| 摘要 |
Methods are provided for detecting and correcting dropped writes in a storage system. Data and a checksum are written to a storage device, such as a RAID array. The state of the data is classified as being in a“new data, unconfirmed”state. The state of written data is periodically checked, such as with a timer. If the data is in the“new data, unconfirmed”state, it is checked for a dropped write. If a dropped write has occurred, the state of the data is changed to a“single dropped write confirmed”state and the dropped write error is preferably corrected. If no dropped write is detected, the state is changed to a“confirmed good”state. If the data was updated through a read-modified-write prior to being checked for a dropped write event, its state is changed to an“unquantifiable”state.
|
