Configuring network for OpenVZ container

To live-migrate a container that is accessible via network, it should possible to access this container with the same IP address regardless of on what node it runs, source or destination. The easiest way to do it is by attaching the container and both nodes to the same network segment via bridging.

To do so, you should first have your hosts' main links (eth0 below) attached to a bridge, and next -- attach the container to this bridge on the source node.

Configuring bridge on host

It's just and example of how to do it on Fedora-20, you can skip this section and do it in your own way.

• Disable NetworkManager and enable legacy networking

 # chkconfig NetworkManager off
 # chkconfig network on
 # service NetworkManager stop
 # service network start

• Configure bridge and attach eth0 device to it.

Create config file /etc/sysconfig/network-scripts/ifcfg-br0

 DEVICE=br0
 TYPE=Bridge
 BOOTPROTO=dhcp
 ONBOOT=yes
 DELAY=0

To /etc/sysconfig/network-scripts/ifcfg-eth0 add line

 BRIDGE="br0"

Then do

 # service network restart

After this you should have eth0 be in bridge (seen by the brctl show br0) and br0 have the ip address (seen by ip a l).

Configure CT networking

Next step is to attach the container to the bridge. To do so, the veth device should be used.

• Assign veth to container

 # vzctl set $id --netif_add eth0,$ct_mac,$host_veth_name,$host_mac,br0

Any of $ct_mac, $host_mac and $host_veth_name can be empty, in that case vzctl will generate the value itself. Plz note, that if $host_mac is arithmetically less that the mac on your eth0 link after starting CT you would see some lag in networking due to br0's mac would change.

• Make sure you have eth0 config file in CT's /etc/sysconfig/network-scripts/ifcfg-eth0.