From: Gustavo Novais (gustavo.novais@novabase.pt)
Date: Tue Oct 25 2005 - 13:27:56 GMT-3
In deed I saw the CDP native vlan mismatch (but only at cdp update time), but I was hoping to see also both ports STP instances move to an inconsistent port state (much faster than CDP). Are the STP BPDU's transparently moving between vlans, even though there's a mismatch? (VLAN 30 does not exist in VTP B, nor there is any stp instance related to it)
For earlier tests I did in lab (using same VTP domain for both switches, so matching stp instances) I saw that only traffic belonging to both vlans involved on the mismatch was affected.
What happened in production environment (where there are two vtp domains) is that no traffic would pass, but there was no inconsistent port on either site of the trunk.
-----Original Message-----
From: CCIE KH49279 [mailto:ccie_lab@inetiq.com]
Sent: terga-feira, 25 de Outubro de 2005 17:08
To: Gustavo Novais; 'Group Study'
Subject: RE: What do you think will happen?
Gustavo,
As you already know native vlans under got1q are not tagged. So when one side of a trunk sees the untagged frame, it will assume that this is for the native vlan. If you have native vlans configured as followed
sw1>>vlan30-------------vlan40<<sw2 and they are dumping traffic on the
trunk, you would think that what sw1 puts out will be seen by sw2 as belonging to vlan 40. The reality is if you look at your CDP messages you will see a native vlan mismatch. When this happens the interfaces will go into blocking mode.
You can do this same thing between a router and switch, when you do this, you will not get the error message, and you will be able to pass traffic.
wayne
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of Gustavo Novais
Sent: Tuesday, October 25, 2005 10:27 AM
To: Group Study
Subject: OT: What do you think will happen?
Hi
One question concerning switching.
if you have a 802.1q trunk joining two different VTP domains (SW1-VTP domain A-------------802.1q trunk---------------SW2-VTP domain B), you must have
(allow) matching vlans on both ends of the trunk.
So you will have VTPA-vlan1---- VTPB-vlan1, VTPA-vlan50----VTPB-vlan50 and so on. Besides that each switch has its own PVST+ database (meaning that on switch 1 has N stp instances and switch 2 has P stp instances.
What will happen if by mistake on SW1 side we change the trunk's native vlan to a vlan that does not exist on VTP B, meaning that there is no STP instance on SW2 for that vlan?
I tried to replicate that situation, and the result was that no traffic would pass though any trunk, even those that didn't interconnect SW1 to SW2, I'm still trying to figure out what could have happened... For what I could see there was no spanning-tree change or recomputation.
I know that at GS, we try to understand how to make things work, and how, but at this case I'm trying to understand how things should not work!
Any thoughts?
Thanks
Gustavo
This archive was generated by hypermail 2.1.4 : Sun Nov 06 2005 - 22:00:53 GMT-3