have OSPF 90/110 and iBGP with 200 and but the default route into "r" 
and when the OSPF failed it will use default route via iBGP

[MHM Cisco World]

Part2
as I mention above I clarify why the default from OSPF prefer than route from iBGP
this Part I check the config, 
if you want only default-route to advertise between two sites then no need

 redistribute ospf 

under bgp. 
the neighbor R1/R2 default-information will be enough in this case 
test done in this lab 
R5-R4 is down network fine 
R4-R2 is down network fine 

[Kanan Huseynli · Accepted solution]

Hi,

in general OSPF does not advertise route, but sends LSAs. So, it is possible to receive and send LSA which matches the sam prefix/prefix-length.

In your case, taking into account that do you mutual redistribution between OSPF <> BGP, here is what normally happen:

Site1-RTR1 and Site2-RTR2 (WAN RTRs) should receive OSPF route which originated from FW and redistribute into BGP (locally originated BGP route). After they advertise this route to each other, but they prefer only its own locally originated (redistributed) route as best, so don't re-redistribute back to OSPF.

When failure happens on Site1 for example, Site1-RTR1 stops receiving default route from FW and stops redistributing it into BGP. In this case, on Site1-RTR1 only one

BGP 0.0.0.0/0 default route

should exists which is iBGP route received from Site2 and installs it into RIB.. If you configured

default-information originate

in OSPF, then  Site1-RTR1 begins to redistribute BGP route into OSPF (actually, this is not redistribution, just advertisement of route when you have default route in RIB), so Site1-RTR2 (core) and fw received this OSPF default originated by Site1-RTR1.

[MHM Cisco World]

PART2 how prevent Loop in mutual redistribute
since you select solution without consider the LOOP of mutual redistribute then I think you know how prevent this LOOP
thanks 
MHM 

[Kanan Huseynli]

@MHM Cisco World ,

better to know technology then to write (I dont' say you are, but most VIP guys work only for "points" and do copy-paste, try to override others' answer which is very sad).

In the given topology there is no case for problems which happen in mutual redistribution. Look carefully, OSPF domains are not joined to each other. Between site1 and site2 there is no OSPF. Topology is like

OSPF_Site1 <> BGP <> OSPF_Site2 

do you see loop here?)

My answer is quite enough and true.

[MHM Cisco World]

Are you run such as this topolgy before?

I dot think so

The LOOP can happened in this case 

Without copy paste lol..

He inject defualt router recieve from internet via redistrubte so defualt route will be inject as E2 

He then redistrubte the ibgp into opsf and inject as E2 

Here if he not care and config 

Metric type 

Metric 

Some prefix filter 

The loop happened.

Note:- two way lsa for same prefix' this make ospf endless send receive lsa for same prefix that not correct 

[Kanan Huseynli]

---

@MHM Cisco World wrote:

Are you run such as this topolgy before?

I dot think so

The LOOP can happened in this case 

Without copy paste lol..

He inject defualt router recieve from internet via redistrubte so defualt route will be inject as E2 

He then redistrubte the ibgp into opsf and inject as E2 

Here if he not care and config 

Metric type 

Metric 

Some prefix filter 

The loop happened.

Note:- two way lsa for same prefix' this make ospf endless send receive lsa for same prefix that not correct 

---

Strange comment...Why do you think that I never run such or more advanced network?

Secondly, how Site1-RTR1 can inject iBGP based default route if AD of iBGP is 200 and OSPF external  (AD 110) from firewall is preferred?

In normal case, neither Site1-rtr1 nor site2-rtr2 will inject OSPF route (via iBGP redistribution) , because they already have OSPF route within its site.

If you believe loop, explain how it may happen exactly in this case (again, this topology does not have mutual redistribution issues, because OSPF domains are not joined).

[MHM Cisco World]

https://costiser.ro/2013/08/19/redistributing-internal-bgp-ibgp-into-an-igp-why-is-it-dangerous/#.ZCM3eOxRWdM

I am so glad to explain where is loop' check this link.

@SandevChopra07800 I will complete lab and check failure.

Thanks.

MHM

[Kanan Huseynli]

Totally different case, where well-known

bgp redistribute internal

is used.

If you want to make loop, you can create anytime (even with static route). The purpose is to explain author his case, but not create loop in his topology, lol.