VLAN Trunking and Routing Qs

Answer to the first question is yes. Although for redundancy purposes, it is best to have two central switches and HSRP configured to mitigate outages during trunk, module, or switch failures. Yes, in my opinion it is wise to match your trunk statements and ensure you have consistent configs. If for no other reason, its easier to recognize the configuration and if/when new resources come in, they don't have to ask themselves what the original designer was thinking.

What do you mean one instance of stp per vlan? Spanning tree runs on a per vlan basis (one root, blocks where needed, etc). I am not aware of a reason why you would have more than one instance per vlan.

Trendkill, I may have misunderstood the 3750 configuration guide section on configuring vlan trunks. What I took to be a limitation may, in fact, be normal behavior. Here's the exact wording:

IEEE 802.1Q Configuration Considerations

The IEEE 802.1Q trunks impose these limitations on the trunking strategy for a network:

In a network of Cisco switches connected through IEEE 802.1Q trunks, the switches maintain one spanning-tree instance for each VLAN allowed on the trunks. Non-Cisco devices might support one spanning-tree instance for all VLANs.

On a per switch instance, I suppose that could work, although not sure how that works if you have different trunks with different vlans/ pruning. Either way, haven't done a lot of non-cisco, so one instance per vlan is status quo.

cisco specific.

original 802.1q was "standard" with 1 spanning tree.

other switch manuacturers just like cisco found this painful in practice (i worked on Nortel gear for a while) and would implement more than 1, up to 1 per vlan.

this made for even more problems when you hooked the different systems together, so eventually we got a multiple spanning tree standard. now all we have to do is get all manufacturers to implement it (pretty much done) and then use it by default (which as usual cisco do not do).

802.1 specifies MST which lets you have 1 spanning tree per set of vlans - and you get to choose which vlans are in which. Logical limit of this is one per vlan.

note you can still get caught. typically you can choose x vlans from the total 4k or so - often x is 64 or 256.

but you may have a lower (much lower) limit on the number of spanning trees, and may need to derate that number if you tweak the timers....

I am not aware

you only need multiple instances if your VLAN carries more than 1 topology - Xylan tin used to have policy based vlans where the topology varied by protocol.

it managed to generate some whole new classes of network problem that no one had seen before......

1 other instance was where you had token ring bridging where spanning tree might be used within transparent switches, but a separate spanning tree controlled the "single route explorer" topology for source routing. Again this has died with along with the Token Rings that spawned it.

So generally - 1 spanning tree at most per vlan.