interesting VTP issue on 3560

[mtaib]

I do have interresting issue with VTP server and client. my test network is somewhat like this :

vtp srvr fa 0/1---dot1q trunk----fa 0/13 vtp clnt

I have configure the server as follows :

int fa 0/1

encapsulation dot1q

mode trunk

vtp mode server

domain name 'funtime'

Additionally, configure some test vlans on server.

And the client resembles the same excpet 'vtp mode is client'.. everything looks normal. A typical

VTP configuration.

Client was not learning those vlan and VTP infos correctly. Sh vtp status, sh vlan, sh int trunk commands shows everything ok. So i decide to debug VTP events and packets.

i got the attached output when i disable/enable the trunk interface in client which is fa 0/13. somehow ithe client is communicating with VTP server. and even the client recieves vlan info. but can't update its vlan database.

Two lines particularly noticeable in the attached file.

1. rev 1 greater than domain funtime rev 0

Domain funtime currently not in updating state

2.MD5 digest failing

Both of these Switches are 3560s. server with L3 image (12.1(19) and client with L2 image (same ver)

Let me know if somebody can help on this. how to fix this VTP communication issues.

MT

[Prashanth Krishnappa]

Enable fa 0/13, just create a test VLAN on the VTP server and see if it gets propagated and thereby take care of the issue. If it does, you can delete the test VLAN.

[mtaib]

I did so. Vlans are not updating on client vlan database. But when i so a debug on vtp packets and events, i could see client switch is recieving vlan advertisements. Those created vlans are included in the debug output, but switch is not accepting them

Regards

MT

[alfredshum]

What's the output of "show vtp status" in both your VTP server and client?

[mtaib]

They are fine. Show vtp stat command on server show the right domain name(funtime) and 'SErver' mode. Client shows the same domain and 'client'

I have done all basic troubleshooting.

Show int trunk ----- show internface is trunking on dot1q, both side.

Show vtp stat------shows right domain, right modes

Show vlan-- on server it lists right vlan i have configured, but on client, doesn't show any vlan updated its database.

When i debug VTP events and packets. i got the attached output in last message. If you read that carefully you will see something is recieved by client including vlans created on server (VLAN0016 and VLAN0017) but, for some reason client switch couldn't update its own VLAN database.

Regards

[tarun209]

Hi

Refer to the link below for solving your problem

http://www.cisco.com/warp/public/473/21.html#ts_vtp_cfg_rev

Thanks & Regards

Tarun G

[mtaib]

Finally, I managed to solve the problem. Removed Client switch from Original VTP domain, configured it as a server for another VTP domain (test) and added another L2 switch on the new VTP (test) as client. At this stage, VTP rev. # increased to 1 and VTP ads were fine (previously rev # remained always 0). VLANs were advertised over the trunk.

Next step, Re-joined the 'test' VTP server to my initial VTP domain (funtime) as client. Immediately the client started learning all vlans over the trunk.

I don't know how this works. But all what I figured out is, it is a play of revision number.

Regards

MT

[milan.kulik]

Genarally,

when you are in a VTP revision number problem like this, I'd recommend:

1. Check VTP revision number on the server and the client.

2. If the client's rev. number is equal or higher then server's one, you are in trouble.

The easiest fixing way is:

Change the VTP domain name to "something_else" on the client. Then change it back to "your_VTP_domain".

This should make the VTP revision number 0 on the client and the switch should update the VLAN database via VTP updates from the VTP server.

Another possibility is adding and removing a new empty VLAN on the VTP server (even several times) until the server VTP revision number exceeds the client revision number.

Regards,

Milan