Description of problem: When you install openstack via packstack, on a fresh default install of F20, the default install (which installs neutron) breaks NetworkManager (I'm not sure why atm). The same configuration installed with nova-network (--os-neutron-install=n) does not have this issue, and it's not mentioned in the RDO quickstart docs (which gets users to install neutron by default now) Version-Release number of selected component (if applicable): I used the Havana packages from F20: # rpm -qa | grep -e packstack -e neutron python-neutronclient-2.3.1-2.fc20.noarch python-neutron-2013.2.1-1.fc20.noarch openstack-neutron-openvswitch-2013.2.1-1.fc20.noarch openstack-neutron-2013.2.1-1.fc20.noarch openstack-packstack-2013.2.1-0.25.dev936.fc20.noarch How reproducible: Always. Steps to Reproduce: 1. Install F20 (bare metal box w/16G ram, selected development tools) 2. sudo su - 3. yum -y update 4. shutdown -r now (reboot to ensure all updates are active) 4. yum -y install openstack-packstack 5. packstack --allinone --os-heat-install=y --os-swift-install=n --os-ceilometer-install=y --os-heat-cloudwatch-install=y --os-heat-cfn-install=y 6. reboot Note I installed heat/ceilometer and disabled swift, but I don't think that is related to the problem. Actual results: After reboot, the system em1 device is up for a few seconds, then loses it's DHCP lease and cannot recover, the interface is broken. Expected results: This is pretty confusing for users IMO, as the default configuration is broken, and apparently requires you to make em1 not network-manager controlled, or disable network-manager completely I guess. It seems like either neutron should learn how to play nicely with network-manager, or packstack should reconfigure the network for the user, or at the least, we should document the required config in the quickstart prerequisites. Additional info: Workaround for me was to edit /etc/sysconfig/network-scripts/ifcfg-em1 adding NM_CONTROLLED="no"
Actually, looks like it's necessary to completely disable NetworkManager: # systemctl disable NetworkManager # systemctl stop NetworkManager Pretty inconvenient for laptop test environments, anyone aware of a workaround?
Is em1 attached to an OVS bridge? Can you provide the messages logfile to see what NM is doing?
This probably belongs to packstack. I guess there were some moves lately to avoid running NetworkManager with Openstack, right?
Ihar, Packstack is not touching and won't touch NM, but I will check if this is not a configuration problem.
Tested RHOS-4 on RHEL-6.5, tested RHOS-5 on RHEL-7 and tested RDO Havana on Fedora 20. I did not reproduced the issue in any of the mentioned setup.
The needinfo request[s] on this closed bug have been removed as they have been unresolved for 1000 days