Repairing Broken Sectors

1. Mark the device as failed in the array
   

   mdadm <raid dev> -f <problem device>

2. Remove the failed device from the array
   

   mdadm <raid dev> -r <problem device>

3. Make the drive read all its sectors to detect all the problems
   

   dd if=<problem device> of=/dev/null conv=noerror

4. With the problems known to the drive, erase all its data so it can do sector re-mapping
   

   dd if=/dev/zero of=<problem device> bs=8192

5. Verify the drive has no bad sectors now
   

   smartctl -t long <problem device>

6. Wait for test completion
   

   smartctl -l selftest <problem device>

7. Re-create the partition table
   

   parted <known good RAID member> unit s print  # Learn what a correct geometry looks like
   parted <problem device> mklabel gpt mkpart primary <start>s <end>s  # Configure the correct geometry

8. Make sure the system is aware of the newly created partition
   

   partprobe <problem device>

9. Add the device into the array
   

   mdadm <raid dev> -a <problem device>

10. Monitor RAID rebuild progress
    

    cat /proc/mdstat

Forcing Drive Detection

1. You have to know which SCSI host (bus) to scan, but once that is known:
   

   echo "- - -" > /sys/class/scsi_host/host#/scan