On Thu, Aug 23, 2012 at 5:17 PM, Tim Spriggs <[email protected]> wrote:
> Hi All,
>
> I have been struggling a little with Open vSwitch under Debian
> Wheezy. I'm not sure if this is because I'm very green with Open
> vSwitch or some other reason. What I would like my simplified network
> setup to look like:
>
> Bridge "br0"
> Port "eth1"
> Interface "eth1"
> Port "br2"
> tag: 2
> Interface "br2"
> type: internal
> Port "br3"
> tag: 3
> Interface "br3"
> type: internal
> Port "br0"
> Interface "br0"
> type: internal
>
> ie: eth1 is a tagged port on the physical switch with vlans 2 and 3
> passed over the wire. br2 and b3 would be corresponding ovs bridges
> for each vlan.
>
> The system itself needs to access the network over its tagged
> interface (lets say vlan2) and thus we can either configure br2 with
> an IP or we can add an interface (veth?) to the bridge and then
> configure the veth pair.
>
> Since /etc/init.d/openvswitch-{controller,switch} both have $network
> in the Required-Start line, /etc/init.d/networking is run before
> openvswitch has a chance to create its bridge interfaces. This seems
> to mean that I can't set parameters on the device as I usually would.
> For example, we turn of autoconf and accept_ra by default so an
> example entry to make br2 into a useful interface in
> /etc/network/interfaces may look like this:
>
> iface br2 inet dhcp
> post-up sysctl -w net.ipv6.conf.$IFACE.autoconf=1
> post-up sysctl -w net.ipv6.conf.$IFACE.accept_ra=1
>
> When I try to use "allow-hotplug br2" or "auto br2" the post-up lines
> are run before ovs-* is started and just error out. br2 is never
> configured with a dhcp address or an IPv6 address.
>
> ---
>
> This is the point that I hope someone will chime in and say, "Tim,
> you're doing it wrong," or, "hmm, that is *supposed* to work..."
>
> In the event that I'm not alone, one really nice way of dealing with
> this might be to set some kind of post-hook inside of openvswitch.
> Something on the order of, "on_create=/sbin/ifup br2" or "ifup=true"
> that is per-interface. The idea being that when ovs creates an
> interface, it calls the standard system utilities to configure it. To
> be clean, there should be a de-configuration step as well (ifdown) so
> that state is kept.
>
Does this help for your use case?
http://openvswitch.org/pipermail/dev/2012-May/016972.html
The scripts are all there. But, someone needs to call it.
ifup --allow=ovs $list_of_bridges
>
> In IRC, switchcade mentioned
> http://openvswitch.org/pipermail/discuss/2011-October/005891.html
> which implements yet another init script. My feeling is that
> openvswitch is creating/destroying interfaces, it should be
> responsible for making sure they are configured appropriately as well.
> Adding another init script is more of a work-around type solution
> here.
>
> Thanks,
> -Tim
> _______________________________________________
> discuss mailing list
> [email protected]
> http://openvswitch.org/mailman/listinfo/discuss
>
_______________________________________________
discuss mailing list
[email protected]
http://openvswitch.org/mailman/listinfo/discuss
