Description of problem: When I power down a guest in virt-manager then I cannot connect to the host machine for 20-50 seconds. The guests are connected using bridged networking. Version-Release number of selected component (if applicable): kvm-83-164.12 kmod-kvm-83-164.12 kernel-2.6.18-194.8.1.el5 Host OS: CentOS 5.5, x86_64 Guest OS: CentOS 5.4, Windows XP How reproducible: Always Steps to Reproduce: Create and install a guest with bridged networking. Next power down the guest OS. Actual results: The current connections to host are broken and it is even impossible to ping the host for 20-50 seconds. Expected results: No impact on host networking. Additional info: Here are described similar problems in Debian and Ubuntu: http://kerneltrap.org/mailarchive/linux-kvm/2010/4/21/6261001/thread https://bugs.launchpad.net/ubuntu/+source/qemu-kvm/+bug/584048 In /var/log/messages I see: Jul 6 17:21:39 daisy kernel: br0: port 2(vnet0) entering disabled state Jul 6 17:21:39 daisy kernel: device vnet0 left promiscuous mode Jul 6 17:21:39 daisy kernel: br0: port 2(vnet0) entering disabled state
It seems that probable cause is described in bug #571991
I was able to fix this problem applying patch from bug #571991
The issue here that somehow, tun got to be the first interface on the bridge. So bridge gets the mac from tun, when we disconnect this tun mac changes. In addition to the patch suggested, we could connect eth0 first, and this issue will go away. Since libvirt does the bridge setup, this needs to be fixed there. Changing component appropriately.
*** This bug has been marked as a duplicate of bug 571991 ***