Abstract: Provided are methods, data processing systems, recovery components and computer programs for recovering from storage failures affecting data repositories. At least a part of the recovery processing is performed while the data repositories are able to receive new data and to allow retrieval of such new data. Although new data items may be received into the repository and retrieved therefrom during recovery processing, updates to the data repository which were performed before the failure and which are then restored to the repository by the recovery processing are restored within a recovery unit of work and are inaccessible to processes other than the recovery process until successful completion of the recovery unit of work. The recovery processing ensures that the recovered repository is consistent with the state of the repository at the time of the failure, but is available for addition and retrieval of new data items before completion of the recovery processing.

Abstract: A method of moving records, stored on a recovery log, which are associated with units of work which have waited for a long time is described. Units of work which have waited for a long time, such as those which are in-doubt have their records copied from a recovery log to a secondary log. When the earliest-written record for this in-doubt unit of work has been copied, a new record called a fork record is written on to the recovery log which has two effects: to subtract the log records currently on the recovery log for this unit of work from the recovery log and to refer to the copied records written on the secondary log. The new record achieves this using two pointers, a pointer to point to the most-recently written record on the recovery log for this unit of work and a pointer to point to the most-recently written record on the secondary log for this unit of work.