Collapsed Core with OSPF requirements

[gg3390]

Hello,

Looking for some advice regarding our collapsed core configuration. 

We have 2x 6500's, with approx 50 L2 trunks to various access switches. SVI interfaces on the two cores are configured with HSRP, and each access switch has L2 trunk to each core. Essentially 2 VLANS being trunked to each access switch - 1 voice and 1 data. We are using STP odd/even numbering to utilise both uplinks so that voice is blocked on one core uplink while the other uplink has voice unblocked - this way both uplinks are operational.

Each core has an uplink to a HSRP enabled MPLS link, which most of our core routes point to - there is no routing protocol at present here. 

I can draw up a simple diagram and config snippets if it assists.

Between the two cores is a L2 link, and a seperate L3 routed link. Due to some legacy networks, there are a large number of static routes on the cores pointing to different networks and firewalls, and at present the network is working well. I would like to configure OSPF on the cores so that routes are learned dynamically, but I've some confusion on some aspects, which I hope somebody can assist with. My questions are as follows:-

1. Is the L2 link between the cores necessary for correct operation of STP?

2. Assuming, for example, the only L3 links between the cores is G1/1 on each, when enabling OSPF this is the only interface that needs to be taken out of passive mode?

3. Is a single area sufficient in OSPF - I saw some comments that suggest a seperate area for each SVI interface?

Thanks in advance, and appreciate the assistance.

[Jon Marshall]

1) It depends on which vlans are on which access switches but basically having that L2 link means HSRP traffic does not have to go via the access switches (assuming your L2 interconnect between your core switches is a trunk and not blocking which it shouldn't be). 

It is a standard design in a setup like yours. 

2) Yes but you may want another one for redundancy although I am not entirely sure what that routed link is doing to be honest. 

In a design like yours you can use OSPF across the L2 trunk link, especially if that link is actually an etherchannel,  and peer using two dedicated SVIs and then make the user vlans passive. 

That said you can use a L3 interconnect if you like. 

3) I have no idea why you would use a separate area per SVI, makes no sense to me. One of the limitations of OSPF is you cannot summarise within an area so you need to take that into account but I doubt you will need to from the sounds of it. 

All of the above however needs to be fitted with your environment ie. it is not entirely clear why you have both a L2 and a L3 link between your core switches and what the L3 link is doing (again I am assuming the L2 link is a trunk link and passing HSRP traffic between the core switches. 

So take that into account when reading the above. 

Jon

[paul driver]

Hello

hope you don’t mind me joining this discussion?

Just wanted to share a couple of points

1) stp isn’t an issue in a L3 access design

2) enable passive interface default and then only allow ospf on the interfaces you wish for become a ospf adjacency 

Also I don’t think this hasn’t been mentioned but  L3 access design  means no extended vlans across lan estate - so if your spanning you vlans now doing then L3 access will not be applicable to implement unless you amend your design 

3)  Regards  a single core /collapsed distribution for ospf - having your access switchs to be in different areas{even stubs possibly totally stubs) isn’t a bad idea-  it would cut down on convergence times and also minimise route tables and switch cpu/memory resources.

[Jon Marshall]

Hi Paul 

I may be misunderstanding the question but I can't see where the OP is suggesting a L3 access layer ? 

Jon