VTP pruning.

Unanswered Question
Aug 8th, 2010

In the CCNP SWITCH 642-813 Official Certification Guide by Hucaby, about VTP pruning, you can find:

"Broadcast and unknown unicast frames on a VLAN are forwarded over a trunk link only if the switch on the receiving end of the trunk

has ports in that VLAN."

Suppose that you have 3 switches and this simple topology:   SW1 ----- SW2 ----- SW3 where ----- is a trunk.

Suppose SW1 and SW3 have ports in VLAN 10 but SW2 not.

If I have to interpret the proposition literaly, both SW1 and SW3 prune VLAN 10 from the trunk to SW2 because SW2 has not ports in VLAN 10. I hope this is false. The algorithm through with pruning is deployed must be more detailed and I have a simple idea about this. Is it possible to find proposition "not very logically correct" also in an official certification guide?

Or do you think I am in error?

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Giuseppe Larosa Sun, 08/08/2010 - 05:24

Hello Speculor,

SW2 will not prune itself from vlan 10 until SW3 starts to do it, from the point of view of SW2 at least a client exists that wants to receive broadcast, multicast and unknown unicast for vlan 10 and it is behind SW3

In other words SW2 CAM table contains at least one MAC address in Vlan 10 that points to interface to SW3 (a trunk) so SW2 CAM table for vlan 10 is not a leaf: there is at least one entry in two different ports.

the book refers to a switch that hasn't other switches downstream, alias a switch at the access layer if you think of the campus design guidelines this may be clear in the context or not.

Hope to help

Giuseppe

speculor_cisco Sun, 08/08/2010 - 05:58

Hello Giuseppe, my name is Matteo, from Bologna.

So, the preposition makes sense only if you consider a particular scenario, isn't it?

I wondered how is possible to modify the preposition in order it makes always sense.

Do you think that putting "CAM entry" instead of "port" is sufficient?

I am not sure. Do you know where it is possible to have more information about the algorithm itself?

I had had an idea. Suppose that all switches in a LAN send out all their trunks the set of VLANs locally configured and that,

when a switch receives these sets, it begins to send the union of them. You will see that the only trunks over which these

messages, fixed the VLAN, are sent in both direction must remain active in that VLAN, the others can be pruned. I have tried a few times and it works.

I am not considering the information in the CAM tables, may be the algorithm works also with these entries but I am not sure.

If the algorithm must work also when the CAM tables are empties, then the algorithm can not use those informations.

In my case, which is not so complex, it is necessary a convergence time. If I am not in error, I have read that pruning is deployed with the exchange

of messages between the switches. I do not know if, with the only information in CAM tables and without exchange messages, it

is possible to deploy pruning. Thanks.

Giuseppe Larosa Sun, 08/08/2010 - 06:18

Hello Matteo,

that speculor reminded me of latin language ... 

a VTP prune message exists  sorry if I have been unclear

the message travels in upstream direction towards network core

basically all devices should agree on the fact of using pruning or not.

if pruning is enabled the VTP messages are used to request pruning ( that is to tell "I'm non interested in this vlan X user traffic")

In real world what is important is that VTP pruning can only save bandwidth on inter switch links but it does not provide help in scalability of spanning tree protocol.

This misunderstanding about what VTP pruning does has caused problems in some environments.

Low end cisco lan switches may support 64 or 128 STP instances. Another limit is present for the max number of vlans in VTP database.

The possible effects are:

devices with incomplete vlan database automatically reverting to VTP transparent mode to protect themselves from execessive vlan number.

devices that automatically disable STP for some vlans

the  second one is more dangerous

to really help STP scalability you need to allow only used vlans on a trunk both sides with

switchport trunk allowed vlan

if you do so the use of VTP pruning becomes less important as you have already "manually pruned" the list of vlans permitted

Hope to help

Giuseppe

speculor_cisco Sun, 08/08/2010 - 07:01

Hello Giuseppe and thanks for your informations. I forgot that in the certification guide is written too:

"VTP pruning occurs as an extension to VTP version 1, using an additional VTP message type. When a Catalyst switch has a port associated with a VLAN, the switch sends an advertisement to its neighbor switches that it has active ports on that VLAN. The neighbors keep this information, enabling them to decide whether flooded traffic from a VLAN should use a trunk port."

I was confused about the terms "keep this information", as the author did not write about a possible flooding of this information.

Now, also through this discussion, I have remembered that the message is a multicast and that, may be, for the author it was obvious that the

information was flooded. Without flooding this information there would be some problems with a correct pruning deployed. If these messages

are flooded in the network, then the solution should work, and the explanation is also simpler then the one I gave before. Do you agree?

Giuseppe Larosa Sun, 08/08/2010 - 07:25

Hello Matteo,

yes the message has a well known multicast destination and a well known encapsulation and this provides for easy propagation in the campus network

the following document best practices for IOS switches provide details for all L2 protocols and other useful info

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

see table 1

Hope to help

Giuseppe

Actions

This Discussion