tuning spanning-tree

[global_trekker1]

I'm in the process of migrating my network from a hub-and-spoke with no redunancy to a fully redundant fiber ring. Due to geographical challenges of the campus and the resources I have available, I will have up to 10 switches in a single chain. I understand that good spanning-tree design recommends no more than 7 switches in a chain. This is my question. Is it possible to tune the spanning-tree timers so that the network is stable? I also have plans to convert to rapid-pvst.

I've posted my network diagram.

When the network's stable the largest loop will be 8 switches. If either of the links off of SW1 fail, the largest loop will be 10 switches.

Thanks in advance

Mike

[Giuseppe Larosa]

Hello,

the answer is yes.

With traditional STP 802.1D the default timer values are set for a 7 switch chain

hello timer = 2 sec

listening, learning = 15 = 7*2 +1

max age timer = 20 seconds = 7*2 +3*2

you need to change the values so that you satisfy the same cases.

For example listening, learning state timers should be 21 sec

You can use the following document as a guide

http://www.cisco.com/en/US/docs/switches/lan/catalyst3550/software/release/12.2_44_se/configuration/guide/swstp.html#wp1150592

.

see table 15-4

the key point is to allow enough time for a network topology change to propagate in the network.

In the same document you can also find the commands to move to RSTP+

Hope to help

Giuseppe

[Francois Tallet · Accepted solution]

That's would be the right moment to use the "root" macro. This macro is wrongly interpreted as a way of specifying the root bridge. In fact, its main use is that you have a diameter option that will do the timer tuning for you. Try "spanning-tree root vlan X primary ?"

Also, the timers requirement for STP are *really* conservative. There are considering that a bridge can take a second to realize a BPDU has been received, lose an additional second to transmit its new BPDU and that up to 2 BPDUs in a row can be lost at any stage of the propagation. Those thing don't really happen any more, so I don't think you would have any problem with the default timers in fact;-)

RSTP is better because BPDUs are generated locally by all the bridge (STP BPDUs were generated by the root bridge and relayed by the other). So the age in the BDPU became a hop count and there is no problem running your network with Rapid-PVST and absolutely no tuning. And anyway with Rapid-PVST, even if you tuned the timers, this would not have much impact as the protocol reconverges without using them!

Regards,

Francois