Re: NetworkManager and STP

[Peter Jones <pjones redhat com>]

On Tue, 2006-03-14 at 10:41 -0500, Dan Williams wrote:

> What's the situation here?  NM only cares if the machine's ethernet card
> has a link or not, which is reported by the driver for the card itself.
> Once the driver reports that it has a link, NM will attempt to acquire a
> DHCP address on that port.
> 
> Are you saying that the switch takes a long time to actually start
> passing traffic from the machine on which NM is running, even though the
> port is active?

Yes, that is exactly the problem with the braindead protocol known as
"spanning tree".

Arguably it's a usage problem -- STP is designed so that you can plug
have a multiple bridges in the same collision domain.  That is, two
cables to your upstream, one to each switch, and then a cable between
the two switches.

The way this is supposed to work is that you configure STP for each of
the ports you're using to communicate with other switches, and thus you
get high availability out of it.  What actually happens is that network
admins get lazy, and instead of planning their network, they turn STP on
for *every* port, so they can have a randomly-connected network.  And
then everybody has to play the STP game, not just bridging devices.

-- 
  Peter