[c-nsp] STP, native VLAN and trunks

[Christian Zeng]

Hi,

today a quite interesting discussion developed here at work regarding
per-VLAN STP, native VLANs and trunking.

Given the following switch topology:

A------------B
|            |
|            |
|            | 
D------------C

each switch has .1q trunks to its neighbours configured. The native VLAN
used on all trunks is the same as well as some VLANs used for
production traffic.

The discussion was about using a different native VLAN ID at least at
one trunk because otherwise a STP topology change could harm trunk
operation and affecting the production VLANs, too, even if PVSTP runs
(blocked native VLAN = trunk failure?).

At the first look, I agreed to this. Digging deeper, more questions
arised and I'm not sure if the different native VLAN will make much
sense.

Why should the recalculation of STP for the native VLAN have any effect
on forwarding frames for any other VLAN over the trunks?

Another question is: When there is a VLAN defined as "native" for a
trunk, this VLAN is excluded from the list of allowed VLANs on that
trunk and it is not used at any other place in the switch, why is there
no STP instance created?

spanning-tree vlan 800,840 priority 8192
!
interface GigabitEthernet2/5
 switchport trunk encapsulation dot1q
 switchport trunk native vlan 840
 switchport trunk allowed vlan 800
 switchport mode trunk
     
#sh span vlan 840

Spanning tree instance(s) for vlan 840 does not exist.

Why should I include the native VLAN on the list of allowed VLANs? 

Thanks,


Christian