Root switch conflict in PVST+

Answered Question
May 19th, 2008

I have 40+ switches including 4 CatOs 6513 switches in VTP server mode with more than 100 vlans.

Most of the access switches don't have all vlan configured in their access ports. when I issue SHOW SPANTREE VLAN 4 on the access switch which is vtp client ( CONSIDERING no access port configured on this switch with vlan 4) it claims itself as root switch for vlan 4.

When I issue same command on the other switch having access port configured in vlan 4, shows actual root switch ( on of the core switch) for vlan 4.

Please anyone help me out from this.

Appreciated.

I have this problem too.
0 votes
Correct Answer by Francois Tallet about 8 years 8 months ago

I guess that when you do "show spantree vlan 4" on your access switch, you are indeed seeing some ports active in vlan 4 (whether they are access or not is irrelevant).

If you switch is showing itself as a root for vlan 4, it simply means that it does not receive the superior information generated by the root bridge on the core. This, probably because you manually pruned vlan 4 from the uplink leading to this core bridge. On the bridge where you are using vlan 4, you probably enabled vlan 4 on the uplink, hence the correct root information.

That should not cause any problem. Your network is de facto partitioned for vlan 4, it is normal to have one root bridge per partition.

Regards,

Francois

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (1 ratings)
Loading.
glen.grant Mon, 05/19/2008 - 15:06

Check to see if it is a smaller switch like a 2950 which only supports like 64 stp instances per switch , if you try to feed it more than it will put itself in transparent mode and it will then think it is the root for anything on that switch . If you are running into something like this "manually" prune off unneeded vlans off the trunk (good practice anyway) .

subharojdahal Mon, 05/19/2008 - 15:58

Thanks for your reply

Yes, I have configured some of the switch uplink (trunk port) to allow only vlans which are configured on local switch access port ( local VLANs).

But I doubt that even i mannually pruned the vlan still switch claims root bridge for that vlan that are not configured in access port of the switch.

So, Any other idea to resolve the issues.

Correct Answer
Francois Tallet Mon, 05/19/2008 - 16:13

I guess that when you do "show spantree vlan 4" on your access switch, you are indeed seeing some ports active in vlan 4 (whether they are access or not is irrelevant).

If you switch is showing itself as a root for vlan 4, it simply means that it does not receive the superior information generated by the root bridge on the core. This, probably because you manually pruned vlan 4 from the uplink leading to this core bridge. On the bridge where you are using vlan 4, you probably enabled vlan 4 on the uplink, hence the correct root information.

That should not cause any problem. Your network is de facto partitioned for vlan 4, it is normal to have one root bridge per partition.

Regards,

Francois

subharojdahal Mon, 05/19/2008 - 17:21

Could you please clear my concepts about inter-vlan routing in multilayer switch?

There are four 6500 switch.

Switch A, B are HSRP pair and stp root for 20 VLANs and C and D are core switch, HSRP pair and STP root for more than 40 VLANs.

routing protocols is eigrp

RAPID PVST+ is enable in each core switches.

Current topology

A<--->B C<---->D

B<--->C

My objective is to make full redundant link in four switch. If I add link as

A<--->D

A<--->D

B<--->D

B<--->

and make all newly added link as trunk similar to old one.

Does it work ?? If not what should I do?

I would appreciate.

paul-y Tue, 05/20/2008 - 07:37

I may not understanding your setup correctly but why would you want to tie A,B and C,D together unless you want to extend your vlans across to those layers.

paul-y Tue, 05/20/2008 - 08:17

I read your previous post and understand what you are trying to do now.

Seems like you are running collapsed core setup. You can use L3 links to connect them together and use EIGRP for load-balancing between AB and CD communication if you can't afford the downtime to consolidate AB vlans on CD.

subharojdahal Tue, 05/20/2008 - 10:33

Your solutions still works if I want to expand my some of the vlan across the switch.

Lets say I want some of the nodes on vlan 134 on switch A and some any other switch.

Could you sugeest me the exact links needed among those switches ?

paul-y Tue, 05/20/2008 - 11:05

With only L3 links between block AB and CD, how are you going to expand any of your vlans from AB to CD without L2 trunk between them ?

Without understanding how your network traffic flows, it is hard to give you the exact link setup.

A <--> B

C <--> D

A <--> C

B <--> D

Assuming AB and CD are both running HSRP, and A , C are both primary gateway.

You can fully mesh them if you have the interface to spare incase of C fails and D takes over and traffic doesn't flow from D --> B --> A.

subharojdahal Tue, 05/20/2008 - 11:28

Paul

Thanks for the question. You guys are absolutely right. But My question is little different. Please go through it once.

Switch A and B are connected.

B is connected to D, D is conneced to C

physical topology like

A C

B D

The switch C and D are connected with WAN VLAN. When I issue SHOW IP ROUTE command in MSFC card of each switches, I found route for each and every VLAN as well as default route.

The Switch C and D are connected with firewal or router ( which i dont wanna bother ) via vlan 16 (WAN VLAN). Each MSFC has default route that forwards packet to IP address of VLAN 16 ( WAN VLAN.

The curret setup is running properly.

My question is if I add redundant link among those swithces, configure its trunk port exactly same as before and let the EIGRP do its work, THEN WHAT IS THE EFFECT IF DOING SO ?? THAT IS MY QUESTION ???

As each trunk link are in native vlan 100 and I have knowledge of root bridge for each vlan.

Switch has CatOS and MSFC (IOS) native mode in all switch.

paul-y Tue, 05/20/2008 - 11:54

Let me make sure I understand this correctly.

C <--L3--> Net

D <--L3--> Net

C <--L2 Trunk--> D

B <--L2 Trunk--> D

A <--L2 Trunk--> B

Please let me know if I understand this correctly.

subharojdahal Tue, 05/20/2008 - 12:32

Yes

C and D have connection to WAN VLAN with IP block size 16 ( /28).

I could see EIGRP route in all of the MSFC ( all switches).

Please let me know if you need more information.

paul-y Tue, 05/20/2008 - 13:34

I would assume C is primary gateway for internet and D is backup and you don't need to extend AB vlans to CD or vice versa.

A <--L3--> C

A <--L3--> D

B <--L3--> C

B <--L3--> D

A <--L2 Trunk--> B

take out B <--L2 Trunk--> D

I don't think extending vlans from AB to CD is a good idea while AB is also aggregating AB vlans. This is hard to scale in my point of view.

You can leave the B <--L2 Trunk--> D to extend the vlan, however, traffic flow will need to be well planned to avoid unpredictable outcome.

subharojdahal Wed, 05/21/2008 - 05:56

Your solution sounds great.

At the same time i have couple of vlan like voice and device management vlan that span over the core switch.

If I consider your solution, do you have ways out to streamline the flow of traffic in those VLAN that spans across the switch.

paul-y Wed, 05/21/2008 - 06:26

Technically spanning AB and CD vlan across each other is a very bad practice if they aren't in the same switch block design. If it is a must,

I would do :

Remove all L3 links between AC AD and BC BD.

Create a square loop topology where

A <--L2 Trunk--> C

C <--L2 Trunk--> D

D <--L2 Trunk--> B

B <--L2 Trunk--> A

All vlan routing terminates on both CD and make AB a pure L2 switch. This design keeps it simple to manage. The L3 approach will make network management much harder as you have to deal with EIGRP between 4 routers.

Hope this helps.

subharojdahal Wed, 05/21/2008 - 06:58

If I have truely L2 link ( trunk link) among the switches, and EIGRP routing enabled in MSFC of those switches, I am seeing EIGRP learned route from other switches while issueing SHOW IP route command.

Does it mean the connection is not L2 rather its layer ?

I DONT see any vlan interface configured in swithc A and B that are also in switch C and D. But still the switch A and B are learning EIGRP route from C and D.

May be I am messed up with Layer 2 and Layer 3 link?

paul-y Wed, 05/21/2008 - 07:06

I would recommend to go through all the vlan trunking you have between those switches. EIGRP is probably communicating through a particular vlan in the trunk. You may run into unpredictable traffic flow where you have a particular vlan in CD that is L3 terminated on AB and then AB routes the traffic through another vlan that interconnects back to CD for L3 routing.

Actions

This Discussion