MST with multiple regions problem

[dariopalermo]

Hi all,

in a large ring network topology (up to 34 different sites), how should the mst be configured? Single region or multiple?

We're going with multiple regions, with each site's switch being the root for it's local region and having 1 additional instance for it's local VLANs. Something like:

spanning-tree mode mst

spanning-tree mst configuration
 name RING_n30
 revision 10
 instance 1 vlan 530, 580

the switches are connected on vlan 104 and node 33 is the CIST:

spanning-tree mst configuration
 name RING_n33
 revision 10
 instance 1 vlan 533, 583
 instance 2 vlan 500

spanning-tree mst 0-1 priority 28672

On vlan 5000 we have the uplink to another whole network, with the other switch using pvst.

We're having troubles with this setup: shutting down one ring port on node 33, breaking the ring, will cause the 21th switch in the chain to go nuts. max hops are left to the default value of 20 but it's not important with multiple regions. We increased the max age to 40 but still no changes. On the 20th switch we can see the port going to the 21th flapping between forwarding and blocking state.

We could switch to a single region with 34 mst instances (one for each site, to set the site's switch to be the root for the local vlans, leaving vlan 104 on mst0 and setting node33 to be the root for instance 0, like it is now) and set the max hops to 36 (in a site we have 1 switch not directly connected to the distribution switch), but I think the current setup is more convenient and something has to be done to tune the the inter-regions operations.

Any advice?

Thanks

[Peter Paluch]

Dario,

Whether it's reasonable to have your network consist of 34 different MST regions, or just a single MST region, depends on what your needs are. Having a separate region allows you to have a separate set of instances and their roots in each region - whether that is of advantage to your situation, only you can tell. I can even imagine questioning the choice of having such a large switched topology. Is there no option of making the inter-region links to be routed instead of switched?

Nonetheless, your current problem intrigues me. You are quite correct that spanning-tree mst max-hops influences internal (intra-region) network diameter while spanning-tree mst max-age influences the "outer" diameter if each region is treated as a single switch. You have also said that even increasing the MaxAge to 40 did not remedy the issue.

My first question is whether you have increased this timer on the CIST Root Switch, or even better, on all switches in your network to keep a consistent timer configuration. Keep in mind that it is the root switch that asserts its timers, so in your case, at least the switch that has become the root of the entire CIST must be configured with the MaxAge of 40 so that this setting is asserted over other switches. To keep configuration consistent, however, I strongly urge you to configure all your switches with this setting - even if it means configuring over 34 switches.

My second question is concerned with getting some debug data from the link between the 20th and 21st switch (the link that is flapping). Can you please post the output of show spanning-tree interface interface-id detail from both switches while the link is in stable state?

Just wondering, if the ring is closed, is this link between the 20th and 21st switch the blocked link?

Looking forward to hearing from you

Best regards,
Peter

[dariopalermo]

Hi Peter

I increased the timer on all switches (29 at this time, 5 are to deployed in the next few months). Using ciscocmd helps a lot ;)

The spanning tree blocks the port between the 14th and 15th switches (as it is supposed) when the whole ring is operational.

Tonight I've got one switch offline (probably a power related issue in a problematic site), so I cannot get the info you asked me to get. I'll try tomorrow.

Regarding the routing solution, I'm planning a migration process that will not disrupt connectivity during the process. Still, I'm curious about why the current setup doesn't work correctly!

Bye,

[dariopalermo]

On the 20th switch (1c6a.7a7c.5c00): 

Port 25 (GigabitEthernet1/1) of MST0 is root forwarding
   Port path cost 20000, Port priority 128, Port Identifier 128.25.
   Designated root has priority 28672, address 1c6a.7a7c.af80
   Designated bridge has priority 32768, address 1c6a.7a7c.8200
   Designated port id is 128.26, designated path cost 160000
   Timers: message age 5, forward delay 0, hold 0
   Number of transitions to forwarding state: 534
   Link type is point-to-point, Boundary RSTP
   BPDU: sent 2368, received 106572

On the 21th (1c6a.7a7c.8200):

Port 26 (GigabitEthernet1/2) of MST0 is designated forwarding
   Port path cost 20000, Port priority 128, Port Identifier 128.26.
   Designated root has priority 28672, address 1c6a.7a7c.af80
   Designated bridge has priority 32768, address 1c6a.7a7c.8200
   Designated port id is 128.26, designated path cost 160000
   Timers: message age 0, forward delay 0, hold 0
   Number of transitions to forwarding state: 476
   Link type is point-to-point, Internal
   BPDU: sent 106602, received 2364

Nothing strange, right?

Bye

[Peter Paluch]

Hi Dario,

I apologize - I needed to drop out from this thread for an entire month. I am sincerely sorry.

You have probably moved on with this problem but I just wanted to ask if there is anything new or interesting you have discovered during solving of this problem.

Regarding the output you have provided above, there is something strange there: Notice that the 21st switch tells you that the designated BPDU claims that the Message Age is 0. That is your clue into this instability.

Best regards,
Peter

[dariopalermo]

Hi Peter, no need to apologize, we all have our things to do! ;)

Yes, I moved on... I had to set all the switches in a single MST region (with max hops set to 36). I had to stabilize the network because is already a production network and we were having too much issues.

Thanks anyway!

Bye