CST-RSTP+ question (BPDUs over the 802.1q trunking) for Francois Tallet :-)

[cristip]

Hi Francois and everybody else

Please have a look at the attached Visio and see if you could plese help me with an aswer

Thank you

Cristian

[Francois Tallet]

Hi Christian,

I'm honored that you are asking me directly, but you should not;-) because you are turning down other members who could have a better answer or who would me more available!

There is a big unknown that prevents me from answering your question. Is the CST implemented by third party bridges running an IEEE standard (STP, RSTP, MST), or is it a Cisco MST region?

I'm going to assume that this is a region implemented with real (third party) IEEE bridges.

PVST bridges run one instance of STP per vlan. The CST region just runs one instance (... the CST!). The vlan 1 on the PVST region interact with the CST. That's why PVST is sending untagged BPDU (not over vlan 1, it's vlan 1 BPDUs, sent untagged): to interact with the CST bridge that just send and receive untagged BPDUs. So basically, vlan 1 will see all the bridges in the network. All the bridges will elect a unique root bridge that can be located in either PVST region or the CST region.

Now, for all other vlans except vlan 1, only the PVST bridges are running a spanning tree. That means that from the perspective of vlan 2 for example, it's as it your two PVST+ region where connected by a hub (the CST region looks like a hub). The PVST+ BPDUs are sent tagged and are thus treated like data traffic by the CST region.

As a result, the PVST+ region will see each other through the CST region. Each vlan instance will elect a unique root (provided there is connectivity through the CST in their respective vlan). The root can only be in either PVST+ region.

Hope this helps;-)

Francois

[cristip]

Hi Francois

Everybody else is free to try an answer, no question about this.

Yes your assumption was correct.

I totally understand what you are saying but because the way I understood the situation is confirmed now I want to ask you if in fact this interoperability can not lead to loops considering that the VLANs in the three structures can comunicate using the 802.1q trunks

[Francois Tallet · Accepted solution]

The configuration shoud not lead to loops, or at least, not lead to permanent loop.

The simplest is keep assuming that the CST region is a group of STP unaware bridges (some kind of hubs). PVST is able to compute a loop free topology taking a hub into account. Now, there is a problem with STP and hub that can come into play here.

Suppose you have the following simple topology:

Bridge A 2--- HUB--.

1 |

| |

1 |

Bridge B 2|-- HUB--.

Bridge A & B are redundantly connected via a link A1-B1 and via another link A2-B2. The link A2-B2 goes through 2 hubs. Suppose port B1 is blocking, everything is fine.

Now, the link between the two hub breaks.

Bridge A 2--- HUB--.

1

|

1

Bridge B 2--- HUB--.

Port B1 has to move to forwading. However, bridge B could not detect the failure of the path through the hubs. It has to wait for max-age before it can consider putting port B1 as a root port, replacing the previous root port B2. Performance are not that good, but so far so good. B1 eventually goes to forwarding and connectivity is re-established between A and B.

Now suppose that the link between the two hubs comes back up. Again, bridge A and B have no clue that this link came up (they don't see a link change, as the link coming up is behind the hubs). The problem is that when the link between the hub become operational, ports A1 A2 B1 and B2 are forwarding. There is a short bridging loop until B2 receives a BPDU from A2 and blocks B1 again.

This kind of problem is exactly what may happen in your network. Your PVST+ region is redundantly connected to the CST region. If the CST region splits, the PVST+ region will not see a redundant path any more through the CST region. It will open its ports. Now, if the CST region heals (and that's something that can happen automatically because the CST region is actually running STP!), then you will have a short bridging loop between the CST region and the PVST+ region, at least until the PVST+ region sees its BPDUs going through the CST region.

Regards,

Francois

[cristip]

Excellent exemple, I anticipated the existence of possible loops but I couldn't imagine the analogy with the big hub for the CST cloud.

Thank you again

Cristian