System recovery re-establishes the state of each business process as it was when a stop or termination of the runtime instance occurred. When the runtime instance is restarted, the recovery process identifies the following state information when the engine went down:
Which business processes have been started because of the reception of a notification event but have not finalized all activities for the business process
What activities within a process have occurred
What is the state of each activity that is in progress
What activities within a process have not started yet
When the runtime instance is started up again, the recovery process reads in the business process information applicable to the open processes, routes any events that were previously captured but not routed, and routes any additional events to the open business processes. This enables completion of the business processes that were initiated by incoming event messages.
Warning! The recovery process can only be performed successfully on the same package that is present on the runtime instance when the engine goes down. If a new package is deployed to the runtime instance, recovery cannot be guaranteed.
Note: Automatic start up of the engine assumes you have not lost data in the input queues.
When an engine goes down, its state may be somewhere within the capture/route or analysis transaction boundaries.