Can a DROther be ABR in OSPF?

[network_geek]

Hi All,
While labbing with OSPF I was just wondering if a DROther can become ABR? I am attaching reference image for everyone's ease. What I am trying to do is floating R5's loopback back into area 0 but it doesn't work. Secondly, if I tweak the configurations and make R1 the DR then everything works fine. In the case of yes or no, please do elaborate the reason as well.

Any help here would be appreciated.

[pman]

Hi,
Can a DROther be ABR in OSPF?

If both the DR and the BDR are disconnected from the switch then there will be a recalculation and both the DR and the BDR will be calculated and selected from the remaining routers.

R1 can be BRother in area 0 and also ABR.
In addition in area 1 R1 can be either DR or BDR for the subnet between R1 to R5

Regarding ping that does not work,

from where to where you tying to ping?

please attache configuration of router 5.
please attache the next output from R1 show router ip route 

[network_geek]

Hi @pman,

Thank you for your reply but I don't think that was what I was looking for. I want to know if a DROther running completely fine be also an ABR? And possible explanation behind a yes/no as well. Thanks.

[paul driver]

Hello

R1 has to be the ABR for area 1 otherwise Area1 would be isolated, Also it doesnt have to be a Designed Router either, DR/BDRs are applicable on broadcast/non broascast connections, and as your topology shows those two rtrs are directly connected so they can be a point-to-point ospf peer as such no DR/BDR election is applicable as such drothers arent either.

[Richard Burts]

I have a couple of points about your question:

- In your drawing there is only a single link between the OSPF areas. That means that only R1 can be ABR. No other router could be ABR because no other router has connection to 2 areas. By definition an ABR connects 2 areas. And so any router connecting to a single area can not be ABR.

- It is not entirely clear whether you are describing R1 as DR for its connection to R2 (which seems likely to be the case) or for its connection to R5. It is quite possible for R1 to be DR on one and to be BDR on the other. Or it could be DR on both. Or it could be BDR on both. And none of these have any relationship to whether it is ABR.

- DR/BDR have to do with processing on a single multi access segment and about who forms full adjacency with all other routers. DR and BDR form full adjacency with all routers on the segment, which non DER/BDR form full adjacency only with DR and with BDR. And processing on that single multi access segment has nothing to do with whether the router is ABR or not.

[Meddane]

Is there a recommendation to ensure that the ABR will be the DR?
The answer is NO. There is no relation between a DR and ABR. an ABR is just there to inject intra area routes with a Type-3 LSA into others areas. Keep in mind that the intra-area routes are carried either via Type-1 LSA or Type-2 LSA, since the DR is the only router allowed to generate a Type-2 LSA, we can formulate the question more specifically as follow: is it recommended that an ABR should be the DR that is responsible about the Type-2 LSA? the answer is no.

The Type-2 LSA carries the subnet of the shared segment between R1, R2, R3 and R4 (let's say 10.1.1.0/24). This subnet is calculated by ANDing operation logic between the Link State ID (which is the IP address of the DR in this segment and the Mask field in the Type-2 LSA, if we look at your topology, regardless who is the DR, from R5 's perspective, R5 sees this shared subnet (10.1.1.0/24 behind the ABR since OSPF INTER AREA design behaves as a distance protocol. Finally R5 will calculate the cost of this inter-area route 10.1.1.0/24 and its calculated cost is always the same regardless if this ABR who originates the Type-2 LSA or not (in other words it's a DR or a drother), it does not matter. A DR has nothing to do with the area concept or inter-area routing, The DR applies only to a single segment, not the entire area.while the ABR is core component for inter-area routing. We can have many segments in one area so we can have many DR in one area.

Finally the DR and ABR are completely separate concepts.

[network_geek]

Hi @Richard Burts: Thank you for your detailed response. I am attaching an updated topology diagram which might make the scenario much more clearer to you. Link b/w R1 and R5 is point-to-point. R4 is the forceful DR for this topology. I want to bring inter-area routes from area 1 into area 0.

Thanks.

[Richard Burts]

Thank you for the updated drawing. It does clarify that there would be elections for DR and BDR on the R1 interface to R2, R3, R4 and no election on the interface to R5. 

As I read through the discussion again I realize that in my first response I focused on DR and BDR but that your question was about DRother. So my response was slightly off the mark. My main point is still true. There is no relationship between election resulting in DR, BDR, and DRother and ABR. So it is quite possible for a router that is DRother to be ABR.

The original post has 2 questions. The first was about whether it is possible for DRother to be ABR. We have dealt with that point fairly fully. There is not anything about being DRother that would prevent R1 from being ABR. I would now like to consider the second question which is about advertising the loopback of R5 into area 0. If in the original configuration the loopback of R5 was not advertised into area 0 it suggests that either the R5 loopback was not correctly learned by R1 or that for some reason R1 was not advertising into area 0 (which might suggest that there was some issue with R1 being active in area 0.

You say that you made some tweaks in the config to make R1 the DR and then it works. To understand this we would need to see the original config of R1 and to know exactly what those tweaks were.

[Meddane · Accepted solution]

@network_geek 

Is there a recommendation to ensure that the ABR will be the DR?
The answer is NO. There is no relation between a DR and ABR. an ABR is just there to inject intra area routes with a Type-3 LSA into others areas. Keep in mind that the intra-area routes are carried either via Type-1 LSA or Type-2 LSA, since the DR is the only router allowed to generate a Type-2 LSA, we can formulate your question more specifically as follow: is it recommended that an ABR should be the DR that is responsible about the Type-2 LSA? the answer is no.

The Type-2 LSA carries the subnet of the shared segment between R1, R2, R3 and R4 (let's say 10.1.1.0/24). This subnet is calculated by ANDing operation logic between the Link State ID (which is the IP address of the DR in this segment and the Mask field in the Type-2 LSA, if we look at your topology, regardless who is the DR, from R5 's perspective, R5 sees this shared subnet (10.1.1.0/24 behind the ABR since OSPF INTER AREA design behaves as a distance protocol. Finally R5 will calculate the cost of this inter-area route 10.1.1.0/24 and its calculated cost is always the same regardless if this ABR who originates the Type-2 LSA or not (in other words it's a DR or a drother), it does not matter. A DR has nothing to do with the area concept or inter-area routing, The DR applies only to a single segment, not the entire area.while the ABR is core component for inter-area routing. We can have many segments in one area so we can have many DR in one area.

Finally the DR and ABR are completely separate concepts.

[Joseph W. Doherty]

"Can a DROther be ABR in OSPF?"

As far as I know (and would expect), the answer is yes.  (For the same reasons noted by other posters.)

As to your issue of R5's loopback is visible in area 0, when R1 is DR but not when DROther, correct?  If so, that's curious.

DR elections only happen with "loss" of the DR, and OSPF would need to re-converge (which depending on what you're doing could be delayed by Cisco's OSPF back-off timers).  So, this is true after about a half minute wait, too?

[network_geek]

Hi everyone!

Thank you for sparing out a moment and guiding me further. Indeed it works fine without any issue. Thank you!

[Joseph W. Doherty]

That's great, but did you identify the original cause of the issue?  Reason I ask, just so we can add to our "toolbox" of "this might be happening" to help others.

[network_geek]

Hi @Joseph W. Doherty: A good root cause analysis doesn't let the curiosity in us die. Well if you peek at the topology I had shared earlier, R1's LSDB had all the ingredients it needed to work fine. For example, the LSDB for area 1 was same and that for area 0 as well. Type-3 LSAs had been built with correct sequence number and stored in it, but for some alien reason it discarded all those LSAs when computing its SPF and it also did not forward it to DR(R4). There it occured to me that there might be the limitation of DROther not being able to become ABR. Upon capturing the packets I couldn't see anything coming out of it(on the multi-access segment). But I can see LSAs coming from R5 into R1.

There I decided as a last resort to restart the OSPF process and it all started working fine. So I am guessing there might be a bug which was causing, although I couldn't find any bug with my OS.

Thank you all. Cheers.

[Joseph W. Doherty]

Very interesting!

Using a DROther, for an ABR, is possibly a bit rare, and often, in my experience, rare usage cases are more likely to bump into bugs.

Thanks for the update.

[network_geek]

@Joseph W. Doherty: That is what makes networking so fun and interesting, and yes sometimes really frustrating.

I have some more side-cases which from time to time I will keep posting here in relevant threads. I hope you land up on my post some other day as well!

@All: Thank you so much for contributing.