Abstract: A method and apparatus for managing customer topologies on packet networks are disclosed. For example, the method creates at least two event correlation instances for at least one customer topology, where a first event correlation instance resides in a primary availability management server, and a second event correlation instance resides in a secondary availability management server. The method also creates a test node for the first event correlation instance, where the test node provides at least one test message. The method then receives at least one response generated by the first event correlation instance that is responsive to the at least one test message, where the at least one response is received by the second event correlation instance. The method then performs a fail-over to the second event correlation instance from the first event correlation instance if a failure is detected from the at least one response.