MST on ProCurve 2810-24G

[Len Perdic]

Just playing around with MST and not having much luck. Here is my config...

conf
vlan 10
name VLAN1013
tagged 1-10
vlan 20
name VLAN1723
tagged 11-20
vlan 1
no untagged 1-20
ex
spanning-tree config-name "TEST"
spanning-tree config-revision 1
spanning-tree instance 1 vlan 10
spanning-tree instance 2 vlan 20
spanning-tree
spanning-tree priority 0

When I plug a cable from port 1 to port 11, port 1 shows 'forwarding' but 11 is 'blocked'?

Am I missing the point here?

TIA,
Len

[Pieter 't Hart · Accepted solution]

>>>
When I plug a cable from port 1 to port 11, port 1 shows 'forwarding' but 11 is 'blocked'?
Am I missing the point here?
<<<

Yes you are!

All STP techniques (STP, RSTP, PVSTP MST) designate a switch in the network as "root".
when STP detects a loop, one (and only one) port in the loop (fuhrthest away from the root) will be set as blocked.
The port opposite the blocked port will be forwarding!

As you got a single switch, this switch is the root.
As this switch sees itself connected to itself (by looking at the root-ID) it will block one (single) poort to break the loop.
(The port opposite the blocked port will be forwarding!)

[Ralf Krause]

Hi Len,

yes, you are ... ;-)

Your discovery is based on the nature of MST standard.

Although you have configured different instances for the two vlans, and although any of those vlans is configured on any port exclusively, from a calculations point of view it is not totally seperated from each other.


I'll try to explain:

You have instance 1 containing vlan 10.
You have instance 2 containing vlan 20.

MST performs the spanning tree calculation three time (in this case:
- Once for the IST instance
(which is not relevant in your setup, since
all ports only have configured vlans 10 or 20)
- Once for instance 1 (representing vlan 10)
- Once for instance 2 (representing clan 20)

However, MST does each calculation over the whole physical box!
Meaning: Although vlan 10 e.g. is not configured on port 11, port 11 is part of instance1's spanning tree calculation, nevertheless.

Weird, isn't it?

The trick here is to tweek the spanning tree parameters (priorities, costs, etc.) in a way that makes sure that port is blocked, which does not contain a particular instance's vlan anyway.


In your example:

For instance 1, tweek the parameters in a way that port 11 (in general: ports 11-20) will be blocked.

For instance 2, tweek the parameters in a way that port 1 (in general: ports 1-10) will be blocked.

Then there will be blocked ports, but without any influence to traffic, because those ports are blocked, which have no vlans of the particular instance assigned any way, and consequently don't have traffic destined for those ports.



One last thing:

Be careful with the show commands. "show spanning-tree" without any options only show the IST and CST (often also called CIST) instances.

If you want to have a look on the port status for a particular instance, please use "show spanning-tree instance ").

Hope that helps,
Ralf

[Richard Brodie_1]

As Ralf says, there is a key difference between MST and PVST.

In MST spanning tree is logically 'below' the VLANs. You make spanning trees and map VLANs onto them. You can't affect the spanning tree topology with VLAN filters.

It's a lot harder to get your head around than PVST. You trade efficiency for complexity.

[Len Perdic]

Gentlemen, thank you for the excellent feedback!

There really is more to this than with PVST! I can see it's going to take a bit of playing around to tweak.

What brought about the playing around is this:

In our data centre we have multiple switches with multiple VLANs, the concern is that someone may inadvertantly create a loop in a downstream switch.

[Galen Richards]

Ralf,
You mention "For instance 1, tweek the parameters in a way that port 11 (in general: ports 11-20) will be blocked."

How would you tweak the parameters to make that happen? Eg. On what instance/port would you change the cost or priority and what would you change it to?