OSPF Peering across multiple VLANs

[NettableWalker]

What is the best practice for controlling adjacencies between multilayer switches with multiple VLANs.

I have a LAN to explore that has several 6500's, none of them using Passive-interface so the Show IP OSPF Neighbors shows many adjacencies. Is this a problem, or something that is best avoided?

I favour using layer 3 links between the switches for routing, is this the best option?

 

[unclerico]

The answer to both of your questions is "it depends". How many is "several" 6500's?? Can you put together a diagram of the topology??

I hate all Uppercase... I don't want my groups to seem angry at me all the time! =)
- ColdFlame (vbscript forum)

 

[NettableWalker]

I've actually got many many LANs that i see this in as i'm working on a network refresh on a global company. I am thinking about say 2x 6500's with up to 10 VLANs, all of which have an OSPF adjacency across them.

Is this a bad thing, or just a waste of resources and extra work for the switch and unnecessary complexity in the routing tables?

 

[unclerico]

I'm assuming that the 6500's are set up in some sort of HA config with HSRP, VRRP, or GLBP?? Unless they had a huge amount of routers/multilayer switches set up in a single area I would say things are probably ok. Although, I can't say for sure since I'm unaware of the topology. How many routes are in the routing table of one of the 6500's??

I hate all Uppercase... I don't want my groups to seem angry at me all the time! =)
- ColdFlame (vbscript forum)

 

[NettableWalker]

Yes, One of the bigger sites has 8x6500, there is HSRP on the VLAN interfaces, just a mix and match of layer2 and 3 in a sub-optimal configuration. There are loads of routes in the IP routing table (maybe 300) as the site uses BGP distributed into OSPF from their MPLS WAN. Most LAN routes show as 4x load balanced OSPF routes. Most switches show about 10 adjacencies from the same neighbor switch because of all the VLAN interfaces not being set to passive.

Is any of this a problem, or just considered "messy"?

I can find nothing about this issue in any CCNP/CCIE text on OSPF, so i am not sure if it's just me being obsessive about neatness in my configs

 

[unclerico]

Wow, 8 x 6500's?? That's a large implementation. In terms of adjacencies, neighborship should form based on the router-id, which should be manually configured with the router-id <ip add> in the router config or it will pick the highest available address on a loopback, or the highest address on a phyical interface. Weird that each switch would form multiple adjacenices with the same switch. Can you post a portion of the sh ip ospf neighbors??

I hate all Uppercase... I don't want my groups to seem angry at me all the time! =)
- ColdFlame (vbscript forum)

 

[Jsteve]

Most likely you are peering each vlan across a trunk link, which is not necessary. Only one peer is needed. You can eliminate some of the peering by adding "passive-interface" for the "Stub" vlan LAN segments. It really depends on what each vlan is used for in your network, like most of the guys stated it's hard to say without knowing your topology. I wouldn't say that having extra peering is bad, but with "passive-interface" on the stub networks you will get several benefits

1. Save on CPU processes with less peering
2. LSA won’t be sent onto the LAN segment (less traffic and more secure)
3. Smaller LSDatabase sent to neighbor

As for the L3 links, there are much better in my opinion, however there are some cases when you need a trunk link and you have no chose.

 

[burtsbees]

Are any of the routes being summarized?

Burt