10-12-2006 12:43 AM - edited 03-03-2019 02:19 PM
Can anyone advise me in the following situation?
We have 4 routers connected in a 'diamond' as follows
R1 at site 1, R2 at site2, R3 at site 3, R4 st site 4.
R1 connects to R2
R1 aLso connects to R3 via a seperate link
R2 connects to R4
R3 also connects to R4 via a seperate link
R1 is in area 0
R4 is in area 1
R2 and R3 are the ABRs between areas
The idea is that if any link fails the sites will remain connected
once ospf recalculates.
Under normal circumstances traffic between site 1 and site 4 should be
routed via R2 and only go via R3 in the event of a failure in between R1
and R2.Traffic between site 1 and 3 should be routed direct between R1 and R3
and go via R4 and R2 in the event of a failure betwwen R1 and R3
The problem we have is suboptimal routing between site 1 and site 4.
Traffic between R1 and R4 is being routed by a higher cost path (via R3)
rather than the lower cost path via R2. The reason I believe is summarization
on the ABRs.
R2 is summarising 10.1.0.0/16 into area 0
R3 is not summarising so is advertising all subnets in area 1 to area 0
As the summarised routes and the non-summarised routes are all in the routing
table on R1 the path being chosen is based on longest prefix rather than cost and hence
highest cost path is being used?
I can adjust the summary addresses but we have discontiguous subnets at site 3 and 4
ie
site 4: 10.1.1.0, 10.1.2.0, 10.1.3.0, 10.1.4.0, 10.1.10.0. 10.1.11.0 - 10.1.99.0
site 3: 10.1.5.0, 10.1.6.0, 10.1.7.0, 10.1.8.0, 10.1.9.0, 10.1.100.0 - 100.1.199.0
Options as I see it:
Re-addressing subnets to allow a single summary from R2 and R3 into area 0 for appropriate
subnets off these routers
Using numerous area 1 range commands on r2 and r3 to summarise networks appropriately from
each router
Distribute list on R1 to prevent certain subnets advertised by r2 or r3 from being added
to the routing table on r1
Can anyone advise if my assumptions are correct and suggest a solution?
Thanks
10-12-2006 04:53 AM
David
If R2 is summarizing and R3 is not then that certainly explains why traffic is routed via R3. You are certainly correct that the routing decision considers longest match first thing and only compares cost when it has more than one path for the particular prefix that it is considering.
In terms of options I certainly agree that re-addressing the network so that subnets are contiguous and can be efficiently summarized would solve the problem and be good in the long run. Since we do not know details of the topology of your network we can not assess the amount of effort required to do this. But it probably is the cleanest and most efficient routing solution.
It seems to me that another option is to remove summarization from R2. If both R2 and R3 are advertising detailed routes then you should be able to manipulate the metrics to achieve the routing solution that you need. Since we do not know details about the topology of your network we can not assess the impact of doing this. But given the little that we do know (4 routers, 2 areas) I am not sure that the impact would be great.
I think that you might also consider an alternative of configuring summarization on R3 equivalent to the summarizatino done on R2. If both routers are ABRs between area 0 and area 1 they should both know the same routes. So I do not see that you would lose any connectivity if you did summarize on both routers.
I like the alternative of the distribute list the least. One thing I dislike about it is that I believe that it could cause lack of connectivity. If you use the distribute list to block some routes from R3 and then there is a problem between R1 and R2 the result would be that R1 would not have routes to at least some of the destinations in area 1.
HTH
Rick
10-12-2006 05:52 AM
Thanks Rick
Looks like re-addressing in the favorite then depending on the effort involved which I'll have to assess. If its too much in the short term I'll go the re-summarisation route and avoid distribute lists just in case!
Cheers
Dave
10-13-2006 06:32 AM
before deciding which protocol you want to use, the first thing is design your network with continuous ip address space which is important to all routing protocols. and then you can summarize your network properly.
as you mentioned, your network seems not huge, so you can disable summarization feature.
on the other hand, as you know, OSPF is link state protocol, so distribute list is not suitable to it comparing with distance vector routing protocol such as RIP or EIGRP.
Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: