OSPF Route Updates - Strange

[StarTAC]

hi all...

i am configuring OSPF and i need to use it to annouce only my various interface IP addresses.. that is, if i've subnet 192.168.0.1/27 on fa0/0, i want to announce only 192.168.0.1, by doing this on OSPF:

router ospf 1
network 192.168.0.1 0.0.0.0 area 0

however, when the OSPF neighbors pick up this update, instead of only having 192.168.0.1/32 in the routing table, they see 192.168.0.1/27, which is basically the whole subnet..

i don't want to advertise the networks using OSPF.. i want to do it using iBGP instead, while OSPF only advertises the links/interface IPs for reachability..

i can only get my desired effect when i subnet an IP as a /32, which i can only do on loopback interfaces.. and not physical interfaces, for obvious reasons..

i have tried using outgoing prefix filter lists, to specify how the networks/IPs should be advertised, to no avail..

has anyone done this before, and can help me...?.. i need OSPF to only announce the interface-specific IP, and not the whole network subnetted on that interface..

al help appreciated..

 

[baddos]

Is there a reason for this? It's better to announce your internal networks through an IGP than a EGP. OSPF is a much better protocol for internal use than BGP.

Have you tried announcing it as a /30 instead? It might not be possible as a /32 but a /30 should definately work.

 

[StarTAC]

BGP is a much more resillien routing protocol when it comes to high loads, than any IGP..

BGP has an IGP side to it.. iBGP.. iBGP is what i'll use to advertise my internal networks... i just use OSPF to advertise the links for reachability, so that the iBGP mesh can be formed, either via full mesh or route-reflection..

OSPF can advertise a /32, no problem.. that's how i advertise my loopback interfaces across my network.. and the routing table does come up with a /32.. which defines that IP address alone..

i was thinking, and i suppose the solution to the problem would be advertise each loopback IP on each of my routers, ONLY.. and not use OSPF to advertise the IPs on the other physical interfaces...

for instance, say i have:

Loopback 0 - 192.168.0.65/32
Ethernet 0 - 192.168.0.1/28
Ethernet 1 - 192.168.0.145/29 e.t.c

the i say..

OSPF advertise - Loopback 0 only.. 192.168.0.65 0.0.0.0
iBGP avdertise - Ethernet networks.. 192.168.0.0/28 and 192.168.0.144/29..

that way, if other routers want to ping 192.168.0.1 or 192.168.0.145, iBGP says:

192.168.0.1 via 192.168.0.65 [the Loopback OSPF has announced...]..

wouldn't that solve the problem, as well as reduce OSPF's routing table..?..

 

[baddos]

BGP wasn't designed to be used as an IGP. By default, BGP won't announce a subnet if it's not in a IGP.

BGP is designed for large amounts of subnets like on the internet. OSPF can handle over 2k subnets, so I'm not sure why you would want to ditch it.

Also you have to look at compatability for other devices. A lot more routers and routing devices support OSPF than do BGP.

BGP's route path selection isn't good for an IGP either. With the OSPF algorithm, the router will pick the best path most of the time without the need for tweaking. With BGP, your going to have to setup some nice route-maps and mess the weights and local pref to get it right.

Not too mention a full mesh iBGP enviroment is terribly complex. This is why they have deisnged route reflectors and such. It's way more efficient to flood a multicast announce than maintain all the different TCP connections for every iBGP peer.

 

[tschouten]

BADOS is completely correct, BGP is something you want to use when you need to use multiple ISPs. Perform Load Balancing and route directing. It is designed for that purpose in mind. BGP is often complex even in its simplest implementation. It was designed to function with IGPs not perform as one.

OSPF on the other hand was designed for interior use and does it quite well. It was designed for large complex networks it will handle high loads of traffic.

I'm not saying what you want to do can't be done. Just saying it wasn't designed with that intent and you are going to face more problems.

 

[tschouten]

The only thing I can think of to help is Create the loopbacks with the intended /32 address. Let OSPF use the loopback as it's Primary address. It should work...

 

[StarTAC]

thanks all for your responses..

however, after alot of testing and research, i still feel iBGP is a better bet for me for my internal routing..

Baddos is right, BGP won't announce a route if it's not in the IGP, by default.. but Baddos, that's why the "no sync" command was made in the first place...

BGP can handle more routes than OSPF can, any day.. throw anything at it, it will take it..

for compatibility, my entire network is Cisco-based.. running the latest 12.2 IOS.. so, whether BGP, OSPF, IS-IS, i'm keeping it within my network, and i don't intend to use any other vendor for a very long time, or ever.. so i don't think that should be a problem either..

you're right, full mesh BGP is complicated.. but route-reflectors solve this, so i won't drill further..

when my network expands to include over say 100 routers and over 2000 routers, my BGP implementation is already in place, and it will handle more than sufficiently, the amount of load thrown at it.. try redistributing your BGP into OSPF, and u'll see what i mean.. ..

all i need from OSPF is discover all my router's loopbacks, create a link state database, and map my entire network.. for reachability.. this will also include alternate paths to the loopback IPs.. i believe this all i need from OSPF.. to maintain dynamic reachability of my network at all times..

then, BGP does the workhose, and advertises all my networks.. it's a fairly simple configuration that's secure, and works.. i really see no reason why iBGP can't be used for my internal announcements, afterall, that's what iBGP does anyway..

Tschouten, thanks for your tip, i was thinking the same thing..

thanks all..

 

[baddos]

I think for a good example of what I'm trying to explain would be to look at an ISP's network. They have very larger networks with a lot of subnets. But they don't run BGP EVER at their core. BGP isn't needed in your core, but rather on your edge. BGP is way too resource intensive and complicated to setup and adminster than OSPF and shouldn't be used as an IGP.

The "no synchronization" command was designed to all BGP to announce routes if the were removed from the IGP, not if they never existed.

BGP isn't an IGP, just ask any vendor.

Because you run all Cisco, doesn't mean you can run BGP on all your Cisco gear.

 

[SF18C]

Why not use EIGRP if you have all Cisco gear? BGP seems to be overkill as an IGP; I guess that's why it is an EGP. My understanding of iBGP is to keep the AS in-tune to the eBGP connections and ensure that all routes learned from the IGP are propagated to eBGP. BGP can very taxing on memory and CPU cycles, not to mention just the amount of bandwidth BGP uses on itself. The biggest thing I remember from Cisco studies on BGP is all the times Cisco stressed how and when not to use it. EIGRP is has rapid convergence, supports VLSM and reduces bandwidth. Both EIGRP and BGP are advanced distance vector type protocols. In my organization (U.S. Army, Europe) we have well over 100 routers and we use EIGRP (and OSPF on the occasional non-Cisco gear).

Just a thought! Good luck and keep us updated on your progress!!!

SF18C
CCNA, MCSE, A+, N+ & HPCC

"Tis better to die on your feet than live on your knees!"

 

[StarTAC]

hi Baddos, just because it's not being done, doesn't mean it can't be done..

BGP isn't really that hard to configure.. in fact, it's cleaner in it's filteration and traffic manipulation than any other IGP i have tried.. true, u can filter OSPF and EIGRP updates, but BGP, iBGP or eBGP, allows me to engineer traffic, restrict neighborship formations, and that kind of thing...

BGP isn't going to really do much, simply advertise the internal networks my various routers know about.. all i do is aggregate them and advertise a whole network...

perhaps if u guys have some free time, u can seutp some labs and see how well it really works out.. honestly speaking, i'd rather ran my internal network of 3000 or so networks on iBGP than OSPF, because BGP is simply built for that...

thanks for all your opinions though, they've really helped.. if any of u guys tries it, do let me know what you think..

good luck..

 

[tschouten]

In all honesty, I can see your point. However, you're taking a give product intended for a certain purpose and using it for a different one. This in itself means your are going to run into some problems somewhere along the line.

It does beg the question of how cool would that be, and I must admit the idea of it intrigues me? The only problem I can see is what bados said, cpu and memory intensive! You would have (HAVE) to make sure all equipment was fully loaded with memory and use primarily higher end equipment.

Beyond that, I can see it being done if you truly needed all that filteration and design internally.

Would be neat to attempt in a lab....hmmm

 

[StarTAC]

i must admit, it does need some memory and CPU.. but am fully stacked up on that, and shd have no problems....

bandwidth shouldn't be a problem for obvious reasons..

thank you all..

 

[oj88]

You can try using the command ip ospf network point-to-multipoint on the ethernet interfaces. This will give you the results you want without having to resort to BGP.

Just remember that one of the conditions for two routers to form neighborship using OSPF is that they must have the same OSPF network type on the common interface they're connected to.

In this scenario, I have the following topology followed by their configs:
e0/0---R1---s1/2-->>--s1/2---R2---e0/0-->>--F0/3---S1---L0

where R1 & R2 are routers while S1 is an L-3 switch.

R1:
interface Ethernet0/0
ip address 192.168.100.1 255.255.255.128
ip ospf network point-to-multipoint
!
interface Serial1/2
ip address 192.168.0.1 255.255.255.252
clockrate 128000
!
router ospf 1
log-adjacency-changes
network 192.168.0.1 0.0.0.0 area 0
network 192.168.100.1 0.0.0.0 area 0
!


R2:
interface Ethernet0/0
ip address 192.168.200.1 255.255.255.128
ip ospf network point-to-multipoint
!
interface Serial1/2
ip address 192.168.0.2 255.255.255.252
!
router ospf 1
log-adjacency-changes
network 192.168.0.2 0.0.0.0 area 0
network 192.168.200.1 0.0.0.0 area 0
!


S1:
ip routing
!
interface Loopback0
ip address 192.168.250.1 255.255.255.0
!
interface FastEthernet0/3
no switchport
ip address 192.168.200.2 255.255.255.128
ip ospf network point-to-multipoint
!
router ospf 1
log-adjacency-changes
network 192.168.200.2 0.0.0.0 area 0
network 192.168.250.1 0.0.0.0 area 0
!


And the resulting routing table are as follows:

R1:
192.168.200.0/32 is subnetted, 2 subnets
O 192.168.200.1 [110/781] via 192.168.0.2, 00:02:19, Serial1/2
O 192.168.200.2 [110/791] via 192.168.0.2, 00:02:19, Serial1/2
192.168.250.0/32 is subnetted, 1 subnets
O 192.168.250.1 [110/792] via 192.168.0.2, 00:02:19, Serial1/2
192.168.0.0/30 is subnetted, 1 subnets
C 192.168.0.0 is directly connected, Serial1/2
192.168.100.0/25 is subnetted, 1 subnets
C 192.168.100.0 is directly connected, Ethernet0/0


R2:
192.168.200.0/24 is variably subnetted, 2 subnets, 2 masks
C 192.168.200.0/25 is directly connected, Ethernet0/0
O 192.168.200.2/32 [110/10] via 192.168.200.2, 00:03:46, Ethernet0/0
192.168.250.0/32 is subnetted, 1 subnets
O 192.168.250.1 [110/11] via 192.168.200.2, 00:03:46, Ethernet0/0
192.168.0.0/30 is subnetted, 1 subnets
C 192.168.0.0 is directly connected, Serial1/2
192.168.100.0/32 is subnetted, 1 subnets
O 192.168.100.1 [110/781] via 192.168.0.1, 00:03:47, Serial1/2


S1:
192.168.200.0/24 is variably subnetted, 2 subnets, 2 masks
C 192.168.200.0/25 is directly connected, FastEthernet0/3
O 192.168.200.1/32 [110/10] via 192.168.200.1, 00:04:48, FastEthernet0/3
C 192.168.250.0/24 is directly connected, Loopback0
192.168.0.0/30 is subnetted, 1 subnets
O 192.168.0.0 [110/791] via 192.168.200.1, 00:04:48, FastEthernet0/3
192.168.100.0/32 is subnetted, 1 subnets
O 192.168.100.1 [110/791] via 192.168.200.1, 00:04:48, FastEthernet0/3


I hope this helps.


Orlando Palomar Jr
CCIE# 11206, CCNP
CIPT Operations Specialist
Phil-Data Business Systems, Inc.

 

[StarTAC]

hi oj88.. thanks for the tip..

but configuration of OSPF is not really a problem for me.. i have already been able to successfully use it as an IGP, and i understand exactly, the way this link state routing protocol works, right down to the last detail..

but i just want to take it further, and have OSPF maintain reachability for the iBGP mesh.. that's all.. iBGP handles internal advertising of my own and client networks...

good luck..