Spanning Tree Issues and VTP Pruning

wayne.goulden · ‎09-09-2003

I have two 6509's at the core and 3500XL's at the access which are dual stacked with gigastacks. Spanning Tree Stackport is configured on the gigastack links due to a dropped backbone architecture.

I've recently noticed that if I add, remove or reload a 3500 switch Spanning Tree seems to re-calculate all VLAN's causing a momentary loss of connectivity across the network. All VLAN's are trunked to all switches and VTP pruning then strips back unused VLAN's. The trunk command on the 6509 is:

set trunk 3/6 nonegotiate dot1q 1-1005

I have completed an Output Interpreter which tells me that VLAN 1 is present on all trunks and should be removed from the trunk. Can this be confirmed?

The issue with VTP is that even though all VLAN's are allowed on a trunk, when VLAN'ing an access switch the VLAN doesn't appear on the 'show trunk'command of the 6509 as in the forwarding state, hence cannot get those users on the network.

There is an identical set-up at another site but the problems are not being experienced. Any ideas?

Many Thanks

Wayne

milan.kulik · ‎09-12-2003

Hi Wayne,

could you explain your topology in details?

What do you mean by "3500XL's at the access which are dual stacked"?

Is there a trunk from one core cwitch to a 3500XL, then stacking connection to second 3500XL and another trunk from the second 3500XL to the second core switch?

STP should not "re-calculate all VLAN's causing a momentary loss of connectivity across the network". You should never loose connectivity across the network if one switch reboots (the only exception it root switch reboot).

Are the trunks configured correctly?

Do you see VLANs on your trunks using sh vlan, sh trunk, sh int .... switchport commands?

VLAN1 should NEVER be removed from trunks. VTP, CDP, CMP and other Cisco service protocols are always sent through VLAN1 on trunks. They should be even if you clear VLAN1 from a trunk (i.e. only user data should be removed) bit I've noticed several bugs in IOS regarding it.

If I understand correctly even when you assign a port on your access switch to some VLANx, VLANx is not shown on the trunk as running. It seems like your VTP prunning is not working correctly. I'd check VTP domain names, passwords, etc.

Try sh vtp status and sh vtp count to see if your access switches are communicating with the VTP server correctly. You can also try to add a new testing VLAN on the server to see if it's propagated to the clients correctly.

HTH,

Milan

mschooley · ‎09-12-2003

who is your root, you may want to make sure your 2 6509's are root, set bridge priority on one to a low number, i.e 128, and the other to 256, the 128 will be your root bridge, if a remote bridge is removed, shouldn't bring down whole topolgy

wayne.goulden · ‎09-15-2003

In response to the two replies so far...

My topology is two 6509's at the core which are etherchanneled. At the access I have dual stacked 3500XL's. One 3500XL trunks to one core switch and the 2nd 3500XL trunks to the other core switch. The two switches are then uplinked with gigastacks. I have Spanning Tree Stackport configured on the inter switch uplinks and spanning tree timers are default across the whole network.

One 6509 is the root bridge with a priority of 8192 and the 2nd 6509 is the secondary with a priority of 16384. I have checked all VTP settings across the whole network and all is OK.

When applying the command "switchport access VLAN XXX" on the access switch the device shows on the switch as up and is present in the MAC address table. When looking on the core switch at the trunk, that VLAN hasn't appeared in the forwarding state, hence no conectivity. In one or two cases a PING from the MSFC to the device has allowed connectivity, only for a stactically assigned IP though. This has only happended on about 3 switches. Everywhere else is fine.

I'm running 6.2(2) CatOS on the 6500's and 12.0(5.4)WC(1)for the 3500's. The MSFC's are running 12.1(7)E.

Thanks for your help so far

cminch · ‎09-15-2003

I was just curious, you are not running in VTP Transparent mode are you?

wayne.goulden · ‎09-15-2003

All switches are VTP clients and the core is the Server. All switches show the correct VTP revision number. Some additional information, only one or two VLAN's are affected by this. If I assign a VLAN to a port then it will not show in the forwarding state on the core trunk. If I try a different VLAN then that will trunk and work. The VLAN that didn't work is however OK on every other switch on the network and will trunk OK.

A reboot of the switch in some instances allows it to work again, however, I'd like to know whats causing it. I was wandering whether a faulty GBIC or gigatstack could cause this to happen?

Thanks

milan.kulik · ‎09-15-2003

How many VLANs are configured in your network?

Don't forget only 64 STP instancies are allowed on Cat3500s.

The effect is when adding 65's VLAN one VLAN will stop running STP (one of them, not necessarilly the new one) and you can get to trouble with similar symptoms.

You can prevent this problem via switchport trunk allowed vlan .... command on your Cat3500.

Regards,

Milan