VTP auto configuration and Dell switches

bdedek · ‎01-12-2010

Today I was reconfiguring a 2950 switch and had erased the flash and reloaded the ios. (vlan.dat erased). When I reloaded the switch, to my surprise it had learned my VTP domain and had joined it as a server. All my vlan information had been propagated to this new switch. Here lies the problem...The switch was uplinked to a non managed Dell switch (non trunk). The unmanaged Dell was uplinked to another Dell 5224 switch, and it was uplinked into a Cat 5509. None of these links were defined as trunks. When the Cisco switch came up, and it learned the VTP domain, one of the intermediate Dell switches, the 5224, dynamically changed its uplink port to the cat 5509 to a dot1q trunk. This caused the devices on the switch to drop while the 5509 STP reconfigured the port and stabilized. When I turned the Cisco switch off, the port returned to a non-trunk uplink, and caused the outage again. Is there anyway to avoid this, or prevent it from happening again in the future?

Thanks alot...

Ganesh Hariharan · ‎01-12-2010

Hi,

When inserting a vtp server with a higher config revision number, the other switches will delete their configuration information and take the VLAN information from the inserted switch. The only way to get the deleted information back is to add the missing VLANs and delete the unwanted VLANs. To avoid this you should set the switch you're inserting into the network to transparent mode because that resets the configuration number, then switch it back to client or server mode. Another way of resetting the configuration number is to change the domain name to something else, like "test", then change it back.

Another problem can happen when you are inserting a switch with a different VTP domain name.

VTP can affect DTP (Dynamic Trunking Protocol) - switches will not form trunks unless they have matching VTP domain names.

Hope that helps out your query !!

Regards

Ganesh.H

Giuseppe Larosa · ‎01-13-2010

Hello bdedeck,

VTP is Cisco proprietary and also DTP is cisco proprietary it is not immediate to understand why the non cisco device in the middle changed its port mode to trunking mode as you have seen.

DTP: protocol that negotiates the trunking state of a port between two cisco switches not supported on cisco routers as VTP. and should not be talked by dell switches as far as I know

for technical details of these protocols see

http://www.cisco.com/en/US/products/hw/switches/ps700/products_white_paper09186a00801b49a4.shtml

Also Cisco uses different STP BPDUs format on access port ( 802.1D standard) and on trunk ports (cisco proprietary) cisco specific format is passed by non cisco switches so device at other end receives: standard BPDU by non cisco switch and Cisco BDPUs this can cause the switch to disable the port and so the issue seen in the network

In a multi vendor context like yours you should consider to use VTP transparent mode on all devices specially if non cisco devices are in the middle otherwise you are at the risk of creating new vlans that are not learned in non cisco devices in the middle.

To be noted a VTP server learns and implements vlans learned in VTP advertisements like a VTP client.

VTP uses multicast messages with a L2 multicast destination and travel on vlan1/native vlan on trunks so VTP messages could travel through the dell switch and could reach your device and VTP db has been populated again.

To be noted a VTP server learns and implements vlans learned in VTP advertisements like a VTP client.

Ganesh: if I've understood correctly original poster was surprised to see vlan database not empty and the explanation is above.

The real problem was however the change on state that caused the out of service.

Hope to help

Giuseppe