4:28 p.m.
On 04/28/2011 04:35 PM, Laine Stump wrote:
On 04/28/2011 09:56 AM, Robert Šmol wrote:
> For my needs I would like to see either vdeswitch or openvswitch
> integrated in the project. Is there any preference from the
> users/developers on which one of those I should start looking into?
Are you volunteering to write some code? :-)
There have been requests for both, and both should eventually be done,
so I guess pick whichever one you like better :-) (since there's
already a patch for VDE, albeit with different XML than we'd like,
that may be quicker to finish.)
You may want to follow development of the work I'll point out below,
though, otherwise you could waste a lot of time.
OK, I'll try to start with vde. I've subscribed now to libvirt-list as well.
>
> Also I would like to know how this should be integrated, from the
> mails I can see devs prefers to have virtual switch present as a
> network. Can someone describe how the integration should be done?
> Support for additional features like vlan, QoS?
Once we get into these kinds of questions, it's probably best to
switch to the developers' list (libvir-list(a)redhat.com) to get better
exposure to people writing the code. I'm Cc'ing this message there
(and switching the Reply-To: as well).
In particular, I recently send an RFC to libvir-list proposing an
expansion of the <network> XML to remove the hard-coded connection
between guest interface and host interface for macvtap (type='direct')
interfaces:
https://www.redhat.com/archives/libvir-list/2011-April/msg00591.html
and in a followup message showed how we could expand on this work to
provide support for openvswitch (or vde):
https://www.redhat.com/archives/libvir-list/2011-April/msg00640.html
I'm planning to have this code finished and tested within the next two
months at most (hopefully considerably earlier). If you're interested
in implementing a <network type='openvswitch'> or <network
type='vde'>
you should subscribe to libvir-list (if you aren't already) so that
you can follow along.
I am now trying to compile the stuff, but I get:
...
CCLD libvirt_driver_nwfilter.la
libtool: link: require no space between `-L' and `/usr/lib'
make[3]: *** [libvirt_driver_nwfilter.la] Error 1
make[3]: Leaving directory `/home/rsmol/Code/libvirt/src'
make[2]: *** [all] Error 2
make[2]: Leaving directory `/home/rsmol/Code/libvirt/src'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/rsmol/Code/libvirt'
make: *** [all] Error 2
...
But I'll figure out eventually.
>
> Robert
>
> On 04/06/2011 04:13 PM, Robert Šmol wrote:
>>
>>
>> On 04/06/2011 04:00 PM, Nicolas Sebrecht wrote:
>>> The 06/04/11, Robert Šmol wrote:
>>>> On 04/06/2011 10:34 AM, Nicolas Sebrecht wrote:
>>>>> The 05/04/11, Robert Šmol wrote:
>>>>>
>>>>>> I would like to ask about status of VDE and libvirt. I found a
>>>>>> thread from
>>>>>>
http://www.redhat.com/archives/libvir-list/2008-February/msg00030.html.
>>>>>>
>>>>> AFAIK, vde_switch is still unsupported by libvirt and nobody
>>>>> seems to
>>>>> work on it these days.
>>>>>
>>>>> Most people wanting vde_switch (or open vswitch) enough leave
>>>>> libvirt.
>>>>>
>>>> On #IRC I got point to message:
>>>>
>>>> http://www.mail-archive.com/libvir-list@redhat.com/msg32081.html
>>>>
>>>> It seems there is a patch ready. I will test it against 0.9.0. What
>>>> is the procedure to get the patch reviewed and eventually get into
>>>> libvirt?
>>> The patch wasn't merged but the why is a question to the
>>> maintainers. I
>>> guess the way it was implemented did not convinced. This is what I
>>> understand from the whole thread, at least.
>>>
>> From the thread it seems the maintainers would like to have a
>> network based on VDE. To me it also looks like better idea (create
>> vde network and then just attach the domain to a network.
>>
>
--
stereoIT - Open Source Technology Partners
http://www.stereoit.com
mobile: +420 776 762378