Why is OSPF always advertising a (false) route?

[gabor_fejer]

Hello All,

I have encountered a really odd problem (at least it is odd  for me :) ) and I  would like to ask your help to get understand it.
So, there are two routers (primary and secondary)  connected to a switch.If the primary WAN link goes down then the fail-over does not work. Actually, it partly works.
Some apps do work but the rest does not. When a local user starts a traceroute towards to another location on his computer,  traffic always goes towards the primary router. So, the 2nd hop is always the primary router even though it's WAN link is down.

In case of the servie is having the above mentioned issue, we see the following routes on the switch coming from the R1 router:

SW1#sh ip route | in 10.10.10.102
O E1    10.0.0.0/8 [110/101] via 10.10.10.102, 1d11h, Vlan10

This is the only one route that being advertised by the primary router. Switch received routes for secondary router as well.

On R1 if we check the routing table we can see this (when fail-over is taking place):

R1(config-router)#do sh ip route | inc 10.0.0.0
      10.0.0.0/8 is variably subnetted, 20 subnets, 10 masks
O        10.0.0.0/8 is a summary, 00:05:11, Null0

So, the SW1 can receive the route because R1 sending  it. Ok. If I am not mistaken, OSPF marks this route  with a NULL0 in order to avoid loop if  it does not know how to forward the packets. (unknown next-hop).  So that, because of this,  the traffic is going to the "black-hole" in case of fail-over. IF WAN link is up , we receive 10.0.0.0/8 from WAN cloud.

OSPF config:

router ospf 10
 summary-address x
 summary-address y
 summary-address z
 summary-address xx
 summary-address yy
 summary-address zz
 summary-address xxx
 summary-address yyy
 summary-address zzz
 summary-address 10.0.0.0 255.0.0.0
 redistribute bgp 1 metric 100 metric-type 1 subnets
 network 10.10.10.80 0.0.0.15 area 0
 default-information originate metric 10
 default-metric 100

On the R2 OSPF config is similar but the metric is higher. (10->20, 100->200)

What I do not understand is that , why is OSPF always advertising only  that particular network? We have a lot of summary routes defined in OSPF and the mentioned one is the only one that being advertised if fail-over happens.

Any idea, what does cause this behaviour?

Thanks a million for your help!

Cheers,

Gabor 

[a.alekseev]

if you are getting the 10.0.0.0/8 from WAN cloud do not summarize it.

[gabor_fejer]

Routers learn other subnets from WAN and they are summarised in OSPF as well but they don't cause any issues. Why ? I still do not understand it.

[a.alekseev]

summary-address 10.0.0.0 255.0.0.0

because of this you have 

O        10.0.0.0/8 is a summary, 00:05:11, Null0

in you routing table on both routers R1 and R2

they both advertise this route to the SW1 but  R1 is better

SW1#sh ip route | in 10.10.10.102
O E1    10.0.0.0/8 [110/101] via 10.10.10.102, 1d11h, Vlan10

so in the situation when R1 lost wan connection you have a black hole.

[gabor_fejer]

Yes, I know it. But, Why does it work this way ? Why only one route ?

I know how to solve the problem, I am interesting in the route-cause that triggers this behaviour.

[Hitesh Vinzoda]

Hi there,

Blame RFC and its default behavior and is expected. To avoid this you may use below command under OSPF

router ospf x

 no discard-route internal

end

HTH

Hitesh

[gabor_fejer]

Thanks for pointing this out. It makes sense....

[Hitesh Vinzoda]

Hi,

Pick up the destination IP which isn't accessible and paste output of "sh ip route x.x.x.x" from R1, R2 and SW.

Thanks

Hitesh

[gabor_fejer]

Hello,

I can not see any suspicious thing. R1 and R2 knows the route via WAN BGP.

There is no assymetric routing between R1 and R2 or something like that.

Unfortunately, I can not send output from Sw1 as it does not managed by me.

R1#sh ip route 10.33.72.76
Routing entry for 10.33.72.0/21
  Known via "bgp 1", distance 20, metric 0
  Tag 2, type external
  Redistributing via ospf 10
  Last update from 10.73.110.65 1w4d ago
  Routing Descriptor Blocks:
  * 10.73.110.65, from 10.73.110.65, 1w4d ago
      Route metric is 0, traffic share count is 1
      AS Hops 2
      Route tag 2
      MPLS label: none
  
  
R2(config-router)#do sh ip route 10.33.72.76
Routing entry for 10.33.72.0/21
  Known via "bgp 1", distance 20, metric 0
  Tag 2, type external
  Redistributing via ospf 10
  Last update from 10.73.133.83 00:00:39 ago
  Routing Descriptor Blocks:
  * 10.73.133.83, from 10.73.133.83, 00:00:39 ago
      Route metric is 0, traffic share count is 1
      AS Hops 2
      Route tag 2
      MPLS label: none