[libvirt] Network Interface: How to disable virbr0 correctly?

[Matthias Bolte]

Sorry for the late response.

2011/3/8 Jake Xu <jake at demonwaremail.net>:
> On Tue, Mar 8, 2011 at 11:20 AM, Matthias Bolte
> <matthias.bolte at googlemail.com> wrote:
>>
>> 2011/3/8 Jake Xu <jake at demonwaremail.net>:
>> > Hi Cole,
>> > Thanks for the wiki link. It would be so useful if the ESX driver
>> > supported
>> > those commands. It seems like the ESX driver does not support most of
>> > the
>> > network or interface configuration.
>>
>> What do you mean by " those commands"?
>>
>> The problem with implementing ESX networking support in libvirt is
>> that ESX's and libvirt's network model don't match directly and it's
>> not entirely clear how two map those two in a useful way.
>>
>> Can you be more specific about what you need in terms of network
>> management?
>>
> Sorry if I did not make it clear. For commands, I meant the net-list,
> net-destroy, etc commands in virsh.
> I am actually using the libvirt Python-bindings to write scripts which
> automate the VM creation process. All I need in terms of network management
> is that I want it to have two interfaces eth0 and eth1 with static ip
> addresses(gateway, subnet, etc), so I don't need any other extra
> configuration such as virtbr0 provided by libvirt/libvirtd daemon. Is there
> any way to prevent libvirtd daemon being installed on the guest VM?

Okay, I knew what commands you meant, I was asking what you want to do
with them. You want to assign static IP addresses to the guests by
defining a fixed MAC to IP address mapping in the virtual network
definition.

In general that would be the right approach. Unfortunately this is not
how it currently works for ESX. Although an ESX server can run it's
own DHCP server for it's virtual machines (at least I guess so from
the documentation) it's not enabled by default and I couldn't figure
out how to enable it yet. By default virtual machines use what ever
DHCP server is in the same subnet as the ESX server itself.

Another problem seems to be that the ESX DHCP server configuration
doesn't allow for fixed MAC to IP address mappings.

The best thing you can currently do to assign fixed IP addresses to
virtual machines is to configure your external DHCP server to do so.

>>
>> > Regarding the eth1 interface, I am not sure why the autostart.xml could
>> > affect bringing up eth1. Before I made changes to the autostart.xml or
>> > renamed the folder completely, the VMs have all interfaces installed:
>> > eth0,
>> > eth1, virbr0 (even though the virbr0 and its iptable rules seem to be a
>> > bit
>> > difficult to get rid of). After I renamed the autostart folder, I was
>> > able
>> > to remove the virbr0 interface and iptable rules, but the eth1 can't be
>> > brought up successfully as described in the opening post.
>> > Thanks,
>> > Jake
>>
>> Are you installing libvirt inside the VMs and running libvirtd? Why
>> are you doing that?
>>
>> Matthias
>>
> No. I am running libvirt with its Python-bindings on a Ubuntu PC, and
> remotely creating VMs on ESX server. Since I am creating VMs on ESX server,
> I don't really need libvirtd. It would be nice if I can disable libvirtd
> before it's being installed on VMs.
> Thanks so much for your help,
> Jake

"Disable libvirtd before it's being installed on VMs"? So you're
installing libvirt inside the VM, but on the other hand you said that
you're not doing that. I'm confused.

Also having libvirtd running on your Ubuntu PC has nothing to do with
networking on the ESX side.

Matthias