VALN locally assigned and VTP prunning and Domain

[zillah2004]

I have got this scenario like this

Switch1:

vtp domain DOGCAT

vtp password CISCO

vtp pruning

vlan 3,5,6,8,10,26,33,52

Switch2:

vtp domain DOGCAT

vtp mode client

vtp password

Requirement:

VLANs not locally assigned should not be received over any trunk links throughout the VTP domain.

Explanation:

By default, all VLANs are allowed to be sent over any trunk link in the VTP domain. Therefore, broadcast frames and frames destined for unknown unicast addresses will be sent over all trunks throughout the domain. This behavior is undesirable when one or more switches throughout the VTP domain receive traffic for VLANs that they do not have locally assigned and are not in the transit path for.

In order reduce this unnecessary traffic VTP offers a feature known as pruning.

VTP pruning allows a switch to communicate with its directly connected

neighbors about what VLANs they have locally assigned and are in the transit path for.

Therefore VLANs that are unnecessary can be “pruned” off of the

interface. VTP pruning can be enabled by issuing the vlan database or global

configuration command vtp pruning.

I am looking to understand the concept of :

1- Can we consider vlan (3,5,6,8,10,26,33,52) locally ? If so why do we consider them localy,,,do we consider them locally because we define them under vlan command in global configuration ? or becaues they are in the same domain (DOGCAT) ?

2- ",,,, frames destined for unknown unicast addresses,,," ?!

3- " transit path for" ?

[amit-singh]

Hi,

1. Locally as you set it up in the vlan datbase of the switch i.e vlan created locally on the switch, of cousre they have to be in the same VTP domain for the VTP advertisement propogation.

2.",,,, frames destined for unknown unicast addresses,,," This situation occurs when a switch purges it mac-address table and by the time it learns the mac again, source continues to send the traffic to a particular destination, these frames travel the whole vlan.

3. Transit path for vlan traffic.

HTH,

-amit singh

[zillah2004]

If i have got another switch in the above scenario

Switch3:

vtp domain DOGCAT

vtp password CISCO

vtp pruning

vlan 13,15,100

Does vlans (13,15,100) consider locally assigned for switch3 only ? though they will be advertised to switch1 and switch2.

Now as we know that switch1 will advertise its valns (3,5,6,8,10,26,33,52) to switch3,,,,,

When I issue the command "prun vlan" global configuration in case of switch3,,,,which vlan will it be pruned off ?

does it prune off the valns (3,5,6,8,10,26,33,52) that have been advertised form switch1 to switch3,,,because they are not locally assigned under global configuration of swtich3 ?

[glen.grant]

Any vlan not assigned to at least one active port and in a up/up status will be pruned . You would not be assigning vlans under global mode on a client in a server /client setup as the vlans are propgated from the server to the client , the only thing that needs to be done on the client end is to assign your ports to whatever vlan you want them in .

[zillah2004]

Any help

[Georg Pauwen]

Hello,

I think your confusion might come from the term ´locally assigned´. To clarify this, conssider the following scenario:

You have 3 switches, Switch1, Switch2, and Switch3. All 3 switches are in the same VTP domain, so the configuration would look like this:

Switch1

vtp mode server

vtp domain CISCO

vtp password cisco

vlan 3,5,6,8,10,26,33,52

Switch2

vtp mode client

vtp domain CISCO

vtp password cisco

Switch3

vtp mode client

vtp domain CISCO

vtp password cisco

Now assume Switch2 and Switch3 are each connected to Switch1 by a trunk link.

VLANs can only be created on the VTP server, Switch1 in this case. The VLANs created on Switch1 are all propagated to Switch2 and Switch3, regardless of wether there is a port on those switches in up/up state assigned to one of those VLANs. Let´s say Switch2 has all ports assigned to VLAN 6 (that is, all ports are configured as following):

interface FastEthernet0/1

switchport access vlan 6

In this case, there would be no real need for all the other VLANs to be propagated to Switch2, because the only VLAN it needs is VLAN 6. This is where VTP Pruning becomes helpful. You enable it globally on the VTP server (Switch1 in this case), and all trunk links are checked for which VLANs are needed at the other end of the trunk connections, which are Switch2 and Switch3.

With regard to your specific question: VLANs 3,5,6,8,10,26,33,52 will be pruned if there are no ports assigned to those VLANs on the client switches (as stated in the example above). As a matter of fact, you cannot even create the VLANs on Swicth2 or Switch3, since they are in VTP client mode.

The only way Switch2 and Switch3 would be able to get the VLANS anyway is when they are configured as VTP servers as well. That might actually be your problem: the default VTP mode is server, that is why you can create VLANs and enable vtp pruning globally. Keep in mind that the VTP server always has the entire VLAN database, and that nothing is pruned off links between VTP servers...

Does that make sense ? Let me know if you still have doubts about the concept...

Regards,

GP

[zillah2004]

Thanks GP for this detailed and easy explaination

What does the word "locally assigned" mean ?

[Georg Pauwen]

Hello,

locally assigned would be when a switchport is configured for a specific VLAN, to come back to the example above:

interface FastEthernet0/1

switchport access vlan 6

In this case, VLAN 6 is locally assigned (but, as Glen stated, only when the port is up/up, that is, when a user is actually connected to the port)...

HTH,

GP

[zillah2004]

Thanks GP,and Thanks for Glen

[zillah2004]

Is there ant special meaning for phrase " tansit path " ?

[Francois Tallet]

I'm sorry that Cisco docs are so cryptic! Putting aside this document, here is my take on a simple vtp pruning explanation:

Suppose that all the vlans are allowed on all the trunks in your network (which is the default). Suppose that you only have two hosts A and B in vlan X in the whole network. This means that vlan X spans the whole domain, but only two stations are using it. If host A sends a broadcast it will be flooded on all the trunks of your network. However, it only needs to be flooded to a single destination B (nobody else cares about vlan X). All the trunks that forward this broadcast and that are not in the direct path from A to B are just wasting bandwidth.

VTP pruning (like its IEEE counterpart GVRP) is a protocol that create a subtree linking all the hosts interested in vlan X. It will thus identify exactly (what the document called transit path) the unique path between A and B where broadcasts needs to be flooded in vlan X and will prevent flooding on any other trunk. Of couse if there are more hosts in vlan X the subtree created by VTP pruning will adapt so that every one of them will be able to receive broadcasts/multicasts/unknown unicasts (anything that is flooded).

Last detail: how do you know if a host is connected to vlan X on a given switch? By default, this is derived from the existence of an access port. An access port in vlan X will advertise its desire to receive flooded traffic in this vlan through the trunks via VTP pruning. I guess that's what they meant by "locally assigned vlans".

Regards,

Francois