2960X switch injected into old network and deleted all existing VLANs.

[Bouki]

 H All,

We do have split MAC wireless design and we purchased APs that can only support Poe+ for full functionality so we had to purchase 12 switches 2960X which support both Poe or/and Poe+.

One of busiest distribution Centre needed an early injection of two new switches (2960X) to the old existing network 3(*)50 switches for both Data and VOIP usage

Our network is segregated into 18 VLANs on the existing network so the assumption is that the VTP revision is higher than 1 and injecting the new switch which most probably running on VTP revision 0 or 1 with either VTP mode won’t affect the old existing network VLAN database on  which  is for sure has a higher VTP revision.

We do run our switches on VTP mode transparent for security reason but sometimes we change the core switch to be on VTP mode server to replicate the VLAN database (Lazy way) to the new switches and reconfigured back both new and old to Transparent, I want to emphasize the word NEW, yes these new injected switches are new.

Well let me tell you something guys, after injecting the switch (2960) it deleted all my existing VLANs caused me a downtime of 20 minutes on the wired network and 35 on the wireless, checked the VTP status on the affected core switch after the disaster the VTP revision was 0, usually a switch with the higher revision will delete the existing one on slave switches (As it becomes the VTP master) and impose  its database but that was  was not the case with me, there were no VLANs been added and brought VTP revision to 0 which matching the new switch.

Can someone please shed some light? I noticed that there are already two cases like mine not answered.

Thanks.

[Predrag Jovic]

When you change VTP mode from transparent to sever or client revision number will be changed to 0. So, you should before adding new switch on old server add and remove one VLAN (VLAN that is not currently in use to increase revision number). It is also advisable on new switches to change vtp mode from server to transparent before connecting it to network so revision number when changed back to client or server  on new device will be reset to 0.

You actually had 2 switches with VTP revision number 0. I am not sure why old switch decided that new switch database with revision 0 should be accepted instead the current one with the same revision number.

[Bouki]

Hi crni000000,

I appreciate the quick reply. I completely agree.

I was just about to update my post, I forgot to mention that the core switch which has been changed to server a while ago and we have been working on the Vlan database for sometimes now without resetting the VTP revision with VTP mode transparent, I am pretty certain that the server (Core Switch) has a higher revision.

Regards

[Predrag Jovic]

I never heard for such case, but, who knows. My team is redesigning/refreshing typically 1 - 2 k Cisco switches per year and we never had such case. If your devices support VTPv3 I guess that would be the safest way to go.

[Bouki]

Well, make it two cases if you click on below link and I am sure I have seen a third one.:

https://learningnetwork.cisco.com/thread/77032

[Bouki]

Hi All,

Notice 3 Cases having the same issue with one think in common, the Injected switch is 2960X, coesidence? .... you tell me.

[Predrag Jovic]

It is not that I don't believe you. I just, fortunately, did not experienced it so far (and I hope it wil stay that way).

:)

We are using all the time tested versions of IOS in our environment (we are testing new IOS version prior to implement it in production. Which IOS versions are involved in your issue (if you can provide that information)?

Please, involve TAC and open case with them.

[Bouki]

Much appreciated cmi00000. I will certainly open TAC case.

I will send the details to you once I have it as I have limited access to the network now.