The present invention relates to an apparatus and method for monitoring and correcting data errors in a computer system, in particular transient data errors in computer systems having very limited tolerance for deteriorations in performance. The method comprises the steps of: writing a set of data to a plurality of separate databases (50, 80), the databases including at least a first database (50) and a second database (80), reading the set of data from the first database and applying an error correction code (ECC) to detect possible multiple bit errors in the set of data which was read from the first database, and in case an error is detected, the error is reported to an error management unit (30). Furthermore, the error management unit manages a switch of means from which data are read from the first database to the second database, and applying the error correction code for correcting single bit errors in the previously read set of data. The writing means corrects data in the first database by writing the set of data from the second database to both of the first and second databases, and reports to the error management unit that the error is corrected.