MST PVSTP interoperation

[skyraven_ro]

Hello,

I've read the "understanding MSTP article" from cisco's website and I have several uncertainties.

These have deepened even more after performing several experiments.

I've setup a test scenario with the following configuration:

C1 Cisco PVST ---- C2 MST ---- C3 MST non-cisco switch (doesn't know PVSTP)

C1 has vlans 1-10 for which it must necessarily be root bridge

C2 must be root bridge for all other vlans

C3 will transport 1-10 + the other vlans

C1 can not be migrated to MST while C2 if possible should interoperate with the non-cisco MST enabled switch.

What I have done:

Setup 1:

C2 mst root bridge for all vlans

C3 learnt of C2 being the root bridge

C1 the PVST also learnt this (as far as I've read all communication with the PVST is done via the IST-CST instance)

Although this worked just fine, unfortunately it wasn't what I was searching for.

Setup 2:

C1 -- C2

C1 lower priority for vlans 1-10 (disabled spanning tree on the other vlans / or removed them from the trunk to C2 if not required to be present there)

C2 reported:

SPANTREE-2-ROOTGUARD_UNBLOCK: Root guard blocking port ...

And the port was shown as blocked in both the IST0 and the other MSTIs.

I've also tried the alternate configuration (not recommended) from:

http://www.cisco.com/warp/public/473/147.html#alternate_configuration

without any luck.

Disabling PVSTP on the C1 interface to C2 of course made the C2 port to be removed from blocking as expected.

I have several questions in regard to this:

a) Why are both IST0 and the boundary ports for the MSTIs placed in Blocking ?

b) According to that article shouldn't there be a way to have the PVST be root bridge for all the instances present on it ?

c) What alternate setup could there be possible to achieve the redundancy desired while maintaining C1 root bridge for vlans 1-10

C2 root bridge for the others

C3 interoperability with C2 (C3 only knows MST and RSTP)

Any advice would be greatly appreciated.

Thanks,

Mihai

[Francois Tallet]

Hi Mihai,

The code was designed especially to prevent what you are trying to do:-( The problem is that C2 is only running one instance at the boundary to C1, the CIST. So for each of its ports leading to C1 it can only block all vlans or forward all the vlans.

If C1 is root for certain vlans, C2 will have to block one of its port to C1 for those vlans. This means that C2 can only block ALL its vlans to C1, considering the rule stated above.

On the other hand, if C2 is the root for the CIST, it will need to put both its ports to C1 in forwarding, which means put ALL the vlans to forwarding on both ports.

You clearly see the contradiction, and that's what the inconsistency you are getting is trying to show.

Why do you need C1 to be root for some vlans? Is that for some load balancing issues? Because you can achieve load balancing without having C1 being the root.

The only solution to your problem seems to have C2 run PVST considering your constraints. It would be much better if you could move C1 to MST and have it participate in the same region as C2 of course...

Regards,

Francois