Fault Management

The current neutron implementation has several high availability and
rescheduling features to enhance its reliability, availability and
serviceability. However, it does not implement a mechanism to directly report
stateful fault information via a fault management interface, which limits
reporting and automated recovery capabilities.

This spec proposes to add a fault management driver to be implemented in
neutron that will perform the OS-level fault management implementation to
generate alarms, report issues when something goes wrong, and also to get,
clear the fault states for ports, agents, hosts, provider networks etc. This
enables another external entity to monitor the faults of different severity
levels occurred in neutron and make fault handling/scheduling decisions for the
related services.