When does Spanning Tree block traffic?

[mvandorp]

Hi All,

In a larger network, we have numerous VLANs. STP is used to prevent loops, though there shouldn't be any.

We are planning on removing a switch, but this switch appears to be the root bridge for a specific VLAN. We don't want this VLAN to go to blocking/learning/listening when removing this switch.

The questions:

*) Will ports go to blocked when changing the root bridge (configuring a better priority on another switch in the same VLAN)?

There is no topology change (yet), so the BPDUs will not have the 'topology change' bit set.

*) When there is a change in topology (i switch off the device), what ports will go into blocking?

I am not really in a position to test, but need to verify before actually making the changes.

Thanks for any answer,

Marcel

[Kevin Dorrell]

If the switch you want to remove is the root, I'm afraid you are going to have to grit your teeth and suffer the outage. Whether or not you elect another bridge to be root first. Not even rapid-pvst is going to help you much for a change of root.

While you are about it, you might like to consider carefully where you actually want the root, so that you have less disruption next time.

There is one thing that concerns me about your posting; You say that STP is used to prevent loops, "though there shouldn't be any". If that is the case, then you don't have any redundant links. In that case, when you remove the root switch you will have a discontinuity in your network: your network will be broken in two. Unless the root switch you are using is actually at the edge, in which case one could ask why you have an edge switch as root.

Kevin Dorrell

Luxembourg

[mvandorp]

Well.... Because I'm implementing redundancy, I need to make changes in the network, and then I find out about those strange places of root bridges.

(I'm an external consultant at this particular site). We now do have a better STP plan, but we need to migrate to the new situation.

This particular switch is indeed at the edge, but as it is a very old switch, it used to be a more centrally deployed one. Spanning tree is quite default for a large part of the network.

FYI, redundancy can also be implemented with routers and firewalls in failover configuration (as we do here). In layer 2 perspective, those are not loops...

Fortunately a lot of STP instances are configured fine, it is just those exceptions that makes live complicated :*)

Can you explain why ports will go to blocked, when changing the root bridge? There is no change in topology, just a few changes in port roles (a designated port can become a root port, or vice versa. With no loops, there are currently no blocked ports).

Thank you for your reply,

Marcel

[Kevin Dorrell]

Marcel,

That is indeed one of the strange features of the STP algorithm. In the absence of any priority settings, it is the lowest MAC address that becomes the root. The lowest MAC address is often the oldest switch, and more often than not one that has migrated out to the edge.

When you change the root bridge there is effectively a toplogy change. There may not be the normal TCN, but there is a new root election and so there is a full convergence, and that takes time. Not only does a port change role, but it has to be listening for a completely different root. Until the Spanning-Tree has stabilised, nobody knows where anything is.

Kevin Dorrell

Luxembourg

[paul.matthews]

Kevin is right - there will be an impact. Removing the root bridge is probably the biggest topology change you can make.

You may be able to minimise the impact by first setting the priority low on the current root, and then slightly higher on the next bridge in the line, though I won't make promises that there will be no impact!

[mvandorp]

I was afraid of it, that's why I asked this specific question. It is hard to get some info on bitlevel about this; it's al generic behaviour.

Well, I think we are going to decide to do this change in a maintenance window, but thanks a lot for the feedback!

Marcel

[paul.matthews]

About the only thing we can be certain of with network is that things can be uncertain!

ANY change should relly be done in a maintenance window relly, no matter how sure you are about what will happen.

Paul.

[Francois Tallet]

I like when I can make some shocking statements like this;-)

If there is currently no blocked port in your network (that's something that I think I have understood from your description) then the STP reconvergence that will result from the removal of the root bridge will not introduce (even temporarily) a blocked port in your network. You should not experience any direct loss of connectivity in your network.

With STP, a port can only move to blocking when it receives better information on a port that is not its root port. That cannot happen without redundancy.

Even though I don't expect anyone to sue me, I'd rather introduce some fine prints;-):

- this is only true for STP (RSTP or MST would introduce some sync), only if you have no blocked port. If you have some blocked ports, they may move slowly, which means that you might experience some temporary loss of connectivity.

- you will have a topology change. This should just result in the aging time to be lowered. It could introduce a little bit of flooding (should not be an issue, specially with STP).

[paul.matthews]

Frank, I wouldn't be so confident!

Imagine a three bridge network A-B-C. A is currently the root. Prio 4000. B has prio 32768 and C has 16000. Lose A, and suddenly the best root as far as B is concerned is out the other side.