Restrict STP Diameter

[anilfernandes]

STP Limitation problem

Scenario

Core Dist Access

4500 ----- 4500 ----3550 (Multiple daisy-chained)

I have manually controlled the vlans on the trunk between the 2 4500s(based on Cisco documentation to control STP diameter). However I find that the 3550s still run STP for all vlans in the domain as they have trunks configured and recieve all vlans. I would ideally not want STP running for all vlans in the domain as only the ones allowed on the trunk between core and dist are needed on the 3550.

I can think of only two solutions for this -

* Run a separate vtp domain for Core and Dist + Access

* disable stp on access 3550 switches for the vlans not required in the access layer. Since the Core-Dist trunk is already filtering out unwanted vlans. This will reduces STP instances on 3550.

Is there any other method to resolve this.?

Thanks - Anil

[glen.grant]

The 3550's should be no different than the 4500's just manually prune off the vlans that are not needed on the trunk links between your 4500 and the 3550. I am assuming that you are running transparent mode between the switches . If you want an easy way to handle this is to make the lead 4500 a server and the other 2 as clients and when you prune off the link between the 4500's only those vlans will be allowed down to the 3550's also .

[Kevin Dorrell]

Anil,

I don't know the solution to this, but I am curious to know how it works in practice.

So, you have pruned a VLAN from the trunk between the 4500 and the 3550. On the 3550's, are there any ports at all that are carrying the unwanted VLAN? Including any further downstream trunks? That is, on a 3550, if you do a show spanning-tree for one of the pruned VLANs, is the interface list totally empty? And there is really still an STP instance running for the pruned VLAN? If that is so, the 3550 cannot have a root port for that VLAN. In that case, who does the 3550 believe is the root for the VLAN?

Kevin Dorrell

Luxembourg

[anilfernandes]

Hi,

I have manually pruned all the unwanted vlans off the trunk between the 4500s.

The 3550s are daisy chained with trunks between them. There are no other connections from the access layer 3550s.

My observation - All vlans in the vtp domain are propagated to the 3550s through vtp. The 3550s have trunks between them. Since these trunks are not manually pruned, they are part of all vlans (in the vtp domain) and hence run an STP instance for all.

From what I am seeing - one must manually prune all trunks on the access 3550s to include only the required vlans. Pruning it off the trunk between the 4500s only does not help. Clearing it off all the 3550s would be cumbersome as there are many access switches involved.

Hence the query on my options.

Thanks,

Anil

[Kevin Dorrell]

Anil,

If you have a trunk out of any box that is carrying an unwanted VLAN, then there will be an STP instance running in the box for that VLAN. I think the only way to prevent this is to trim the VLAN from the box completely.

But I thought with VTP it was possible to have the VTP do this automatically, by making the VLAN prune eligible. That way it would get pruned automatically from all trunks to the box, and (I think) the STP instance would disappear. No need for manual pruning with VTP v2, isn't that so?

Kevin Dorrell

Luxembourg

[anilfernandes]

Thanks Kevin.

My understanding is that prune eligible will only reduce the flooded traffic by not sending traffic if the switch on the recieving end does not have ports in those vlans.

All vlans will be learnt via vtp at the access switches (through vtp on vlan 1) and hence will run STP for all vlans thus increasing the STP instances.

Isn't this the behavior . I may be wrong here.

Regards,

Anil

[Kevin Dorrell]

You are right. I just tried it, and it behaves as you said. I think I misunderstood pruning. Oh well, back to the books.

Kevin Dorrell

Luxembourg