[c-nsp] MST Problem on Cat 3560 Switch

[Andrew Fort]

Alexandra Alvarado wrote:
> Hello,
> 
> I have a MST network composed of too many catalyst 3550 (Version
> 12.1(14)EA1a) switches, and until now all was ok, but recently I replaced
> the root switch with a cat 3560 Version 12.2(25)SEB2) switch.  The problem
> is that this switch generates too many topology changes, (approximately 100
> in a day), the majority of this changes are not seen on the peer switch.
> Example:
> 
> Switch1 -------------  Switch2    ---------------- switch3
> 
> In The connection between switch2 and switch3, you can see topology changes
> on switch2 but there are no changes on switch3.
>
> What could be the cause of this problem?

Though this is tangential and probably not the cause of your problem, 
we've run across a few MSTP issues that we've seen on peer switches of 
the switches causing the fault, recently.

Get TAC onto it, we've been quite impressed with the level of clue 
w.r.t. STP issues.  That having been said, STP still sucks for SP's, but 
hey :).

Here's two of the recent bugs we've hit which have caused bizarre (not 
usually outage inducing) behaviour.  These are recently found problems 
in the 3550 code:

CSCed65818 (temp STP loop when running RSTP (MST or RapidPVST))
CSCeh16869 (MST does not sync on Desg port)

The second one obviously can cause some grief; but IIRC it has only been 
shown to occur when the 3550 is the root bridge.

Also, a question :)
How do you do maintenance on 100 MSTP configurations, incidentally, 
without causing massive convergence events (boundary ports and the 
like)?  I'm assuming you pre-configured so you never have to touch it; 
is this the case?

-andrew