This test reproduces a customer-reported problem where our Neutron
driver's Felix status watching was not resilient to the etcd server
being restarted. The impact of that was that the Neutron driver
thought (after a following timeout period) that all the Felixes were
dead, and so refused to schedule any new VM instances.
This test reproduces a customer-reported problem where our Neutron driver's Felix status watching was not resilient to the etcd server being restarted. The impact of that was that the Neutron driver thought (after a following timeout period) that all the Felixes were dead, and so refused to schedule any new VM instances.