Sanity check on design doc please

[Jon Marshall]

It's probably me but could someone have a look at this link in Cisco's campus network design doc -

http://www.cisco.com/en/US/docs/solutions/Enterprise/Campus/HA_campus_DG/hacampusdg.html#wp1108489

scroll down to figure 41 where the distribution switch interconnect is blocking and both access layer switch uplinks are forwarding. If vlan 2 is on both access-layer switches does this not form a L2 loop ?  The diagram shows both switches forwarding on vlan 2 on both links but to me that seems as though a packet such as a broadcast could origiinate on one access-layer switch and then end up back there.

If each access-layer switch had it's own vlan then i could understand the logic but not with both switches having the same vlan.

Am i missing something ?

Incidentally this is linked to the question about GLBP and STP in case anyone wants some more context.

Jon                  

[Giuseppe Larosa · Accepted solution]

Hello Jon,

from what I can see the network design of Figure 41 is based on the fact of having the link between distribution switches blocked

>> Figure 41     GLBP with STP Blocking Distribution-to-Distribution Link

But you are right if both access layer switches serve the same Vlan another link should be blocked or a loop is formed. it is not enough to block on distribution to distribution link.

I guess you are right as you see they have used different colors, so the right figure should be use different client Vlans on different access layer switches.

So this should be a variant of the design with no L2 connection between distribution switches ( I don't remember if it is called U or V design).

To do so the link between the two distribution switches has been assigned a rather high STP cost on the non root side so that Dis2 (the one on the right in the picture)  reaches the root bridge via access layer switch.

This STP configuration changes allows both uplinks to be in forwarding state allowing access switch end users to access the GLBP  AVFs but as you noted should require the deployment of different client vlans on different access layer switches.

EDited : I was fouled at the beginning then when I took a piece of paper to draw the diagram I understood your doubts

Hope to help

Giuseppe

[Jon Marshall]

Thanks for the response Giuseppe but i am still not getting it. From the diagram -

if a broadcast packet is received on Access A it is forwarded to both to dist A and dist B.

1) Dist A would forward it to Access B which would then forward it to dist B which would then forward it back to Access A.

2) Dist B would also receive the packet from Access A. Dist B forwards it to Access B which forwards it to dist A and back to Access A.

If all the links are forwarding for the same vlan then what stops the above happening ?

*** Edit - sorry i missed your last underlined bit or i was a bit too fast typing my response. So if the vlan is same on both access switches can you actually see it working ? 

Sorry, just reread your post and it looks like we agree about the vlans on the access-layer switches. Thanks for checking that for me because i thought i was either missing something or i really didn't understand STP !

Jon

[Giuseppe Larosa · Accepted solution]

Hello Jon,

I was still modifying my answer now I agree with you the document is wrong another link would be blocked if both access layer switches are in the same client vlan

This STP configuration changes allows both uplinks to be in forwarding state allowing access switch end users to access the GLBP  AVFs but as you noted should require the deployment of different client vlans on different access layer switches.

If it is the same vlan another link of Dis2 would be blocked so the use of both uplinks would be true only for one access layer switch.

Hope to help

Giuseppe

[Jon Marshall]

Giuseppe

Thanks again, glad it's not just me

Jon