Software RAID (mdadm) - RAID 0 returns incorrect status for disk failure/disk removed
 

Problem Description: mdadm does not update the /proc/mdstat when a drive fails or removed for RAID 0 case. When mdadm --detail /dev/md0 is executed, the information is not updated. It is as if there were no drive failures or drive removal.

I have tried testing drive fail/drive removal for RAID 5 and /proc/mdstat and mdadm --detail returns the expected result (it is displayed in /proc/mdstat that the drive failed).

I would like to ask why the behavior of drive failure/drive removal is not as expected in RAID 0 case.

Ideas deeply appreciated.

My suspicion is that this is because a RAID0 can suffer no failures.

When one disk of a RAID1 or RAID5 fails, the remaining drives continue operating normally. But as soon as one drive of a RAID0 fails, Linux can no longer perform any operations on the array.

If Linux does nothing, instead of marking the drive as failed, then there is a chance you can repair/replace/recover the failed drive and restart the array.

Thanks for the reply.

If this is the case, how can the user check what drive failed or confirm the status of the RAID 0 when either one of its drives fails? Is there a way wherein a DeviceDisappeared event or Fail event will be sent through mdadm --monitor --mail when a RAID 0 mdadm array fails or a way wherein the /proc/mdstat file will be updated?
I am using mdadm v2.6.7.1 (ubuntu distribution)

Yes, you're exactly right. Running mdadm in monitor mode should produce a DeviceDisappeared event. See this section of the man page:


http://man-wiki.net/index.php/8:mdadm#MONITOR_MODE

Thanks. Got the DeviceDisappeared working.