ospf design

Unanswered Question
May 8th, 2009
User Badges:

The design is like this:

the backbone is composed of two core ( link with etherchannel), this link is the backbone;

several distribution L3 swiths with 2 uplinks ( one for core1 , one for core 2;

this ( core 1, core 2 , distribution 1 make area 1 and area 1 is a totally stub area ) and so on for others distribs;

my question is :

what is the issue if the link between the cores goes down because the backbone is going to be split

thanx for answers

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
lhoyle Fri, 05/08/2009 - 06:19
User Badges:

My first guess would be that spanning tree would take are of it for you. Go to each core and do "sh spanning-tree blockedports". One side to each distribution stack is probably blocked.



Giuseppe Larosa Fri, 05/08/2009 - 07:43
User Badges:
  • Super Silver, 17500 points or more
  • Hall of Fame,

    Founding Member

Hello Fab,

if I understand correctly you have only the etherchannel in area 0.


When all the member links are down you have a backbone divided in two parts.


The impact is that if there are wan edge routers connected only to core1 core2 will be unable to know of them and viceversa.


You can easily provide backup paths in area 0:

make some dist to core links L2 trunks.

Have an additional Vlan defined only at layer2 on distrib and at layer3 on core devices and put the subnet in area 0.


add an ip ospf cost high on both SVI on core and you have your backup.


You can do this on a pair of distrib nodes uplinks and you should be fine.


Hope to help

Giuseppe



lamav Fri, 05/08/2009 - 14:16
User Badges:
  • Blue, 1500 points or more

Hi:


Just to add a bit to what Giuseppe rightly pointed out, a failure of all the inter-switch links that are in Are 0 between the two core switches will cause an isolated area scenario, not just a partitioned area.


A partitioned area is one in which a non-backbone intra-area link has failed and the area itself becomes split into two separate sub-areas. Assuming that each sub-area still has a connection to its ABR, the network may not be too adversely effected. One thing that will happen is that all the routes that would have been exchanged between those intra-area routers will go from being intra-area routes to inter-area routes because they will have to go through one of the ABRs to get to the other sub-area.


When the backbone area gets partitioned, this creates a scenario of isolated non-backbone areas. This is a much more serious situation; one in which one side of the network attached to one core router will be completely isolated from the other non-backbone areas that connect to the other core in Area 0. In short, they will not know of each other.s existence; they will not exchange routing information; and they will not be able to communicate between each other. You have effectively created two OSPF domains, each with one core router acting as Area 0.


HTH


Victor

fd_case17 Tue, 05/12/2009 - 12:40
User Badges:

I used a loopback assigned in area0 on the 2 core

so when I shutdown the inter-core interface , the IA routes still appear in RIB ,

with 2 uplink ( one for core1 one for core 2)


but I notice that the backbone is marked as Inactive with sh ip ospf



Actions

This Discussion