Switches, Hubs, and Modems
cancel
Showing results for 
Search instead for 
Did you mean: 

How to isolate loop errors with 4108

Frank Sauerbier
Frequent Advisor

How to isolate loop errors with 4108

Hi all,
we have many student labs connecting to the campus network by 4108. These 4108-segments are connected to a 5400-Mesh. Now we want to prevent network downs if a student builts a loop e.g.
At the moment if this happens the hole segment and the mesh is infected and nearly the hole campus network goes down.

What is the best way? RSTP? Broadcast Control?
Please any suggestions.

Thanks
Frank
7 REPLIES
Matt Hobbs
Honored Contributor

Re: How to isolate loop errors with 4108

Definitely enable RSTP as your first step towards preventing loops.

You could also disable auto-MDIX which will prevent most loops as the student would need a cross-over cable. This option is available on the newer products but I'm fairly certain it can be done on the 4100's.
Frank Sauerbier
Frequent Advisor

Re: How to isolate loop errors with 4108

Hi Matt,
I am not sure about RSTP. I thought I have to assign RSTP to choosen ports - or can I enable it for all ports? Doesn't this effect the normal port behaviour?

Frank
Matt Hobbs
Honored Contributor

Re: How to isolate loop errors with 4108

Hi Frank,

RSTP should be enabled on all ports. On the 4108 you simply need to type the command 'spanning-tree'.

By default RSTP treats all ports as 'edge' ports so when a port comes online it immediately starts forwarding traffic - therefore there should be no difference to that of a port without RSTP (by the way, on the 4108 you can only enable it globally anyway).

On your switch to switch links (uplink ports) you should (not mandatory but best practice) set:
'no spanning-tree edge-port'

The uplink ports will then go through the full spanning-tree learning phase before the port starts forwarding traffic. This is more useful when you have redundant links going to other switches.

Matt
RobB_8
Advisor

Re: How to isolate loop errors with 4108

I was just looking at the exact same question at one of my sites. I have a 4108 core switch with a bunch of secondary closets with stacks of 2524's connected together with stacking modules. One of the techs caused a loop in one of the closets and it took down the entire site. This is the spanning tree config on the 4108(all the referenced ports are uplinks to closets), the closets do not have spanning tree.
spanning-tree
no spanning-tree A1 edge-port
spanning-tree A1 priority 4
no spanning-tree A2 edge-port
spanning-tree A2 priority 4
no spanning-tree A3 edge-port
spanning-tree A3 priority 4
no spanning-tree B1 edge-port
spanning-tree B1 priority 4
no spanning-tree B2 edge-port
spanning-tree B2 priority 4
no spanning-tree B3 edge-port
spanning-tree B3 priority 4
no spanning-tree C1 edge-port
spanning-tree C1 priority 4
no spanning-tree C2 edge-port
spanning-tree C2 priority 4
no spanning-tree C3 edge-port
spanning-tree C3 priority 4
spanning-tree priority 1
Is this the best scenario... I have heard conflicting opinions on where to put RSTP... Some say the core, some say the edge, some say either core or edge but not both.
Matt Hobbs
Honored Contributor

Re: How to isolate loop errors with 4108

I would recommend enabling it as close to the edge as possible, and on the core. If you're running a routing protocol to your edge switches then it's not so important to run it in the core, but not many people run routing to the edge switches in ProCurve based networks.

Also make sure you have the latest firmware for the 4108, at some point in time they added support for RSTP to detect loops on the same port. So in Rob's case, since it was probably only occurring on one of the links of the 4108, it should have at least been able to block that port and left the rest of the network unaffected.
RobB_8
Advisor

Re: How to isolate loop errors with 4108

Thanks Matt... very interesting... I am curious about your insight on a couple more points...
1) I am on firmware 7.70 because I am not prepared to eliminate CDP transmit yet... but the bug must have been previous to 7.70, as I did not find it in any of the newer versions in the release notes.
2) In a scenario like mine what would you recommend on things like Priority, Edge, P-P, and Mcheck?
Traditionally:
Priority: I would not touch for workstation (keep at 8), for uplinks/downlinks 4, for WAP's 12.
Edge: yes for workstations, no for uplinks/downlinks
Point to Point: do not ususally touch
Mcheck: don't usually touch
Thanks
Matt Hobbs
Honored Contributor

Re: How to isolate loop errors with 4108

I might have to take what I said back about RSTP detecting loops on the same port. This fix:

RSTP (PR_99049) â Switch does not detect and block network topology loops on a single
port. For example, the port connects to a hub that has a loop or the port connects to an inactive node via IBM 'Type 1' cable.

Has been included in most other products such as the 2600/3400/5300. The 4100 and 6108 though seem to have missed out on it. Possibly because they are and always will be on the 07 branch... e.g. G.07.99, H.07.83.

Or possibly because they were never affected, but I'd be surprised if that was the case.

As for the values, point-to-point, mcheck, port priorities.... I leave at the default settings unless I have a specific requirement to change them which is extremely rare.

The only one I will change is on my core switch - I will set the spanning-tree priority to a higher value such as 1.
'span priority 1'. Which will ensure it is the root of the spanning-tree (unless of course something has priority 0 and/or a lower mac-address).