Solved: configure cisco 3750 mstp with other rstp

yoram12345 · ‎03-02-2009

Hi

I have a triangle topology of 2 cisco + other switch :

2 cisco are configured as mstp with vlan 10 on instance 1

the other switch is configured rstp

the problem is that i get 2 root in the topology

and the also when i change teh priority of the other switch nothing is changed

Is there a specific config to support the cisco both for mstp in one port and rstp on the other one ?

can anybody suggest me a solution

Thanks

Francois Tallet · ‎03-03-2009

You're not giving me enough information to answer for sure, but I think you've been hit by the "sync" mechanism, which is common to RSTP and MST.

During its reconvergence, those protocols might have to sync the non-edge ports in order to prevent transient loops. Designated port connected to a peer RSTP or MST bridge send a proposal and quickly gets back an agreement. If a port is edge (i.e. configured for portfast), it is connected to a host and cannot introduce a loop, so it is not synced.

The problem is that if a port is edge but not configured as such, it is blocked by the sync and takes 30 seconds to come back to forwarding. With RSTP and MST, it is thus very important to make the proper "portfast" configuration on the port leading to hosts. That's what I was alluding to at the end of my previous answer, as I did not know exactly what kind of behavior you observed. I guess you were doing some form of ping from a host connected on a port that did not have portfast configured on it. In that case, it is perfectly expect to have 30 seconds of traffic interruption.

If this was not the case, then please come back with additional information. The vlan to instance mapping, RSTP or MST should be irrelevant for the discussion;-)

Regards,

Francois

View solution in original post

Yudong Wu · ‎03-02-2009

You can only config one mode by using "spanning-tree mode". why don't you convert them to use the same mode?

yoram12345 · ‎03-02-2009

The thing is that the other switch support only RSTP , where the cisco need to be configured as MSTP .

This is a simulation of a network where some of teh switches w\can work only MSTP and some other vendors only RSTP.

I understood from cisco manual that it can be done , isnt it ?

Giuseppe Larosa · ‎03-03-2009

Hello Yoram,

the other switch is implementing a single instance of 802.1W rapid STP.

what happens for vlans associated to instance 0 ?

they may be able to communicate only in instance 0 not on instance 1.

try this way.

Actually informations for instances > 0 are carried inside the IST instance (0 instance) BPDUs as record fields.

For this reason I would try with instance 0.

probably the mono instance device is ignoring the record field for instance 1 and saw two root bridge are elected.

Hope to help

Giuseppe

yoram12345 · ‎03-03-2009

hi Giuseppe,

The thing is that when i map the vlan to instance 0 , the swithcing time of traffic is 30 sec .

But on instance 0 it was < 1sec

Francois Tallet · ‎03-03-2009

Hi Yoram,

The thing with MST is that you can only do vlan load balancing within a region. Here, you are redundancy between the region and an external switch to the region. You don't have redundancy *in* the region. All this to say that you can certainly configure some vlan to instance mapping (if this is a first step to a more complex network), but at that stage it is of absolutely no use because you won't have vlan load-balancing.

I guess that in your setup, both MST switches are in the same region. From the external RSTP bridge, they will look like a single virtual RSTP bridge. If the MST region is the root, then this virtual bridge will be root. In any case, a link will be blocking between the RSTP bridge and one of the two MST bridges (never between the MST bridges).

RSTP and MST interact seamlessly: everybody runs instance 0. In your MST region, the vlans mapped to instance 1 are following instance 1 on the link between the MST bridges, they are following instance 0 on the link going to the RSTP bridge.

If you want a behavior that is exactly RSTP in your whole network, make sure that the MST bridges are not in the same region. Just put a different region name. Actually, the IEEE suggest putting the mac address of the bridge in the region name by default, so that MST bridges act like RSTP bridges by default. We did not do that. So if you put your MST bridges in different region, all the vlans will be following instance 0 on all the link, and it will look like plain basic RSTP.

Now, if convergence time is not good (I think that what you were suggesting in another email), then you'll have to give us more information on how you are measuring it (between what port and what port? what are the states of the ports on the path between the hosts that are trying to communicate to each other? etc...).

Regards,

Francois

yoram12345 · ‎03-03-2009

hi Francois ,

Thanks for the info.

We solved the swithcing time by the following ;

MSTP was configured on vlan 0.

we used debug level and noticed that the ports to the teq are blocked.

so we used portfast to prevent that.

Can u expalin this ?

By the wasy what is the correct definition of MSTP instance when interact with RSTP ?

I assume the MSTP should be instance 0 isnt it ?

Yudong Wu · ‎03-03-2009

When portfast is enabled on the port, the port will be a edge port, so it won't be blocked.

Yes, MST should use instance 0 to interact with RSTP.

Francois Tallet · ‎03-03-2009

You're not giving me enough information to answer for sure, but I think you've been hit by the "sync" mechanism, which is common to RSTP and MST.

During its reconvergence, those protocols might have to sync the non-edge ports in order to prevent transient loops. Designated port connected to a peer RSTP or MST bridge send a proposal and quickly gets back an agreement. If a port is edge (i.e. configured for portfast), it is connected to a host and cannot introduce a loop, so it is not synced.

The problem is that if a port is edge but not configured as such, it is blocked by the sync and takes 30 seconds to come back to forwarding. With RSTP and MST, it is thus very important to make the proper "portfast" configuration on the port leading to hosts. That's what I was alluding to at the end of my previous answer, as I did not know exactly what kind of behavior you observed. I guess you were doing some form of ping from a host connected on a port that did not have portfast configured on it. In that case, it is perfectly expect to have 30 seconds of traffic interruption.

If this was not the case, then please come back with additional information. The vlan to instance mapping, RSTP or MST should be irrelevant for the discussion;-)

Regards,

Francois

yoram12345 · ‎03-03-2009

hi Francois,

Thanks, you have ansered my all questions

Yoram

yoram12345 · ‎03-04-2009

hi francois,

Only one question.

why the switching time was 1 sec when i configured the traffic vlan on port 1 (without defining the portfast to the host).

There is a different behaviour when the vlan is on instance 0 or 1.

when it was instance 0 i needed to configured it on port fast to the host as u stated .