I'm a rookie so apologies in advance for any mistakes. I'm redesigning my LAN and thinking of a collapsed core/dist--access design with L3 links from core/dist to access. I don't need to span VLANs across switches, want to segment broadcast domains, and not deal with STP across switches and into core/dist.
I currently have a 6509 and 4x 3750 Stackwise stacks all linked (hub&spoke) w/ L2 Gig EtherChannel trunks. I'd like to use the 6509 and a new 3750G as my core with a L2 GEC trunk between them; can't afford another 6k/4k and like the StackWise capabilities of the 3750G for growth. Production servers and SAN will be connected to the core switches on spanned VLANs with multiple network connections from each server/SAN to the 6509 and 3750G. I would also use the existing 3750 stacks as access layer with L3 links back to the core, GEC to the 6509 and a single fiber connect to the 3750G to form a triangle. BTW I don't plan on using any routing protocols just static since: I don't have that many networks, I've segmented to minimize # of routes and I only have Std/Base IOS images. I also don't plan on running VTP which hopefully won't have any impact.
My questions are:
1) Any changes/recommendations? Will this even work?
2) According to Cisco design guidlines for L3 access-to-dist interconnections (see RST-2032 pg 69) the connection between the dist switches should be L3. In my design, it's an L2 GEC trunk between the 6509 and 3750G to span VLANs. I'm breaking away from best practice guidelines but I don't know of any other way to span VLANs besides trunking.
3) I will have SVI interfaces on the core switches on the VLAN with the access-to-dist connections but would prefer the traffic from the access layer to primarily use the GEC connections to the 6509, so non-equal cost pathes. I was thinking of using HSRP between the SVIs and designating the primary/secondary via priority but how do the access switches know to send traffic out which L3 interface? Straight metrics? Is there a better way? Also, does this mean I should avoid using 'mls ip cef load-sharing full'?
4) Reverse this situation, on each 3750 stacks do I need to run HSRP between the 2x L3 interfaces to ensure that the core can get to the 3750 stack?
5) I plan to run a subset of the RPVST toolkit, especially loopguard and bpduguard since there are dumb switches on the floor, on the access switches. Any gotchas?
6) Does the default VLAN have to be the same across all switches? I'm thinking about CDP, VTP, and other mgmt traffic that travels over the default VLAN. Will it still be carried over L3 links?
Thanks in advance.