ospf design and routing loop issue

Hello INE Users:

This is my first post at INE. Currently I am preparing for the CCIE Lab Exam. I came across below design of OSPF and want to lab it up and see its behavior. Below is the design:

This is a 9 node topology:[View:/cfs-file.ashx/__key/CommunityServer.Components.UserFiles/00.00.01.83.99/nssa_5F00_area.zip]

2 areas are configured: Area 0 and Area 1 (NSSA)
REDIST-1 & REDIST-2 are the two routers where redistribution is happening for static routes in ospf
REDIST-1 & REDIST-2 are connected to ROUTER-6 & ROUTER-7 via area 1 interface (passive)
All the routers except ROUTER-6 & ROUTER-7 are part of Area 0 and Area 1 (NSSA) which means that each of these routers is a ABR/ASBR
All the routers except ROUTER-6 & ROUTER-7 have IPVLAN interfaces configured. .1 interface is part of Area 0 and has a high cost. .2 interface is part of Area 1 (NSSA) and has a low cost. 
Translator role for Type7 to Type 5 conversion is done by REDIST-2 - this is selected based on highest router id

I would like to take help from INE community and it would be good if we can discuss on this design and its pros and cons if any. Also if this design would call for any routing loop issues for the redistributed routes. 

I have attached the topology as well for better understanding.  

Thanks
Prakash

Comments

  • Welcome to this forum!!

    All the routers except ROUTER-6 & ROUTER-7 have IPVLAN interfaces configured. .1 interface is part of Area 0 and has a high cost. .2 interface is part of Area 1 (NSSA) and has a low cost.

    Few things are not very clear from your description.

    All the other routers, they are forming adjacancies in which area 0 or 1 ? If its 1 then all area 0 are disconnected and you need a virtual link in order to fix that. If its area0 then why do you need to have area 1 nssa on all the routers, in that case it just required on REDIST routers.

    For better response provide more clear picture of your topology.

  • Thanks dcancerian.

    I thought I had attached the topology to my original post but looks like it had some problem. I have uploaded it on this link http://i.imgur.com/rvLKq.png

    So in this topology you will see .1 and .2 interfaces. .1 interfaces are in area 0 and .2 in area 1. Cost of area 1 interfaces is less than area 0. Also forgot to mention one point that redistribution is happening through a route map where we are setting the redistributed routes to be E1. 

    ospf adjacancies are in both area 0 and area 1. We can simplify the design by doing some changes to the topology but the point is that this design is being used in production nodes currently with a very big service provider. I am just trying to understand it better as to the reasons they have chosen this design and problems it can have. 

  • Hi praks!

    Welcome to ieoc!

    I thought I had attached the topology to my original post but looks like it had some problem. I have uploaded it on this link http://i.imgur.com/rvLKq.png

    Thanks for the drawing!

     

    So in this topology you will see .1 and .2 interfaces. .1 interfaces are in area 0 and .2 in area 1. Cost of area 1 interfaces is less than area 0. Also forgot to mention one point that redistribution is happening through a route map where we are setting the redistributed routes to be E1.

    Hmm....looks like almost every interface participates in area 0 and area 1 when I read yor drawing? This will not work. The other numbers on the drawing...what do they stand for?

     

    Regards!

    Markus

     

     

     

  • Markus,

    Could you please share your thoughts as to why this design will not work? I am mostly trying to see if there is any flaw in the design. Also the other numbers are the slot and the port numbers. Not much of importance, I just put it there for easy understanding.

    Thanks
    Prakash 

  • Well I am referring to the passive interfaces you mentioned. I dont get the drawing as its quite unusual for me to look at (no offense here). Passive interfaces cannot build adjacencies. Probably you mean the passive interfaces are connected to routers to which you create static routes and then redistribute them into ospf?

    2nd thing is that I am not able to pull the area design out of your drawing concerning the routers and interfaces. A router can be part of multiple areas (ABR) bot I cannot see it in the drawing how you mean it.

    The design maybe works depending on what the real tpology is :).

     

    Regards!

    Markus

  • Could you please share your thoughts as to why this design will not work? I am mostly trying to see if there is any flaw in the design. Also the other numbers are the slot and the port numbers. Not much of importance, I just put it there for easy understanding.

    To me this topology design diagram is not very clear so its very difficult make any comments.

  • To me this topology design diagram is not very clear so its very difficult make any comments.

    Same here - Had a hard time interpreting it as well

  • Well I am referring to the passive interfaces you mentioned. I dont get the drawing as its quite unusual for me to look at (no offense here). Passive interfaces cannot build adjacencies. Probably you mean the passive interfaces are connected to routers to which you create static routes and then redistribute them into ospf?

    2nd thing is that I am not able to pull the area design out of your drawing concerning the routers and interfaces. A router can be part of multiple areas (ABR) bot I cannot see it in the drawing how you mean it.

    The design maybe works depending on what the real tpology is :).

     

    Regards!

    Markus

     

    Sorry for the confusion. Yes those are passive interfaces so no OSPF adjacency on those interfaces. Just the static are getting redistributed back in OSPF. 

    I tried again to put the topology in a simpler way and I have uploaded it here http://i.imgur.com/WQKrD.jpg

     

     

  • That drawing is better :). Thanks for that!

    Never seen a design like this before.

    What is your intention to use 2 areas on all routers in the backbone?

    Usually you would use area 0 in the backbone and other areas directly connected to area 0 for other purposes like separating remote sites from the headquarter or whatever.

    Well if the interfaces would ran in VRFs then I maybe could understand it as you want to raise the size of the whole topology but this is completely new to me.

    Here is a design link for OSPF AREAs:

    http://www.cisco.com/en/US/tech/tk365/technologies_white_paper09186a0080094e9e.shtml#t14 

     

    Regards!

    Markus

     

     

  • Hi Prakash

    This is a better topology design diagram.

    Any network design is essentially based on business needs so unless those needs are not identified its very difficult to make comments on any design. Here are some general comments on the current design only based on the topology diagram.

    This design seems to be highly redundant and provides a flexibilty of redistrubution on any router if needed. There could potentially be no loop in the topology because redistribution is happening one way i.e from static to ospf on two routers so each routers will have routes (via either of the two redistribution point depending on the metric value) to reach any netowok outside this domain.

    However it has a flaw of generation of huge amount of redundant summary LSA on every router since all routers are ABR which takes more CPU cycle. Also since all routers are either part of area 0 and area 1 therefore LSA flooding comprises of all routers for each area which is bandwidth consuming.

    There could be better design then this but again it highly depends on the business needs of your organization


  • @Markus

    Yes you are right some of the routers in that topology do have interfaces part of vrf

    @dcancerian

    Problem we are facing is that static routes for same subnets are being redistributed back in ospf on both the redistribution points (R1 & R2). Doing this is creating a incorrect path selection where some of the routers are selecting high cost path towrads some of the redistributed routes instead of a low cost path. 

    Now for all routers in the topology from R1 to R7 will have 4 LSAs in Area 1 (4 E1 routes - 2 Type7 and 2 Type5). Forwarding address for these LSAs is non-zero and is set to IP address of interface on R8 & R9. 

    For e.g.:

    R7 router will have 4 LSAs for 192.168.10.0/24 (subnet redistributed via static route on R1 & R2). 1 Type5 & Type7 LSA will have forwarding address of R9 interface and next hop will be R2 interface in Area 1. Other Type5 & Type7 LSA will have forwarding address of R8 interface and next hope will be R1 interface in Area 1.

    From R7s perspective, R2 is the short path with a cost of 10 whereas path via R1 is a high cost path with a cost of 100. This cost is calculated based on the forwarding address cost. Still for this subnet I see it is selecting the high cost path via R1 which is really odd for me? This incorrect path selection at times causes loops, so loop is just a symptom, real cause is the incorrect path selection. 

    This is not happening for all the subnets being redistributed, only a few. I am just trying to look at this from design side if there is any flaw which is causing this. Ofcourse if thats not the case then it could might as well be a bug in the vendor's code. 

    Note: This is not a cisco network. 

Sign In or Register to comment.