looping in mstp

[quest]

hi,
i am facing a problem when configuring mstp in set of moe than 4 switches

S2 S3

| / |

S1------------------------S4



Consider a setup as shown (derived from the 5 mesh)

In a 4 switch setup (S1 connected to S3, S2 not connected to S3 (connection removed for simplicity), all bridge priorities P2), initially S1 is the root with all 3 of its port being designated. We change priority of S2 to P1 (P1
Now when we change priority P1 (of S2) back to P2 (to make it the original configuration), S2 transmits a BPDU reflecting this. Now S1 in turn transmits it own BPDU (with P2:S1), to S2, S3 & S4.

But as soon as this happens, I also noticed that S4 in turn transmits a BPDU back to S1 with information saying it has access to P1:S2 (got from its link between S4 & S3), with a cost 60000 ( i.e via S3-S2-S1). this is because the old message from S3 (containing P1:S2) is recived at S4 & is not aged out.
This thus makes the link of S1 connected to S4 as a root (originally it was supposed to be designated). S1 in turns transmits a BPDU (with P1:S2) back to S2 & S3. S3 again forwards this BPDU to S4, & S4 back to S1 (each with incremental root cost). Thus we keep getting BPDUs with the TC bit set, causing a lot of instability. This goes on till the message ages out. After aging out the convergence occurs almost instantaneously.

This scenario is aggrevated when more switches are present in the network.

When I reduced the max age of the message, the convergence happened much faster (since the aging out happened sooner).

Is there any solution to this problem or is there a way i can optimize the situation.

Thanks in advance,
Kurian.

[quest]

oops....I forgot to attach my email id.
For any clarifications do mail me at kurian_abby@yahoo.com

[Matt Hobbs]

What switches and firmware version are you using? There have been a lot of spanning-tree related fixes in the latest firmware releases, E.10.23, M.08.86, I.08.87 & H.08.86... if you're not using the latest, upgrade and test again.