S.M.A.R.T. list on disk3 is looking bad, but how bad is it really

Software 5.0-rc16c

3 HDD's with no parity drive (it's preclearing)

The 3 2TB disks are all connected to a user share.

I've notised that the content of disk 3 isn't showing in the share folder (not on my Win8 pc or on my WDTVlive box)

If i choose to view disk 3 i can see the content and even play the video files.

+

Roughly 4TB of data are divided on the 3 disks as follows (all disks are "highwater")

disk1 Browse /mnt/disk1 WDC_WD20EARX-00PASB0_WD-WMAZA9033613 (sde) 1953514552 27°C 2 TB 249.65 GB free

disk2 Browse /mnt/disk2 WDC_WD20EARS-00MVWB0_WD-WCAZA4671510 (sdd) 1953514552 29°C 2 TB 307.44 GB free

disk3 Browse /mnt/disk3 WDC_WD20EARS-00MVWB0_WD-WCAZA3663146 (sdc) 1953514552 *         2 TB 1.55 TB free      

As i understand highwater, the 4TB should be divided Disk1 1,5 TB Disk2 1,5 TB Disk3 1TB. (but that's not the case)

is seems like disk3 is the boy nobody want's to play with.

I've tried to....... Hold the phone.

This is embarresing.

In the share settings if put disk1:disk2:disk:3

And not disk1:disk2:disk3 as it should be.

Now i get the full content in the share folder view.

BUT is this the reason for the highwater error?

I've looked at the http://tower:8080/myMain?seq=24&view=smart&sort=num (mymain smart view)

And it tells me there is a "current pending sector=2" What is that? and can i fix it? and is that the reason for the wrong highwater level?

If you didn't allow disk3 for a while then it would have less data. The split level can also cause unequal data filling.

The pending sector is another issue you need to address. I believe you need to swap the drive out and run a preclear or something similar that writes to the drive and see if it clears and is stable.

It's getting worse for disk3  :'(

Current pending sectors=2

offline uncorrectable=2

multi zone error rate=43

I'm a newbie, so i don't know how bad this is.

I'm currently running the 'check parity run' (after the first parity run with the disk just mounted to the array)

After this i plan to preclear my 2 ST2000DM001 drives and get them in the array, would it be a good idea to transfer the content of disk3 on to my computer at take disk3 offline (or even out of the system for good)  and sign another drive in it's place?

From the un-official unRAID Manual:

http://lime-technology.com/wiki/index.php/Un-Official_UnRAID_Manual#Included_and_Excluded_disk.28s.29

"The included disks(s) parameter defines the set of disks which are candidates for allocation to that share. All disks may be used by the user share when the Included disk(s) parameter is left blank. Specify the disks to include here. For example, set the included disk(s) to "disk1,disk2,disk3" to allow the share to only use disk1, disk2 and disk3."

It looks like you want to include all your disks so you can leave the include disks field blank. However it says you should use a comma as the separator if you do want to specify what disks to include.

Sorry, I'm getting carried away with my markup 

Since you're not running with parity yet and there is some problem with your disk 3. I think you could just copy the data over to another disk for the time being to allow you to get it out of the system to run some tests or PreClear it.

As far a high water goes, I had mine screwed up for a long time and my system is very uneven. I seem to be make up the space slowly now that I have set a different level. Even so, I still don't quite get it! 

It may not seem like it, but you're on your way to becoming an unRAID expert!

I am running with parity, with no remarks from the system about sync errors or other bad stuff, so everything seems healthy.

But the s.m.a.r.t. list on disk3 is bugging me.

Pending sectors are unreadable and will interfere with unRAIDS parity protection. Any disk with pending sectors needs to be rebuilt. If you have a (pre-cleared recommended) spare replace disk3 with the spare and then run pre-clear on the disk. The pending sectors should be zero after the pre-clear and the disk becomes the new spare. Otherwise, rebuild the disk on to itself: Un-assign the disk and Start the array. Stop the array and assign the disk. Start the array and the rebuild will begin. After the rebuild the pending sectors should be zero.

(a)  Write a few GB of data to disk3;  then delete it [write directly to disk3, not to a share].

(b)  Now look at the SMART report and see if there are still any pending reallocations.  [Note:  a few reallocated sectors is fine;  but pending sectors should "go away" when you write to the disk.]

©  Assuming the SMART data is now fine, run a parity check and confirm all is okay (do a correcting check so if there are any glitches they'll be corrected --- if the result isn't zero errors, then run it again and confirm it's zero the 2nd time)

If the pending sectors count clears; and the parity check is okay; then you're fine.

The pending sectors are allocated sectors. Writing new files will not clear them The existing files need to be rewritten.

(a)  Write a few GB of data to disk3;  then delete it [write directly to disk3, not to a share].

 

(b)  Now look at the SMART report and see if there are still any pending reallocations.  [Note:  a few reallocated sectors is fine;  but pending sectors should "go away" when you write to the disk.]

Gary, I really appreciate the help you have given so many users, and the friendly and respectful way you do it.  But in this particular instance, you are not correct, as a Pending sector is only remapped or released by a direct write to that specific sector.  Since you usually have no idea where the Pending sectors are, it is extremely unlikely that writing a few gigabytes to the drive will hit them.  You either have to write to the entire drive (as Preclear does), or create a process using something like badblocks to create a list of all the questionable sectors and then rewrite them with dd or something else (losing their previous contents).

We have talked a few times about using the non-destructive write mode of badblocks to perform a SpinRite-like refresh of the disk surface.  It hasn't been tested, and it isn't fully documented, that I've been able to find.  One person claims it reads sectors, writes a pattern to them, reads that back, then writes the original contents back.  If true, that would be perfect, but I cannot find any confirmation.  So this would be experimental if anyone wants to try it.  I would certainly make sure all important files are backed up elsewhere first.  If you want to try it:    (keep in mind this is going to take a long time!)

* --- Pseudo SpinRite refresh of drive surface ---

* If you don't have a current SMART report, obtain one

* Make sure drive is not mounted.  If part of the array, make sure array is stopped

* At command line:  badblocks  -svn  /dev/sda    (change sda to appropriate drive)

* Obtain a final SMART report and compare

Rob & dgaschk ... Appreciate the info -- I didn't realize that a disk wouldn't update its pending reallocations without a specific refresh of those sectors that were pending.

Seems the best solution (at least for those of us who have Spinrite) is to simply run Spinrite on the drive 

But in the absence of that, the badblocks process seems like it should work okay.

However, in this case, I'd suggest something a bit different that avoids the need to work with the Linux commands and doesn't require stopping the array:    Since disk3 is sparsely populated (1.55TB free on a 2TB drive),  I'd simply copy ALL of the data from disk3 to a drive on another machine (perhaps your Windows PC, or an external drive);  then DELETE all of the files on disk3;  and then copy everything back.    This can be (in fact has to be) done with the array running, and certainly ensures that no data remains on a pending sector.  It may, in fact, not clear the pending sectors (if the data doesn't happen to be written to them) -- but the next time anything is written to those sectors they will be reallocated then, so they're no longer an issue.

Of course if KB36 happens to have a copy of Spinrite, I'd simply boot to that and let Spinrite refresh the entire disk.

Pending sectors will prevent the rebuilding of a different failed disk. If a disk has pending sectors then no other disk can be rebuilt. No disk can have any pending sectors in order for the parity system to work. Rebuilding the disk onto itself is the simplest method of clearing pending sectors. Rebuilding onto a spare and then running pre-clear on the disk is preferable.

The pending sectors have wanished.

I've been moving the files from my Readynas + Dlink box. I thourght that i've left out disk3 in all my user share. But it seem like one of them missed.

So about 500 GB of data was written to disk3. And now they are all gone.

all as in

Current pending sectors=2

offline uncorrectable=2

multi zone error rate=43

Gone.

I plan to copy the current files on disk3 to a external drive. Preclear the last 2 disks from my dlink nas and sign one of them to the disk3 spot.

Would that be ok, or is there another way?

If the pendings are all gone, then you can simply keep using the disk.  But if you'd prefer to use one of your other disks you can simply replace it and let the system rebuild the data onto that new disk.

... Copying everything somewhere else to back it up isn't a bad idea -- but as long as parity is good, a rebuild should work with no problem.