Abstract: A data processing system having a pair of mirrored storage units maintains a state record of the mirrored pair in system memory. In order to be able to determine state when the system is re-initialized, this state information is also stored on each storage unit of the mirrored pair, and in an alternate location. When the state changes, the operating system writes the new state to those storage units which are still functioning, and to the alternate location. In order to prevent ambiguous situations, only certain defined state transitions are permitted. When the system is re-initialized, it attempts to read the state information stored on the storage units. If either unit can not be read, the system substitutes the state retrieved from the alternate state record for the state that would have been read from the non-responding unit. This pair of states from the two units index an unique entry in a state derivation table containing the resultant state.