Errors with Soft RAID 5 Array, no disks off line

[vamythguy]

I have a 4 disk soft RAID-5 array and I'm receiving a ton of these messages
in the system logs:

Dec 31 17:04:38 tibeaux smartd[2384]: Device: /dev/sda, 3 Offline
uncorrectable sectors
Dec 31 17:04:39 tibeaux smartd[2384]: Device: /dev/sdd, 48 Currently
unreadable (pending) sectors

But cat /proc/mdstat reports:

Personalities : [raid6] [raid5] [raid4]
md0 : active raid5 sda[0] sdd[3] sdc[2] sdb[1]
      735351936 blocks level 5, 64k chunk, algorithm 2 [4/4] [UUUU]

unused devices: <none>

Best way to fix this, or is it not a problem?

[Pedro Fernandes Macedo]

vamythguy wrote:
>  I have a 4 disk soft RAID-5 array and I'm receiving a ton of
> these 
>  messages in the system logs:
>
>  Dec 31 17:04:38 tibeaux smartd[2384]: Device: /dev/sda, 3 Offline 
>  uncorrectable sectors
>  Dec 31 17:04:39 tibeaux smartd[2384]: Device: /dev/sdd, 48 Currently 
>  unreadable (pending) sectors
>
>  Best way to fix this, or is it not a problem? 
It is a problem. This means that both
drives are starting to develop bad 
sectors. Since you have raid 5, you can replace the disk and let the 
array rebuild itself. Or keep using those disks, but add a hot spare 
disk to the array.
I usually replace the disks as soon as I see smartd complain about them, 
but if you wait too much , you risk loosing your data.. Last time I 
replaced a disk in such condition, it had over 1600 uncorrectable 
sectors, making it impossible to transfer the data to the new disk via 
DD, pvmove, cp... Luckily it was just a proxy/gateway/vpn server, so a 
reinstall wasnt difficult.

--
Pedro Macedo

[vamythguy]

On 12/31/06, Pedro Fernandes Macedo <webmaster(a)margo.bijoux.nom.br&gt; wrote:
>  vamythguy wrote:
>  > I have a 4 disk soft RAID-5 array and I'm receiving a ton of these
>  > messages in the system logs:
>  >
>  > Dec 31 17:04:38 tibeaux smartd[2384]: Device: /dev/sda, 3 Offline
>  > uncorrectable sectors
>  > Dec 31 17:04:39 tibeaux smartd[2384]: Device: /dev/sdd, 48 Currently
>  > unreadable (pending) sectors
>  >
>  > Best way to fix this, or is it not a problem?
>  It is a problem. This means that both drives are starting to develop bad
>  sectors. Since you have raid 5, you can replace the disk and let the
>  array rebuild itself. Or keep using those disks, but add a hot spare
>  disk to the array.
>  I usually replace the disks as soon as I see smartd complain about them,
>  but if you wait too much , you risk loosing your data.. Last time I
>  replaced a disk in such condition, it had over 1600 uncorrectable
>  sectors, making it impossible to transfer the data to the new disk via
>  DD, pvmove, cp... Luckily it was just a proxy/gateway/vpn server, so a
>  reinstall wasnt difficult.
>
>  --
>  Pedro Macedo
>
>  --
>  fedora-list mailing list
>  fedora-list(a)redhat.com
>  To unsubscribe: https://www.redhat.com/mailman/listinfo/fedora-list
>  

I suspected as much.  Two questions:

1) Do the different messages imply they are dieing in different ways?
2) Is there any way to salvage either drive, maybe by quarantining the bad
sectors or something?

[Bruno Wolff III]

On Sun, Dec 31, 2006 at 21:11:02 -0500,
  vamythguy <vamythguy(a)gmail.com&gt; wrote:
>  
>  1) Do the different messages imply they are dieing in different ways? 
The drive with only 3 bad sectors you may want to continue using, but the one
with 48, you should strongly consider replacing as fast as you can.

>  2) Is there any way to salvage either drive, maybe by quarantining
> the bad
>  sectors or something? 
If you write over the bad sectors, the drive will remap them to spare sectors.
This won't happen unless you either get a successful read or write over the
bad sectors.

In theory you might not have lost any data, but because the one drive is
offline you can't be sure you can know what value goes into the bad sectors
of the drive that is still in the array.

You should first try to back up what you have now. Then you can try to figure
out which files are possibly corrupt
(see: http://smartmontools.sourceforge.net/BadBlockHowTo.txt) and decide
if it is worth trying to fix them. You may be able to use the data on the
failed drive to fix the files.

Once you have receovered everything you can, you should run badblocks
on the drives (using a livecd is probably easiest) to try to find other
bad blocks. Then run smartctl -t long on the drives to see how bad things
are. A few reallocated blocks don't necessarily warrant tossing a drive.
You need to balance your budget versus the cost of replacing the data you
might lose and the extra likelyhood of failure suggested by a drive having
reallocated sectors.

For the future, you want to be running smartd so that you are warned about
bad sectors while only one drive has them so that you can repair the array
before there are two drives with errors.