Hi,

 

I randomly run "smartctl -H" command on the console for all the drive in the array and smartctl reports one of my drive is going to fail. Based on the unRAID main screen, everything is green. Should I replace the drive right now? It seems to me that unRAID will not show any warning until the drive fails in certain way? Does unRAID use any information from smartctl?

 

Here is the screen capture:

 

root@Tower:/boot/packages# smartctl -H /dev/sdc

smartctl version 5.38 [i486-slackware-linux-gnu] Copyright © 2002-8 Bruce Allen

Home page is http://smartmontools.sourceforge.net/

 

=== START OF READ SMART DATA SECTION ===

SMART overall-health self-assessment test result: FAILED!

Drive failure expected in less than 24 hours. SAVE ALL DATA.

Failed Attributes:

ID# ATTRIBUTE_NAME          FLAG    VALUE WORST THRESH TYPE      UPDATED  WHEN_FAILED RAW_VALUE

  5 Reallocated_Sector_Ct  0x0033  013  013  036    Pre-fail  Always  FAILING_NOW 3499

 

 

Thanks,

--Tom

YES.

 

You should disconnect or unassign the disk from the array.  If you do it correctly, the disk will still APPEAR to be present with all of its data intact, when in truth its contents are being simulated using the other disks in the array + parity.

 

I'd likely keep the machine powered down until I got a replacement, but the truth is the risk is likely about the same either way (keeping it running or keeping it turned off).

 

Replace the disk with one >= the size of the failed disk.

 

 

unRAID will show errors when it is unable to read a given disk.  But, it will reconstruct the data from parity+other data disks and then write the data back to where it was unable to read.  If the write is successful, it will not take the drive out of service.  Behind the scenes, on the disk itself, the SMART software will mark a sector it is unable to read as bad and "pending reallocation" on the next "write" to that sector.  When that occurs, a different sector is written to, and the SMART report shows an increment to the "Reallocated Sector Count". Since the reconstructed data was written back to the drive, and it was apparently successful, you are fine until the drive runs out of spare sectors to re-assign for ones it finds are bad.

 

Many drives have no reallocated sectors, some only have a few and the number never changes over time.  Yours apparently has had 3499.

 

unRAID will take a drive out of service when a "write" to it fails.  Other than that, it will just show a "read" error count, fix the error by writing the data back to the disk, and continue to serve your files.  Apparently, so far, none of the "writes" have failed. .. but... you are exhausting the pool of spare sectors on the disk available for re-allocation. 

 

Only the manufacturer knows how big that pool of spare sectors might be.  It might be 3500.  It really does not matter, you are above the threshold they think is safe to continue to use the disk.  They think failure is imminent.

 

Do you have any sectors "Pending Reallocation" ?  Those would be where a bad sector was identified (could not be read), but could not be reallocated            (either no write has subsequently occurred to that same sector, or no more free sectors exist).

 

As already advised, you should un-assign or disconnect that drive. Then purchase a new drive and assign it to the same slot as soon as possible.  If you have any critical data on that drive, copy it to another drive now.    When you replace the drive with a new drive, the old drive's data will be re-built into the new drive when you press the "Start" button.  Whatever you do, DO NOT press the "restore" button.  It would cause the data on that un-assigned drive to be forgotten and cause a full parity check to be started based on the remaining drives. 

 

Joe L.

Hi,

 

Thanks for the writeup and explanations. I will replace the hard drive and try to send it back to manufacturer since it is still under warranty.

 

--Tom

For extra protection in case anything goes wrong in the rebuild, I'd suggest hanging onto the failing drive until the rebuild completes, you are able to run a parity check successfully, and you are confident all is well.

Hi bjp999,

 

Thanks a lot. That's a good advice. I sure have time to hang on the disk before I finish rebuilding the array.

 

--Tom