Preclear not successfull [newbie]

[niven]

Hello.

 

I ran a preclear the other day on six hard drives, some of the drives didn't successfully preclear.

 

I bought two new Seagate ST3000DM001 3TB drives, they precleared successfully! the other four drives; Samsung SpinPoint F4EG 2TB all failed the preclear.

the preclear command i used was;

 ./preclear_script.sh -A /dev/sdX.

 

I attached the Samsung preclear results and the Seagate results, I don't know what to look for so if more info is needed I can provide the entire syslog.

 

The Samsung drives were bought in 2010-2011, and i have used them with a storage-box from ONNTO Data-Tale with RAID 5.

 

I'm new to this and I don't know if it is the drives that doesn't work, or if the storage box made some settings or did something to the data on the drives.

Anyways i attached the results for one failed drive and one successful drive , so if someone can see what goes wrong or have any ideas etc. That would be really great !

 

- niv

preclear_results.txt

[binhex]

hi niv, ok firstly your using the right command to pre-clear those disks, some people do more than 1 cycle but that is a personal preference.

 

basically the pre-clear report is telling you that it expected to read all zeros in the post-clearing-read phase.  It did not, this could be due to a bad disk, or a bad disk controller port, or bad RAM in the server, or even a marginal power supply that allows noise on its supply lines to the disk to influence the returned values.

 

so my course of action would be as follows:-

 

- you need to upgrade the firmware on those samsung drives as they have a known issue with the older firmware causing data corruption, so you need to get them up to date if they aren't already, this is def priority #1

 

- check memory on the server using memtest (built into unraid, can be accessed on boot) let it run at least one full cycle

 

- switch the controller the samsung disks are connected to and re-run pre-clear - could be a faulty controller

 

- change the psu to rule this out

[niven]

Thanks for the reply binhex.

 

some people do more than 1 cycle but that is a personal preference.

Oh i thought one cycle would be enough since it takes about 20 hours +, but would the second cycle be faster?

 

- check memory on the server using memtest (built into unraid, can be accessed on boot) let it run at least one full cycle

Im running the memtest now, what should i look for ? do i need to do anything else than just let it run? I tried to read up on it in the wiki but i couldn't for some reason find the section... (bad eyes I guess).

 

Edit: Ran one cycle now and it says: Pass complete, no errors.

        Waiting for second cycle to complete.

 

- you need to upgrade the firmware on those samsung drives as they have a known issue with the older firmware causing data corruption, so you need to get them up to date if they aren't already, this is def priority #1

I don't know if the firmware on those drives are up to date, but i will try to update the firmware and see if that has any effect. Its not only one drive that failed but four... So I don't think that's a coincidence. The other Seagate 3tb drives precleared just fine.

 

- change the psu to rule this out

Wouldn't this affect the Seagate drives if this were the problem? Since they precleared just fine is it safe to rule this one out?

[Joe L.]

Thanks for the reply binhex.

 

some people do more than 1 cycle but that is a personal preference.

Oh i thought one cycle would be enough since it takes about 20 hours +, but would the second cycle be faster?

no, it would be the same.    Google bathtub-curve to learn why many prefer multiple pre-clear cycles  (I would do one at a time so you can judge how it is going after each pass)

 

- check memory on the server using memtest (built into unraid, can be accessed on boot) let it run at least one full cycle

Im running the memtest now, what should i look for ? do i need to do anything else than just let it run? I tried to read up on it in the wiki but i couldn't for some reason find the section... (bad eyes I guess).

Let it run overnight if possible, expect NO errors.  Any memory errors and all bets are off for anything else on the server.

- you need to upgrade the firmware on those samsung drives as they have a known issue with the older firmware causing data corruption, so you need to get them up to date if they aren't already, this is def priority #1

I don't know if the firmware on those drives are up to date, but i will try to update the firmware and see if that has any effect. Its not only one drive that failed but four... So I don't think that's a coincidence.

If they have buggy firmware on the disks themselves, it could be the issue.  I would not trust them for any data if they cannot be zeroed.

The other Seagate 3tb drives precleared just fine.

 

- change the psu to rule this out

Wouldn't this affect the Seagate drives if this were the problem? Since they precleared just fine is it safe to rule this one out?

A poor power quality supply, or one overloaded with the newly added drives could cause errors.  Optimally you want a single rail high amperage power supply.  What specific power supply are you using? 

 

Additionally, if you were pre-clearing multiple drives at the same time it could be you ran out of free "low" memory.  You should have used the "-r" "-w" and "-b" options to the preclear script to limit their memory usage.  Lastly, some disk controllers have been reported t have issues with multiple concurrent disk use.  You might have one of those that have buggy firmware/hardware.

 

In any case, the errors in the post-read phase would only show as random errors in parity checks.  (and you'll pull your hair out trying to figure those out)

[binhex]

nope im afraid a second cycle would be just as fast as the first one, so yes your talking considerable amounts of time to do multiple pre-clear cycles, but some people swear by it doing either 2 or 3 cycles on every drive before trusting the drive 100%.

 

on the memtest your really just looking for any errors, if i remember rightly it will be quite obvious if you get errors as i think the text comes up red.

 

i cannot emphasis strongly enough that you really need to make 100% you are running the latest firmware on those samsung drives, it is very possible that the preclear issues you are seeing are due to drives running an older firmware version. the firmware is difficult to find nowadays as those drives are not manufactured any more and seagate has now taken over samsungs drive production, i have pushed up a copy of the firmware i use, as i have the same make and model drive:-

 

http://sharesend.com/tjf6rg79

 

your right that it probably would of affected the seagate drives if it was the psu, however it is still a possibility, so just bear it in mind, its the last thing to change if nothing else works and your 100% sure the drives are ok.

[niven]

Thanks for the reply Joe L.

 

no, it would be the same.    Google bathtub-curve to learn why many prefer multiple pre-clear cycles  (I would do one at a time so you can judge how it is going after each pass)

Will do! Really want to learn as much as possible since I've heard a lot of good things with unRAID, and I want to use it for the future!

 

Let it run overnight if possible, expect NO errors.  Any memory errors and all bets are off for anything else on the server.

Ok, I will let it run over the night then, i just need to [ESC] and it will reboot? I don't need for it to finish or anything?

 

If they have buggy firmware on the disks themselves, it could be the issue.  I would not trust them for any data if they cannot be zeroed.

Yes i think they might have buggy firmware.. I actually ordered some new 3 TB from Seagate and WD now to see if they will work. It's not the end of the world if I can't use those Samsung drives I will find some other use for them maybe put them back into the storage box and use them for minor backups etc Also it would be more comforting to use disks that aren't known to have issues.

 

What specific power supply are you using? 

Im using this one: CORSAIR HX 650W PSU ATX12V V2.31, 80 Plus Gold, Modular, 2x 6+2-pin PCIe, 8x SATA, 4x Molex its brand new.

 

In any case, the errors in the post-read phase would only show as random errors in parity checks.  (and you'll pull your hair out trying to figure those out)

don't want to loose any hairs, I need them all while im still young

[niven]

Thanks for the reply binhex.

 

nope im afraid a second cycle would be just as fast as the first one, so yes your talking considerable amounts of time to do multiple pre-clear cycles, but some people swear by it doing either 2 or 3 cycles on every drive before trusting the drive 100%.

Ok good to know I guess it's better to be safe than sorry.

 

i cannot emphasis strongly enough that you really need to make 100% you are running the latest firmware on those samsung drives, it is very possible that the preclear issues you are seeing are due to drives running an older firmware version. the firmware is difficult to find nowadays as those drives are not manufactured any more and seagate has now taken over samsungs drive production, i have pushed up a copy of the firmware i use, as i have the same make and model drive:-

 

http://sharesend.com/tjf6rg79

 

Thanks! I will fix them to the latest firmware.

[itimpi]

Im using this one: CORSAIR HX 650W PSU ATX12V V2.31, 80 Plus Gold, Modular, 2x 6+2-pin PCIe, 8x SATA, 4x Molex its brand new.

That is a good Single rail power supply that can power a reasonable number of drives (I would guess around 20) without issue.

[niven]

So the Memtest has completed over the night; 18 cycles with no error, is it safe to presume there is no memory errors?

 

You should have used the "-r" "-w" and "-b" options to the preclear script to limit their memory usage.

 

also another question when preclearing future drives; do i just add all of the "-r" "-w"  "-b" to the command I used? will it be something like this?

 

./preclear_script.sh -A -r -w -b /dev/sdX.

 

I found this in the wiki page about preclear;

If you are pre-clearing multiple drives it is recommended you also use the optional "-r" "-w" and "-b" options. These set the size and number of blocks of data read and written to the disks during processing. Although not needed when clearing a single drive, If you do not use these options the preclear script may run out of memory when concurrently processing multiple large drives. Suggested values are

preclear_disk.sh -r 65536 -w 65536 -b 2000 -A /dev/sdX

is this what i should use for multiple drive preclears? are the values standard or do i need to adjust them? Also the command I used was ./preclear_script.sh, is this the same as ./preclear_disk.sh?

 

So the script I should use to preclear my drives would be; 

./preclear_disk.sh -r 65536 -w 65536 -b 2000 -A -c 2 /dev/sdX

if i wanted to do 2 cycles on let's say 3 new drives? Will this prevent out of "low" memory issues?

 

Also the Samsung drives I have were manufactured before December 2010, so they have not been upgraded to the latest firmware. I guess this is most likely what's causing the issues. Or like Joe L. suggested that I ran out of "low" memory since i precleared six drives this way.