On Sun, Sep 11, 2011 at 7:18 PM, Nathan Eisenberg
<[email protected]> wrote:
>> - LAG eliminates the need for spanning-tree.  Avoid STP in your network
>> if at all possible... Or rather, always design L2 networks so that
>> spanning tree has nothing to do, i.e. there are never multiple paths.
>> Note that each LAG is considered a single interface for STP purposes.
>
> Eh, I think you're dismissing STP way too readily.
>
> LAG eliminates the need for spanning tree *for those two ports*.  You still 
> need it if there's any possibility your users/noc crew/janitors might do 
> something silly like plug both ports on a Polycom VOIP phone into the wall 
> jacks.  Half of the reason that STP is a really good idea on access networks 
> is that it prevents BAD alternate paths.  I can't tell you how many networks 
> I've audited where STP has been keeping things running without anyone even 
> knowing about it.
>

Yes and no. I've seen roughly equally as many issues caused by STP as
I have scenarios where it prevented a problem. What I've seen in the
last week alone is a good illustration of the last decade plus - one
layer 2 loop on unmanaged switches (which don't have STP) that melted
down the network, and one instance where a brand new Cisco Small
Business switch with default settings caused 50-75% packet loss on the
entire network by causing some bad interaction with STP on the rest of
the network (exactly what I unfortunately couldn't troubleshoot,
customer just wanted STP off on the new switch and everything was fine
then).

Ideally you want STP on, especially in large networks where it can be
extremely time consuming to track down layer 2 loops, but it causes
issues about as much as it saves you from issues. And I would
definitely avoid any scenario where it's required for your network to
function properly if at all possible.
_______________________________________________
List mailing list
[email protected]
http://lists.pfsense.org/mailman/listinfo/list

Reply via email to