Todd,
 
It actually was never a problem.  libvirt(Centos5.4) and openvswitch 0.90.7 
work fantasic together no scripts/glue required.
I have the following in my libvirt XML config file.  It tells KVM to bridge to 
the openvswitch using the network device name netwin0.
<interface type='bridge'>
      <mac address='54:52:00:5f:a8:72'/>
      <source bridge='brovs1'/>
      <target dev='netwin0'/>
</interface>
 
Then in the openvswitch config I had the port defined already for vlan access 
so when KVM bridged in and created the port, VLAN access was applied and 
traffice flowed.
  bridge.brovs1.port=bond1
  bridge.brovs1.port=brovs1
  vlan.brovs1.trunk=4
  vlan.brovs1.trunk=5
  vlan.brovs1.trunk=6
  vlan.netwin0.tag=6
But with openvswitch 1.0.0 this method does not work as I can't 
predefine/configure the port.
Is there any chance of having OpenVswitch save Port configuration in the 
database?  That way when the port is initialized the config would be applied?
 
Thanks again..  I guess I'm off to have a look at the INSTALL.KVM to get things 
working with 1.0.0
 
Cheers
 
Mark
 
 
 
 


>>> Todd Deshane <desha...@gmail.com> 23/05/2010 10:26 >>>
Hi,

Comments inline.

On Fri, May 21, 2010 at 4:02 PM, Dan Wendlandt <d...@nicira.com> wrote:
>
>
> On Thu, May 20, 2010 at 1:47 AM, Mark Persohn
> <mark.pers...@superchoice.com.au> wrote:
>>
>> I have looked at trying to run a script as prt of the VM startup but since
>> we are running libvirtd/qemu in bridge
>> mode it will not run scripts!
>
> I do not have any hands-on experience with libvirt and KVM, but from talking
> to other people who have used OVS with libvirt, I was under the impression
> that libvirt provides a mechanism by which you can bring up VM interface
> devices using an arbitrary script.
>
> The libvirt docs
> (http://libvirt.org/formatdomain.html#elementsNICSEthernet ) mention:
>
> Generic ethernet connection
>
> Provides a means for the administrator to execute an arbitrary script to
> connect the guest's network to the LAN. The guest will have a tun device
> created with a name of vnetN, which can also be overridden with the <target>
> element. After creating the tun device a shell script will be run which is
> expected to do whatever host network integration is required. By default
> this script is called /etc/qemu-ifup but can be overridden.
>
>   ...
>   <devices>
>     <interface type='ethernet'/>
>     ...
>     <interface type='ethernet'>
>       <target dev='vnet7'/>
>       <script path='/etc/qemu-ifup-mynet'/>
>
>
>     </interface>
>   </devices>
>   ...
>
> I am guessing such a script might let you override libvirt's mechanism for
> adding a VM interface device to the bridge and instead let you directly
> configure it using OVS (e.g., ovs-vsctl set port netwin0 tag=6 ).
>
> As I recall, you're not the first person to have this problem so however we
> end up resolving it, we should document this better on the OVS site.
>

I haven't explicitly tried libvirt + open vSwitch yet either, but I
had read this:
http://www.mail-archive.com/libvir-l...@redhat.com/msg18133.html 

I did however run across this blog that gave me the idea to just put
the config in place in a "just in time" manner:
http://veejoe.net/blog/2010/02/network-virtualisation/ 

But, even though that trick works, it is clearly a hack. I think
libvirt will (if they haven't already) add support for gracefully
handling open vSwitch and similar things. Especially since qemu is
discussing supporting (maybe they already do support) various network
backends (vepa, vswitches, macvtap, etc.).

If someone tests explicitly whether libvirt supports the script option
or not please post the results to the list.

Thanks,
Todd

-- 
Todd Deshane
http://todddeshane.net 
http://runningxen.com 

!DSPAM:4bf8527b14849641513966!






Please consider the environment before you print this email
For email conditions, please see our website  
http://www.superchoice.com.au/email_conditions.htm
_______________________________________________
discuss mailing list
discuss@openvswitch.org
http://openvswitch.org/mailman/listinfo/discuss_openvswitch.org

Reply via email to