So, what I think is fact that when port is initially added by os-vif, it triggers OvsdbMonitor event: http://logs.openstack.org/32/602432/3/check/tempest-full/4b76af0/controller/logs/screen-q-agt.txt.gz#_Jan_16_21_35_54_776433 and such port is added to events['added'] list thus neutron-openvswitch-agent process this port.
That is causing issues and some errors in logs as port is actually not configured yet (ofport=-1) so it isn't actually processed. But for some reason it's indeed added to port_info["current"] and if it happens, such port will be not processed later probably as it already is in "current" list.
I think that we should skip events for ports with ofport=-1, there is no need to trigger ovs agent when such port is added to bridge IMO.
So, what I think is fact that when port is initially added by os-vif, it triggers OvsdbMonitor event: http:// logs.openstack. org/32/ 602432/ 3/check/ tempest- full/4b76af0/ controller/ logs/screen- q-agt.txt. gz#_Jan_ 16_21_35_ 54_776433 and such port is added to events['added'] list thus neutron- openvswitch- agent process this port. "current" ] and if it happens, such port will be not processed later probably as it already is in "current" list.
That is causing issues and some errors in logs as port is actually not configured yet (ofport=-1) so it isn't actually processed. But for some reason it's indeed added to port_info[
I think that we should skip events for ports with ofport=-1, there is no need to trigger ovs agent when such port is added to bridge IMO.