HP Switch - Spanning Tree

[Chuckster]

I've had Spanning Tree enabled for some years and only now am I looking in to this a little deeper.

 

Comparing a few switches I noticed that the Switch Priority is set at 32768.

 

What is the ideal number for the core and edge switches?

 

In the Cli, the description for Priority is:

 

Set the device STP priority (the value is in range of 0-61440 divided into steps of 4096 that are numbered from 0 to 15, default is step 8). Not applicable in RPVST mode.

[ass17]

Lower the value on a switch would become the ‘root bridge’ switch. So a value of either 0 or 4096 suffice for the core. I don’t use 1 as some switches don’t like it when PVST (per vlan spanning tree).

 

Take a read of this which might help:

 

https://www.networkworld.com/article/2223757/cisco/cisco-subnet-9-common-spanning-tree-mistakes.html

[FN-GM]

32768 is the default priority. I am guessing all your switches have the default priority as well. When this happens the switch uses the lowest mac address as the tie breaker to determine the root bride. The switch with the lowest mac address is usually the oldest. This can result in an old switch in a cabinet on the edge of the network being the spanning tree root bridge. This means all your network traffic will go through this old switch and take a very inefficient path.

 

In short it is good to reduce this priority.

 

Depending on your spanning tree version you might need to do this for every VLAN.

Edited June 22, 2018 by FN-GM

[ass17]

32768 is the default priority. I am guessing all your switches have the default priority as well. When this happens the switch uses the lowest mac address as the tie breaker to determine the root bride. The switch with the lowest mac address is usually the oldest. This can result in an old switch in and cabinet on the edge of the network being the spanning-tree root bridge. This means all your network traffic will go through this old switch and take a very inefficient path.

 

In short it is good to reduce this priority.

 

Depending on your spanning tree version you might need to do this for every VLAN.

 

[emoji1303]

[Chuckster]

Say, for example, I reduce the priority on my core switch to 0 and another core switch to 1. What would the priority levels be within each of those VLANS on per core switches?

 

32768 is the default priority. I am guessing all your switches have the default priority as well. When this happens the switch uses the lowest mac address as the tie breaker to determine the root bride. The switch with the lowest mac address is usually the oldest. This can result in an old switch in a cabinet on the edge of the network being the spanning tree root bridge. This means all your network traffic will go through this old switch and take a very inefficient path.

 

In short it is good to reduce this priority.

 

Depending on your spanning tree version you might need to do this for every VLAN.

[FN-GM]

Say, for example, I reduce the priority on my core switch to 0 and another core switch to 1. What would the priority levels be within each of those VLANS on per core switches?

 

You will need to set the priority increments of 4096. So you can set one as 0 and another of 4096.

 

You set the priority on the VLAN rather than the switch. If you have multiple VLANS you can set them all to be a priority of 0.

[Chuckster]

Currently, my core switch 1 is now priority 0. Core switch 2 has a priority of 1 (4096).

 

Will there be any adverse affects if I was to set every VLAN priority to 0 in each switch despite the switch is set to priority 0 and the other 1?

[FN-GM]

Will there be any adverse affects if I was to set every VLAN priority to 0 in each switch despite the switch is set to priority 0 and the other 1?

 

No - in fact it should be set for each VLAN as there is an instance of spanning tree running for each of them.

[Bruce123]

FN-GM wrote:

This can result in an old switch in a cabinet on the edge of the network being the spanning tree root bridge. This means all your network traffic will go through this old switch and take a very inefficient path.

 

Are you sure this is correct?

 

Just because a switch has been elected a root bridge in a spanning tree instance it doesn't necessarily follow that all traffic will go via that switch.

 

The main issue (if my understanding is correct) would be if this switch was rebooted or removed, it would trigger a re-election and all ports on all switches participating in spanning tree will stop forwarded frames (and go through blocking > listening > learning for stp) until convergence is complete. Depending on how long it takes to converge, users will experience a momentary loss in connectivity. The same issue applies when manually modifying the priority value, it's best done out-of-hours and with a serial cable connected to the switch, in case, for whatever reason, you need to change it back to what it was (*).

 

Thanks,

 

Bruce.

Edited July 10, 2018 by Bruce123

[FN-GM]

FN-GM wrote:

 

Are you sure this is correct?.

 

I am sure

 

 

The main issue (if my understanding is correct) would be if this switch was rebooted or removed, it would trigger a re-election and all ports on all switches participating in spanning tree will stop forwarded frames (and go through blocking > listening > learning for stp) until convergence is complete. Depending on how long it takes to converge, users will experience a momentary loss in connectivity. The same issue applies when manually modifying the priority value, it's best done out-of-hours and with a serial cable connected to the switch, in case, for whatever reason, you need to change it back to what it was (*).

 

That is correct. The re-election will happen when the machine is rebooted and then again when it comes back online. However if you're running rapid spanning tree you probably wouldn't notice it happen.

[Edgehill]

This is not correct, not all traffic will go through the root bridge. if traffic is going upstream so to speak it would only go as far as needed and not through the root and back..

But if you leave priorities as the default and a less than ideal switch is elected it would potentially still be inefficient for anything that does need to go through the root bridge.

With default STP convergence is 30 to 50 seconds and changing priorities will cause a re-election, so yes this should be done out of hours.

Edited July 10, 2018 by Edgehill