Description of problem: When we have a baremetal node managed by OpenStack, it uses OVS for networking. Installing OpenShift seems to work to some degree, however uninstallation of OCP on this host will introduce breaking change with stopping/uninstalling Open vSwitch. Version-Release number of selected component (if applicable): Running against 3.5 when discovered using openshift-ansible, still an issue in current versions as well. How reproducible: Every time we uninstall OCP on a OpenStack managed baremetal node. Steps to Reproduce: 1. Have OpenStack managed cluster of baremetal nodes that are using OVS (as default) networking. 2. Install OpenShift on baremetal nodes 3. Uninstall OpenShift Actual results: Baremetal node drops due to OVS tasks in uninstall playbook. Expected results: OpenStack and OpenShift should be able to coexist.
Are you disabling the openshift SDN? If we simply made the installer not remove openvswitch when you've disabled the openshift sdn would that suffice?
The problem is that they use the same service, so stopping the service brings down both.
Ok, so if we just do nothing with openvswitch whenever openshift_use_openshift_sdn=false will that solve your problem?
We still do use the openshift sdn on top of openstack, so we are not disabling it.