VLAN spanning-tree root and VLAN L3 SVI

[tojackson]

I have a traditional core/access switches connected, Catalyst 4506 with 802.1Q uplink trunks to two core 6504-E switches. Spanning-tree roots for VLAN's were on core switch with the active HSRP/SVI. Shutdown HSRP/SVI on the cores and moved the SVI to the L3 access switches, but we left the spanning-tree root on the cores... Is this causing the clients/ports in the VLAN on the access switches to do a double/triple hop over the uplink for ingress/egress?  Gig1/1 is the trunkport/uplink to the core, wanted to move L3 functions to the access switch to reduce spanning VLANs across the core network.              

Current configuration : 253 bytes
!
interface Vlan196
description Tower I - 8th Floor VLAN
ip address 10.200.196.1 255.255.255.0
ip access-group 115 in
ip helper-address 164.103.160.150
ip helper-address 172.20.135.201
no ip redirects
no ip unreachables
ip pim sparse-mode
end

!

I0504506A8#sh spann          
I0504506A8#sh spanning-tree vlan 196

VLAN0196
  Spanning tree enabled protocol ieee
  Root ID    Priority    196
             Address     6400.f1ee.c140
             Cost        4
             Port        1 (GigabitEthernet1/1)
             Hello Time   2 sec  Max Age 20 sec  Forward Delay 15 sec

  Bridge ID  Priority    49348  (priority 49152 sys-id-ext 196)
             Address     0015.f960.9ac0
             Hello Time   2 sec  Max Age 20 sec  Forward Delay 15 sec
             Aging Time 300

Interface        Role Sts Cost      Prio.Nbr Type
---------------- ---- --- --------- -------- --------------------------------
Gi1/1            Root FWD 4         128.1    P2p
Gi1/2            Altn BLK 4         128.2    P2p

[Tom Watts]

Hi Tojackson, I guess this depends on how stuff is interconnecting. It's obvious gi1/1 is forwarding and gi1/2 is blocking. So from the furthest access switch, what path must it take to reach gi1/1? That is the number of hops involved for normal traffic.

Now, if you're concerned about a specific VLAN and you need gi1/2 forwarding to reduce travel time for other traffic, you may employ RPVST to have that specific VLAN and cost to go to gi1/2.

In some part of the network I support we have a pair of Cisco 7606 which feeds in to a 4507R and off the 4507R we have a ring of 2955 with even 10-12 L2 switches on the ring. The consequence of multiple layer 2 hops is not of much concern and our spanning tree stops with the 4507 since we're not concerned about broadcast storm on the routed interfaces on the 7600.

-Tom
Please mark answered for helpful posts

[tojackson]

tom,

Thanks for the reply, but Im only concerned about the one VLAN/SVI. Clients are directly connected to the 4506 access switchports and they're in VLAN196. The spanning tree root is on the upstream core switch, but the SVI is on the local acess switch with the directly connected clients. The L2 hop count concerns me due to the possibility of additional round-trip delay. Even with the SVI/Default Gateway is on the local access switch w/clients, does the L2 path cause frames to go over the uplink/trunk to the core/root and then back down the uplink/trunk to the SVI/Default Gateway for the subnet and then back aging to leave the subnet?

[Tom Watts]

Your access layer shouldn't have any direct access to the 6500's so this shouldn't be a problem.

You should be able to physically figure out how your traffic flows based on what is blocking. If you're worried about traffic routing from access client to the 6500 then back down to the 4500 over any link then that is your problem. The spanning tree between the 4500 and 6500 should be irrelevant since it should be servicing redundancy for those 2 products. If you're running further downstram access switch in to the 6500 by passing the 4500, you probably shouldn't do that.

I just saw the Visio, it will take the prefer path on the "left side". It won't do anything round-about unless there is some downstream switch connecting in to 6500's.

-Tom
Please mark answered for helpful posts

[tojackson]

Tom,

My concern is that the L2 path is different than the L3 path and is adding additional L2 hops.  L2 path would have the L2 traffic going up to the 6500 on the non-blocking L2 port, and returning back down that same L2 link to get to the L3 default gateway. Could this casue an issue with drops and performance eventhough the 1GB link is not heaviliy utilized?