[ovs-discuss] determining the network interface attached to a bridge

Ben Pfaff blp at nicira.com
Wed Sep 12 22:11:18 UTC 2012

On Wed, Sep 12, 2012 at 04:53:44PM -0500, ovs.to.anomalyst at xoxy.net wrote:
> 1) I am working on upstart scripts for ovs and trying to automate the
> task of creating a configuration script tasked with getting the bridge
> up and assigned an address via DHCP on boot.
> How does one determine the network interface being used as the
> upstream bridge interface?.

OVS is just a switch.  It doesn't distinguish "upstream" interfaces
from other interfaces, in the same way that any network switch you buy
at the store doesn't.  (10/100 Mbps switches might have a port labeled
"uplink" but normally that just means that the rx/tx lines are

> Without having added any additional VM ports to the bridge yet, I have this:
> $ sudo ovs-vsctl show
> f23e06e3-733a-46c3-b340-fc53d757ba42
>     Bridge "br1"
>         Port "eth2"
>             Interface "eth2"
>         Port "br1"
>             Interface "br1"
>                 type: internal
> Which has eth2 as the UBI, but has nothing explicit marking it as
> such. Is it correct to assume that the 1st port/interface is always
> the UBI?

No, not necessarily.

> 2) Is it possible to connect a libvirt/qemu VM directly to a vanilla
> ovs bridge or is brcompat required?

brcompat is not required.

> 3) How do I obtain git access to add the .conf files  I have created
> to amke them available for public consumption. (I am pretty much a git
> noob).

Email us a patch as described in SubmittingPatches.

> BTW "http://openvswitch.org/git/openvswitch" gives a 404 error. That
> URL is referenced in
> "http://openvswitch.org/cgi-bin/gitweb.cgi?p=openvswitch;a=summary"

How odd.  I'll ask for this to be fixed.  In the meantime, you can use
Git protocol instead.

> 4) May I suggest a revision to ovs-vsctl to put an asterisk or tilde
> in the list to reflect the UBI that was assigned? E.G.
> 	Interface* "eth2"

OVS doesn't have such a concept.

More information about the discuss mailing list