i'm running a raid 5 array set up with mdadm on my fileserver. when one of the drives began to log uncorrectable sector errors, and reads were failing (burning programs would fail on some images on the disk, video was corrupted) the drive didn't drop out of the array. this brings to mind some questions i can't find the answers to:

- why didnt the drive drop out of the array?
- why doesn't the array use the parity to pass along good data to the reading applications instead of letting reads time out (?) or sending corrupted data? (when i manually failed the drive, good data came out of the degraded array, as i suppose you'd expect)

thanks in advance,
jonathan

hmmm, what do the logfiles say ?

maybe:

kernel: raid5:md0: read error corrected..........


I think disks only drop out of the array, if the whole device fails (is gone).