On 04/17/2011 09:47 AM, Ulli Horlacher wrote: > > I will integrate the other lxc basic commands as soon the phys nic > renaming bug is solved.
The bug is coming from the kernel and the network namespace life cycle. When the network namespace dies, the physical nic is moved back to the host. Unfortunately, the last process of the container can exit making lxc to rename 'devX' to the original name but that may fail because the network namespace could be still alive. That happens, for example, when there are TCP orphan sockets waiting for a timer to expire before being destroyed (FIN_WAIT1, FIN_WAIT2, ...). So the physical network device may be moved back after lxc exits, leading to the network device name inconsistency. I think the problem should be solved from the kernel, because it renames the netdev to devX when there is a name collision. As we changed the name 'eth1' in the host to 'eth0' in the container, when the container exits, the kernel detect the conflict name with the 'eth0' from the host and rename container's 'eth0' to 'dev%d'. As a workaround, I think setting lxc.network.name = eth1 in the container should fix the problem (assuming /etc/network/interfaces file is modified for this name). So no name conflict will appear when the container exits. I will dig to push in the mainstream kernel a fix but I am not sure that will be accepted. Thanks -- Daniel ------------------------------------------------------------------------------ Benefiting from Server Virtualization: Beyond Initial Workload Consolidation -- Increasing the use of server virtualization is a top priority.Virtualization can reduce costs, simplify management, and improve application availability and disaster protection. Learn more about boosting the value of server virtualization. http://p.sf.net/sfu/vmware-sfdev2dev _______________________________________________ Lxc-users mailing list Lxc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/lxc-users
