In case this helps anyone - restarting the neutron_server container on all controllers, and then restarting the neutron_openvswitch_agent container on all compute nodes, solved the issue.
Some other services, such as nova_compute, had the same issue. Restarting all nova services on all controllers, and then nova services on all compute nodes, solved the issue.
In case this helps anyone - restarting the neutron_server container on all controllers, and then restarting the neutron_ openvswitch_ agent container on all compute nodes, solved the issue.
Some other services, such as nova_compute, had the same issue. Restarting all nova services on all controllers, and then nova services on all compute nodes, solved the issue.
Eric