A method and apparatus used in a storage network facilitates the protection of data in, and replacement of, storage devices about to fail before the failure happens. In a network that includes a set of storage devices organized as a non-redundant (for example RAID 0) array, a storage device about to fail in the non-redundant array can be replaced by another storage device, typically from a pool of spares. The method includes detecting a condition of a first particular storage device in the non-redundant array. Conditions which are detected according to various embodiments indicate that the first particular storage device is suffering events indicating that it is likely to fail, or otherwise suffering from reduced performance. The conditions are detected for example, by the receipt of a signal from the storage device itself, or by the monitoring of statistics concerning the performance of the storage device. The method further provides for selecting a particular spare storage device, which can be used in place of the first particular storage device. In response to detecting the condition, data stored in the first particular storage device is migrated to the second particular storage device, and the second particular storage takes the place of the first particular storage device in the non-redundant array. The first particular storage device can then be gracefully removed from the network without loss of service to the clients computers.