Handling physically broken disk

If a disk is physically broken, it cannot be recovered by auto recovery or manual recovery. Do not keep broken disks in the file system and schedule to delete them from the file system.

Deleting disks when auto recovery is not enabled (check this by mmlsconfig restripeOnDiskFailure):

Deleting NSD disks from the file system can trigger disk or network traffic because of data protection. If your cluster is busy with application IO and the application IO performance is important, schedule a maintenance window to delete these broken disks from your file system. Follow the steps in the Starting the disk failure recovery section to check if a disk is physically broken and handle the broken disks.

Deleting disks when auto recovery is enabled (check this by mmlsconfig restripeOnDiskFailure):

When the IO operation is being performed on the physically broken disks, IBM Storage Scale marks the disks as non-functional. Auto recovery suspends the disks if it fails to change the availability of the disk to Up and restripes the data off the suspended disks. If you are using IBM Storage Scale 4.1.0.4 or earlier, deleting the non-functional disks triggers heavy IO traffic (especially for metadata disks). On IBM Storage Scale 4.1.0.4, mmdeldisk command has been improved. If the data on non-functional disks have been restriped, the disk status will be Emptied. The mmdeldisk command deletes the non-functional disks with the Emptied status without involving additional IO traffic.