When the database detects an internal error that prevents it from continuing, it stores diagnostic information and reinitialize itself. This event is called a crash (or reset). After the database reinitializes, it rolls back any incomplete transactions or requests in progress. This process is called a recovery.
When an AP on an AP/1012 or 3600 system is taken down or encounters an internal error that prevents it from continuing, it shuts down. This event is called an AP reset. Any incomplete transactions or requests in progress for sessions connected through that AP are rolled back. Sessions connected through any other APs are unaffected.