spanning tree confusion

[sysad43]

I need some help understanding my spanning tree results.

I have 2 routers connected to each other, and 3 switches in a loop. So R3-SWA-SWB-SWC-R2 and back to R3.
I have set R3 as the root with the lowest priority 8192 , R2 next lowest 12284, and switches default 32768 priority.
All set to RSTP

When the tree comes up, R2 blocks fwding to SWC and so no loops. Thats good except now all traffic is flowing from R3 to SW3 which is a slower connection. I was expecting one of the switches to block instead thus having traffic flow to my switches through both routers and back. What could be wrong? Or how would I set it to block fwding in the middle of loop, so no loops and efficient traffic, but still redundant if one of the switches fails? Im not ready to move to layer 3 redundancy yet.

In addition, on SWC it appears to be using a higher cost link as the root. The long way around. 

Interface Role Sts Cost Prio.Nbr Type
------------------- ---- --- --------- -------- --------------------------------
Gi1/0/1 Root FWD 4 128.1 P2p
Te1/0/1 Desg FWD 2 128.51 P2p

[Joseph W. Doherty]

Could you further explain the router interfaces?  As usually, they would be L3 interfaces and not participate in STP nor close a L2 loop.

Are they L3 switches running all L2 interfaces?

If so, STP topology should reflect least cost path, from each node, to the root.

In your topology, if all link costs the same, SWC should transit R2.

[sysad43]

They are all layer 3 switches, yes. The two routers are layer 3 switches. Thats why the confusion. I did just move everything from PVST to rapid-pvst, Do I need to somehow force a recalc of the cost?

I have several switches choosing the higher cost as the root.

SWC - Te101 is connected to R2
Interface Role Sts Cost Prio.Nbr Type
------------------- ---- --- --------- -------- --------------------------------
Gi1/0/1 Root FWD 4 128.1 P2p
Te1/0/1 Desg FWD 2 128.51 P2p

R2 - Twe101 is connected to SWC

Interface Role Sts Cost Prio.Nbr Type
------------------- ---- --- --------- -------- --------------------------------
Twe1/0/1 Altn BLK 2000 128.1 P2p
Twe1/0/2 Altn BLK 20000 128.2 P2p
Twe1/0/3 Altn BLK 2000 128.3 P2p
Twe1/0/6 Desg FWD 2000 128.6 P2p
Twe1/0/11 Desg FWD 2000 128.11 P2p
Twe1/0/12 Root FWD 2000 128.12 P2p

So its chooseing the longer, higher cost as the root?

[sysad43]

I think the issue may be R2 is using long path cost method instead of short. I dont have anything faster than 10g so is there any danger in changing it from long to short? Or better to change everything to long?

[Joseph W. Doherty · Accepted solution]

As long and short use different metrics for the same bandwidth, you'll want all switches to you the same costing method.

If all devices support long, I would recommend that for possible bandwidth growth.  (Even though you only have 10g, what if you wanted to Etherchannel them?)

[sysad43]

Thanks, that did fix it. Its now pathing as expected.

[sysad43]

Fun follow up. Came in this morning with fixed spanning tree paths and users were complaining about slow speed. I traced it a bad 10g fiber connection which was not being used because of the path cost was wrong. But now its right, but the cable is bad.