IBM Midrange System DS4000 Series Hardware Manual page 433

5. Check the data. If possible, avoid making any changes on the volume until it is clear that
there is no data corruption, that is, mount the file system as read-only and perform fsck or
chkdsk without attempting to fix errors. If data appears to be corrupt, then contact your
IBM Support representative before taking any further actions.
6. After confirming data integrity, we can reconstruct the remaining failed disk(s) in the array.
When complete, the associated logical drives return to an Optimal state.
The Reconstruct option is available through the Storage Manager Subsystem
Management window by highlighting the drive and then selecting Advanced 
Recovery  Reconstruct Drive.
Reconstructing a failed disk results in data being regenerated on it from the remaining
disks in the RAID array.
7. Reassign the hotspare drives.
Note: Extreme care should be taken when selecting the Revive Drive option on an
assigned disk. The function behaves differently depending on whether another drive is
sparing for the drive being revived. If a hotspare has taken over, then reviving the drive will
force a copyback to start. If there is no hotspare in use, then it will return the disk into an
Optimal state as a member of the array, irrespective of whether it contains valid data or not.
Sometimes, following a power or channel incident, some disks return to a Replaced state
even though there is no reconstruction or copyback in progress. In this case, the disk needs to
be changed to a Failed state before it can be revived or reconstructed.
To fail a disk through the Storage Manager Subsystem Management window, highlight the
drive and then select Advanced  Recovery  Fail Drive.
Checking RAID redundancy
We have already mentioned earlier in the Preventative Maintenance section that Media Scan
includes an option to perform a RAID redundancy check. By default, the redundancy check is
normally disabled for all logical drives. If enabled, Media Scan schedules to complete the
check as a low priority task within the preset duration period (default of 30 days). However,
there is also an option to perform an immediate data redundancy check on an array from
Storage Manager.
Chapter 7. Advanced maintenance, troubleshooting, and diagnostics
415