On Tue, Sep 30, 2008 at 04:57:15PM +0400, Anton Protopopov wrote:
2008/9/29 Evgeniy Sokolov <evg(a)openvz.org>
>
> On Mon, Sep 29, 2008 at 04:11:39PM +0400, Evgeniy Sokolov wrote:
>>
>>> 2008/9/29 Evgeniy V. Sokolov <evg(a)parallels.com <mailto:
>>> Patch looks good.
>>> But currently it will work strange because of all drivers (except OpenVZ)
>>> in libvirt require specified <source bridge=""/> - it is
generic.
>>> With patch you need to specify both
>>> <source bridge='eth10'/> - to satisfy common requirements
>>> <target dev='eth10'/> - to create network
>>>
>>
>> That is not right. The <target> element *must* be optional when creating
>> a new domain. If omitted, the driver must generated a suitable target
>> dev according to its desired naming scheme. Only the <source> element
>> can be compulsory
>>
>
> Agree with you. You say about "how should it be". I say about "how is
it
> now".
> Currenty OpenVZ driver don't generate device name - my mistake.
Is vnet$i (where $i is the number of automatically generated interface
name for this container) appropriate?
Possibly, though its not terribly consistent so far.
Xen - vifNNN.MMM (NNN dom ID, MMM = nic number)
QEMU - vnetNNN (NNN = global host nic number)
LXC - vethNNN (MMM = global host nic number)
We can't change the format Xen uses, but it could be desirable to make
LXC, QEMU and OpenVZ all use vnetNNN scheme by default for auto assigned
nics names on the host side.
Daniel
--
|: Red Hat, Engineering, London -o-
http://people.redhat.com/berrange/ :|
|:
http://libvirt.org -o-
http://virt-manager.org -o-
http://ovirt.org :|
|:
http://autobuild.org -o-
http://search.cpan.org/~danberr/ :|
|: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|