[libvirt-users] Some missunderstanding in the wiki page

Hi, I'm not in the list and I just would like to comment it to the comunity. So, if someone want to contact me, make CC to me in the reply. The wiki page [1] talk about networking, and show two typical configurations: "virtual network" and "shared physical device". The shared physical device has an error or something that could make confused, I think. What I want to say? In my case, I have a box with several nics, and I want to speciffy that ethX should be used exclusively but virtual domain X. The documentation is clear in this: The NAT based connectivity is useful for quick & easy deployments, or on machines with dynamic/sporadic networking connectivity. More advanced users will want to use full bridging, where the guest is connected directly to the LAN. The configuration (I have follow only the debian/ubuntu steps) shows that br0 has assigned an IP. I did the mistake to try to assign the same IP in the guest network configuration, I understand that, and obviously it doesn't work (duplicate IPs). So, to me, it would me more clear if the br0 interface has no IP. In my case, I have two nics: eth0 the interface for the host and eth1 for the guest. The eth1 part in my network/interfaces is: ... auto eth1 iface eth1 inet manual auto br0 iface br0 inet manual bridge_ports eth1 bridge_stp on bridge_maxwait 0 bridge_fd 0 ... and then in the guest you configure the IP as you want. In the host, eth1 (or whatever interface you bridge) has no IP, so it's not used, and the guest uses completely. What do you think? Regards, Leo [1] http://wiki.libvirt.org/page/Networking

On 30/06/10 10:35, Leopold Palomo-Avellaneda wrote:
The configuration (I have follow only the debian/ubuntu steps) shows that br0 has assigned an IP. I did the mistake to try to assign the same IP in the guest network configuration, I understand that, and obviously it doesn't work (duplicate IPs). So, to me, it would me more clear if the br0 interface has no IP. In my case, I have two nics: eth0 the interface for the host and eth1 for the guest. The eth1 part in my network/interfaces is:
... auto eth1 iface eth1 inet manual
auto br0 iface br0 inet manual bridge_ports eth1 bridge_stp on bridge_maxwait 0 bridge_fd 0 ...
and then in the guest you configure the IP as you want. In the host, eth1 (or whatever interface you bridge) has no IP, so it's not used, and the guest uses completely.
The common configuration that the wiki is documenting is presumably the case where the host has a single ethernet interface that is shared by the host and the guests. In that configuration the bridge does have an address, which is the address of the host machine. Basically a bridge is like a virtual ethernet switch inside your machine, where one port is connected to each device that is enslaved to it and another port is connected to the kernel's IP stack - that port appears as the brN interface and can have an IP address (for the host) assigned to it. Tom -- Tom Hughes (tom@compton.nu) http://compton.nu/

A Dimecres 30 Juny 2010, Tom Hughes va escriure:
On 30/06/10 10:35, Leopold Palomo-Avellaneda wrote:
The configuration (I have follow only the debian/ubuntu steps) shows that br0 has assigned an IP. I did the mistake to try to assign the same IP in the guest network configuration, I understand that, and obviously it doesn't work (duplicate IPs). So, to me, it would me more clear if the br0 interface has no IP. In my case, I have two nics: eth0 the interface for the host and eth1 for the guest. The eth1 part in my network/interfaces is:
... auto eth1 iface eth1 inet manual
auto br0 iface br0 inet manual bridge_ports eth1 bridge_stp on bridge_maxwait 0 bridge_fd 0 ...
and then in the guest you configure the IP as you want. In the host, eth1 (or whatever interface you bridge) has no IP, so it's not used, and the guest uses completely.
The common configuration that the wiki is documenting is presumably the case where the host has a single ethernet interface that is shared by the host and the guests. In that configuration the bridge does have an address, which is the address of the host machine.
and then you can assign _another_ IP to the guest or must assign the same?
Basically a bridge is like a virtual ethernet switch inside your machine, where one port is connected to each device that is enslaved to it and another port is connected to the kernel's IP stack - that port appears as the brN interface and can have an IP address (for the host) assigned to it.
but, my main doubts is about to have a sever with several nics and several guest. I think that a good approach (if it's possbile) is to have a an specific nic for each guest. No? Regards, Leo

On 01/07/10 09:11, Leopold Palomo-Avellaneda wrote:
A Dimecres 30 Juny 2010, Tom Hughes va escriure:
The common configuration that the wiki is documenting is presumably the case where the host has a single ethernet interface that is shared by the host and the guests. In that configuration the bridge does have an address, which is the address of the host machine.
and then you can assign _another_ IP to the guest or must assign the same?
A different IP or the two will fight during ARP resolution with both the host and the guest trying to claim the IP address.
Basically a bridge is like a virtual ethernet switch inside your machine, where one port is connected to each device that is enslaved to it and another port is connected to the kernel's IP stack - that port appears as the brN interface and can have an IP address (for the host) assigned to it.
but, my main doubts is about to have a sever with several nics and several guest. I think that a good approach (if it's possbile) is to have a an specific nic for each guest. No?
Sure if you've got spare nic's in the host or you need maximum performance then that may well be a good idea. The single nic case that I described is more for the average person running a few vms on their workstation. Tom -- Tom Hughes (tom@compton.nu) http://compton.nu/
participants (2)
-
Leopold Palomo-Avellaneda
-
Tom Hughes