Connecting three 2510G-48 switches

[swedude8]

Hello!

I would like to get some tips from the HP community regarding how to best configure the connectivity of three HP 2510G-48 switches in a ring configuration (A-B, B-C, C-A) using double copper wiring (so really A=B, B=C, C=A).

The idea is to have no single point of failure, so any site X can reach any other site Y regardless of wiring failure or switch failure. I will be using numerous VLANs. I will not be using any type of routing (this is done elsewhere).

Right now I have two of these 2510G-48 switches connected using double copper wiring and without using any trunk configuration (e.g. A=B wiring). I use spanning-tree to avoid broadcast storm type issues.

** When connecting three switches as described above, is it mandatory to use trunks or would it suffice to keep using spanning-tree? I assume the whole circular logic will be way more complicated in a A=B, B=C, C=A environment, as the amount of loops increase from A=B e.g. 1 to MANY.

Note: I use command-line configuration methods only, so no web clicking.

Best regards,

Ulf

Sweden

[Chrisd131313]

Hi Swedude8,

Based on your scenario I would trunk each of the dual links between the 3 switches to provide aggregated bandwidth (trunk port1,port2 trkx trunk) and then enable spanning tree on the 3 switches to block one of the links to stop the loops.

It is up to you if you decide to let spanning tree decide where it wants to block or if you state which trunk you want spanning tree to block on. As you already have spanning tree enabled I am assuming that you have one of your two switches already designated as your root? If you want to specify which trunk you want spanning tree to block then you can do this with the following command "span trkx priority y" where trkx is the trunk between the two switches and where y is a number between 4-7. By default a trunks priority is set to 4, so any trunk priority below 4 will be seen as a potential blocking point.

Hope this makes sense and if anyone else whats to add to this please jump in :)

[paulgear]

Hi Ulf,

In addition to what Chris said, i would add:

1. It makes no difference to the switch whether you have two individual links or one trunk link. It's not "way more complicated" - the only issue is how much bandwidth you want; combining the links into trunks makes more sense from a speed perspective.

2. Make sure you set your switch priorities so that you know which one is the root bridge. Setting link priorities is not enough to guarantee a particular STP topology.

[swedude8]

Hello,

Thank you for your reply! I understand I should set root bridge priorities (at least).

Currently the A=B switch ports (47 and 48) both have numerous tagged VLANs. Is it possible to configuration-wise SMOOTHLY move ports 47+48 from the current tagged configuration to a link aggregation-type configuration? It seems I lose all VLAN configuration for ports 47+48 when I create a trunk Trk1 on these ports. This will most likely cause a service disruption, e.g. I will not be able to access (at least one of) these switches once the A=B link is down. I do not have an out-of-band network -- there are terminal servers attached to the switch console port but in the end these also rely on operational inter-switch connectivity.

The fallback solution would be to do the trunk configuration change, via console, simultaneously and on-site. Obviously I prefer to not do this.

Any ideas most welcome!

Kind regards,

Ulf

Sweden

[Chrisd131313]

Hi Ulf,

Because you are currently using Spanning Tree you should be able to add the blocked port to the trunk first on both switches. Then configure the VLAN tagging on the trunk, then when you have a maintenace window or scheduled downtime you can then add the other port in to the trunk on the upstream device first, you will lose connectivity for a couple of seconds, but you can then add the downstream device port in to the trunk and it should all be back up and running without you having to visit site.

In theory this shoulk work!! But if it was me, I would arrange a visit to the site and do it there, just in case anything goes wrong.

UPDATE: I can confirm that if you create a trunk with one port member and set your VLAN tagging, when you add in the 2nd port it will not lose the VLAN configuration on that trunk. The only bit I am not sure of is how distruptive it will be, but from previous times I have changed a port to trunk it is usually a couple of seconds.

One other point to take in to account is how spanning tree will react. I have a sneaky suspicion that as soon as you create the trunk it will put that trunk to priority 4, which could - or likely to - cause an unexpected topology change (or even a loop) - which might disconnect the link that is currently in use. Without seeing your spanning tree config it would be hard to say exactly what would happen.

HTH.

[swedude8]

Hi Chris!

Thank you for your input. I never thought of moving one port at a time into the Trk1 link aggregation!

I have no other spanning-tree configuration, other than just enabling spanning-tree. Probably a bad idea. I notice however (via 'show spanning-tree') one of the A=B switches is a designated root. I guess the two switches sort things out by themselves. I recon this may not be a good idea when introducing a third switch. What would be a proper way to configure some sort of Spanning Tree hierarchy?

I notice port 48 has a spanning-tree state of 'Blocking' on the non-root switch. Does this mean this switch is the 'downstream' switch?

Kind regards,

Ulf

Sweden

[Chrisd131313]

Hi Ulf,

If you have just enabled Spanning Tree without any configuration changes the root is assigned to the switchwith the lowest bridge identifier...

Every switch running an instance of MSTP has a Bridge

Identifier, which is a unique identifier that helps distinguish

this switch from all others. The switch with the lowest Bridge

Identifier is elected as the root for the tree.

The Bridge Identifier is composed of a configurable Priority

component (2 bytes) and the bridge’s MAC address (6 bytes).

The ability to change the Priority component provides

flexibility in determining which switch will be the root for the

tree, regardless of its MAC address.

So if Spanning Tree is in its default state then it make the switch with the lowest MAC address root. You can change this by issueing this command on a specific switch...

spanning-tree priority <0-15> (default is 2)

the multiplyer is x4096 so the number x4096 gives you the switch priority you see in "show spanning-tree".

This will then set the switch as the root (if yu set it to 0 or 1, or reducethe priority of the other switches belwo 2 - but for security your root should always be 0 so no other superior BPDUs can over ride your root - there are other means of guarding against this, but I would suggest reading your switches advanced traffic management guide relating to STP). Usually you will set this on one of your Core switches which has the larger port speeds. But this is a little irelevant if all ports on your switches are the same 1Gb or 10Gb. If so I would normally set the root on the switch which connects to our MPLS network. (There maybe a better option and I am sure one of the other guys in here can shed some light on that).

When I say downsteram it is relative to your connection. If your connection to the network is on switch A then you would make the chagne on switch B first so when connectivity is lost you are still able to change switch A's port config so switch B comes back online. If you change Switch A's port setting before Switch B's then you have no way of configuring the port on switch B without physical access, or a being connected directly to switch B. Hope that makes sense!! apologies if that was poorly explained!

It might be worth checking the priority setup on port 47 and port 48, because I would hazzard a guess they are both set to 128 currently, therefore, If you create a trunk with port 48 the priority of this link will change to 64 taking presidence, you might want to increase the priority of port 47 to above 64 so when the trunk is created spanning tree doesn't failover to the new trunk with the higher priority and no VLAN tagging!!

spanning-tree 47 priority <0-15> (default is 8 for ports and 4 for trunks)

the multiplyer is x16 so the number x16 gives you the port priority you see in "show spanning-tree".

Hope this Helps.

[paulgear]

Hi Ulf,

Getting familiar with the basics of STP would really help you understand what's going on in this scenario and plan your changeover. There's plenty of good training material out there from HP, Cisco, and Juniper, and even Wikipedia has a pretty good explanation of spanning tree.

As Chris said, the way switches choose the root bridge is by a rather arbitrary mechanism (MAC address), and you really need to set switch priorities manually if you want to control which links are blocked. (A minor quibble: no HP switch i've seen has defaulted to priority 2 - all the models i've worked with default to 8 x 4096 = 32768.) Moving one link at a time into the trunk (and setting its priority to be the same as a normal port first as Chris suggested) is a good way to ensure minimal cutover time, but be prepared for several seconds to a minute of downtime during reconvergence.

[Chrisd131313]

I stand corrected, the default switch priority is 32768 not 8192, I checked the wrong switch in my environment :P Thanks for correcting that Paul.